UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Investigating the epidemiology and work disability impacts of anxiety and depression disorders after… Jones, Andrea 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2020_may_jones_andrea.pdf [ 1.76MB ]
Metadata
JSON: 24-1.0387452.json
JSON-LD: 24-1.0387452-ld.json
RDF/XML (Pretty): 24-1.0387452-rdf.xml
RDF/JSON: 24-1.0387452-rdf.json
Turtle: 24-1.0387452-turtle.txt
N-Triples: 24-1.0387452-rdf-ntriples.txt
Original Record: 24-1.0387452-source.json
Full Text
24-1.0387452-fulltext.txt
Citation
24-1.0387452.ris

Full Text

INVESTIGATING THE EPIDEMIOLOGY AND WORK DISABILITY IMPACTS OF ANXIETY AND DEPRESSION DISORDERS AFTER MUSCULOSKELETAL INJURY USING LINKED HEALTH DATA  by  Andrea Marie Jones  M.Sc., McMaster University, 2010  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Population and Public Health)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  December 2019 © Andrea Marie Jones, 2019   ii The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: Investigating the epidemiology and work disability impacts of anxiety and depression disorders after musculoskeletal injury using linked health data  submitted by Andrea Marie Jones in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Population and Public Health  Examining Committee: Dr. Mieke Koehoorn Supervisor  Dr. Christopher McLeod Supervisory Committee Member  Dr. Ute Bültmann   Supervisory Committee Member Dr. Catherine Backman University Examiner Dr. Mary De Vera University Examiner  Additional Supervisory Committee Members:  Supervisory Committee Member  Supervisory Committee Member   iii Abstract  Anxiety and depression disorders are common after lost-time musculoskeletal work injury, and may be a modifiable determinant of work disability. Despite this, little evidence exists on the epidemiology or impacts of mental health in the work injury population. This dissertation generated new information on the descriptive epidemiology and work disability impacts of anxiety and depression disorders among workers with lost-time musculoskeletal work injury. Accepted lost-time claims for spine or upper limb strain or sprain work injury were extracted for workers in the Canadian province of British Columbia from 2000 to 2013. Anxiety and depression diagnoses and physician mental health services were identified using physician billing, hospital discharge and prescription data. Analytic approaches included prevalence estimates; multinomial, Cox, and instrumental variable regression models; direct adjusted survival curves; and model stratification by gender. Approximately 1 in 10 men and 3 in 10 women had a recent or current anxiety or depression disorder at the time of injury. Both pre-existing and new onset anxiety and depression disorders were associated with a lower probability of sustained return to work, and pre-existing anxiety was associated with a higher probability of lost-time recurrence after initial return to work. The effect size of the association between pre-existing anxiety and sustained return to work was greater for men than for women, and pre-existing depression was associated with a lower probability of sustained return to work for men but not women. Lastly, instrumental variable methods offer a promising analytic approach to estimate the effects of mental health treatment, based on physician treatment preference, on work disability outcomes for workers with a mental disorder. These findings offer important and novel contributions to our understanding of the prevalence, impacts, and management of anxiety and depression disorders in a lost-time musculoskeletal work injury context. Workers’ compensation benefits and programs intended to improve return to work after musculoskeletal injury should take pre-existing anxiety and depression disorders into consideration, in addition to new onset disorders attributable to the injury; and, gender specific strategies may be warranted to optimize return-to-work outcomes.   iv Lay Summary Anxiety and depression disorders are common after musculoskeletal work injury. This is concerning as anxiety and depression are the leading causes of disability worldwide. The purpose of this research was to describe the occurrence of anxiety and depression disorders among workers with musculoskeletal work injury, and to investigate the impacts of these disorders on work disability outcomes. Approximately 1 in 10 men and 3 in 10 women were found to have a recent or current anxiety or depression disorder at the time of musculoskeletal work injury, and both pre-existing and new onset anxiety and depression disorders were associated with longer work disability. Collectively these results support the inclusion of anxiety and depression disorders as part of the disability management plan for workers with musculoskeletal work injury.     v Preface The work presented here is that of the candidate (Andrea Marie Jones) with supervisory guidance from members of the dissertation committee (Drs. Mieke Koehoorn, Chris McLeod, and Ute Bültmann). Identification, design, and performance of the research program The candidate was the lead researcher responsible for the identification of the research objectives; review of the literature; design of the studies; preparation, manipulation, and analyses of the data; interpretation of the results; and writing of the first and final versions of the chapters. All members of the dissertation committee provided input at all stages of the research process from initial conception to editing the final version of the dissertation. Contributions from additional collaborators are as follows: 1) stakeholders from WorkSafeBC provided input on the development of the research objectives and interpretation of the findings, 2) Ms. Suhail Marino provided assistance with preparing the Data Access Request to Population Data BC, 3) Population Data BC served as a research liaison, securing permission and data access from data stewards at the British Columbia Ministry of Health, PharmaNet, and WorkSafeBC, and 4) Ms. Lillian Tamburic provided assistance with preparing the data for analyses.  Ethics approval Ethics approval for this research was obtained from the University of British Columbia’s Behavioral Research Ethics Board (H15-02150). Disclaimer All inferences, opinions, and conclusions drawn in this dissertation are those of the author, and do not reflect the opinions or policies of the Data Stewards.     vi Table of Contents Abstract .......................................................................................................................................... iii Lay Summary ................................................................................................................................. iv Preface............................................................................................................................................. v Table of Contents ........................................................................................................................... vi List of Tables ............................................................................................................................... viii List of Figures ................................................................................................................................ xi Acknowledgements ....................................................................................................................... xii Chapter 1: Introduction ................................................................................................................... 1 1.1 General overview ........................................................................................................... 1 1.2 Rationale and purpose .................................................................................................... 2 1.3 Research objectives ........................................................................................................ 3 1.4 Study design overview ................................................................................................... 4 1.5 Thesis outline ................................................................................................................. 4 Chapter 2: Background ................................................................................................................... 5 2.1 Musculoskeletal conditions ............................................................................................ 5 2.2 Strain or sprain work injury ........................................................................................... 5 2.3 Conceptual models of disability and work disability ..................................................... 6 2.4 Mental Health............................................................................................................... 13 2.5 The Importance of sex and gender to the research objectives ..................................... 21 2.6 Conceptual framework, rationale, and research hypotheses ........................................ 23 Chapter 3: Databases and case definitions .................................................................................... 30 3.1 Jurisdictional Context .................................................................................................. 30 3.2 Data files ...................................................................................................................... 31 3.3 Data Extraction, Linkage, and Cleaning ...................................................................... 34 3.4 Anxiety and depression case definitions ...................................................................... 37 Chapter 4: Prevalence, timing, and risk factors of anxiety and depression (Research Objective 1)....................................................................................................................................................... 56 Background ........................................................................................................................... 56 4.1 Methods........................................................................................................................ 57 4.2 Results .......................................................................................................................... 63 4.3 Discussion .................................................................................................................... 82   vii Chapter 5: Impacts of pre-existing and new onset anxiety and depression on sustained return to work (Research Objective 2) ........................................................................................................ 90 5.1 Background .................................................................................................................. 90 5.2 Methods........................................................................................................................ 93 5.3 Results .......................................................................................................................... 97 5.4 Discussion .................................................................................................................. 106 Chapter 6: Impacts of pre-existing anxiety and depression on return to work and lost-time recurrence events (Research Objective 3) ................................................................................... 112 6.1 Background ................................................................................................................ 112 6.2 Methods...................................................................................................................... 113 6.3 Results ........................................................................................................................ 118 6.4 Discussion .................................................................................................................. 124 Chapter 7: Impacts of physician mental health services on return to work for workers with pre-existing anxiety (Research Objective 4) ..................................................................................... 128 7.1 Background ................................................................................................................ 128 7.2 Methods...................................................................................................................... 131 7.3 Results ........................................................................................................................ 139 7.4 Discussion .................................................................................................................. 150 Chapter 8: Discussion ................................................................................................................. 157 8.1 Summary of results and contributions ....................................................................... 157 8.2 Methodological limitations and considerations ......................................................... 161 8.3 Implications and considerations for future research, policy, and practice ................. 163 8.4 Future research recommendations ............................................................................. 166 8.5 Summary .................................................................................................................... 169 References ................................................................................................................................... 170 Appendix ..................................................................................................................................... 191    viii List of Tables Table 1: Anxiety and depression case definitions previously applied to Canadian health claims data ................................................................................................................................................ 40 Table 2: Case definition for anxiety disorders in the Ministry of Health datasets ....................... 49 Table 3: Case definition for depression cases in the Ministry of Health datasets ........................ 50 Table 4: Anxiety and depression case status for the year prior to injury: A comparison of the primary and alternative case definitions (column percentages reported) using BC claims for spine or upper limb strain or sprain, 2000-2013 .................................................................................... 54 Table 5: Socio-demographic, injury, clinical and work factors among lost-time upper limb or spine strain or sprain claims in BC from 2000 to 2013 ................................................................ 64 Table 6: Prevalent anxiety and depression cases by gender and time period (row percentages reported) among lost-time upper limb or spine strain or sprain claims in BC from 2000 to 2013 66 Table 7: The proportion of anxiety and depression cases prevalent in the year before lost-time upper limb or spine strain or sprain work injury also prevalent in the year after injury, and vice versa in BC from 2000 to 2013 ..................................................................................................... 67 Table 8: Adjusteda odds ratios (OR) with 95% confidence intervals (CI) for risk factors of prevalent anxiety and depression in the year before lost-time upper limb or spine strain or sprain work injury in BC from 2000 to 2013, multinomial regression ................................................... 71 Table 9: Adjusteda odds ratios (OR) with 95% confidence intervals (CI) for risk factors of new onset anxiety and depression in the 3 months after lost-time upper limb or spine strain or sprain work injury in BC from 2000 to 2013, multinomial regression ................................................... 76 Table 10: Summary table depicting the adjusteda associations between risk factors and anxiety and depression disorders that were i) prevalent in the year before lost-time upper limb or spine strain or sprain work injury and ii) new onset in the three months following injury in BC from 2000 to 2013 ................................................................................................................................. 79 Table 11: Summary of review articles that examined the role of anxiety and depression as risk factors for return to work outcomes after musculoskeletal injury ................................................ 91 Table 12: Final study sample of lost-time upper limb or spine strain or sprain work injury claims in BC by anxiety and depression case status in the year before injury and sustained return to non-modified work within 365 days (yes/no), 2009 to 2013 ............................................................... 97 Table 13: Unadjusted and adjusteda results for Cox regression models, probability of sustained return to non-modified work after lost-time upper limb and spine strain or sprain work injury in BC from 2009 to 2013 .................................................................................................................. 99 Table 14: Unadjusted and direct adjusteda results for survival curves by anxiety and depression status in the year before injury, time (days) to sustained return to non-modified work after lost-time upper limb or spine strain or sprain work injury in BC from 2009 to 2013 ....................... 101   ix Table 15: Effect modification of the relationship between anxiety and depression (based on the year before injury) and sustained return to non-modified work by gender on the multiplicative and additive scales controlling for potential confoundersa, lost-time upper limb or spine strain or sprain claims in BC from 2009 to 2013 ...................................................................................... 102 Table 16: Frequency and timing (days) of new onset anxiety and/or depression disorders and impact on sustained return to non-modified work (SRTW) on the additive scale after lost-time upper limb or spine strain or sprain work injury in BC from 2009 to 2013 ............................... 105 Table 17: Frequency, timing and duration of first recurrence events for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 .............................................................. 119 Table 18:Unadjusted estimates of the association between anxiety and depression in the year before injury and return to work and recurrence events lasting 1 day or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 ............................................... 122 Table 19: Adjusteda estimates of the association between anxiety and depression in the year before injury and return to work and recurrence events lasting 1 day or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 ............................................... 122 Table 20: Summary table depicting adjusteda associations of anxiety and depression in the year before injury with return to work and lost-time recurrence events lasting 1 day or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 ............................. 123 Table 21: Distribution of study variables across levels of actual treatment for the full study sample of workers with pre-existing anxiety in addition to lost-time upper limb and spine strain or sprain in BC from 2000 to 2013 ............................................................................................. 140 Table 22: Comparison of the prevalence differencea across levels of actual treatment to the prevalence difference across the highest and lowest quartiles of the instrumental variables (IV)..................................................................................................................................................... 142 Table 23: Unadjusted proportion of workers that received actual treatment by quartiles of the instrumental variables ................................................................................................................. 145 Table 24: Characteristics of the first-stage instrumental variable (IV) regression models......... 146 Table 25: Instrumental variable two stage least squares linear regression estimates of the impacts of minimal adequate treatment, counseling, and prescription services on disability days (log-transformed) ................................................................................................................................ 148 Table 26: Conventional linear regression estimates of the impacts of minimal adequate treatment, counseling, and prescription services on disability days (log-transformed) ............................... 149 Table 27: Median number of each type of mental health service event in the year before injury by those that had at least one ........................................................................................................... 191 Table 28: The proportion of claims with at least one of each type of mental health service event in the year before injury by anxiety and depression case status (based on the primary case definitions) in the year before injury .......................................................................................... 191   x Table 29: Median number of each type of mental health service event in the year before injury by assigned case status (based on the primary case definitions) in the year before injury .............. 192 Table 30: Unadjusted estimates of the association between anxiety and depression in the year before injury and the return to work and recurrence events lasting 7 days or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 ..................................... 193 Table 31: Adjusted estimates of the association between anxiety and depression in the year before injury and the return to work and recurrence events lasting 7 days or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 ..................................... 193    xi List of Figures Figure 1: A multistage model of cognition, disability, and affect. Adapted with permission from Pincus and Williams(26) ............................................................................................................... 10 Figure 2: Conceptual framework for impacts of anxiety and depression on return to work for lost-time musculoskeletal work injury and influencing factors .................................................... 24 Figure 3: Visualization of the relationships examined by research objectives 2 and 3 ................ 27 Figure 4: Visualization of the relationships examined in research objective 4 ............................ 29 Figure 5: Time frames used to identify index events and secondary validation events in the anxiety and depression case definitions ........................................................................................ 48 Figure 6: Construction of the study sample consisting of lost-time upper limb and spine strain or sprain workers’ compensation claims in BC from 2000 to 2013.................................................. 59 Figure 7: Visual description of the sustained return to non-modified work outcome variablea using four hypothetical return to work trajectories ....................................................................... 94 Figure 8: Number of claims according to the number of return to work and recurrence events for men (left) and women (right), for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 (% = percent of total claims in the study sample or percent of the claims at risk of the specific return to work event) .................................................................................... 117 Figure 9: Analytic framework for research question on the association between treatment for anxiety and return to work outcomes after musculoskeletal injury ............................................ 130 Figure 10. Construction of the study sample consisting of workers with lost-time upper limb or spine strain or sprain work injury and a prevalent anxiety disorder in the year before injury in BC from 2000 to 2013 ....................................................................................................................... 133 Figure 11: Timeline for identifying prevalent anxiety in the year prior to injury and receipt of mental health services among workers with lost-time upper limb or spine strain or sprain work injury in BC, Canada................................................................................................................... 134      xii Acknowledgements Due to the support and contributions of many people and agencies, completion of this dissertation was feasible. First and foremost, I cannot thank my supervisor, Dr. Mieke Koehoorn enough for having guided and supported me through my PhD program and dissertation project from start to finish. I am also extremely grateful for my committee members, Dr. Chris McLeod and Dr. Ute Bültmann for their thoughtful and insightful feedback throughout the project. The School of Population of Public Health and the Bridge Program created welcoming and enriching environments for me to learn and develop as a student and health researcher. I am thankful for the many wonderful students, staff, and faculty that I got to meet and become friends with along the way.  My research was conducted at and supported by the Partnership for Work Health and Safety. My colleagues there were instrumental in helping me through many stages of the research process, not to mention supportive in all areas of work and life. I consider myself extremely fortunate to have been surrounded by such a great group. Lastly, my PhD training was generously supported by a Bridge CIHR Strategic Training Fellowship, WorkSafeBC Research Training Award, Centre for Research on Work Disability Policy Student Trainee Award, and UBC Faculty of Medicine Graduate Student Initiative Award. Receipt of this support allowed me to undergo my PhD and complete this dissertation on mental health, work injury, and work disability. Lastly, to Brad and my family – so much love – and thanks for humouring me   1 Chapter 1: Introduction 1.1 General overview In the 2017 Global Burden of Disease Study, musculoskeletal conditions were the leading contributor to global disability followed by mental disorders.(1) While the causes of musculoskeletal conditions are varied and multi-factorial, contributing factors can include the activities and demands of paid work. In fact, it is estimated that across the globe, 37% of low back pain (the single most leading cause of disability)(1) is attributable to work.(2) In Canada, workers’ compensation systems are responsible for providing benefits, medical care, and rehabilitation for work injury and illness. In the Canadian province of British Columbia (BC) (representing one of ten workers’ compensation systems in the country), musculoskeletal conditions are the most common type of compensated work injury and illness.(3) Musculoskeletal strains or sprains comprise a large proportion of these conditions as they are the most common compensated injury sub-type, and the largest contributor of compensated lost-work days.(3) From 2013 to 2017 in British Columbia (BC), there were 147,955 lost-time claims for musculoskeletal strain and sprain alone (not including other types of musculoskeletal conditions), accounting for 20% to 25% of all claims and $1.7 billion in disability costs.(3) Given this, management of musculoskeletal work injury, particularly strain or sprain, is a major issue for the BC workers’ compensation system. In addition to the direct burden of musculoskeletal conditions, there is the added burden of comorbid conditions. Musculoskeletal conditions, especially those characterized by chronic pain, often co-occur with depression; and there is growing evidence of a similar co-occurrence of musculoskeletal conditions with anxiety.(4,5) Further, musculoskeletal conditions and mental disorders are both risk factors for work disability.(6) The inter-relationships between musculoskeletal conditions, mental disorders, and work disability have meaningful implications for how workers’ compensation systems and health care practitioners approach medical care, rehabilitation, and return to work management for musculoskeletal work injury. While this is an issue of growing concern, current disability management and healthcare practices may not be congruent with, or sufficiently attentive to, injured workers’ mental health needs concurrent with musculoskeletal work injury. Arguably, there is a lack of evidence to inform how concurrent   2 mental disorders should be addressed, managed, or treated in the context of compensation and rehabilitation for musculoskeletal work injury.  1.2 Rationale and purpose Prior research documents the presence and progression of anxiety and depression symptoms after musculoskeletal work injury, but little is known about the frequency of pre-existing mental disorders, or risk factors for mental disorders in the population of workers with work injury. In addition, the literature on the impacts of mental disorders on return to work after musculoskeletal injury has had mixed results and population-level analyses of these relationships are lacking. Lastly, among workers with musculoskeletal work injury who also have a mental disorder, little is known about their recent or ongoing mental health treatment, and how this might affect return to work.    There is also a greater need to understand the role of gender in work disability, especially in regards to mental health. Evidence demonstrates that workers’ compensation experiences and outcomes vary by gender,(7,8) as does the risk of anxiety and depression disorders in the general population.(9) However, little is known about how men’s mental health experiences following musculoskeletal work injury might differ from women’s and how this impacts return to work. More knowledge in this area can help guide the development of gender sensitive strategies for mental health in work disability management and prevention policy and practice The overall purpose of this thesis was to explore the distribution of common mental disorders among workers with musculoskeletal work injury, and the impact of these disorders on return to work outcomes. For mental health, anxiety and depression disorders were the focus of examination as they are the most frequently occurring mental disorders in injured worker populations.(10–12) This is also true in the general population where 1 in 10 people are affected by one or both in a given year.(13) Likewise, upper limb and spine strain or sprain injuries were chosen as a focus of interest as they are the most frequently occurring compensated lost-time work injury claims and a significant source of work disability.   3 1.3 Research objectives Four research objectives were chosen to address specific gaps in the literature. Among workers with lost-time upper limb or spine strain or sprain work injury, the specific research objectives were to: 1. Document the descriptive epidemiology of anxiety and depression disorders including:  a. the one-year period prevalence of anxiety, depression, and co-morbid anxiety/depression for the year before and the year after injury, as well as the cumulative prevalence for all years prior to injury (back to 1991); and b. the timing of anxiety and depression disorder onset relative to the injury; and c. socio-demographic, injury, clinical, and work-related risk factors for: i. prevalent anxiety, depression, and co-morbid anxiety/depression in the year prior to injury; and  ii. new onset anxiety, depression, and co-morbid anxiety/depression in the three months after injury. 2. Investigate the impacts of pre-existing anxiety and depression disorders (alone and co-morbid) prevalent in the year before injury, as well as new onset anxiety and depression disorders that develop during the return to work process, on sustained return to work. 3. Investigate the impacts of pre-existing anxiety and depression disorders (alone and co-morbid) prevalent in the year before injury on specific return to work events that occur as a part of the overall return to work process, namely, entry to non-modified work (from lost-time) and recurrence of lost-time (after initial return to non-modified work). 4. Investigate the applicability of instrumental variable methods based on physician treatment preferences to examine the effects of anxiety treatment (pharmacotherapy and counseling) on work disability outcomes, using a study sample of injured workers with pre-existing anxiety disorders. 5. Investigate if measures of disease frequency or measures of association between exposures and outcomes, as specified for the objectives above, differ for men and women.     4 1.4 Study design overview Four population-level, observational studies were conducted, one for each research objective outlined above. All four studies used longitudinal and retrospective linked administrative data from the province of BC’S Ministry of Health and workers’ compensation system (WorkSafeBC) obtained via Population Data BC (www.popdata.bc.ca). Accepted claims for lost-time upper limb or spine strain or sprain work injury and return to work outcomes were identified from the workers’ compensation data, while anxiety and depression disorders were identified from the Ministry of Health data using diagnoses from physician and hospital visits, and anti-depressant and anxiolytic prescriptions from dispensing events. Other socio-demographic, injury, clinical, and work-related variables were collectively identified from the administrative datasets including registry files provided by the Ministry of Health and claims files provided by WorkSafeBC. Study sample selection criteria, variable definitions, and statistical analyses are described in detail in Chapters 3 through 7. 1.5 Thesis outline This thesis is organized into eight chapters. This chapter provides an overall summary of the dissertation research. Chapter 2 provides a summary of the literature on musculoskeletal conditions; work disability, anxiety and depression disorders; the importance of gender to the research objectives; and lastly, a conceptual framework, rationale and hypotheses for the research objectives. Chapter 3 describes the methodology and the research setting that are common to each research objective. This includes the jurisdictional context for the research; the data files, extraction, linkages and data cleaning processes used to construct the study database; and a review of existing case definitions to identify anxiety and depression disorders using health claims data for research, and a summary of the anxiety and depression case definitions adapted for this research. Chapters 4, 5, 6, and 7 constitute the research chapters of the thesis that correspond to the 1st, 2nd, 3rd, and 4th research objectives (outlined above) accordingly. Each of the research chapters includes a background, methods, results, and discussion subsection. Lastly, Chapter 8 provides an overall discussion of the dissertation research, including a summary of the results and contributions to current knowledge, strengths and limitations of the work, a discussion of the implications, and recommendations for future research.  Supplemental information is contained in Appendices to these aforementioned chapters.   5 Chapter 2: Background 2.1 Musculoskeletal conditions Musculoskeletal conditions, including injuries and diseases of the musculoskeletal system, are one of the most significant workforce health issues and a major contributor to the total burden of disease. They can limit people’s ability to perform activities of daily living and occupational work activities through pain and functional limitation, and ultimately decreased quality of life. Low back pain is the most common type of musculoskeletal condition and one of the most common reasons for seeking medical care in high income countries.(14) At any point in time, 12% of adults worldwide report low back pain severe enough to limit activity for more than a day, and 23% of adults report experiencing this in the course of a month.(15)  At a societal level, the economic cost of musculoskeletal conditions is substantial. In 2010, the economic burden of musculoskeletal disease in Canada was $8.7 billion per year, including $2.0 billion in lost productivity costs.(16) Likewise the total economic burden of injuries, including those affecting the musculoskeletal system, was $18.6 billion per year, including $3.8 billion in lost productivity costs.(16) Worldwide, musculoskeletal disorders account for 21% of the total years lived with disability, second only to mental and behavioral problems.(17) Due to aging populations and increasing prevalence, musculoskeletal disorders are expected to remain a predominant source of disability for decades to come in Canada and worldwide.(17) 2.2 Strain or sprain work injury In BC, musculoskeletal conditions are the most common type of compensated work injury or illness, and over half of these are due to strain or sprain injuries.(18) Strain includes damage to muscle or tendon tissue while sprain includes damage to ligaments that connect muscles and tendons to bone. These can occur due to overuse (e.g. repetitive motion) or due to more acute events such as a fall. From 2012 to 2016 in BC, there were 148,755 new lost-time workers’ compensation claims for strain or sprain that constituted 57% of all lost-time claims, 58% of all registered work days lost due to work injury or illness, and over $1.8 billion in short and long-term disability benefit costs.(18)  Over half of all work strains and sprains in BC are spine related.(18) Upper limb strains and sprains are also common in workers’ compensation populations,(19) although BC statistics   6 describing their relative frequency compared to other work injuries are lacking. Risk factors for these injuries span the biomechanical, psychosocial, and socio-demographic domains. A 2010 systematic review of 63 longitudinal studies found that heavy physical work, awkward postures, smoking, high body mass index, being a women, co-morbidities, low job control, low work satisfaction, and high psychosocial job demands were consistent risk factors for musculoskeletal work injury commonly identified across studies of the spine (including the neck and low back), and upper limb.(20) Lifting was also a risk factor for the spine (but not upper limb), repetitive work was a risk factor for the upper limb (but not the spine), and prolonged computer work was a risk factor for the elbow/forearm and the wrist/hand (but not the spine or shoulder).(20) 2.3 Conceptual models of disability and work disability Conceptual models of work disability are useful tools to help understand how and why work disability develops after injury, as well as to guide work disability management and prevention including but not limited to a workers’ compensation context. They can also be used to guide analyses intended to identify causes of work disability. For example, they can be used as a tool to help identify potential confounders or modifiers of a relationship of interest. For the current research, existing models of work disability can be drawn upon to understand the context within which individual factors such as mental co-morbidity might affect return to work (a measure of work disability) after lost-time upper limb or spine strain or sprain work injury. In the rest of this section, concepts of disability are described as an introduction to the topic, followed by a description of conceptual models of work disability and return to work most relevant to the research objectives.  2.3.1 Concepts of disability and work disability Concepts of disability vary across cultures and time. In Western culture, early concepts of disability were largely medical in nature with disablement thought to directly correspond with pathology.(21) Following this, relational concepts of disability were introduced that differentiated disability from non-relational measures like pathology and functional limitation. In 1965, the social model of disability was introduced by Nagi that viewed disability as the result of a gap between a person’s capabilities and the demands imposed by the sociocultural and physical environments.(22) In this way, a person may experience varying levels of disability in different environments depending on how well the environment is adapted to their needs. In the 1990s,   7 Verbrugge and Jette (1994) developed the disablement process model that built on the social model of disability in several ways including an explicit subdivision of social roles within which disability can be experienced. Examples of such social roles include activities of daily living, paid and unpaid work activities, social activities, and leisure activities.(23) In this latter framework, work disability is a specific subtype of disability. There are many definitions of work disability including those based on conceptual or legal elements. For this dissertation, work disability is conceptualized as a health-related limitation in the performance of, or access to, paid work or employment, including work entry, productivity, sustainability or retention, and advancement. In line with the social model of disability (22), non-health related factors like the overall economic, social, and political environment, the health care and workers’ compensation systems, the occupational environment, and job demands, amongst other factors, are hypothesized to influence the work disability process. Lastly, while some definitions of work disability include unpaid work, the focus of this dissertation is on work performed within the labour force. Thus, the ‘work’ aspect of work disability in this dissertation refers to paid work or employment unless otherwise indicated.  2.3.2 Conceptual models of work disability In two interdisciplinary review articles conducted by Schultz et al. in 2000 and 2007, five categories of work disability models specific to musculoskeletal injuries were identified.(24,25) These included biomedical, insurance, labour relations, psychiatric, and biopsychosocial models. Brief summaries of these five categories are provided here. Some additional models not included in the reviews by Schultz and colleagues, but relevant to the current research objectives, are also included. 1. Biomedical models Biomedical models assume that physical pathologies (e.g. the size of a soft tissue lesion) are the primary or sole determinant of work disability for physical injury.(24,25) Correspondingly, treatments are chosen to target the underlying physical pathology or its associated symptoms. (24,25) While this can be suitable for acute or early stage uncomplicated physical injury, it fails to address the multi-dimensional nature of work disability and the influence of non-biological and non-physical factors.(24,25) For this reason, amongst others, biomedical models are no   8 longer the de facto disability model used by workers’ compensation systems in Canada and other similar countries. 2. Insurance models Insurance models as described in the literature posit that while some people with injury claims have a legitimate need for disability benefits due to injury, others exploit insurance systems through dishonesty for personal gain.(24,25) These models have applications for insurance practices and procedures, but their application is limited to injuries where objective measurement of injury severity and pathology is feasible.(24,25) Overall, they are arguably not that useful for understanding the multi-faceted nature of work disability. Drawbacks include a lack of scientific methods for distinguishing between legitimate and illegitimate claims, an over focus on possible gains associated with a successful claim (e.g. financial benefits of insurance) accompanied by an under focus on personal loss due to injury (e.g. social or occupational role impairment).(24,25) 3. Labour relations models In labour relations models, the role of the workplace in disability management is highly emphasized.(24,25) A key tenet of these models is that employers and workers mutually benefit from the recovery and return to work of injured workers.(24,25) These models use a disability management approach that considers the socio-political context of the workplace and includes the employer as a key stakeholder.(24,25) A rehabilitation plan is used that integrates physical treatment with return to work goals and processes, rather than a sole reliance on traditional medical treatment. This approach can promote return to work especially in the context of large organizations with sufficient resources and buy-in from management, but has limited application for clinically complex injuries, or for small employers, and can result in loss of privacy for the worker.(24,25) 4. Psychiatric models 4 a. Traditional According to traditional psychiatric models, pain is either physical or psychological in origin. (24,25) When an injured person’s pain reaction is considered disproportionately large given the physical pathology (e.g. tissue damage), excess pain is considered to have a psychological source, usually in the form of a diagnosable psychiatric disorder.(24,25) Psychiatric models   9 include treatments for physical injury and for psychiatric symptoms, but the latter are usually reserved for people with a diagnosed psychiatric disorder presumed to have arisen from the complications of injury.(24,25) These models benefit people with medically recognized and clinically high levels of injury-related psychological symptoms.(24,25) However, they also over-pathologize the psychological aspects of injury and disability, oversimplify pathways between pain and disability, under recognize sub-clinical mental symptoms and mental disorders that do not arise from the physical injury, and under emphasize system level (e.g. the workplace) and non-psychiatric individual-level determinants of work disability. 4 b. Cognitive models relating pain and depression Cognitive models relating pain and depression comprise another category of psychiatric model for understanding the frequent co-occurrence of chronic pain and depression as well as their associations with disability. However, in these models, non-relational measures like functional disability are typically examined, rather than work disability or return to work. A specific example of a cognitive model relating pain and depression is the diathesis-stress model by Pincus and Williams (Figure 1).(26) A key feature of the diathesis-stress model is that while all of its components (diatheses, trauma, disability, and negative affect) are connected by circular loops, some connections are stronger than others. At the start of the cycle, diathesis combines with a primary stressor (e.g. injury) to concurrently create a state of disability and negative affect (i.e. depression). Disability then exacerbates the primary stressor (e.g. injury) and also contributes to negative affect. Lastly, negative affect contributes to disability and increases cognitive vulnerability that in turn increases perception of pain and enhances distress. In the model, disability is the primary driver of negative affect (i.e. larger arrow in Figure 1). While this model more accurately captures the complex interactions between depression, injury, and disability than traditional psychiatric models, it has not yet been generalized to other mental health conditions and symptoms like anxiety. Further, it is not specific to occupational injury or work disability and, on its own without further development or integration with other research disciplines, does not currently offer salient evidence-based strategies to guide return to work or work disability prevention.   10  Figure 1: A multistage model of cognition, disability, and affect. Adapted with permission from Pincus and Williams(26)  5. Biopsychosocial models Biopsychosocial models use a multi-dimensional, interdisciplinary approach to injury diagnoses, treatment, and management.(24,25) A key tenet of these models, which is supported by empirical evidence, is that the physical pathology of injury is not the sole source of disability, and with the exception of extremely severe injuries, it is not the primary source of disability.(24,25) Empirical evidence supports this tenet, as the best prediction models for work disability and occupational injury include variables from a wide bio-psychosocial spectrum.(27) Biopsychosocial models overlap conceptually, to some degree, with each of the other models already discussed above, but tend to adapt a broader perspective for understanding work disability with a wider range of influencing factors. A number of biopsychosocial models exist, and while no two are exactly alike, there are many commonalities and shared characteristics, as each successive model builds on the strength of others. A select few with relevance to the current work are highlighted here. 5 a. Acute/sub-acute/chronic phase of low back pain  For at least as early as 1987, management guidelines for reducing work disability due to low back pain have adapted a phase based approach to emphasize the developmental nature of work disability, and to guide intervention timing.(28) A common example is the acute, sub-acute, and chronic phase approach developed by Spitzer et al. (1987),(28) that was later adapted by the Sherbrooke model in the 1990’s.(29) This model is based on a combination of evidence from high quality intervention studies and the ‘return to work curve’ for low back pain that depicts the proportion of workers who remain off work over time since injury. The acute phase includes the first 30-days post-injury when the rate of return to work is high with approximately 50% of   11 workers returning to work. This is followed by the sub-acute phase (30- to 60-days post injury), when the rate of return to work decreases and risk of long-term disability increases. The chronic phase starts at 60-days post injury and is characterized by a low probability of return to work. Based on a number needed to treat principle, conservative treatment and symptom monitoring are recommended for the acute phase when the rate of return to work is already high, while for the subacute phase when the rate of return to work slows and risk of long term disability increases, interventions that combine clinical management with an occupational intervention (e.g. work modification) are recommended.  5 b. Eight phase model of low back pain In 1994, Krause and Ragland introduced another phase-based model of work disability for occupational low back pain that included eight unique phases as opposed to three with further considerations for the social determinants of disability.(30) Like the prior model, phases were defined according to time since injury. According to Schultz et al. (2007), a drawback of the three and eight phase based models is that neither adequately account for the recurrent nature of work disability due to musculoskeletal injury (including low back pain), and their application is limited to short term management of an acute low back pain episode, with less relevance to episodic trajectories, other musculoskeletal conditions, or even the long-term chronic phase of low back pain.(25) 5 c. Disability prevention management model for low back pain This paradigm for low back pain by Loisel et al. (2001) adapts the acute, sub-acute, and chronic phased-based approach described above but also integrates an ecological/case management approach.(31) This model explicitly groups the biopsychosocial determinants of work disability into four systems: personal, workplace, healthcare and compensation systems. It encourages a shift from a disease treatment paradigm to a disability management one that includes stakeholders from each of the four systems. In 2005, the model was expanded to include determinants of work disability that exist upstream of the aforementioned systems, including the overall societal context, culture, and politics.(32) 5 d. Developmental conceptualization of return to work   12 The developmental conceptualization of return to work as proposed by Young et al. (2005) describes return to work as an evolving process involving four key phases: off work, work re-entry, retention, and advancement.(33) Unlike the other phase-based models previously discussed it is not limited to low back pain and has a more generalized application for all lost-time work injury and illness. In this model, return to work is a process involving a series of events, transitions, and phases, as well as interactions with other individuals and the environment. This model makes few clinical recommendations, as these tend to vary by injury type. Rather, it emphasizes key time points for assessment of work capacity and performance, and assessment (or reassessment) of work goals, based on the workers’ return to work phase. 5 e. Readiness for return to work model The readiness for return to work model by Franche and Krause (2002) integrates two prior models: 1) the eight phase model of disability by Krause et al. (outlined earlier) and 2) the readiness for change model from the field of health promotion.(34) The readiness for change model describes five stages that a person progresses through during the process of behavior change. These five stages were used to explain progression through the eight disability phases outlined originally by Krause and colleagues (1994). Stages from the readiness for change model include: pre-contemplation, contemplation, preparation for action, and maintenance. In regards to readiness for change, stage placement and progression (or regression) are considered a function of a person’s self-efficacy, decision balance, and other change processes specific to return to work (e.g. involvement and action of a health care professional or the workplace). While this model proposes an interesting framework to understand and manage work disability, it has not been empirically tested. Summary of the conceptual models of work disability Some of the models reviewed here include mental health as a risk factor for work disability after physical injury, and the readiness for return to work model explains in depth how mental disorders such as depression might affect psychological determinants of return to work like self-efficacy. The diathesis-stress model captures the complex inter-relationships between depression, injury, and disability, but its generalizability to work disability in a workers’ compensation context and other mental disorders remains unknown. While the acute/sub-acute/chronic model of low back pain and the eight-phase model of low back pain both provide specific time-based   13 recommendations for the clinical and workplace-based management of low back pain, recommendations for mental symptom management after musculoskeletal injury are similarly lacking. The readiness for return to work model describes possible benefits and services to address mental symptoms after lost-time musculoskeletal injury (e.g. ongoing pain management and counseling) but studies to evaluate the effectiveness of these benefits and services are lacking.   For the purposes of this dissertation, no single existing model adequately captures the inter-relationships of mental health, musculoskeletal work injury, and work disability, including the timing of these inter-relationships. Likewise, no single existing model adequately accounts for the contextual or individual level factors that might affect these aforementioned relationships in a workers’ compensation context. For these reasons, the models reviewed here were used to inform the development of a conceptual framework intended to guide the current research. The conceptual framework developed for this study is described further in section 2.6, after the sections on mental health (section 2.4) and the importance of sex and gender (section 2.5), as information from these two sections also informs the development of the framework. 2.4 Mental Health There has been increased interest among workers’ compensation systems to address co-morbidity, in particular mental health disorders, in the recovery and return to work process for physical work injury. This may be due to the high prevalence of anxiety and depression disorders among workers with musculoskeletal work injury, where the prevalence of these disorders is elevated compared to general populations.(10,11,35–37) Most notably, at one month post injury, over 40% of workers with spine or upper limb musculoskeletal injury have clinically high levels of depression symptoms (i.e. a score of 16 or more on the self-reported Center for Epidemiological Studies – Depression Scale), even after workers with depression in the year before injury are excluded from the study sample.(36)  The following section provides an introduction to anxiety and depression disorders, including brief summaries of their incidence, prevalence, and risk factors in the general population. Possible rationales are also provided for why anxiety and depression disorders frequently co-  14 occur with physical work injury. A summary of the literature describing risk factors for anxiety and depression more specific to injured worker populations is provided in Chapter 4.  2.4.1 Anxiety Description Anxiety disorders are characterized by excessive fear or anxiety. According to the American Psychiatric Association, ‘fear is an emotional response to an imminent real or perceived threat, while anxiety is anticipation of future threat’.(9) Different types of anxiety disorders are often co-morbid with each other but can be differentiated from one another based on the source of fear or anxiety (e.g. anxiety about social interaction versus anxiety about physical health symptoms), as well as the reactive emotional response or thoughts.(9) Types of anxiety disorders include selective mutism, specific phobia, social anxiety disorder (social phobia), panic disorder, agoraphobia, generalized anxiety disorder, substance/medication induced anxiety disorder, and separation anxiety disorder. Older classification systems also include posttraumatic stress disorder and obsessive-compulsive disorder. Anxiety disorders tend to be chronic with symptoms that wax and wane between clinical and subclinical levels across the lifetime.(38)  Prevalence and incidence Approximately 10% of adults experience an anxiety disorder in a 12-month period, and over 30% experience one in their lifetime.(39) Among Canadian adults, specific phobia is the most common type of anxiety disorder with a one-year period prevalence of 6% to 8%, followed by social phobia (7%), obsessive-compulsive disorders (2%), generalized anxiety (1% to 3%), and panic disorders (1%).(40,41) The median age of onset for anxiety disorders at 11 years is much earlier than for depression disorders at 30 years.(42) Risk factors  Specific anxiety disorders have their own risk factors, but some general risk factors common to most types of anxiety include: being a women, childhood adversities, temperamental traits like shyness or behavioral inhibition, and family history of mental illness (heritability).(43)   Association with physical work injury There is growing evidence of an association between chronic pain and anxiety, however the evidence base for this relationship is less robust than that of chronic pain and depression.(44)   15 There are two theoretical positions explaining the co-occurrence of anxiety with chronic pain. In the first position, pain and anxiety are thought to exacerbate each other and this leads to each condition sustaining itself longer than if the other was not present. In the second, they are thought to have a common cause or shared vulnerability.(44) Unlike depression, the two theoretical positions outlined above do not suggest that chronic pain conditions are a prospective risk factor for new onset anxiety disorders. In the case of physical work injury, trauma or disability related to the injury could invoke anxiety as an emotional response, especially if there is a prior history or an underlying anxiety disorder. In this instance, anxiety could exacerbate pain, slow down physical recovery, or increase a person’s hyper-awareness of possible social challenges that occur in the process of return to work. These latter effects could occur through a heightening of the anxiety response in the autonomic nervous system, increased avoidance behaviours, or hypervigilance and cognitive biases.(44) Alternatively, psychotropic drugs used to treat anxiety are risk factors for unintentional incidents, and this may explain the higher observed prevalence of anxiety among injured workers.(45) 2.4.2 Depression Description Depression disorders are characterized by feelings of sadness, emptiness or irritability, along with somatic and cognitive changes that significantly affect the capacity to function in one or more important hemispheres of life including work, domestic life, social life, or recreation.(9) Types of depression disorders include major depression disorder, persistent depression (dysthymia), disruptive mood dysregulation disorder, premenstrual dysphoric disorder, substance/medication-induced depression disorder, depression disorder due to another medical condition, other specified depression disorder, and unspecified depression disorder. Differences between depression disorders include symptom timing and duration, and presumed aetiology.(9) The most common type of depression - major depression is characterized by episodes of depression symptoms lasting at least two weeks with clear changes to affect, cognition, and neuro-vegetative functions.(9) Major depression disorder usually involves recurring episodes,   16 but a diagnosis can be made based on a single episode. Forty percent of people with major depression episodes recover within three months, 60% recover within six months, and 80% recover within one year.(46) Prevalence and incidence Based on self-reported diagnosed conditions, major depression episodes affect 13% of Canadians in their lifetime; and 5% and 2% of Canadians in the past year, or past 30 days, respectively.(47) The risk of a first time major depression episode increases after puberty and the annual peak prevalence is highest among people aged 15 to 25 years.(47) Risk factors Risk and prognostic factors for major depression include neuroticism (or negative affect), adverse childhood experiences, stressful life events, family history of depression (heritability), and other major non-mood disorders.(9) A 2006 study on the descriptive epidemiology of prevalent major depression in Canada found a higher risk of prevalent depression among women; and among individuals with a chronic medical condition, low income, or who were unemployed.(47) Prevalence also increased with age among men who never married.(47)  Association with physical work injury Depression frequently co-occurs with physical health conditions, especially those that involve chronic pain.(4,5) The relationship between pain and depression is thought to be bi-directional with pain initially preceding depression more often than not, and then depression exacerbating pain, and vice versa once both are present.(4,48) Factors that may play a role in the pain-depression relationship include shared neuroanatomical pathways and neurotransmitters, disability, fatigue, and cognitive behavioral factors such as passive coping and helplessness.(49,50) Depression can also be situational. While some depression episodes occur for no apparent reason, there is increased likelihood of a depression episode after a stressful or serious, negative life event.(51) Lost-time work injury can represent such events and may explain the heightened prevalence of depression in the post injury time period. Alternatively, symptoms of depression and psychotropic drugs commonly used to treat depression are also risk factors for unintentional incidents that may explain a high prevalence of depression among the injured worker population.(45,52)   17 2.4.3 Potential mechanisms linking anxiety and depression with poor return to work The epidemiological studies conducted for this dissertation examine associations between anxiety and depression disorders with return to work events and outcomes among workers with physical injury. Here, potential mechanisms that could underlie these relationships are outlined. This research will not investigate these mechanisms specifically but they do provide possible explanations for why associations might exist, if they are observed in the study findings.  1. Mechanisms related to the personal system 1 a. Increased pain Anxiety and depression may affect return to work after musculoskeletal injury through the amplification of pain. Pathways through which this might occur are not fully understood, but potential biological mechanisms involved in both the production of pain, anxiety, and depression include: 1) shared genetic susceptibility; 2) shared neurobiological molecules involved in central nervous system functioning such as neurotransmitters (e.g. serotonin and norepinephrine), neuromodulators, neurotrophic factors, pro-inflammatory cytokines, or hormones; and 3) shared neuroanatomical regions including the prefrontal cortex, limbic system (e.g. amygdala and hippocampus), hypothalamic pituitary adrenal axis, sleep center, and the spinal cord/dorsal horn.(53,54) Psychological mechanisms such as pain catastrophizing and hypervigilance may also lead to an increased perception of pain among people with anxiety or depression.(55,56) 1 b. Activity disruption Activity disruption has also been suggested as a potential mechanism. As described by Sullivan and colleagues, symptoms of anxiety and depression, such as low self-efficacy, loss of interest in previously enjoyable activities, fatigue and low energy, low motivation, pessimistic views of the future, fear avoidance behaviours, and general withdrawal from social, recreational and occupational activities, could interfere with participating in activities or adhering to treatments that aid with recovery and return to work.(57) For example, depressed workers are more likely to drop out of physical therapy treatment for musculoskeletal work injury than non-depressed workers.(57)  1 c. Resistance to treatment   18 There is evidence that mental symptoms can reduce the effectiveness of treatment for musculoskeletal injury. For example, Slepian and colleagues found that workers with depression or post-traumatic stress symptoms were less likely to respond to physical therapy treatment intended to reduce pain catastrophizing after musculoskeletal injury.(58)  1 d. Beliefs regarding return to work  Multiple studies have found that return to work expectations are highly predictive of return to work outcomes.(59,60) Like with activity disruption, symptoms of anxiety and depression (e.g. worry, fear avoidance, negative outlook, fatigue, difficulty concentrating, and increased pain among others) could decrease workers’ beliefs that they are able to return to work, resulting in longer work disability durations compared to those without anxiety and depression.  2. Mechanisms related to the work or compensation systems 2 a. Greater need for work accommodation In a representative sample of the Canadian population, over half of adults with an anxiety or depression disorder reported the need for job accommodations to continue working.(61) In a community sample of the working population in the Canadian province of Alberta, 84% of adults with an anxiety or depression disorder reported needing workplace mental health accommodation (e.g. weekly meetings, reduced work hours, modified job duties, work at home, access to an employee and family assistance plan), although only 31% had received an accommodation.(62) It is possible that the need for work accommodation among workers with an anxiety or depression disorder is amplified when other physical health or disability issues, such as lost-time physical work injury, are present. An absence of work accommodation following lost-time physical work injury could explain poor return to work outcomes for injured workers in general, but even more so for workers with an anxiety or depression disorder.  2 b. Stigmatization  Lastly, as suggested by Carnide et al. (2015), stigmatization of mental symptoms and disorders, or negative responses to them by employers or other return to work partners, may impede the return to work process – even if the reason for lost-time is due to a physical injury.(63)    19 2.4.4 Measurement of anxiety and depression for research Anxiety and depression symptoms are a normal part of life. Symptoms do not necessarily indicate a disorder as they occur in varying severities and combinations, and may not necessarily induce role impairment. For diagnosis, a specific set of criteria regarding symptom type, number, frequency, and duration must be met. Typically, various emotional, behavioral and physical symptoms are considered, as well as role impairments, before symptoms are considered a disorder or ascribed a diagnosis.  Methods of measuring anxiety and depression for research can largely be divided into three categories: medically recognized disorders, self-reported symptoms, and clinical interviews. The administrative data used by the current study falls under the first category of medically recognized disorders. Each of these three categories are valid. However, they measure different constructs and have unique pros and cons that are briefly summarized here.  Medically recognized disorders are measured using participants’ self-reported diagnoses, patient medical records (i.e. written or electronic notes from the treating clinician), or administrative data usually in the form of health insurance or payment claims data. A benefit of these methods is that diagnoses are clinically verified (although this statement only partially applies to self-reported diagnoses). Another benefit of health claims data and self-reported diagnoses is that data collection is often less arduous than other methods that require long questionnaires or interviews. In addition, self-reported diagnoses, patient records, and health claims data can be used to identify past anxiety and depression episodes that differentiates them from self-reported symptoms and clinical interviews predominantly designed to detect prevalent symptoms or disorders. Drawbacks of measurement methods based on medically recognized disorders include low sensitivity as untreated disorders are not captured, and underestimations of prevalence; and, they are generally not effective for capturing subclinical symptoms. Within this category prone to low sensitivity, medical records typically have the highest sensitivity, followed by claims data, and lastly self-report.(64) While medical records have higher sensitivity than claims data, they often require manual data extraction, or if electronic, can be more difficult to access than claims data (especially at a population level) due to additional privacy and confidentiality concerns.   20 The second category is self-reported anxiety and depression symptoms usually collected through interviews or questionnaires. Current symptom profiles are typically measured on a continuum with ranges proposed to correspond to severity, or a cut-point to describe those with a high probability of having clinically diagnosable depression or anxiety. Examples of tools commonly used in research include the Beck Depression and Anxiety Inventories, Hamilton Rating Scales for Depression and Anxiety, Hospital Anxiety and Depression Scale, and the Centre for Epidemiological Studies Depression Scale,(65–70) although this list is not exhaustive. In clinical practice, these tools can be used for screening but they are not equivalent to diagnostic evaluation. Drawbacks of self-report methods include no clinical verification of diagnoses and poor detection of past episodes, while benefits include higher sensitivity and capture of subclinical symptoms.  The third category includes structured clinical interviews where a trained clinician or mental health professional interviews participants using a guide to make a diagnosis. While these are generally viewed as the gold standard for detecting current disorders, they are also resource intensive and less feasible for large studies, especially those conducted at the population level. They are also not designed to measure anxiety and depression disorder history. For the current research, health claims data were used to identify medically recognized anxiety and depression disorders among workers with compensated lost-time musculoskeletal strain or sprain. Most prior research on workers’ mental health after physical work-injury has been heavily reliant on self-reported anxiety and depression symptoms or diagnoses collected during the post-injury time period, and one or two American studies have used structural clinical interviews. While all methods of anxiety and depression measurement have their strengths and limitations, there is benefit in using different methods such that observed findings for underlying associations where they exist are robust to different measures despite their biases. In particular, self-reported measures of mental symptom timing and severity can be affected by recall and social desirability biases, especially when collected in the post injury time period when injured workers may feel that their eligibility for disability benefits is at stake.   21 2.5 The Importance of sex and gender to the research objectives According to the Canadian Institutes for Health Research, gender refers to ‘socially constructed roles, behaviours, expressions and identities of girls, women, boys, men, and gender diverse people’; while sex refers to biologically determined ‘physical and physiological features including chromosomes, gene expression, hormone levels and function, and reproductive/sexual anatomy’. Both sex and gender are upstream determinants for health.(71) Socially constructed gender roles and their relation to musculoskeletal work injury, mental health, and work disability outcomes are of interest to the current research as gender roles may be modifiable through large-scale changes to social and physical environments via policy, programs, or educational interventions.  The Canadian workforce itself is highly gendered with high concentrations of women in certain industries like healthcare and low concentrations in others like construction. Even within the same job, men and women can have different job tasks and responsibilities, and experience different social expectations (72), that could impact return to work. In the general working population, men have an overall higher rate of compensated work injury than women, but this finding does not extend to all injury types, industries and occupations. In an American study of the West Virginia working population, men had a higher overall rate of work injury claims than women, and this trend extended to almost all industries except for services and agriculture. When industry and occupation were accounted for, and individual injury types were considered, women had higher rates of injury than men for carpal tunnel syndrome, burns, sprains, and fractures.(73) Somewhat similar results were found in a study of the working population of the Australian State of Victoria. In this study, men had a higher overall rate of physical work injury claims than women, but after adjusting for occupation, women had higher rates of musculoskeletal and tendon injury claims.(74)  Experiences of mental health also vary between men and women. In the aforementioned Australian study, women had a higher rate of work disability claims for mental disorders than men and this difference was not fully explained by gender differences in industry or occupation.(74) In the general population, women are approximately two times more likely to experience anxiety and depression in a given year than men, and men are over three times more likely to experience substance use disorders than women.(75) These latter findings persist across   22 populations and measurement methods, and are not fully explained by the limitations of a single measurement method or study.(75) In fact, gender differences in the distribution of anxiety and depression disorders have been described as the most stable finding in psychiatry.(75) Despite this, evidence to explain the higher risk among women is still in its infancy. Potential mechanisms involved in women’s higher rate of depression include: 1) genetic risk and gene-environment interactions; 2) sex hormones; 3) physiological differences in stress response; 4) negative affect, rumination, and body shame; 5) higher rates of interpersonal stressors, violence, and childhood sexual abuse; and 6) societal structural gender inequities.(76) Similar mechanisms have been proposed to explain women’s higher rate of anxiety.(77) Compared to gender differences in injury rates, less is known about gender differences in return to work outcomes. There have been three recent Canadian studies conducted on this topic that are highlighted here, although neither of these examined mental health. In the first study by Koehoorn et. Al (2014), women consistently had more disability days paid than men after lost-time musculoskeletal injury, even after controlling for occupation and injury subtype.(7) The second study conducted by Stock et al. (2016) examined lost-time non-traumatic musculoskeletal disorders and similar to the first study, found longer duration of work absence for women.(78) The third study by Macpherson et al. (2018) compared men’s and women’s return to work trajectories in more detail, and found that sex and gender differences varied by disability duration and province.(8) Following injury, men initially have a higher probability of return to work, but of those still off work after two to four months, women had a higher probability of return to work until approximately 10 months. The finding that sex and gender differences in return to work vary by province suggests that some system-level factors, potentially those related to the compensation or health care systems or labour laws within a jurisdiction, influence men and women differently.  The finding that sex and gender differences in return to work vary over the disability duration suggests some cultural or societal impacts involving disability and masculine and feminine identities or constructs. Lederer et al. (2012) examined gender specific risk factors for poor return to work after long-term disability due to upper limb or spine musculoskeletal work injury in Quebec, although mental health measures were not included.(79) While men and women had several shared risk factors, they also had differences. For example, having a dependent and awareness of a workplace-based occupational health and safety program   23 were risk factors for delayed first return to work among women only, while a high perceived physical workload and higher job security were risk factors among men only. 2.6 Conceptual framework, rationale, and research hypotheses 2.6.1 Conceptual framework A conceptual framework for the impacts of anxiety and depression on return to work after lost-time musculoskeletal work injury, and influencing factors was designed for the purposes of this research (Figure 2). Similar to the diathesis-stress model by Pincus and Williams (1999),(26) both a diathesis (which could be pre-existing anxiety and depression disorders or a vulnerability to mental symptoms), and the stressor of interest (lost-time musculoskeletal injury) have a causal influence on mechanisms known to influence return to work. These mechanisms, some of which are described in the readiness for return to work model by Franche and Kraus (2002) as change processes include emotional symptoms (e.g. sadness or worry), psychological factors (e.g. pain catastrophising or self-efficacy), social interactions, pain, and functional limitations. Many of these change processes overlap with symptoms of depression and anxiety. In this way, the conceptual framework captures all of 1) the development of new, 2) the exacerbation of pre-existing, and 3) the implications of non-exacerbated pre-existing depression and anxiety symptoms, in a post-injury context. Drawing on the disability prevention management model by Loisel et al. (1991),(32) the conceptual framework depicted in Figure 2 also considers other return to work risk factors from four main systems: the personal, work, compensation, and the health care systems; as well as the overall societal context. Of particular importance to the current dissertation, is the influence of sex and gender on the central elements and pathways in the framework. While there are other factors to take into consideration for understanding return to work, those hypothesized as most relevant for understanding the role of mental health are represented in Figure 2.   24  Figure 2: Conceptual framework for impacts of anxiety and depression on return to work for lost-time musculoskeletal work injury and influencing factors 2.6.2 Rationale and hypotheses  Objective 1 Multiple studies have examined the prevalence of anxiety and depression disorders or clinically high levels of anxiety or depression symptoms after musculoskeletal work injury, but little is known about the prevalence of these disorders among workers at the time of injury. Examination of this will provide evidence for workers’ compensation systems on the extent to which anxiety and depression disorders are present in the injured working population and guide decisions about directing resources for the treatment and management of prior disorders versus the prevention of new onset disorders or symptoms (or both). Examination of longitudinal health claims data can provide information on the timing of anxiety and depression disorder onset among injured workers that is less prone to recall and social desirability biases than self-reported measures collected post-injury. This will help provide a more accurate understanding of mental health issues after physical injury. Further, a clear understanding of how anxiety and depression   25 disorders are distributed within the population of workers with lost-time strain or sprain work injury (e.g. differences in the prevalence of these disorders across age groups) can inform the development of evidenced-based prevention, management, and treatment efforts; that in turn can help ensure adequate and appropriate attention is given to groups in greatest need of mental health services and supports. Identification of anxiety and depression risk factors unique to new onset conditions occurring during the post-injury time period can help to better understand who is at risk of developing an injury-related anxiety or depression disorder, as well as the causes of these disorders in a lost-time musculoskeletal work injury context. Lastly, there has been limited examination of risk factors for anxiety and depression among workers with lost-time musculoskeletal work injury, and no examination of gender specific risk factors despite the highly gendered nature of mental disorders, work injury and work disability. Identification of gender specific risk factors for anxiety and depression among workers with lost-time musculoskeletal work injury can inform the development of gender sensitive interventions within workers’ compensation systems.  For the first objective, it was hypothesized that the prevalence of both anxiety and depression would be higher in the year after lost-time upper limb or spine strain or sprain work injury compared to the year before. The risk factor analyses were primarily descriptive and intended to inform hypotheses for future research. These were designed to address the following questions: 1) do risk factors for anxiety alone, depression alone, and comorbid anxiety and depression vary in this population?; 2) do the socio-demographic and clinical risk factors for anxiety and depression in this population vary from other general or clinical populations previously described in the literature?; 3) do men’s anxiety and depression risk factors vary from women’s in this population, especially in regards to the injury-related and work-related risk factors that are relatively under studied in the literature?; 4) do the risk factors for new onset anxiety and depression disorders in the three months following injury vary from the risk factors for prevalent anxiety and depression disorders identified in the year prior to injury? Once again with special attention to injury and work-related risk factors that may be indicative that workplace and injury characteristics independently influence the development of common mental disorders following lost-time upper limb or spine strain or sprain work injury.   26 Objectives 2 and 3 Prior research on the associations of anxiety and depression with return to work after lost-time musculoskeletal injury has been mixed. Common limitations of prior studies include self-reported measures of anxiety and depression, small clinical samples with limited external generalizability, and short follow up time that may not capture the total effects of anxiety and depression on return to work after lost-time. Further, while lost-time recurrence after initial return to work is common for many injuries including low back pain, there is limited evidence on anxiety and depression as prospective risk factors for lost-time recurrence events. A better understanding of these relationships can inform the use and timing of interventions intended to improve return to work for workers with anxiety or depression. Lastly, associations between gender and work disability due to mental disorders are not well understood in general, and even less is known about this topic in the context of physical work injury.   The purpose of objective 2 was to measure the effects of anxiety and depression disorders on the probability of sustained return to work after lost-time upper limb or spine strain or sprain work injury. The primary hypothesis was that both anxiety and depression disorders (pre-existing in the year before injury and new onset during the return to work process) would be associated with lower probability of sustained return to work, specifically return to pre-injury work status without modification or accommodations (herein referred to as return to non-modified work) (Figure 3). It was also hypothesized that the negative effects of anxiety and depression on the probability of sustained return to non-modified work would be greater for men than for women. This is in line with recent epidemiological evidence demonstrating that, at a population level, anxiety and depression disorders are more strongly associated with work disability among men than among women.(80,81) Possible mechanisms for this hypothesized gender difference are outlined in greater detail in Chapter 5. Objective 3 examined specific events occurring as a part of the overall return to work process after lost-time upper limb or spine strain or sprain work injury. The primary hypothesis of objective 3 was that anxiety and depression would contribute to a delayed sustained return to non-modified work through two potential intermediate mechanisms: 1) decreased probability of return to non-modified work (from lost-time) and 2) increased probability of lost-time recurrence (after initial return to non-modified work) (Figure 3). Similar directions of association were   27 expected for both men and women, but like with objective 2, larger effect sizes were hypothesized for men.  Figure 3: Visualization of the relationships examined by research objectives 2 and 3 Objective 4 Evidence suggests that pre-existing mental disorders are a risk factor for reduced resilience to traumatic or difficult life events.(82) Workers with lost-time due to musculoskeletal work injury are at risk of long-term work disability and possibly other negative work disability outcomes like job loss or exit from the labour force. These risks may be heightened for workers with a pre-existing mental disorder from prior to injury, although mental health treatment could mitigate this. Workers’ compensation systems may be able to facilitate return to work after musculoskeletal work injury for workers with a mental disorder using post-injury mental health interventions such as psychological counseling, pain management or work accommodations. However, public health systems also have a responsibility to provide patients with mental disorders with adequate treatment across the life course that subsequently could increase their resilience to challenging life events like work injury. As anxiety disorders are currently undertreated in Canada’s public health system,(83–87) improvements in treatment levels offers a potential target area to improve the lives of Canadians with anxiety, including those engaged in the labour force. The potential benefits of increased treatment levels are multi-fold and could include less anxiety-related work disability among workers. Objective 4 examined the use of an instrumental variable approach based on physician treatment preferences to estimate associations of mental health services with work disability outcomes among people with anxiety in the unique context of lost-time upper limb or spine strain or sprain work injury. Physician mental health services were chosen as the focus of interest as physicians are the de facto service providers for mental disorders in the BC public health system, as public funding for other mental health service providers is extremely limited. A decision was made to Anxiety and/or depression Delayed sustained return to non-modified work Lower probability of return to non-modified work (from lost-time) Increased risk of lost-time recurrence (from non-modified work)   28 limit the focus of objective 4 to pre-existing anxiety disorders (as opposed to pre-existing anxiety and depression disorders as in previous objectives) for the following reasons. First, findings from objectives 2 and 3 suggested that pre-existing anxiety disorders have a stronger association with return to work events and outcomes than pre-existing depression disorders. Second, recommended mental health services for anxiety disorders vary from those of depression disorders (although there are some similarities) and early descriptive analyses of the healthcare data suggested that physicians provide less mental health visits and less pharmaceutical treatment to anxiety patients than to depression patients (Table 27). This suggests that workers with anxiety and workers with depression should be analyzed separately for the purposes of this objective. Finally, future research could adapt the instrumental variable analyses methods for depression if the findings for workers with anxiety indicate that instrumental variables based on physician treatment preferences are a suitable methodological approach to examine associations between mental health treatment and return to work. Findings from objective 4 will inform future research methodology in this area.  Figure 4 visualizes the relationship examined in objective 4. The primary hypothesis for objective 4 was that among workers with pre-existing anxiety prior to injury, receipt of minimal adequate anxiety treatment, anxiety prescription services, and psychological counseling from a physician would each be associated with favourable return to work outcomes. For gender, it was hypothesized that associations between physician mental health services and time to return to work would be similar for men and women in terms of direction and effect size, as there is limited evidence that the aforementioned treatments are more effective for men versus women or vice versa.     29     Figure 4: Visualization of the relationships examined in research objective 4 As the direct relationships between receipt of physician mental health services and return to work were likely to be confounded by anxiety symptom severity (an unmeasured variable in the study data), the anxiety treatment preferences of each worker’s primary general practitioner provider was used as an instrumental variable (described more thoroughly in Chapter 7).  Anxiety treatment preferences of the worker’s primary general practitioner  Receipt of physician mental health services (minimal adequate treatment, prescriptions, and counseling) Return to work    30 Chapter 3: Databases and case definitions The purpose of the following section is to describe the jurisdictional context of the dissertation research, databases used for analyses, data extraction and linkage processes, data cleaning processes, and the case definitions for anxiety and depression disorders. Descriptions of the study samples, the study variables, and the analytic techniques, used to address each individual research objective, are provided in the individual chapters that follow. 3.1 Jurisdictional Context In Canada, injury or illness triggered or exacerbated by an event, activity, or exposure in the work environment, is considered work-related. Whether or not an injury or illness is work-related, determines in part, the social benefit systems responsible for covering the medical treatment and welfare of the injured person. Both workers’ compensation and public health systems are administered on a provincial or territorial level, within guidelines set by federal legislation.(88,89) Canadian workers’ compensation systems are ‘no fault’ insurance systems where workers give up the right to sue their employer in exchange for guaranteed wage-loss, medical and rehabilitation benefits in the case of work-related injury or illness. Meanwhile, public health systems oversee medically necessary health services for non work-related injury and illness.(88,89) These two health systems operate in parallel and share many of the same providers and resources.(90) Indeed, an injured worker can be treated within the public health care system while the cost of these services covered is by the workers’ compensation system.  In BC, the Ministry of Health oversees the administration of universal public health insurance for all BC residents (also known as the Medical Services Plan [MSP]). This includes medically required services provided by physicians, hospitals and supplementary health-care practitioners, and diagnostic procedures. By law, all BC residents are required to be registered in the plan and also to pay a monthly premium to help finance it. During the study period, monthly premiums were based on adjusted net income and paid by residents or their employer through a billing system separate from other tax payments. Under certain circumstances, the monthly premiums may be waived or reduced (e.g. unexpected change in financial situation due to illness, disability or loss of job).    31 The BC Ministry of Health also administers PharmaNet, a province-wide network that links all of BC’s pharmacies to a central data system. All dispensed medications and medical supplies from community and hospital outpatient pharmacies in BC are entered into PharmaNet as are some instances of medications provided by physicians to patients during an office, clinic, or emergency department visit. While dispensed medications and medical supplies are captured in the PharmaNet database, it is important to note that, for most residents, they are not covered by the public health care system and are usually paid out of pocket or by private health insurance. WorkSafeBC oversees the administration of workers’ compensation insurance for BC workers and is funded by employers through insurance premiums. In the event of a work injury or illness, WorkSafeBC provides compensation and support for lost wages and relevant medical treatment, as well as rehabilitation and return to work processes, for workers’ with an accepted claim. However, based on independent studies, it is estimated that the BC compensation system under-reports work-related serious injury by 10% to 15% (91) and that 18% of filed claims are not initially accepted.(92) Another study estimated that 40% of work-related injury and illness in Canada is not reported to workers’ compensation.(93) Evidence suggests bias in claim acceptance rates in Canada by injury type, occupation, and socio-demographic characteristics. For example, a higher likelihood of claim acceptance is associated with higher worker seniority compared to lower worker seniority, female gender compared to male gender, and acute physical injuries (e.g. cuts or puncture wounds) compared to non-acute or physical injuries with less easily verified etiology (e.g. allergy or irritation).(92,94) From 2000 to 2013 (the time period covered by this research), 93% to 95% of workers in BC were registered for workers’ compensation insurance through WorkSafeBC, and in more recent years, coverage rates have increased to 97%.(95) Self-employed individuals and small workplaces (e.g. 1-2 workers), and some federal workers covered under federal insurance schemes (e.g. federal police, military), were exempted from coverage under WorkSafeBC during the study time period.  3.2 Data files For the current study, administrative data files from the province of BC’s Ministry of Health and Workers’ Compensation Board (WorkSafeBC) were linked at the individual, person-level to answer the research questions. The data was obtained via Population Data BC, a multi-university, data and education resource that supports access to, and linkage of, data on BC’s 4.6   32 million residents for research on human health, well-being, and development.(96) To ensure privacy and confidentiality, all aspects of the research were conducted in accordance with federal and provincial privacy legislation governing access to data. Short descriptions of the data files and key variables used for this research are outlined here. WorkSafeBC files – included the claim file, claim cost summary file, employer file, employer classification file, employer operating location file, employer operating location classification file, return to work file, average earnings decision file, and firm level data file. The claim file was used to obtain information on the injury (date, primary and secondary International Classification of Diseases, ninth revision (ICD-9) codes, accident type, body part type, and nature of injury), claim eligibility (for compensation benefits), type of benefits received (health care, short term disability, long term disability, vocational rehabilitation, or fatality benefits), and occupation and industry. The average earnings decision file was used to obtain information on the worker’s pre-injury annual earnings and shift schedule. The return to work file was used to obtain information on return to work events and outcomes. All WorkSafeBC files are available from 1981-01-01 onward except for the following: return to work file available from 2009-01-01 onward, average earnings decision file available from 1989-01-01 onward, and firm level data available from 1997-01-01 onward.(97) The following files were obtained from the Ministry of Health: 1. Medical Registry file (Available from 1986-01-01 onward) - includes all individuals who received and/or were registered to receive services in BC. Fields include date of birth, sex, geo-codes describing the location of residence (first three digits of postal code, neighbourhood income decile and quartile, multiple geographic variables to describe residential location (e.g. health authority area and first three digits of postal code), MSP registration data (start date and number of days registered), and dependents (total number of residents under age 19 registered on the MSP contract for the given year).(98) 2. Medical services plan (MSP) file (Available from 1985-04-01 onward) – contains billing records for medical services by fee-for-service practitioners to individuals covered by MSP. Practitioners include physicians, other practitioners (physiotherapists, massage practitioners, naturopathic physicians, etc.), and out of province practitioners.  For this dissertation, billing data   33 for all services provided by fee-for-service practitioners including those paid for by MSP or workers’ WorkSafeBC were included in the file by Population Data BC. The file contains the following fields: service date, practitioner number, practitioner specialty code, type of service code, service units, fee item (indicates the insured service for which the practitioner was paid), paid amount, explanatory code (describes what happened to the claim), date paid, ICD-9 diagnostic codes, and claim type (insurer responsible for paying for the claim). Prior to 1991, no diagnosis codes are available. From 1991 to 1994, a single variable with an ICD-9 diagnosis was included. From 1995 onward, one variable to indicate the primary ICD-9 diagnosis associated with a billing record, along with four other variables to indicate secondary ICD-9 diagnoses were included, although only a primary diagnosis is required for billing (i.e. secondary diagnoses can be left blank).(99) MSP and WorkSafeBC both adhere to the 9th revision of the ICD coding system but do not distinguish between versions of the 9th revision (e.g. CM or AM). Practitioners are only required to enter the first three digits of the codes, diminishing the ability to detect or use subtle differences in diagnoses between the ICD-9 versions.  3. Hospital discharge abstract file (Available from 1985-04-01 onward) – contains data on hospital discharges, transfers and deaths of in-patients, and day surgery patients from acute care hospitals in BC. Fields include: admission date, discharge date, ICD-9 or ICD-10 diagnoses codes, and diagnosis type (relates each diagnosis to the patient’s hospitalization, e.g. primary or complication). From 1991 to 2000, a primary ICD-9 diagnosis and 15 secondary ICD-9 diagnoses were included. From 2000 onward, a primary ICD-10 diagnoses and up to 24 secondary ICD-10 diagnoses were included.(100) 4. PharmaNet file (Available from 1986-01-01 onward) – contains data on dispensed medications and medical supplies. For the current study, only medications with documented pharmacological treatment of anxiety or depression were included in the extract. Fields include: specialty code (practitioner specialty), drug identification number (unique drug ID assigned by Health Canada), service date, dispensed quantity (number of drug units), dispensed days (number of days until the dispensed medication runs out), and directions (number of drug units to be taken per day by the patient). (101)   34 In the current study, anxiolytics and anti-depressants were defined using the World Health Organizations Anatomical Therapeutic Chemical (ATC) classification system that classifies drugs based on the organ or system they act on, as well as their therapeutic use, chemical properties, mechanism of action, and route of administration.(102) In the PharmaNet extract provided by Population Data BC, drugs were classified using Health Canada Drug Identification Numbers (DIN) but not ATC codes. To account for this, drug products currently or previously approved for use in Canada with ATC codes beginning with N05B (anxiolytics) or N06A (anti-depressants) and their corresponding DIN were downloaded from the Drug Product Database publically available through the Health Canada section of the Government of Canada website.(103) This was then used as a crosswalk (ATC-DIN) to add ATC codes to the PharmaNet extract.  3.3 Data Extraction, Linkage, and Cleaning 3.3.1 Data extraction and linkage The data necessary to conduct the research was extracted, de-identified, and linked by a trained programmer at Population Data BC. For de-identification, names and other personal identifiers not necessary to conduct the research were removed from the data and replaced with a unique identifier that also acted as a crosswalk across data files. For linkage, both deterministic and probabilistic techniques were used and have been described elsewhere.(104,105) In general, Population Data BC achieves a linkage rate above 95% at the individual level across databases.(104) Prior to linkage and analyses, Population Data BC constructed an extract of all workers’ compensation claims based on the following criteria: 1. Injury date occurring between 01/01/2000 and 12/31/2013; 2. Age 18 to 64 years at the time of injury;  3. At least one of the following: a. A primary or secondary ICD-9 code indicating upper body soft tissue injury (338 pain not elsewhere classified, 710-719 arthropathies and related disorders, 720-724 disorders of the back, 725-729 rheumatism excluding the back, 831-834 dislocation of upper limb, 840, 841, 842, 846, 847); or   35 b. Using the body part code, indication that the injury affected one of the following body parts: neck, elbow, wrist, fingers, hand, other upper limb, back, shoulders, or trunk; or c. Using the nature of injury code, indication of tendinitis, tenosynovitis, back strain, or other strain. Using the earliest injury date from each claim as the index date, all available data from the pre-injury time period (as far back as the data would allow) and two years following injury, were extracted for each person identified in the data extract from WorkSafeBC, Medical Services registry and health records, hospital discharge abstract, and the PharmaNet files. Inclusion of historical data in the extract allowed for flexibility in defining the pre-existing co-morbidity and prior claim variables during analyses. As the data was limited to two years of follow up time after the earliest injury date associated with a claim, workers’ compensation events and payments associated with a claim, but occurring more than two years after the initial injury date, were not captured in the data extract. Follow up time was determined at the level of the claim and not the level of the worker. In this way, workers could have multiple claims in the extract, including multiple claims with the same diagnostic code. In BC, when workers seek compensation for a work-related injury or illness that is possibly related to a previously compensated injury or illness, it is up to the case manager to adjudicate the etiological nature of this relationship. The case manager then makes an informed decision as to whether or not the subsequent injury or illness should be registered under the prior claim or as a new one. After data extraction, de-identification, and linkage, the study database was accessed and analyzed by the researcher using a secure research environment maintained by Population Data BC. Analytic cohorts were defined from the study database for each research objective. More refined inclusion and exclusion criteria for the analytic cohorts are described in subsequent chapters. 3.3.2 Data cleaning The study database was checked for potential data entry errors, duplicates, and entries not relevant to the research objectives. Below is a description of the key data cleaning processes instrumental to conducting the research:      36 1. Claims data  First, to ensure that only accepted claims remained in the study database, claims with a value of ‘rejected’ or ‘disallowed’ for the claim eligibility ID variable or ‘withdraw’ for the claim active status variable, were removed.   2. Medical services data Over 30% of the entries in the medical services plan information file were for administrative billing purposes only (e.g. reverse billing or credit adjustments) and were not representative of a unique medical service or visit. To resolve duplication issues, claims with exact matches for the following variables: study id (unique specific id applied to each individual in the data), service date (the date on which the service was rendered), practitioner number (used to identify the practitioner who rendered the service), fee item (numeric code to identify the service rendered), and claim type (identifies the type of practitioner and the source of payment for the fee item) were grouped together and referred to as ‘fee item groups’. The total costs associated with a fee item group was calculated as the sum of the costs of all fee items in the group. Items belonging to fee item groups with a total sum cost of 0 or less were then deleted.  After accounting for reverse billing and credit adjustments, the medical services plan information file was restricted to entries for physician visits. Diagnostic procedures, other medical procedures, no charge referrals, and services from non-physician practitioners were removed from the file for the purposes of this study (with one exception in the analytic study on physician mental health services when visits, procedures, and diagnostic tests were all used to identify the general practitioner that provided the largest amount of each workers’ primary health care). Physician visits were identified using the specialty (describes a practitioner’s specialty e.g. psychiatry or general practitioner) and service code (e.g. e.g. regional examinations, consultation, counseling, institutional visits) variables. 3. PharmaNet data Records from the PharmaNet file that included a record of administrative processing but no record of a dispensed pharmaceutical were removed. These represented 0.28% of the records in the PharmaNet file.   37 Data entries with exact matching values for the following variables were considered duplicate records and deleted from the database: study id, service date (dispensing date), DIN, dispensed quantity (number of units dispensed), and dispensed days (the time period to be covered by the medication). These represented less than 0.01% of the PharmaNet records. 4. Hospital discharge data In the hospital discharge abstracts, from 1985 to 2007 when ICD-9 was the primary coding system, questionable diagnoses were prefixed with a Q. To prevent misclassification of medical conditions, these were removed from the hospital data and not considered in the construction of the study variables. These represented less than 1% of the total hospital diagnoses from this time period. 3.4 Anxiety and depression case definitions 3.4.1 Case definitions reported on in the literature There are no ubiquitous or gold standard case definitions for anxiety and depression in administrative data.(106) The validity of a given case definition can vary across data sources and study samples due to differences in billing and recording processes and the frequency of off-label prescribing, amongst other reasons.(107,108) Internal aspects of a case definition such as the list of diagnoses used to identify the condition of interest, and the number of years of data taken into consideration, can also affect validity.(109) Reported measures of validity for anxiety and depression case definitions vary widely. However, an appropriate balance of sensitivity and specificity can be developed for research purposes. A 2005 systematic review found that most validation studies draw positive conclusions on the use of health claims data for the identification of psychiatric research. (106)  To facilitate the development of case definitions for anxiety and depression in this study, a search of the peer review and grey literature was conducted. Case definitions for anxiety and depression previously applied to Canadian health claims or similar data sources were identified, with particular emphasis on those previously applied to BC health data. Findings from the search are summarized in Table 1. Four studies each with a unique case definition for depression previously applied to BC data sources were identified. Data sources for these BC studies included health claims data, electronic medical records, and primary care sentinel surveillance   38 data.(108,110–112) No measures of validity were reported for any of the BC studies examining depression; and no studies using BC health data to identify anxiety were found. Multiple measures of validity based on comparisons with patient health records, self-reported diagnoses, and the Composite International Diagnostic Interview (CIDI) were reported for anxiety and depression case definitions using health claims data from other Canadian provinces.(107,113–116) Compared to patient health records, that are more detailed, sensitivity for case definitions using Canadian health claims can be as high as 65% for anxiety and 76% for depression, and specificity can be as high as 100% for anxiety and 97% for depression.(114–116)  In general, anxiety and depression case definitions for health claims data vary based on the following factors: - Event type(s): definitions typically consider one or more of the following types of health service contact events: diagnoses from outpatient physician services, diagnoses from hospitalizations, and pharmaceutical dispensing.  - Diagnostic codes or pharmaceuticals lists: the ICD-9 or ICD-10 codes, or pharmaceuticals used to identify anxiety or depression cases in health claims data can vary across case definitions.  - Primary versus secondary diagnoses: Physician billing and hospital data can contain both primary and secondary diagnoses. To improve specificity, some case definitions consider only primary diagnoses, while others to improve sensitivity, consider both primary and secondary ones.  - Clinician specialty: most case definitions include diagnoses from all clinician types, however some are restricted to diagnoses from general practitioners and psychiatrists, or psychiatrists only. - Number of events: the number of events necessary to meet the case definition is sometimes dependent on the type of event being considered (e.g. either one hospitalization or two physician visits with a diagnosis of depression is often sufficient to meet the case definition). One or two events are most common but some algorithms require more (Table 1). In addition, the length of time allowed to pass between events (time frame), in order for them to both count towards a cumulative threshold necessary to   39 meet the case definition can vary. Generally, as the number of required events increases, sensitivity decreases and specificity increases.(113,115,116)  - Time frame: Typically, a time frame of 1 or 2 years is used to count the number of events (Table 1) and the beginning and end of the time frame are defined using calendar time (e.g. January 1st 2010 to December 31st 2010). Longer time frames are expected to result in higher sensitivity especially when cumulative cases are being identified, however they can also result in lower specificity.(113) A consideration is that identification of cumulative cases across long time frames may not be appropriate for episodic conditions such as depression depending on the nature of the research objectives. A person with depression at some point in a five-year time frame may not actively experience depression for the vast majority of that time – and this may have important implications for some research objectives, but not others. In addition, how frequently people with depression or anxiety would be expected to have a relevant health system contact event must be taken into consideration when determining the time frame and the number of events necessary to meet the case definition. To our knowledge, the shortest time frame for anxiety and depression case definitions (used on Canadian health claims data), that required more than one event to meet the case definition, was one year. This was likely due to a relatively low use of mental health services (e.g. 1 to 2 physician visits in a year) for many true cases (Table 1). - Chronicity: in a chronic approach, once the case definition is met, an individual is considered to have the condition for the remainder of the study (i.e. indefinitely). In an episodic approach, once a case definition is met, the individual is only considered to have the condition for a defined time period, or certain criteria must be met in order for the condition to be considered ongoing. - Prevalent versus incident cases: case definitions for incident cases (or episodes) typically require that the incident case be preceded by a defined time period with no relevant health care events for the condition of interest, but this is not necessary for prevalent cases. For identifying newly treated depression cases, West et al required a time period of 180 days with no antidepressant dispensing events prior to an index date,(115) and Puyat et al. required a time period of 365 days with no physician or hospital diagnoses for depression.(111)    40 Table 1: Anxiety and depression case definitions previously applied to Canadian health claims data  Author Data source Study sample Time frame Event type(s) Case definition Validity Anxiety       Marrie (2013)b(113) Manitoba health claims data Multiple sclerosis patients 2 years All hospital diagnoses (up to 16). Only 1 diagnosis available on physician claims.   ICD-9 a: 300.0, 300.2 ICD-10: F40, F41 Rx: Anxiolytics including N05AB12, N05AB06 ≥1 H or ≥ 2 P or (≥ 1 P and ≥ 2 Rx) Compared to medical records: 42% sensitivity  82% specificity 14% PPV 95% NPV Fair agreement (k=0.23) Marrie (2014)(107)  Manitoba and Nova Scotia health claims data Multiple sclerosis patients 2 years Primary diagnoses only d ICD-9 a: 300.0, 300.2 ICD-10: F40, F41 ≥ 1 H or ≥ 3 P Compared to self reported conditions (Manitoba/ Nova Scotia): 29%/.62% sensitivity  93% / 74% specificity 41%/ 34% PPV 89%/ 90% NPV Fair agreement (k= 0.26/ 0.27) Marrie (2016)(114)  Manitoba health claims data Inflammatory bowel disease patients 2 years All hospital diagnoses (up to 25). Only 1 diagnosis available on physician claims.   Case definition A with no Rx: ICD-9 a: 300.0, 300.2 ICD-10: F40, F41  Case definition B with Rx: ICD-9 a: 300.0, 300.2 ICD-10: F40, F41 Rx: Anxiolytics including N05AB12, N05AB06      ≥ 1 H or ≥ 3 P    ≥ 1 H or ≥ 2 P or (≥ 1P and ≥ 2 Rx)  Compared to self-reported physician diagnoses  (Case def’n A/ Case def’n B): 46%/ 62% sensitivity 85% / 71% specificity 26%/ 19% PPV 94%/ 95% NPV Fair agreement (k=0.23/ 0.16)  Compared to the Composite International Diagnostic Interview: 30%/ 49% sensitivity 88% / 76% specificity 53%/ 48% PPV 74%/ 77% NPV Fair agreement (k=0.21/ 0.25)            41 Author Data source Study sample Time frame Event type(s) Case definition Validity Depression West (2000) (115)  Saskatchewan health claims data General population 2 years Only 1 diagnosis available on physician claims.  ICD-9: 296, 309, 311  Rx: Antidepressants  ≥ 1 P and  ≥ 1 Rx Compared to medical records: 71% sensitivity 85% specificity 86% PPV 70% NPV Marrie (2013)b(113) Manitoba health claims data Multiple sclerosis patients (plus matched controls with no MS) 2 years All hospital diagnoses (up to 16). Only 1 diagnosis available on physician claims.  ICD-9 a: 296.2, 296.3, 298.0, 300.4, 311 ICD-10: F32, F33, F34 Rx: N06AA01, N06AA02, N06AA04, N06AA11, N06AA12, N06AA17, N06AA21, N06AB03, N06AB04, N06AB05, N06AB06, N06AB08, N06AB10, N06AF03, N06AF04, N06AG02, N06AX06, N06AX11, N06AX16, N06AX21, N06AX23 ≥1 H or ≥ 2 P or (≥1 P and ≥ 2 Rx) Compared to medical records: 68% sensitivity  73% specificity 49% PPV 86% NPV Moderate agreement (k=0.45) Puyat (2013) (112) BC electronic medical records General population 1 year All diagnoses ICD-9: 296, 311  Rx: Antidepressants ≥ 1 P or ≥ 1 Rxc Not available. Marrie (2014)(107) Manitoba and Nova Scotia health claims data Multiple sclerosis patients 2 years Primary diagnoses only  ICD-9 a: 296.2, 296.3, 298.0, 300.4, 311 ICD-10: F32, F33, F34  ≥ 1 H or ≥ 4 P  Compared to self reported conditions (Manitoba/ Nova Scotia) 55%/49% sensitivity  94% / 82% specificity 74%/ 53% PPV 87%/ 80% NPV Moderate agreement (k= 0.53/ 0.37)                  42 Author Data source Study sample Time frame Event type(s) Case definition Validity Depression Williamson (2014)(116) Canadian Primary Care Sentinel Surveillance Network data repository General population Not relevant for single events All diagnoses ICD-9: 296, 311  Rx: N06CA01, N06AB04, N06AB10, N06AB03, N06AB08, N06AX11, N06AG02, N06AB06, N06AF04 ≥ 1 P or ≥ 1 Rxc Compared to electronic medical records:  81% sensitivity 95% specificity 80% PPV 95% NPV Chronic Disease Information Working Group (2015)(110) BC health claims data General population 1 year All diagnoses ICD-9: 296, 311, 50b 1CD-10: F32, F33 ≥1 H or ≥ 2 P Not available. Marrie (2016)(114)  Manitoba health claims data Inflammatory bowel disease patients 2 years All hospital diagnoses (up to 25). Only 1 diagnosis available on physician claims.   Case definition A with no Rx: ICD-9 a: 296.2, 296.3, 298.0, 300.4, 311 ICD-10: F32, F33, F34  Case definition B with Rx: ICD-9 a: 296.2, 296.3, 298.0, 300.4, 311 ICD-10: F32, F33, F34 Rx: N06AA01, N06AA02, N06AA04, N06AA11, N06AA12, N06AA17, N06AA21, N06AB03, N06AB04, N06AB05, N06AB06 N06AB08, N06AB10, N06AF03, N06AF04, N06AG02, N06AX06, N06AX11, N06AX16, N06AX21, N06AX23      ≥ 1 H or ≥ 4 P     ≥ 1 H or ≥ 5 P or (≥ 1P and ≥ 7 Rx)  Compared to self-reported physician diagnoses  (Case def A/ Case def B): 77%/ 70% sensitivity 89% / 89% specificity 60%/ 57% PPV 95%/ 93% NPV Moderate agreement (k=0.59/ 0.54)  Compared to the Composite International Diagnostic Interview: 50%/ 47% sensitivity 89% / 89% specificity 65%/ 64% PPV 81%/ 80% NPV Moderate agreement (k=0.42/ 0.39)               43 Author Data source Study sample Time frame Event type(s) Case definition Validity Depression Puyatt (2016)(111)  BC health claims data General population Not relevant for single events Primary diagnoses only ICD-9: 296, 296.2, 296.3, 311  ICD-10: F32, F33, F39 ≥1 H or ≥ 1 P Not available. a ICD-9 codes from physician claims truncated to the third digit b Examined multiple algorithms – the one closest to the current study is presented c The medication criteria alone is insufficient if ICD-9 diagnosis 300 (Anxiety disorders) is present H= diagnosis associated with a hospital visit; P = diagnosis associated with a physician visit; Rx = prescription    44 3.4.2 Case definitions selected for this research Case definitions were developed to address each research objective and are described here. The same set of case definitions for anxiety and depression were used for Chapters 4, 5, and 6, while different definitions were used for Chapter 7. These were informed by the context and objectives of each study as well as the case definitions presented in Table 1 and the frequency of health service events for anxiety and depression in the health claims data (Table 27, Table 28, and Table 29 in the Appendix, p. 191 to 192).  3.4.2.1 Case definitions used for Chapters 4, 5, and 6 1. Event types For the current study, three event types were used: diagnoses from outpatient physician visits, diagnoses from hospitalizations, and pharmaceutical dispensing events (anti-depressants and anxiolytics).  As described in the previous section, diagnostic codes from the physician billing data were limited to those associated with a physician’s visit. Diagnoses associated with outpatient health services delivered by other practitioners were excluded, as the diagnoses of mental health disorders is exclusive to physicians’ scope of practice. Diagnoses from services not primarily defined as patient visits were excluded to prevent false positives due to 1) query-based services such as diagnostic tests (although these are not commonly used for anxiety and depression diagnosis as is the case for other conditions like diabetes), or 2) potentially unconfirmed diagnoses such as those associated with a no-charge referral from a general practitioner to a psychiatrist. Lastly, counts of physician visits were limited to one per worker per day, although in the event that multiple physician visits for a given worker were identified in the data for a single day – all diagnoses from these multiple visits were considered.  2. Diagnostic codes and pharmaceutical list Anxiety:  i. Physician visits with an ICD-9 code for anxiety (300 anxiety, dissociative and somatoform disorders, 308 acute reaction to stress, and 309 adjustment reaction); ii. Physician visits with code 50b that is a diagnostic code unique to BC’s health services plan used to indicate anxiety, depression, or both anxiety and depression together;   45 iii. Hospitalizations with an ICD-9 or ICD-10 code for anxiety (ICD9 codes: see previous list; ICD-10 codes: F4 anxiety, dissociative, stress-related, somatoform and other nonpsychotic mental disorders, F68 other disorders of adult personality and behavior, and F341 persistent mood [affective] disorders); and  iv. Prescriptions with an Anatomical Therapeutic Chemical (ATC) code of N06A (anti-depressants) or N05B (anxiolytics). Anti-depressants such as SSRIs and SNRIs are recommended first-line pharmacotherapy agents for panic disorders, social anxiety disorders, and generalized anxiety disorders. Anxiolytics such as benzodiazepines are efficacious for anxiety but are recommended as second line agents due to side effects, dependence, and withdrawal issues.(117) Recommended pharmacotherapy agents for obsessive-compulsive disorders and posttraumatic stress disorders include anti-depressants, but not anxiolytics based on a lack of efficacy for anxiolytics.(117)  Depression  i. Physician visits with an ICD-9 code for depression (311 depression disorder, not elsewhere classified, and 296 episodic mood disorders); ii. Physician visits with code 50b (as per above); iii. Hospitalizations with an ICD-9 or ICD-10 code for depression (ICD-9 code: see previous list; ICD-10 codes: F3 mood [affective] disorders [with the exception of F341 that was coded as anxiety to be consistent with the ICD-9 coding system); and iv. Prescription fills with an ATC code of N06A (anti-depressants). Anxiolytics are not a recommended pharmacotherapy agent for depression so they were not included in the depression case definition.(118)   The use of unique code 50b for anxiety/depression in the BC medical services data poses a unique challenge for identifying anxiety and depression cases in health claims data not present in other provinces. For the first three analytic chapters, case definitions were chosen based on the assumption that a diagnosis code of 50b is an indicator for both anxiety and depression. This is based on findings from Jones (2002) who found that, of physicians in BC who submitted at least one claim with a diagnosis of either code 50b or ICD-9 code 311 (major depression) in a year, 28% exclusively used code 311 throughout the year (i.e. they submitted no claims with code 50b), and only 7% exclusively used code 50b.(119) The remaining 64% used both. For the current study, the assumption is that the majority of physicians who use both code 50b and 311   46 do so differentially – in that 311 is used to indicate major depression only while code 50b is used to indicate a combination of anxiety and depression symptoms. Further, based on the findings from Jones (2002), the assumption is that approximately 7% of physicians in BC use code 50b non-discriminately for patients with anxiety only, depression only, or co-morbid anxiety and depression. While one option is to exclude code 50b from the case definitions, this was deemed inappropriate for the first three analytic chapters as it accounts for 30% of all anxiety and depression diagnoses in the physician billing data. Three digit ICD-9 codes can include up to two decimal digits for diagnostic specificity (e.g. xxxx.xx, xxx.x, or xxx). In the MSP data, it is common to code diagnoses using the three digits with no decimals, even though a more specific diagnosis may be indicated. To account for this, anxiety and depression case definitions for the current study using just the first three digits are umbrella definitions for a number of anxiety and depression disorders respectively. Notably, ICD-9 code 296 (included in the diagnoses list for depression) includes subcategories for depression disorders and bipolar disorders. Bipolar disorder is considered a type of depression under some grouping systems, but not others, due to its symptoms that include both depressive lows and manic highs. Very few of the depression cases identified in the current study will likely have bipolar disorder as bipolar disorder is relatively rare in the general population (2%).(120) The high specificity values for depression case definitions based on three digit ICD-9 codes presented in Table 1 support the latter statement.  3. Primary versus secondary diagnoses It was determined a priori that the case definitions would include both primary and secondary diagnoses from the physician billing and hospital data. As anxiety and depression are frequently co-morbid with chronic physical health conditions, as well as other mental disorders, it was hypothesized that anxiety and depression diagnoses may frequently present in the secondary diagnoses fields. During analyses, 98% of the anxiety and depression diagnoses identified from the physician billing and hospital data were primary diagnoses, indicating that restricting the case definitions to primary diagnoses only would have minimal impact on validity. Likewise, Marie et al. (2014) found that case definitions for anxiety and depression based on primary diagnoses only, versus all diagnoses (up to three), performed identically using health claims data in the Canadian province of Nova Scotia.(107)    47 4. Clinician specialties No further restrictions were placed on the case definitions based on clinician specialty. During the analyses and after data cleaning, 99% of the anxiety and depression diagnoses associated with a physician visit were from either a general practitioner (79%) or a psychiatrist (21%).  5. Time frame, case definitions, and number of events The current study used a rolling time frame to count the frequency of anxiety and depression related health care events (based on the timing of the individual injury events) (Figure 5) rather than a fixed time frame based on set calendar dates, as is frequent in other studies (e.g. January 1st 2001 to December 31st 2001). First, for each claim, all anxiety and depression related physician visits, pharmaceutical dispensing, and hospitalization events were identified for all years available in the data. Second, an iterative process was begun, whereby one by one, each individual anxiety or depression related health care event identified in the previous step was selected (referred to as the index event when under selection), and a set of rules (outlined in Table 2 and Table 3) were used to determine if the worker from the claim was an anxiety or depression case at the time of the index event. As described in Table 2 and Table 3, to determine if the conditions for anxiety or depression at the time of a given index event (i.e. healthcare event) were met, all other anxiety and depression related health care events from 365 days before and the 365 days after the index event were taken into consideration.  The studies in this dissertation measure the one-year period prevalence of anxiety and depression for both the year before injury and the year after injury. Workers were considered to have prevalent anxiety during the year before injury if they had at least one index event (i.e. anxiety or depression related health care event) occur during the year before injury that met the conditions of the anxiety case definition (Figure 5 and Table 2). The same method was used for the year post injury, as well as for depression. In Chapters 4 and 5, new onset anxiety and depression in the post injury time period (defined as either the first three months following injury, or the time period from injury to sustained return to non-modified work) is also examined in secondary analyses. To be considered at risk for new onset anxiety or depression during the post-injury time period, the worker could not have any depression or anxiety related events in the year prior to their injury.   48  Figure 5: Time frames used to identify index events and secondary validation events in the anxiety and depression case definitions aThe index event refers to the anxiety or depression related health care event currently under examination. Each anxiety or depression related health care event identified in the time periods of interest (one year pre-injury and one year post- injury) was examined as an index event at least once. This was necessary as several anxiety or depression related health care events can occur within a given time period, and some of these may meet the requirements of the case definitions, while others may not. In the figure, an anxiety or depression related health care event occurred at 60 days post injury. To determine if this event met the requirements of the case definitions, secondary validation events occurring in the 365 days before it and 365 days after it were identified and considered. Using a time scale where the injury is day 0, the timeframe used to validate the index health care event occurring at 60 days post injury corresponds to the time spanning between the 305 days before injury (60-365) and the 425 days after injury (60+365).   Injury 0 +365 -365  Time frame to capture index events for prevalent anxiety or depression in the year post-injury Time frame to capture index events for prevalent anxiety or depression in the year pre-injury Index eventa +60 Time in days Time frame to capture secondary events to validate the index event* occurring 60 days post injury +425 -305     49 Table 2: Case definition for anxiety disorders in the Ministry of Health datasets Index eventa Anxiety disorder at the time of the index event– case definition Anxiety diagnosis - hospital Yes Anxiety diagnosis - physician visit Yes, if there is at least one of the following within 365 days before or 365 days after the index event:  Anxiety diagnosis – hospital  Anxiety diagnosis – physician visit  Code 50b   Antidepressant  Anxiolytic Code 50b – physician visit Yes, if there is at least one of the following within 365 days before of 365 days after the index event:  Anxiety diagnosis – hospital  Anxiety diagnosis – physician visit  Anxiolytic  Antidepressantb   Code 50bb Anxiolytic Yes, if there is at least one of the following within 365 days before or 365 days after the index event:  Anxiety diagnosis – hospital  Anxiety diagnosis – physician visit  Code 50b Antidepressant Yes, if there at least one of the following within 365 days before or 365 days after the index event:  Anxiety diagnosis – hospital  Anxiety diagnosis – physician visit  Code 50bb a Each incidence of an anxiety or code 50b diagnosis from a physician visit or hospital admission, or the dispensing of an anxiolytic or antidepressant was considered as an index event. A worker was considered to have anxiety at the time of the index event if the conditions outlined in the table were met bCriteria meets both the anxiety and depression case definitions     50 Table 3: Case definition for depression cases in the Ministry of Health datasets Index eventa Depression disorder at the time of index event– case definition Depression diagnosis - hospital Yes Depression diagnosis – physician visit Yes, if there is at least one of the following within 365 days before or 365 days after the index event:  Depression diagnosis – hospital  Depression diagnosis – physician visit  Code 50b   Antidepressant Code 50b – physician visit Yes, if there at least one of the following within 365 days before or 365 days after the index event:  Depression diagnosis – hospital  Depression diagnosis – physician visit  Antidepressantb  Code 50bb Antidepressant Yes, if there at least one of the following within 365 days before or 365 days after the index event:  Depression diagnosis – hospital  Depression diagnosis – physician visit  Code 50bb a Each incidence of a depression or code 50b diagnosis from a physician visit or hospital admission, or the dispensing of an antidepressant was considered as an index event. A worker was considered to have depression at the time of the index event if the conditions outlined in the table were met bCriteria meets both the anxiety and depression case definitions 3.4.2.2 Additional Considerations for the Case Definitions In BC, fee-for-service physicians provide the majority of the province’s outpatient mental health services included under public health insurance. Such services were included in the MSP file used in this study. However, during the study period some publically insured mental health services were available via community-based clinics with salaried clinicians. The latter were not captured in the data files used for this research due to the aforementioned alternative payment structure. This is not expected to have a major impact on the estimates of prevalence in the current research as Kisely et al. (2009) found that including data from community-based clinics only increased the prevalence estimate for all psychiatric disorders in BC by 1%.(121) Second, the case definitions outlined in Table 2 and Table 3 were chosen to account for the potential low specificity of some health care events for the condition indicated while maintaining a reasonable level of expected sensitivity. For example, in addition to anti-depressants being a recommended first-line pharmacotherapy agent for both anxiety and depression, anti-depressants and anxiolytics are both prescribed for other approved and non-approved uses.(122) Notably, anxiolytics are sometimes prescribed short term for depression, especially in the early stages if temporary anxiety symptoms are high, even though this is not supported by current clinical   51 guidelines.(123,124) Validation studies of diagnoses from BC physician billing data indicate that diagnoses are moderately reliable and a valuable reflection of the primary problem treated;(125) however, there remains the possibility that diagnoses may be uncertain or contain coding error. To account for these issues, none of the following were sufficient to meet the anxiety or depression disorder case definitions on their own: i) one or more anxiolytic dispensing events, ii) one or more anti-depressant dispensing events, and iii) a single diagnosis associated with a physician visit. A single physician visit had to be accompanied by a secondary health system contact event also indicating the same type of disorder (anxiety or depression); and, pharmaceutical dispensing event(s) had to be accompanied by a relevant diagnosis from either a physician visit or hospitalization within 365 days before or after the index event. For a variety of chronic conditions (including depression), two outpatient or one inpatient event(s) occurring within a 365 day time frame is a generally accepted rule for producing a case group with adequate assurance of having the condition of interest.(126)  3.4.2.3 Consideration of alternative case definitions  The case definitions selected for the first three analytic chapters were chosen to increase the probability that those classified as not having anxiety truly did not have an anxiety episode, and that those classified as having anxiety truly did have at least one anxiety episode during the time period of interest(s), and likewise for depression; without having to exclude an unreasonably large proportion of the study sample, although some exclusions were permitted. Multiple strategies were used to achieve this. In addition to the strategies already described, a differentiation was made between claims with 1) no anxiety or depression health care events (referred to as the none group), and 2) anxiety or depression health care events not sufficient for the conditions of the case definitions (referred to as the case-like group for depression or anxiety; e.g. a single diagnosis from a physician visit not accompanied by a relevant secondary event, or the use of anxiolytics or anti-depressants without an accompanying diagnosis of anxiety or depression). Using the primary case definitions, claims were classified into five groups: anxiety only, depression only, anxiety and depression (both), none, and case-like. For most analyses, the case-like group was excluded. A similar approach was used by Mahar (2017) to examine the impacts of severe psychiatric illness on cancer outcomes using health claims data.(127)   52 In addition to the inclusion of a case-like group for the anxiety and depression variable, other key ways that the primary case definitions varied from those already reported on included the use of 1) anti-depressants (in addition to anxiolytics) on the pharmaceutical list for anxiety, and 2) code 50b on the diagnoses lists for both anxiety and depression. External cross validation of the primary case definitions with patient records was outside the scope of this project. However, to examine the effects of including certain events (e.g. code 50b diagnosis) in the primary case definitions, five alternative case definitions were developed. The classification of anxiety and depression by the alternative case definitions was then compared to the primary case definitions (Table 4) using the study sample from objective 1a. The alternative case definitions differed from the primary case definitions in the following ways: Alternative 1  No code 50b criteria: Code 50b for anxiety/depression was removed from the diagnosis list used to identify depression events, and the diagnosis list used to identify anxiety events. Alternative 2  No antidepressant criteria for anxiety: Anti-depressants were removed from the pharmaceutical list used to identify anxiety events. Alternative 3 No pharmaceutical criteria (No Rx): The pharmaceutical data (dispensing of anxiolytics or anti-depressants) was not used to identify anxiety events or depression events. Alternative 4 No code 50b criteria and no prescribing criteria (No Rx and No 50b): as described above. Alternative 5 1P or 1H: One physician or one hospital visit with an appropriate diagnosis is sufficient to meet the case definition. No code 50b criteria and no Rx criteria (as described above). The distribution of anxiety and depression classifications based on the alternative case definitions did not vary dramatically from the distribution based on the primary case definitions (Table 4). The ‘anxiety and depression’ group had the least stable membership, and claims assigned to this group by the primary case definition were often re-classified to other groups by the alternative case definitions. The majority of claims classified as anxiety only or depression only by the primary case definition, were reassigned to these same case groups by the alternative case definitions (51.7% to 96.9% of the time), and a minority were reclassified to the none and   53 the case-like groups. Many of the workers classified as case-like by the primary definition were re-classified as none by the alternative case definitions. Compared to the alternative case definitions, the primary case definitions took into account the greatest amount of information regarding health care utilization for anxiety and depression. Due to this, they were able to identify the greatest number of anxiety and depression cases as well as potential false negatives (i.e. separation of the case-like group from the none group). This is suggestive of a higher sensitivity (identification of a greater number of cases) and also higher negative predictive value for the primary case definitions compared to the alternative ones. Based on these findings and other considerations reviewed above in this chapter, the primary case definitions were consistently used for anxiety and depression classifications throughout the first three analytic chapters (Chapters 4, 5, and 6)    54 Table 4: Anxiety and depression case status for the year prior to injury: A comparison of the primary and alternative case definitions (column percentages reported) using BC claims for spine or upper limb strain or sprain, 2000-2013 Alternative case definition classifications Primary case definition classifications  N (%) New distribution b 294,912 None 211,737 (71.8) Case-like 25,951  (8.8)  Anx only 16,237 (5.5) Dep only 11,187 (3.8) Anx & dep a 29,800 (10.1) 1. No code 50b criteria      None      Case-like      Anx only      Dep only      Anx & dep   100 0 0 0 0   13.3 86.7 0 0 0   4.2 10.3 85.5 0 0   0.97 2.13 0 96.9 0   10.6 14.7 8.7 15.8 50.3   74.3 9.8 5.6 5.3 5.1 2. No anti-depressant criteria for anxiety      None      Case-like      Anx only      Dep only      Anx & dep   100 0 0 0 0   0 100 0 0 0   0 14.5 85.5 0 0   0 0 0 100 0   0 0 0 27.6 72.4   71.8 9.6 4.7 6.6 7.3 3. No Rx criteria      None      Case-like      Anx only      Dep only      Anx & dep  100 0 0 0 0  56.6 43.4 0 0 0  18.5 23.3 58.2 0 0  16.4 13.6 0 70.1 0  8.4 6.7 8.3 21.6 55.1  79.3 6.3 4.0 4.8 5.6 4. No code 50b or Rx criteria      None      Case-like      Anx only      Dep only      Anx & dep   100 0 0 0 0   70.0 30.0 0 0 0   26.8 21.5 51.7 0 0   17.3 16.1 0 66.6 0   34.8 10.6 11.6 21.2 15.9   83.6 5.5 4.0 5.3 1.6 5. 1P or 1H (no code 50b and no Rx)      None      Case-like      Anx only      Dep only      Anx & dep   100 0 0 0 0   70.0 0 20.9 8.5 0.7   26.8 0 70.4 0.3 2.5   17.3 0 0 80.9 1.7   34.7 0 12.8 26.0 26.4   83.6 0 7.0 6.5 2.9 a  Includes workers who met both the anxiety and the depression case definitions during the year prior to injury b The new distribution refers to the proportion of claims in each level of the alternative case definition 3.4.2.4 Case definitions used for Chapter 7 Chapter 7 examined the impact of physician mental health services on return to work among a sample of injured workers with pre-existing anxiety at the time of musculoskeletal work injury. Different methods for identifying anxiety (used as a study sample selection criteria) and depression (included in the analysis as a potential confounder) than those described above were   55 used for multiple reasons. First, the anxiety and depression case definitions for Chapter 7 were chosen based on methodology from a similar study that used BC administrative data to examine mental health care utilization among patients with major depression.(128) This study did not consider pharmaceutical data or code 50b in its case definitions. Second, as anxiety severity was an important unmeasured confounder in the analyses for Chapter 7, it was important that the study sample consisting of workers with pre-existing anxiety be as homogenous as possible in terms of anxiety severity. It has been suggested that some practitioners use MSP specific code 50b anxiety/depression to indicate less severe symptomology (129) and findings from Chapter 5 supports this. Excluding pharmaceutical events from the case definitions can also increase homogeneity of the anxiety case group as this then requires all anxiety cases to have received at minimum two diagnoses from a physician, rather than one diagnoses and one prescription dispensing event which is a less specific requirement. Further, as the analyses for Chapter 7 was conducted more recently than the analyses for the other chapters, there were informed by more recent literature. By the time of the Chapter 7 analyses, a new validation study of depression diagnoses in BC health claims data had been published and it became apparent that other authors in this area are choosing to omit code 50b from their case definitions.(130) For these reasons, the MSP specific code 50b anxiety/depression and pharmaceutical events were excluded from the anxiety and depression case definitions in Chapter 7. In Chapter 7, all workers with one or more diagnoses for depression (ICD-9 code 311 or 296; and ICD-10 code F3 [F341 excluded]) associated with a physician or hospital visit in the two years before injury were classified as having depression. The case definition for anxiety in Chapter 7 was as follows: at least one diagnoses of anxiety (ICD-9 code 300, 308, 309; and ICD-10 code F341, F4, and F68) in the year prior to injury associated with a hospital or physician visit (referred to as the index diagnoses) and at least one additional diagnosis of anxiety (meeting the same criteria as the index diagnoses) in the 12 months preceding the index diagnoses.     56 Chapter 4: Prevalence, timing, and risk factors of anxiety and depression (Research Objective 1) Background Studies of workers with lost-time musculoskeletal injury suggest an elevated prevalence of self-reported depression symptoms and disorders during the post-injury time period compared to the pre-injury time period, and compared to general populations.(4,10,11,35–37,131) Self-reported anxiety symptoms and disorders also appear to be more prevalent after lost-time musculoskeletal injury than in general populations, but evidence concerning the timing of anxiety onset relative to the injury is not conclusive.(10,11,35,37,131) According to cognitive models of clinical depression such as the diathesis-stress model, depression develops in chronic pain patients due to an underlying vulnerability (diathesis) that occurs in combination with a life stressor.(26) In the context of physical work injury, possible stressors that could lead to the development of post-injury depression or anxiety include: injury-related trauma, pain, functional disability, work disability,(26)  other types of social role disability that affect one’s family or social life, financial strain; as well as negative interactions with the health care system, the compensation system, or the employer.(26,132,133)  In general populations, anxiety and depression disorders are more prevalent among women (compared to men), people with lower socio-economic status (compared to people with higher socio-economic status), and people with other physical or mental co-morbidities (compared to people with no or few co-morbidities).(40,47,134) Knowledge of risk factors for prevalent or new onset anxiety or depression in workers with physical injury is limited. O’Hagan et al. (2012) examined a limited set of socio-demographic risk factors for prevalent anxiety and depression following permanent impairment due to physical work injury.(37) Higher education and higher pre-injury personal income were associated with a lower likelihood of self-reported depression symptoms, while older age and being a man were associated with a lower likelihood of a self-reported depression diagnosis.(37) None of the socio-demographic risk factors were associated with self-reported symptoms of anxiety, and associations with anxiety diagnoses were not examined.   57 Prior research on the prevalence and timing of anxiety and depression disorders in workers with physical injury, as well as risk factors for these disorders has been limited by collection of data during the post injury time period only, self-reported timing of anxiety and depression onset, and small clinical samples. Self-reported data collected in the post injury time period may be prone to recall bias, and small clinical samples can introduce selection bias and limit generalizability of findings to inform policies and programming. Further, only a limited set of risk factors for anxiety and depression have been examined in injured workers. To address these limitations, linked administrative data from the province of BC were used to examine the prevalence and timing of, and risk factors for, anxiety and depression disorders in a population-level cohort of workers with accepted lost-time claims for upper limb or spine strain or sprain. Findings may inform anxiety and depression prevention, screening, and treatment practices by health providers and workers’ compensation systems for workers with lost-time upper limb or spine strain or sprain work injury. The specific research objectives were to: 1. Estimate the one year period prevalence of anxiety, depression, and co-morbid anxiety/depression for the year before and the year after injury, as well as the cumulative prevalence for all years prior to injury; and 2. Examine the timing of anxiety and depression disorder onset relative to the injury; and 3. Examine the socio-demographic, injury, clinical, and work risk factors for: i. prevalent anxiety, depression, and co-morbid anxiety/depression in the year prior to injury; and  ii. new onset anxiety, depression, and co-morbid anxiety/depression in the three months after injury. 4.1 Methods A detailed description of the data sources is provided in the previous chapter, but a summary of key points reiterated below. 4.1.1 Study Sample Accepted lost-time claims for upper limb or spine strain/sprain injuries from January 1st 2000 to December 31st 2013 were extracted from the WorkSafeBC claim file using the primary ICD-9   58 code from the claim (840 shoulder and upper arm, 841 elbow and forearm, 842 wrist and hand, 846 sacroiliac region, and 847 back).  The study sample was limited to workers between 19 to 64 years of age at the time of injury, with no missing values for the risk factor variables, and registered in the provincial health services plan for at least 275 days (9 months) in both the year before and the year after injury. A registration period of at least 275 days was chosen to ensure that health system contact events for anxiety and depression would be captured in the data and not missed due to migration in or out of the province, while also allowing for temporary gaps in registration that occur due to missed or delayed payments of monthly premiums. This cut off value has been used in previous studies of these datasets.(111) Of the 317,512 accepted lost-time claims that met the age and injury criteria, 7.1% (n=22,599) were excluded due to insufficient registration (i.e. less than 275 days) in the provincial health services plan, and another 0.9% (n=2,748) were excluded due to missing values for the risk factor variables. The final study sample consisted of 292,165 claims (Figure 6). For the analyses of risk factors for prevalent anxiety or depression at the time of injury, those with an anxiety or depression related health care event in the year prior to injury that did not meet the case definitions (also referred to as the case-like group) were excluded from the study sample (8.8%, n=25,676 excluded; n = 266,489 claims remaining) (Figure 6). This increased the probability that claims classified as not having anxiety or depression in the year prior to injury truly did not have these disorders. These claims were not excluded from the analyses on prevalence, as this would reduce the denominator of the prevalence measure and bias the prevalence estimate upward.  For the analyses of risk factors for new onset anxiety or depression in the three months after injury, the study sample was further restricted to claims with no anxiety or depression in the year before injury (21.2%, n=56,584 excluded; n=209,905 claims remaining) (Figure 6). This ensured that only workers at risk of new onset anxiety and depression were included in these analyses.    59   Figure 6: Construction of the study sample consisting of lost-time upper limb and spine strain or sprain workers’ compensation claims in BC from 2000 to 2013 4.1.2 Study Variables 1. Outcome Variables Using methods described in the previous chapter (Table 2 and Table 3 in Chapter 3), prevalent cases of anxiety, depression, and co-morbid anxiety and depression were identified for 1) the year before injury (including the 364 days prior to injury and the day of injury) and 2) the year after injury. For the risk factor analysis, new onset cases of anxiety, depression, and co-morbid anxiety and depression were also identified during the first three months following injury, but only for workers with no anxiety- or depression-related health care events in the year prior to injury. A three-month time period was chosen, as new onset disorders in this period are more likely to be wholly or partially attributable to the injury or its aftereffects. Thus, new onset anxiety and depression disorders during the three months following injury may have unique risk profiles compared to prevalent anxiety or depression disorders from before injury. 317,512 lost-time claims met the age and injury criteria, 2000 to 2013 22,599 (7.1%) with insufficient medical registration period excluded 292,165 remaining 266,489 remaining 2,748 (0.9%) with missing values for the risk factor variables 25,676 (8.8%) with an anxiety or depression health care event not sufficient to meet the case definitions excluded (i.e. the case-like group) Prevalence analysis Risk factor analysis for prevalent disorders at injury 209,905 remaining 56,584 (21.2%) with prevalent anxiety or depression in the year prior to injury excluded Risk factor analysis for new onset disorders post-injury    60 2. Risk factors Socio-demographic, clinical, injury, and work-related risk factors for anxiety and depression disorders were identified based on a review of the return-to-work and mental health literature and the availability of variables in the administrative data. 2 a. Socio-demographic  Socio-demographic risk factors included i) gender (man/woman), ii) age group (19 to 24, 25 to 29, 30 to 39, 40 to 49, 50 to 59, or 60 to 64 years of age), iii) personal income decile, iv) dependent(s) under 19 years of age in the household (0/1 or more), and v) location of residence (urban/rural). Gender was measured using the sex field from the medical registry file. A drawback of administrative records is that they often lack measures of gender, however, disaggregation of the data based on sex can allow for the analysis of gender issues as the two measures are inter-related.(135) Income deciles were constructed for each year in the study period from the distribution of income for injured workers with a claim in that year. Location of residence was assigned based on the second digit of the workers’ postal code at the time of injury (0=rural, all other possible values = urban).  2 b. Injury  Injury risk factors included i) injured body part (shoulder and arm, sacroiliac region, and other parts of the back and neck), ii) incident type (traumatic event such as a stressful event or an act of violence or force; transportation incident; contact with an external object or substance such as being struck, caught in or rubbed/abraded by an object or exposed to a toxic substance; fall, slip, or trip; overexertion or repetitive motion; and other bodily motion), iii) secondary diagnosis on the claim other than anxiety or depression (yes/no), and iv) year of injury. 2 c. Clinical  Clinical risk factors included i) number of prior workers’ compensation claims in the last five years (0, 1, 2, 3 or more, regardless of the diagnosis), ii) presence of one or more mental disorders other than anxiety or depression (0/ 1 or more) and iii) somatic co-morbidity index score (0, 1, 2, 3, 4, 5 or more). To measure mental co-morbidity (other than anxiety or depression), physician visits and hospital admissions in the year before injury associated with diagnoses from the ICD-9 mental disorders category (290-319) and the ICD-10 Mental,   61 Behavioral and Neurodevelopmental disorders category (F01-F99) were identified (anxiety or depression diagnoses excluded). To measure somatic co-morbidity, all physician visits and hospitalizations in the year before injury were identified. Events associated with diagnoses identical to the diagnosis from the claim, or the diagnoses used to measure mental co-morbidity were deleted. Remaining diagnoses were categorized according to the ICD-9 disease categories and a variable to indicate the number of ICD-9 categories present in the year preceding injury was constructed as the somatic co-morbidity index score. This approach to measuring co-morbidity using Canadian administrative data for people with orthopedic injury was developed by Cameron et al.(136) Based on findings from Gabbe et al, it is comparable to other common co-morbidity measures including the Charlson and Functional Co-morbidity Indexes in predicting disability outcomes following orthopedic injury.(137)  2 d. Work-related  Work-related risk factors included i) firm size (30 or less employees, 31 to 150, 151 to 1000, 1001 to 10000, or 10001 or more employees), ii) shift type (fixed versus other), and iii) occupation (standardized occupation codes categorized as sales and services; art, culture, recreation, sport; business, finance, administration; health; management; natural and applied sciences; primary industry; processing, manufacturing, utilities; social science, education, government; trades, transport, equipment operators). Additional work variables for industry and job demands (strength, body position, and social demands based on the occupational job code) were also considered but not included in the final analyses due to collinearity (associations of χ2>50,000) with each other, and with the occupation variable (which was retained for final analyses). For collinear risk factors, only the risk factor with the lowest AIC value in the bivariable analyses was retained for the final adjusted model.(138) 4.1.3 Analyses All analyses were stratified by gender and conducted using SAS 9.4.(139)  1. Prevalence and timing The one-year period prevalence of anxiety, depression, and co-morbid anxiety and depression was calculated as the number of claims classified as cases during the time period of interest divided by the total number of claims in the study sample. To examine timing, the proportion of   62 claims with a prevalent anxiety disorder in the year prior to injury that persisted into the year after injury, and the proportion of claims with a prevalent anxiety disorder in the year after injury that was also present in the year prior to injury were calculated. The former was performed to examine the relevancy of disorders prevalent in the year before injury to the return to work process, and the latter was performed to allow direct comparison of the findings with other research in this area. The timing of depression was also examined in this manner. As 82,541 of the 292,165 claims in the study sample used for the prevalence analyses (28.3%) were repeat claims belonging to workers who already had an earlier claim selected for inclusion, a sensitivity analysis was conducted to examine the effects of including more than one claim per worker in the study sample on the prevalence measure. When the study sample was restricted to the first claim meeting the inclusion criteria, estimates of anxiety and depression prevalence for men and women were similar to the primary analyses (all differences were less than 1% - data not shown). 2. Risk factors associated with prevalent anxiety or depression at injury Unadjusted and adjusted associations of the risk factors with prevalent anxiety or depression (none, anxiety, depression, and co-morbid anxiety and depression) in the year before injury were assessed using multinomial regression models. Odds ratios and 95% confidence intervals were reported. Elimination of risk factors not associated with the outcome from the final adjusted models was considered, but all risk factors were associated with the outcome at a 95% level of confidence in both the men’s and women’s models.  3. Risk factors associated new onset anxiety or depression after injury Unadjusted and adjusted associations of the risk factors with new onset anxiety or depression (none, anxiety, depression, and co-morbid anxiety and depression) in the three months following injury were assessed using multinomial regression models. Anxiety and or depression from prior to the year before injury was included as a risk factor in new onset anxiety or depression analyses. This was measured using the longitudinal health care data (back to 1991) and the same case definitions as described earlier.     63 4.2 Results 4.2.1 Study sample The distribution of the risk factors by gender is described in Table 5. Age was evenly distributed across claims for workers in their thirties, forties, and fifties, with fewer in the youngest and oldest age categories; and over half were men (60.1%). In the study sample limited to upper body and spine strains and sprains, the most commonly injured body part was the upper back or neck (58.0% of men, 58.8% of women), and over exertion was the most common incident type (51.7% of men, 48.0% of women) followed by falls, slips, and trips (16.6% of men, 18.4% of women) and other bodily motions (17.6% of men, 19.9% of women). Less than half of claims selected for inclusion had no prior workers’ compensation history in the last five years (32.7% of men, 43.4% of women). Notable gender differences were observed for the income, somatic co-morbidity, and occupation variables. Compared to men, women were less likely to be in the higher income deciles (highest decile: 26.7% of men versus 11.4 % of women; second highest decile: 24.6% of men versus 14.5% of women). Somatic co-morbidity was common among both women and men; however, women were more likely than men to have a somatic co-morbidity score of 5 or higher (29.8% of women versus 13.4% of men), and less likely to have a score of 0 (3.8% of women versus 10.8% of men) or 1 (10.6% of women versus 20.1% of men). The most common occupational groups for men were trades, transport, or equipment operators (55.3%), sales and services (14.5%), and processing, manufacturing, and utilities (12.7%); while the most common occupational groups for women were sales and services (38.2%) and health (32.5%).   64 Table 5: Socio-demographic, injury, clinical and work factors among lost-time upper limb or spine strain or sprain claims in BC from 2000 to 2013   Men (N=175,566) n (%) Women (N=116,599) n (%) Socio-demographic    Age group (years)        19 to 24 19,440 (11.1) 9,053 (7.8)      25 to 29 19,568 (11.2) 10,221 (8.8)      30 to 39 45,577 (26.0) 26,804 (23.0)      40 to 49 50,231 (28.6) 38,241 (32.8)      50 to 59 33,925 (19.3) 28,059 (24.1)      60 to 64 6,825 (3.9) 4,221 (3.6) Income quintile        1: lowest 23,246 (13.2) 31,923 (27.4)      2 30,178 (17.2) 26,913 (23.1)      3 31,984 (18.2) 27,530 (23.6)      4 43,214 (24.6) 16,923 (14.5)      5: highest 46,944 (26.7) 13,310 (11.4) Dependents        0 112,744 (64.2) 70,535 (60.5)      1 or more 158,404 (35.7) 46,064 (39.5) Location        Urban 158,404 (90.2) 104,382 (89.5)      Rural 17,162 (9.8) 12,217 (10.5) Injury   Injured body part        Sacroiliac  26,341 (15.0) 14,967 (12.8)      Back/neck 101,907 (58.0) 68,508 (58.8)      Upper limb 47,318 (27.0) 33,124 (28.4) Incident type        Exertion/repetitive motion 30,885 (51.7) 55,934 (48.0)      Traumatic 2,529 (1.4) 4,797 (4.1)      Fall/slip/trip 29,219 (16.6) 21,470 (18.4)      Contact object 10,457 (6.0) 7,213 (6.2)      Transportation 11,686 (6.7) 3,965 (3.4)      Bodily motion 30,885 (17.6) 23,220 (19.9) Secondary claim diagnosis        No 148,640 (84.7) 89,083 (76.4)      Yesa  27,010 (15.3) 27,516 (23.6) Clinical   Somatic co-morbidity        0 19,029 (10.8) 4,461 (3.8)      1 35,301 (20.1) 12,400 (10.6)      2 40,090 (22.8) 20,143 (17.3)      3 34,600 (19.7) 23,578 (20.2)      4 22,996 (13.1) 21,312 (18.3)      5 or more 23,550 (13.4) 34,705 (29.8) Mental co-morbiditya        0 168, 158 (95.8) 112,116 (96.2)      1 or more 7,408 (4.2) 4,483 (3.8)   65  Men (N=175,566) n (%) Women (N=116,599) n (%) Clinical    Prior claims        0 57,425 (32.7) 50,600 (43.4)      1 45,117 (25.7) 29,481 (25.2)      2 29,133 (16.6) 16,430 (14.1)      3 or more 43,891 (25.0) 20,088 (17.2) Work   Firm size        30 or less 54,154 (30.9) 17,963 (15.4)      31 to 150 47,275 (26.9) 23,479 (20.1)      151 to 1000 42,737 (24.3) 26,357 (22.6)      1001 to 10000 28,086 (16.0) 27,762 (23.8)      10,001 or more 3,314 (1.89) 21,038 (18.0) Shift type        Fixed 144,520 (82.3) 84,332 (72.3)      Other 31,046 (17.7) 32,267 (27.7) Occupation        Sales and services 25,368 (14.5) 44,518 (38.2)      Art, culture, recreation, sport 1,312 (0.8) 1,636 (1.4)      Business, finance, administration 8,202 (4.7) 7,085 (6.1)      Health 5,983 (3.4) 37,836 (32.5)      Management 2,620 (1.5) 2,564 (2.2)      Natural and applied sciences 3,711 (2.1) 644 (0.6)      Primary industry 7,030 (4.0) 1,716 (1.5)      Processing, manufacturing, utilities 22,319 (12.7) 5,565 (4.8)      Social science, education, government 1,983 (1.1) 8,097 (6.9)      Trades, transport, equipment operators 97,038 (55.3) 6,938 (6.0) a anxiety and depression not included 4.2.2 Prevalence For both the year before and the year after injury, the prevalence of anxiety was higher than that of depression for both women and men, and the prevalence of anxiety or depression in women was approximately twice that of men (Table 6). The prevalence of both conditions was approximately 2% higher in the year post injury than the year pre-injury. In women, the prevalence of anxiety or depression disorders was 23.4% and 20.7% respectively in the year before injury, and 25.7% and 22.7% in the year post injury. In men, the prevalence of anxiety or depression disorders was 10.4% and 9.3% respectively in the year before injury, and 12.2% and 10.8% in the year post injury. Co-morbidity of anxiety and depression disorders was common in the study sample (Table 6). For men, 3.9% had anxiety only in the year before injury, 2.8% had depression only, and 6.5%   66 had co-morbid anxiety and depression. For women, in the year before injury, 8.0% had anxiety only, 5.3% had depression only, and 15.4% had both.  For women, 60.9% had anxiety and 53.9% had depression at some point prior to injury (back to 1991); and for men, 33.7% had anxiety and 28.8% had depression at some point prior to injury (Table 6).  Table 6: Prevalent anxiety and depression cases by gender and time period (row percentages reported) among lost-time upper limb or spine strain or sprain claims in BC from 2000 to 2013  Nonea Anx only Dep only Anx & depb Total anxc Total depc Men n=175,566       1 yr pre-inj 86.8 3.9 2.8 6.5 10.4 9.3      1 yr post-inj 84.6 4.5 3.1 7.7 12.2 10.8      All years pre-inj 61.3 9.9 5.0 23.8 33.7 28.8 Women n=116,599       1 yr pre-inj 71.2 8.0 5.3 15.4 23.4 20.7      1 yr post-inj 68.5 8.7 5.7 17.0 25.7 22.7      All years pre-inj 33.5 12.9 5.5 48.0 60.9 53.5 a For measures of prevalence, the case-like group was included in the none group b Includes all claims with both anxiety and depression (i.e. co-morbid anxiety and depression) c Includes all claims with the disorder (e.g. anxiety alone and co-morbid anxiety and depression; or depression alone and co-morbid anxiety and depression) 4.2.3 Timing  The majority of anxiety and depression disorders from the year prior to injury were also present in the year after injury (Table 7). Of the men with anxiety in the year before injury, 65.5% also had anxiety in the year after injury; and of the men with depression in the year before injury, 66.9% also had depression in the year after injury. Of the women with anxiety in the year before injury, 72.1% also had anxiety in the year after injury; and of the women with depression in the year before injury, 73.8% also had depression in the year after injury.      67 Table 7: The proportion of anxiety and depression cases prevalent in the year before lost-time upper limb or spine strain or sprain work injury also prevalent in the year after injury, and vice versa in BC from 2000 to 2013  Pre-injury cases Post-injury cases  Freq Proportion persisting (%) Freq Proportion pre-existing (%) Men n=175,566 Total anxa 18252 65.5 21422 55.8 Total depa 16319 66.9 18941 57.7 Women n=116,599 Total anxa 27,261 72.1 29,944 65.7 Total depa 24,211 73.8 26,439 67.6 a Includes all claims with the disorder (e.g. anxiety alone and co-morbid anxiety and depression; or depression alone and co-morbid anxiety and depression) Likewise, the majority of anxiety and depression disorders present in the year after injury were also present in the year before injury (Table 7). Of the men with anxiety in the year after injury, 55.8% also had anxiety in the year before injury; and of the men with depression in the year after injury, 57.7% also had depression in the year before injury. Of the women with anxiety in the year after injury, 65.7% also had anxiety in the year before injury; and of the women with depression in the year after injury, 67.6% also had depression in the year before injury. 4.2.4 Risk factors 4.2.4.1 Unadjusted models for prevalent anxiety and depression in the year before injury  All risk factors considered for inclusion in the final adjusted models were significant at a level of p<0.05 in the unadjusted models for both men and women (data not shown). While there were some significant differences in the odds of prevalent anxiety and/or depression from year to year, no secular trends were observed across the 14-year study period. For this reason, the year of injury variable was not retained for the adjusted models. 4.2.4.2 Adjusted models for prevalent anxiety and depression in the year before injury  The relationships between the independent variables and the outcome observed in the adjusted models were similar to those seen in the unadjusted models, although several of the ORs in the adjusted models were attenuated. 1. Men  1 a. Socio-demographic factors Findings from the adjusted models for prevalent anxiety or depression in the year before injury for men are presented in Table 8. Odds of anxiety only, depression only, and co-morbid anxiety   68 and depression increased with age up until age 40 to 49 years, after which the odds decreased but remained elevated relevant to the reference group of age 19 to 24 years of age. Odds of depression only and co-morbid anxiety and depression decreased with increasing income, and odds of anxiety only was lower in higher income groups compared to the lowest income group. Having a dependent was associated with lower odds of depression only, and co-morbid anxiety and depression compared to no dependents. Living in a rural location was associated with lower odds of anxiety only, and co-morbid anxiety and depression compared to an urban location. 1 b. Injury factors Back or neck strain or sprain was associated with higher odds of anxiety only, depression only, and co-morbid anxiety and depression compared to sacroiliac strain or sprain.  Compared to overexertion incidents, all other incident types had higher odds of co-morbid anxiety and depression. Falls, slips and trips, and transportation incidents were also associated with higher odds of anxiety compared to overexertion incidents. Having a secondary diagnosis to the primary strain or sprain diagnosis on the claim was associated with anxiety, and co-morbid anxiety and depression.  1 c. Clinical factors Large effect sizes were observed for the clinical risk factors. The somatic co-morbidity index score and prior claims variables had a strong positive dose response relationship with the odds of anxiety only, depression only, and co-morbid anxiety and depression. Mental co-morbidities other than anxiety or depression were also associated with higher odds of anxiety only, depression only, and co-morbid anxiety and depression. 1 d. Work factors A positive dose response relationship was observed between increasing firm size and an increasing odds of anxiety only, depression only, and co-morbid anxiety and depression. Compared to men working in sales and service occupations, men in health occupations had higher odds of anxiety only, depression only, and co-morbid anxiety and depression.      69 2. Women 2 a. Socio-demographic factors Among women (Table 8), odds of anxiety only increased with age up until age 40 to 49 years, and odds of depression only and co-morbid anxiety and depression increased with age up until age 30 to 39 years, after which the odds decreased but remained elevated relevant to the reference group of age 19 to 24 years of age. Odds of anxiety only, depression only and co-morbid anxiety and depression decreased with increasing income. Having a dependent was associated with higher odds of anxiety only and lower odds of depression only. Living in a rural location was associated with higher odds of depression and co-morbid anxiety and depression compared to an urban location.  2 b. Injury factors Compared to sacroiliac strain or sprain, back or neck strain or sprain was associated with higher odds of anxiety only, depression only, and co-morbid anxiety and depression; and upper limb strain or sprain was associated with lower odds of depression only and co-morbid anxiety and depression. Compared to overexertion incidents, all other incident types had higher odds of co-morbid anxiety and depression. Traumatic incidents were also associated with higher odds of anxiety only and depression only; and falls, slips and trips, and incidents involving contact with an external object were associated with higher odds of anxiety. 2 c. Clinical factors Large effect sizes were observed for the clinical risk factors. The somatic co-morbidity index score and prior claims variables had a strong positive dose response relationship with the odds of anxiety only, depression only, and co-morbid anxiety and depression. Mental co-morbidities other than anxiety or depression were also associated with higher odds of anxiety only, depression only, and co-morbid anxiety and depression. 2 d. Work factors Compared to women working in sales and service occupations, women working in business, finance and administration, and health occupations had higher odds of anxiety only, depression only, and co-morbid anxiety and depression; and workers in management had higher odds of depression only.   70 3. Key differences in risk profiles for the anxiety and depression case groups The risk profiles for anxiety only, depression only, and co-morbid anxiety and depression were similar to each other. The most notable difference in the risk profiles for the three outcome groups was observed for the somatic co-morbidity index variable in both the men’s and women’s adjusted models. While there was a strong positive dose response relationship between this variable and anxiety only, depression only, and co-morbid anxiety and depression; the effect sizes were greater for anxiety only and co-morbid anxiety and depression than for depression. Some additional gender dependent differences in the risk profiles for the outcome groups were observed and described below. 4. Key differences in men and women’s risk profiles  For the socio-demographic risk factors, the income variable had a negative dose-response relationship with anxiety only among women but not men, and having a dependent was a risk factor for anxiety only among women but not men. Living in a rural area was a risk factor for depression only and co-morbid anxiety and depression among women, and a protective factor for anxiety only and co-morbid anxiety and depression among men. For the injury variables, having a secondary diagnosis on the claim was positively associated with anxiety only and co-morbid anxiety and depression among men, but not women. Clinical risk factors were similar for men and women, but the effect size for the mental co-morbidity variable was larger for men than for women. For the work variables, a positive dose response relationship between firm size and anxiety only, depression only, and co-morbid anxiety and depression was observed for men but not women. Women working in business, finance, and administration had increased odds of anxiety only, depression only, and co-morbid anxiety and depression compared to women in sales and services – but this effect for occupation was not seen in men.   71 Table 8: Adjusteda odds ratios (OR) with 95% confidence intervals (CI) for risk factors of prevalent anxiety and depression in the year before lost-time upper limb or spine strain or sprain work injury in BC from 2000 to 2013, multinomial regression  Men  Women  Anx only Dep only Anx & dep   Anx only Dep only Anx & dep  Socio-demographic         Age group (years)             19 to 24 1 1 1  1 1 1      25 to 29 1.42 (1.25-1.61) 1.38 (1.20-1.60) 1.54 (1.40-1.70)  1.29 (1.14-1.45) 1.48 (1.28-1.70) 1.41 (1.29-1.55)      30 to 39 1.72 (1.54-1.93) 1.77 (1.56-2.00) 2.00 (1.84-2.18)  1.63 (1.47-1.82) 1.78 (1.56-2.01) 1.88 (1.73-2.04)      40 to 49 1.91 (1.71-2.13) 2.03 (1.80-2.30) 2.09 (1.91-2.27)  1.73 (1.56-1.92) 1.76 (1.56-1.99) 1.85 (1.71-2.00)      50 to 59 1.75 (1.56-1.96) 1.76 (1.55-2.00) 1.74 (1.59-1.90)  1.68 (1.52-1.87) 1.73 (1.53-1.96) 1.63 (1.51-1.76)      60 to 64 1.57 (1.34-1.83) 1.27 (1.05-1.54) 1.12 (0.98-1.28)  1.66 (1.44-1.93) 1.38 (1.15-1.65) 1.29 (1.15-1.45) Income quintile             1: lowest 1 1 1  1 1 1      2 0.93 (0.84-1.02) 0.87 (0.79-0.97) 0.83 (0.77-0.89)  0.97 (0.91-1.04) 0.85 (0.79-0.92) 0.86 (0.82-0.90)      3 0.93 (0.85-1.02) 0.84 (0.76-0.93) 0.80 (0.75-0.86)  0.94 (0.87-1.00) 0.81 (0.75-0.88) 0.78 (0.74-0.82)      4 0.93 (0.85-1.02) 0.80 (0.72-0.89) 0.77 (0.71-0.82)  0.84 (0.77-0.91) 0.77 (0.71-0.85) 0.69 (0.65-0.74)      5: highest 0.83 (0.75-0.91) 0.73 (0.66-0.81) 0.66 (0.62-0.71)  0.81 (0.75-0.89) 0.72 (0.65-0.80) 0.64 (0.59-0.68) Dependents             0 1 1 1  1 1 1      1 or more 0.96 (0.91-1.01) 0.86 (0.81-0.92) 0.87 (0.83-0.90)  1.08 (1.03-1.13) 0.91 (0.86-0.97) 0.97 (0.93-1.01) Location             Urban 1 1 1  1 1 1      Rural 0.85 (0.78-0.93) 1.08 (0.98-1.19) 0.89 (0.83-0.96)  0.98 (0.91-1.05) 1.19 (1.10-1.29) 1.07 (1.01-1.13) Injury        Injured body part             Sacroiliac 1 1 1  1 1 1      Back/neck 1.11 (1.03-1.19) 1.13 (1.04-1.23) 1.12 (1.06-1.19)  1.08 (1.01-1.16) 1.07 (0.98-1.15) 1.09 (1.03-1.15)      Upper limb 1.04 (0.96-1.13) 1.07 (0.97-1.18) 1.08 (1.02-1.16)  0.94 (0.87-1.01) 0.91 (0.83-0.99) 0.91 (0.86-0.97) Incident type             Exertion 1 1 1  1 1 1      Traumatic 1.06 (0.88-1.29) 1.17 (0.94-1.47) 1.16 (1.00-1.35)  1.25 (1.12-1.39) 1.11 (0.97-1.27) 1.33 (1.23-1.45)      Fall/slip/trip 1.08 (1.01-1.16) 1.06 (0.97-1.15) 1.17 (1.11-1.24)  1.16 (1.09-1.24) 1.14 (1.06-1.23) 1.16 (1.11-1.22)      Contact object 1.04 (0.93-1.16) 1.01 (0.89-1.14) 1.09 (1.00-1.19)  1.10 (1.00-1.21) 1.04 (0.92-1.16) 1.07 (0.99-1.15)      Transportation 1.13 (1.02-1.25) 1.00 (0.89-1.13) 1.16 (1.07-1.25)  0.93 (0.81-1.07) 0.96 (0.82-1.12) 1.11 (1.00-1.22)      Bodily motion 1.03 (0.96-1.11) 1.04 (0.96-1.12) 1.07 (1.01-1.13)  1.04 (0.98-1.11) 1.03 (0.96-1.11) 1.07 (1.02-1.12)            72   Men  Women  Anx only Dep only Anx & Dep  Anx only Dep only Anx & Dep Injury Secondary claim diagnosis             No 1 1 1  1 1 1      Yes (not anxiety or depression) 1.13 (1.05-1.21) 1.02 (0.94-1.11) 1.06 (1.01-1.12)  1.01 (0.96-1.07) 0.94 (0.88-1.00) 1.02 (0.98-1.07) Clinical        Somatic co-morbidity             0 1 1 1  1 1 1      1 1.64 (1.41-1.90) 1.48 (1.29-1.71) 1.49 (1.34-1.66)  1.74 (1.44-2.12) 1.50 (1.24-1.81) 1.60 (1.39-1.84)      2 2.50 (2.18-2.88) 1.93 (1.69-2.21) 2.24 (2.02-2.48)  2.39 (1.99-2.87) 1.73 (1.45-2.07) 2.14 (1.88-2.45)      3 3.56 (3.10-4.08) 2.31 (2.01-2.64) 3.02 (2.72-3.35)  3.06 (2.55-3.67) 2.06 (1.73-2.46) 2.84 (2.49-3.24)      4 4.43 (3.84-5.10) 2.63 (2.28-3.02) 4.12 (3.71-4.58)  3.71 (3.09-4.45) 2.29 (1.92-2.73) 3.56 (3.12-4.06)      5 or more 7.36 (6.41-8.44) 3.43 (2.99-3.94) 6.39 (5.77-7.08)  5.79 (4.84-6.92) 3.00 (2.52-3.56) 5.47 (4.81-6.23) Mental co-morbidity             0 1 1 1  1 1 1      1 or more 4.50 (4.13-4.90) 4.80 (4.36-5.29) 6.63 (6.22-7.06)  2.98 (2.68-3.31) 3.42 (3.04-3.84) 5.24 (4.85-5.65) Prior claims             0 1 1 1  1 1 1      1 1.09 (1.01-1.16) 1.04 (0.96-1.12) 1.06 (1.00-1.12)  1.09 (1.03-1.16) 1.09 (1.02-1.16) 1.17 (1.12-1.22)      2 1.17 (1.09-1.27) 1.02 (0.93-1.11) 1.16 (1.09-1.23)  1.20 (1.12-1.29) 1.25 (1.15-1.35) 1.30 (1.24-1.37)      3 or more 1.36 (1.27-1.46) 1.32 (1.22-1.43) 1.37 (1.30-1.45)  1.45 (1.36-1.54) 1.33 (1.23-1.43) 1.50 (1.43-1.57) Work        Firm size             30 or less 1 1 1  1 1 1      31 to 150 0.99 (0.92-1.06) 1.00 (0.93-1.09) 1.01 (0.96-1.07)  0.93 (0.86-1.00) 0.91 (0.83-0.99) 0.93 (0.88-0.98)      151 to 1000 1.06 (0.99-1.14) 1.02 (0.94-1.11) 1.07 (1.01-1.13)  0.86 (0.80-0.93) 0.81 (0.74-0.88) 0.83 (0.79-0.88)      1001 to 10000 1.18 (1.09-1.28) 1.10 (1.00-1.21) 1.18 (1.11-1.26)  0.97 (0.90-1.04) 0.94 (0.86-1.03) 0.99 (0.93-1.05)      10,001 or more 1.22 (1.03-1.44) 1.21 (0.99-1.48) 1.29 (1.13-1.48)  0.94 (0.87-1.03) 1.02 (0.92-1.13) 1.07 (1.00-1.14) Shift type             Fixed 1 1 1  1 1 1      Other 0.96 (0.90-1.03) 0.97 (0.90-1.05) 1.05 (1.00-1.11)  0.95 (0.90-1.00) 1.00 (0.94-1.06) 1.01 (0.98-1.06)      73  Men  Women  Anx only Dep only Anx & Dep  Anx only Dep only Anx & Dep Work Occupation             Sales and services 1 1 1  1 1 1      Art, culture, recreation, sport 0.78 (0.56-1.08) 0.83 (0.57-1.19) 0.80 (0.62-1.02)  0.97 (0.79-1.18) 1.10 (0.87-1.38) 1.06 (0.91-1.23)      Business, finance,… administration 1.01 (0.89-1.15) 0.94 (0.81-1.10) 0.95 (0.86-1.06)  1.13 (1.03-1.25) 1.34 (1.19-1.49) 1.28 (1.19-1.38)      Health 1.62 (1.41-1.85) 1.40 (1.19-1.65) 1.34 (1.20-1.50)  1.11 (1.04-1.18) 1.11 (1.03-1.20) 1.11 (1.06-1.17)      Management 1.06 (0.85-1.30) 1.03 (0.80-1.32) 0.98 (0.83-1.17)  1.15 (0.99-1.34) 1.22 (1.02-1.46) 1.04 (0.92-1.18)      Natural and applied sciences 0.95 (0.79-1.15) 1.00 (0.81-1.25) 0.92 (0.79-1.07)  0.77 (0.55-1.08) 0.74 (0.49-1.13) 0.82 (0.63-1.06)      Primary industry 0.99 (0.85-1.14) 0.93 (0.79-1.11) 0.90 (0.80-1.01)  0.68 (0.55-0.85) 0.90 (0.71-1.13) 1.02 (0.88-1.18)      Processing, manufacturing… 0.87 (0.79-0.97) 0.90 (0.80-1.02) 0.85 (0.78-0.92)  0.64 (0.57-0.72) 0.81 (0.70-0.93) 0.79 (0.73-0.87)      Social science, education… 1.33 (1.08-1.64) 1.08 (0.83-1.42) 1.36 (1.15-1.60)  1.27 (1.16-1.39) 1.53 (1.38-1.70) 1.40 (1.31-1.50)      Trades, transport, equipment operators 0.94 (0.87-1.01) 0.97 (0.89-1.06) 0.89 (0.84-0.95)  0.95 (0.85-1.05) 1.09 (0.96-1.22) 1.11 (1.03-1.20) a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, and occupation  74 4.2.4.3 Post injury adjusted models for new onset anxiety and depression  There were 139,100 men and 70,805 women with no anxiety or depression related health care events in the year before injury. Of these, 1.0% (n=1315) of men and 1.9% (n=1357) of women developed anxiety only, 0.4% (n=609) of men and 0.8% (n=577) of women developed depression only, and 0.7% (n=956) of men and 1.3% (n=950) of women developed co-morbid anxiety and depression in the three months after injury.  Many variables exhibited similar effects in the post-injury adjusted models for new onset anxiety and depression (Table 9) as in the pre-injury adjusted models for prevalent anxiety and depression (Table 8). Key differences between the pre-injury and post-injury adjusted models are highlighted as follows. In the post injury adjusted models, the prior anxiety or depression variable (for episodes occurring before the year prior to injury) was strongly, positively correlated with new onset anxiety, depression, and co-morbid anxiety and depression post injury. The other clinical variables (somatic co-morbidity, mental co-morbidity other than depression or anxiety, and prior claims) showed similar directions of effect in the post injury models compared to the pre-injury models, but the effect sizes were notably smaller. In the post injury adjusted models, the secondary diagnosis variable was associated with increased odds of anxiety, depression, and co-morbid anxiety and depression, for both men and women although some of the 95% CIs included the effect estimate of ‘1’. This effect was not seen in the women’s pre-injury adjusted model, and while it was present for the anxiety and co-morbid anxiety and depression outcomes in the men’s model, the effect sizes were smaller. Differences in the effects of age were observed in the post injury models compared to the pre-injury ones. In the women’s post-injury adjusted model, odds of new onset anxiety, depression and co-morbid anxiety and depression were highest among the two youngest age groups (19 to 24 and 25 to 29 years of age), and decreased with age. In the men’s post-injury adjusted model, odds of new onset anxiety and co-morbid anxiety and depression were lowest among the youngest (19 to 24 years) and oldest age groups (60 to 64 years), and elevated among men 25 to 29, 30 to 39, and 40 to 49 years of age; and odds of new onset depression were lower among men from the older age groups (40 years of age and older) relative to the reference group (19 to 24 years of age). For the incident type variable, odds of new onset anxiety and co-morbid anxiety and depression post injury were significantly higher for transportation incidents compared to over exertion incidents   75 for both women and men. This effect was also observed in the pre-injury adjusted model for men, but transportation incidents were not related to anxiety in the women’s pre-injury adjusted model.  Table 10 provides a summary of the relationships between the study risk factors and prevalent and new onset anxiety and depression conditions among the cohort of lost-time upper limb or spine strain or sprain claims in British Columbia.  76 Table 9: Adjusteda odds ratios (OR) with 95% confidence intervals (CI) for risk factors of new onset anxiety and depression in the 3 months after lost-time upper limb or spine strain or sprain work injury in BC from 2000 to 2013, multinomial regression  Men  Women  Anx only Dep only Anx & dep   Anx only Dep only Anx & dep  Socio-demographic         Age group (years)             19 to 24 1 1 1  1 1 1      25 to 29 1.32 (1.02-1.70) 0.95 (0.68-1.33) 1.51 (1.14-2.00)  1.15 (0.88-1.50) 0.84 (0.58-1.21) 0.99 (0.73-1.34)      30 to 39 1.30 (1.03-1.64) 1.13 (0.85-1.51) 1.35 (1.04-1.75)  1.02 (0.80-1.30) 0.94 (0.68-1.29) 0.95 (0.72-1.24)      40 to 49 1.29 (1.02-1.63) 0.73 (0.53-0.99) 1.08 (0.82-1.41)  0.93 (0.74-1.18) 0.65 (0.47-0.89) 0.76 (0.58-0.99)      50 to 59 1.14 (0.89-1.45) 0.55 (0.39-0.76) 0.96 (0.73-1.28)  0.91 (0.71-1.16) 0.51 (0.37-0.71) 0.66 (0.50-0.87)      60 to 64 0.99 (0.70-1.40) 0.60 (0.36-1.00) 0.55 (0.33-0.90)  0.56 (0.36-0.85) 0.38 (0.20-0.71) 0.71 (0.47-1.07) Income quintile             1: lowest 1 1 1  1 1 1      2 1.06 (0.87-1.31) 0.90 (0.68-1.20) 1.03 (0.82-1.30)  1.06 (0.91-1.23) 0.96 (0.77-1.20) 1.10 (0.92-1.32)      3 0.96 (0.78-1.19) 0.88 (0.66-1.17) 0.88 (0.69-1.11)  0.83 (0.70-0.99) 0.68 (0.53-0.88) 0.98 (0.81-1.20)      4 1.05 (0.86-1.28) 0.86 (0.65-1.14) 0.91 (0.72-1.14)  0.95 (0.78-1.15) 0.78 (0.58-1.04) 0.87 (0.68-1.10)      5: highest 0.94 (0.76-1.15) 0.83 (0.62-1.11) 0.74 (0.58-0.94)  0.79 (0.63-0.99) 0.45 (0.30-0.67) 0.78 (0.59-1.03) Dependents             0 1 1 1  1 1 1      1 or more 0.97 (0.86-1.10) 1.00 (0.84-1.20) 1.10 (0.95-1.26)  1.25 (1.11-1.42) 1.00 (0.83-1.20) 0.96 (0.83-1.12) Location             Urban 1 1 1  1 1 1      Rural 0.89 (0.73-1.09) 1.13 (0.87-1.47) 0.95 (0.76-1.20)  1.17 (0.98-1.38) 1.30 (1.01-1.67) 0.89 (0.71-1.11) Injury        Injured body part             Sacroiliac 1 1 1  1 1 1      Back/neck 1.16 (0.98-1.37) 1.22 (0.95-1.57) 0.92 (0.77-1.10)  1.19 (1.00-1.43) 1.06 (0.82-1.37) 0.99 (0.81-1.21)      Upper limb 1.11 (0.92-1.33) 1.21 (0.92-1.60) 0.70 (0.57-0.87)  1.15 (0.95-1.40) 0.87 (0.66-1.16) 0.92 (0.74-1.14) Incident type             Exertion 1 1 1  1 1 1      Traumatic 1.32 (0.90-1.93) 1.19 (0.66-2.15) 1.67 (1.09-2.55)  1.53 (1.20-1.94) 1.33 (0.88-2.00) 1.32 (0.97-1.79)      Fall/slip/trip 1.00 (0.85-1.17) 1.14 (0.91-1.43) 1.07 (0.89-1.29)  0.99 (0.84-1.16) 1.12 (0.88-1.41) 1.11 (0.92-1.32)      Contact object 1.05 (0.83-1.34) 0.98 (0.69-1.39) 1.23 (0.94-1.60)  1.11 (0.88-1.38) 0.98 (0.68-1.41) 0.87 (0.65-1.16)      Transportation 1.64 (1.35-1.99) 1.30 (0.96-1.77) 1.65 (1.31-2.08)  1.54 (1.18-2.01) 0.66 (0.37-1.17) 1.35 (0.97-1.87)      Bodily motion 1.01 (0.86-1.18) 0.95 (0.76-1.20) 0.80 (0.66-0.98)  0.97 (0.83-1.12) 1.21 (0.98-1.50) 0.69 (0.56-0.83)           77    Men  Women  Anx only Dep only Anx & Dep  Anx only Dep only Anx & Dep Injury Secondary claim diagnosis             No 1 1 1  1 1 1      Yes (not anxiety or depression) 1.26 (1.09-1.45) 1.27 (1.03-1.58) 1.18 (0.99-1.40)  1.24 (1.09-1.41) 1.17 (0.96-1.43) 1.15 (0.98-1.34) Clinical        Somatic co-morbidity             0 1 1 1  1 1 1      1 1.30 (1.02-1.67) 1.19 (0.84-1.67) 1.25 (0.94-1.67)  1.19 (0.82-1.72) 0.76 (0.46-1.25) 1.09 (0.71-1.68)      2 1.51 (1.19-1.92) 1.34 (0.96-1.87) 1.31 (0.99-1.74)  1.39 (0.98-1.98) 1.19 (0.76-1.88) 1.30 (0.87-1.96)      3 1.68 (1.31-2.14) 1.54 (1.10-2.15) 1.66 (1.25-2.20)  1.71 (1.21-2.41) 1.05 (0.67-1.65) 1.59 (1.07-2.37)      4 2.19 (1.70-2.81) 1.74 (1.22-2.49) 2.01 (1.50-2.70)  1.84 (1.30-2.61) 1.24 (0.79-1.95) 1.59 (1.06-2.38)      5 or more 2.74 (2.13-3.51) 2.08 (1.46-2.97) 2.37 (1.77-3.17)  2.10 (1.49-2.96) 1.29 (0.83-2.01) 1.87 (1.26-2.77) Mental co-morbidity             0 1 1 1  1 1 1      1 or more 1.54 (1.18-2.00) 1.79 (1.25-2.58) 1.64 (1.22-2.21)  1.31 (0.94-1.83) 1.40 (0.86-2.28) 1.36 (0.92-2.01) Prior claims             0 1 1 1  1 1 1      1 0.91 (0.78-1.06) 1.09 (0.88-1.36) 1.00 (0.84-1.20)  1.09 (0.95-1.25) 1.00 (0.81-1.24) 0.96 (0.81-1.13)      2 1.00 (0.85-1.19) 1.09 (0.85-1.39) 1.03 (0.85-1.26)  1.26 (1.08-1.49) 1.01 (0.77-1.32) 1.06 (0.87-1.29)      3 or more 1.01 (0.87-1.17) 1.09 (0.87-1.37) 1.13 (0.95-1.35)  1.04 (0.88-1.22) 1.35 (1.06-1.71) 1.10 (0.91-1.33) Prior anxiety or depression             None 1 1 1  1 1 1      Anxiety only 2.47 (2.11-2.89) 1.72 (1.3-2.27) 1.84 (1.48-2.29)  2.37 (2.02-2.78) 1.67 (1.26-2.21) 1.21 (0.95-1.55)      Depression only 1.72 (1.34-2.20) 3.75 (2.86-4.9) 2.10 (1.59-2.77)  1.37 (1.05-1.78) 3.37 (2.51-4.51) 1.60 (1.17-2.18)      Anxiety and depression 2.54 (2.23-2.90) 3.02 (2.5-3.65) 3.67 (3.17-4.24)  2.38 (2.08-2.71) 2.70 (2.21-3.30) 3.35 (2.87-3.90) Work        Firm size             30 or less 1 1 1  1 1 1      31 to 150 0.99 (0.92-1.06) 1.00 (0.93-1.09) 1.01 (0.96-1.07)  0.93 (0.86-1.00) 0.91 (0.83-0.99) 0.93 (0.88-0.98)      151 to 1000 1.06 (0.99-1.14) 1.02 (0.94-1.11) 1.07 (1.01-1.13)  0.86 (0.80-0.93) 0.81 (0.74-0.88) 0.83 (0.79-0.88)      1001 to 10000 1.18 (1.09-1.28) 1.10 (1.00-1.21) 1.18 (1.11-1.26)  0.97 (0.90-1.04) 0.94 (0.86-1.03) 0.99 (0.93-1.05)      10,001 or more 1.22 (1.03-1.44) 1.21 (0.99-1.48) 1.29 (1.13-1.48)  0.94 (0.87-1.03) 1.02 (0.92-1.13) 1.07 (1.00-1.14)     78  Men  Women  Anx only Dep only Anx & Dep  Anx only Dep only Anx & Dep Work Shift type             Fixed 1 1 1  1 1 1      Not fixed 1.05 (0.91-1.21) 1.09 (0.88-1.34) 1.03 (0.87-1.22)  1.16 (1.03-1.32) 0.82 (0.67-1.00) 0.91 (0.78-1.06) Occupation             Sales and services 1 1 1  1 1 1      Art, culture, recreation, sport 0.83 (0.42-1.62) 1.71 (0.82-3.54) 0.42 (0.13-1.32)  0.83 (0.49-1.40) 0.55 (0.23-1.35) 0.88 (0.49-1.57)      Business, finance, administration 0.69 (0.51-0.94) 0.55 (0.32-0.93) 0.66 (0.45-0.96)  1.28 (1.01-1.61) 1.13 (0.80-1.61) 0.93 (0.69-1.24)      Health 1.08 (0.79-1.47) 1.51 (0.97-2.34) 0.97 (0.64-1.45)  1.09 (0.93-1.29) 1.03 (0.80-1.32) 1.13 (0.93-1.37)      Management 0.83 (0.52-1.34) 0.94 (0.45-1.95) 0.93 (0.53-1.66)  0.71 (0.45-1.12) 0.78 (0.41-1.48) 0.93 (0.59-1.47)      Natural and applied sciences 1.17 (0.82-1.67) 1.16 (0.66-2.03) 0.60 (0.33-1.09)  0.51 (0.19-1.38) 0.61 (0.15-2.46) 0.33 (0.08-1.33)      Primary industry 0.79 (0.57-1.11) 0.68 (0.40-1.16) 0.71 (0.46-1.08)  0.90 (0.56-1.46) 0.61 (0.27-1.38) 0.83 (0.47-1.45)      Processing, manufacturing, utilities 0.81 (0.65-1.01) 0.96 (0.69-1.32) 1.15 (0.90-1.47)  1.13 (0.88-1.46) 1.06 (0.73-1.54) 1.08 (0.80-1.45)      Social science, education, government 1.31 (0.86-2.00) 0.84 (0.36-1.93) 1.03 (0.58-1.83)  1.35 (1.09-1.68) 1.38 (1.00-1.90) 0.97 (0.73-1.28)      T ades, ransport, equipment operators 0.82 (0.70-0.97) 0.99 (0.77-1.28) 0.97 (0.80-1.18)  0.92 (0.72-1.18) 0.94 (0.65-1.35) 0.89 (0.67-1.19) a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, prior anxiety or depression, firm size, shift type, and occupation  79 Table 10: Summary table depicting the adjusteda associations between risk factors and anxiety and depression disorders that were i) prevalent in the year before lost-time upper limb or spine strain or sprain work injury and ii) new onset in the three months following injury in BC from 2000 to 2013  Prevalent in the year before injury New onset in the three months following injury  Men Women Men Women  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Socio-demographic              Age group (years)                  19 to 24 1 1 1 1 1 1 1 1 1 1 1 1      25 to 29                  30 to 39                  40 to 49                  50 to 59                  60 to 64             Income quintile                  1: lowest 1 1 1 1 1 1 1 1 1 1 1 1      2                  3                  4                  5: highest             Dependents                  0 1 1 1 1 1 1 1 1 1 1 1 1      1 or more        1   1  Location                  Urban 1 1 1 1 1 1 1 1 1 1 1 1      Rural             Injury             Injured body part                  Sacroiliac 1 1 1 1 1 1 1 1 1 1 1 1      Back/neck                  Upper limb                  80  Prevalent in the year before injury New onset in the three months following injury  Men Women Men Women  Anx only Dep only Anx & dep  Anx  only Dep only Anx & dep  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Injury Incident type                  Exertion 1 1 1 1 1 1 1 1 1 1 1 1      Traumatic                  Fall/slip/trip       1           Contact object                  Transportation  1                Bodily motion             Secondary claim diagnosis                  No 1 1 1 1 1 1 1 1 1 1 1 1      Yes (not anxiety or depression)             Clinical             Somatic co-morbidity                  0 1 1 1 1 1 1 1 1 1 1 1 1      1                  2                  3                  4                  5 or more             Mental co-morbidity                  0 1 1 1 1 1 1 1 1 1 1 1 1      1 or more             Prior claims                  0 1 1 1 1 1 1 1 1 1 1 1 1      1         1  1       2       1           3 or more             Prior anxiety or depression                  None       1 1 1 1 1 1      Anxiety only                  Depression only                  Anxiety and depression                81   Prevalent in the year before injury New onset in the three months following injury  Men Women Men Women  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep  Work Firm size                  30 or less 1 1 1 1 1 1 1 1 1 1 1 1      31 to 150  1      1          151 to 1000                  1001 to 10000                  10,001 or more             Shift type                  Fixed 1 1 1 1 1 1 1 1 1 1 1 1      Not fixed     1        Occupation                  Sales and services 1 1 1 1 1 1 1 1 1 1 1 1      Art, culture, recreation, sport                  Business, finance, administration                  Health                  Management                  Natural and applied sciences  1                Primary industry                  Processing, manufacturing, utilities                  Social science, education, government                  Trades, transport, equipment operators              = greater likelihood of the indicated disorder (i.e. adjusted OR>1) with 95% CI that does not cross one  = greater likelihood of the indicated disorder (i.e. adjusted OR>1) with 95% CI that does cross one  = less likelihood of the indicated disorder (i.e. adjusted OR<1) with 95% CI that does not cross one  = less likelihood of the indicated disorder (i.e. adjusted OR<1) with 95% CI that does cross one a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, and occupation. The analyses for new onset anxiety and depression disorders was also adjusted for prior anxiety or depression.  82 4.3 Discussion 4.3.1 Summary of main findings In this linked administrative data cohort study, a notable number of claims had anxiety, depression, or co-morbid anxiety and depression in the year before (men: 13.2%; women 29.8%) and the year after injury (men: 15.4%; women: 31.5%). While there was an increase in the prevalence of anxiety and depression in the year after injury compared to the year before, this difference was small (~2%). Similarly, among claims with no anxiety or depression in the year prior to injury, the incidence of new onset anxiety (1.0% men, 1.9% women), depression (0.4% men, 0.8% women), or comorbid anxiety and depression (0.7% men, 1.3% women) in the three months following injury was small. Examination of anxiety and depression timing relative to the injury, found that the majority of the disorders present in the year prior to injury persisted into the year after injury, and that the majority of the disorders present in the year after injury, were also present in the year before injury. Lastly, multiple socio-demographic, clinical, injury, and work-related risk factors for prevalent and new onset anxiety and depression were found.  4.3.2 Prevalence The prevalence of anxiety and depression disorders in the year following injury were respectively 12.2% and 10.8% for men in our study sample and 25.7% and 22.7% for women. These prevalence estimates for post-injury depression disorders are low compared to other studies of injured workers, however this is likely due to differences in level of impairment, chronicity of work disability, and recovery expectations. Casey et al. (2017) found that among workers with permanent work impairment in Ontario, 56% of women and 47% of men self-reported a diagnosis of depression in the past six months;(140) and based on structural clinical interviews of workers with chronic disabling work musculoskeletal disorders, Dersh et al. (2002, 2006) found that 55% have a depression disorder, and 11% have an anxiety disorder.(10,11) The lower prevalence of post-injury depression in the current study sample is not surprising as it included all workers with lost-time upper limb or spine strain or sprain, including those who returned to work during the acute phase of injury (less than three months). There is limited research on the prevalence of anxiety disorders following work injury although the estimate of 11% by Dersh et al. (2002, 2006) is equivalent to the estimate for men in our study sample (also 10.8%).(10,11)   83   4.3.3 Timing The finding that the majority of anxiety disorders present in the year following injury, were present prior to injury is in accordance with findings from Dersh et al. (2006)(10) However, the finding that the majority of depression disorders present post injury also developed prior to injury is not in accordance with findings from Dersh et al. (2007), O’Hagan et al. (2012), and Casey et al. (2017).(37,131,140) Possible explanations for this discrepancy include: recall bias due to self-reported timing of depression onset and differences in selection criteria in prior studies, and perhaps concerns among injured workers of reporting prior conditions while being adjudicated for receiving insurance benefits. Further, Dersh et al. examined workers off work for four months or more, and O’Hagan and Casey examined workers with permanent occupational impairment.(37,131,140). Injured worker populations with permanent impairment or long term disability may be at greater risk of new onset depression than workers with relatively recent lost-time strain or sprain for which they are expected to return to work. Lastly, O’Hagan and Casey excluded workers with a prior claim while they were included in the current study.   4.3.4 Risk factors 1. Sociodemographic factors Findings concerning the associations of income with prevalent anxiety and depression are similar to those found in the Canadian general population where having low income is associated with higher risk of prevalent generalized anxiety disorder, and major depression.(40,47) The increased prevalence of anxiety among older age groups in the current study sample, particularly those that are middle age, is similar to the age distribution for anxiety found in the general population.(40) However, the age distribution for prevalent depression found in the current study sample, which like anxiety, was highest among men and women between the ages of 40 to 49, appears to vary from the Canadian general population where prevalence peaks prior to age 25.(47) This could be due to low labour force participation by young adults with depression;(141) under treatment of depression disorders in young adults in BC resulting in younger workers from the current study with a depression disorder being misclassified an ‘none’;(142) or a high prevalence of life-stressors among middle age workers (i.e. the ‘sandwich’ generation) such as providing informal   84 care or financial support for family members that are children, young adults, or elderly, as well as other financial or career issues all while maintaining employment. Current findings suggest gender differences in the geographical distribution of injured workers’ mental health in BC. Living in a rural area in comparison to an urban one was associated with i) lower odds of prevalent anxiety among men but not women, and ii) higher odds of prevalent depression among women but not men. Literature on the association of rural living with anxiety and depression (versus urban living) is mixed. There is some evidence that rural living is associated with a lower prevalence of these disorders (143,144) that is contradictory to the current study’s findings on depression for women. However, recent studies of the Canadian and American general populations report no associations.(40,47,145) Discrepancies between the current study and existing literature may be due to use of stratification by gender, or non-generalizability of the study sample to other provincial, non-working, or uninjured populations. There is little literature to explain the current geographical finding, however, many rural areas in BC are economically reliant on male dominated industries like primary resources and, according to sociological analyses, rural living is more strongly characterized by patriarchal male-dominant values and practices than is urban living.(146,147) These gendered contextual factors may have protective effects for working men’s risk of anxiety and increased risk for working women’s risk of depression. In the adjusted pre-injury models, having a dependent in the home was a protective factor for prevalent depression for men and women. This is contrary to other findings that suggest being a parent to a child who is not yet independent is a risk factor for depression, possibly due to the strains of raising children.(148,149) This discrepancy may be due to uncontrolled confounding in the current study by factors such as relationship status, as being married or having a partner is a protective factor for new onset mental disorders and also associated with better mental well-being.(150–152) Likewise, parents may have higher levels of family or other social support compared to non-parents that can have positive effects on mental health.(153) In contrast, having a dependent in the home was a risk factor for prevalent and new onset anxiety in women (but not men). This could be due to the differential effects of parenting strains on men and women, both in a non-disability and disability context. For example, a study of dual-earner mothers and   85 fathers found that strains associated with childcare contributed to psychological strain among mothers but not fathers.(154)  2. Injury factors In the pre-injury models, anxiety and depression in the year prior to injury was the outcome variable. However, due to the chronological timing of events, associations between pre-existing anxiety and depression and the injury variables, suggests that anxiety and depression are risk factors for certain injury types. For example, falls, slips, and trips were associated with pre-existing anxiety for men, and pre-existing anxiety and pre-existing depression for women. In the general population, depression is associated with increased risk of unintentional incidents even after controlling for anti-depressant and anxiolytic use.(52) Cognitive side effects of depression such as difficulty concentrating, fatigue, and insomnia, are suggested mechanisms for the association. Anti-depressants and anxiolytics (particularly benzodiazepines) are also independently associated with increased risk of falls in the elderly, (45) and likely also in the working population although there is less research to support this latter statement.  In the men’s pre-injury model, pre-existing anxiety only and co-morbid anxiety and depression (but not depression only) were associated with having a secondary diagnosis on the claim; but there was no evidence of similar associations for women. In the descriptive analyses, women were approximately 1.5 times more likely to have a secondary diagnosis on the claim. This suggests that men may be less likely to report secondary symptoms in comparison to women overall, but that anxiety might increase the likelihood of reporting secondary symptoms by men, possibly due to lower symptom sensitivity threshold or increased awareness and concern regarding health symptoms.(155,156) A notable finding from the post-injury models is that transportation incidents were significantly associated with new onset anxiety in both men and women; and other traumatic incidents (e.g. stressful acute event or an act of violence or force) were significantly associated with new onset anxiety in significantly in women and non-significantly men, This indicates that anxiety may be a psychological sequelae of these types of more traumatic incidents. Several studies have documented associations between anxiety and traumatic incidents, such as transportation incidents.(157–160)     86 Lastly, in the post-injury model, there was evidence that having a secondary diagnosis on the injury claim was positively association with anxiety, depression, and co-morbid anxiety and depression for both women and men. This suggests that more complex injuries, especially multi-site injuries, may be a risk factor for psychological sequelae for both men and women. 3. Clinical factors In this study, the clinical risk factors were important determinants of both prevalent and new onset anxiety and depression. The somatic co-morbidity index score had a strong positive dose-response relationship with anxiety, depression, and co-morbid anxiety and depression; and having one or more mental co-morbidities (other than anxiety or depression) was also associated with these outcomes, especially in the pre-injury models. Both of these findings are in line with what is consistently reported in the literature based on general populations and other clinical study samples about more complex health profiles associated with anxiety and depression.(161–166)   The number of prior claims was also associated with prevalent anxiety and depression at the time of injury in a positive dose-response fashion. Additionally, over half of the study sample had a prior claim in the last five years. A possible, although uninvestigated explanation, is that mental health and work injury are inter-related and act as a re-iterating feedback loop where one leads to another and vice versa.  This is supported by the literature on depression and pain conditions, where depression is both a cause and consequence of pain (4), as well as evidence that if you have a prior claim you are more likely to incur another.(167) Another possible explanation is that workers with prior claims may have greater exposure to both physical and mental health risks at work, and this may lead to both multiple claims and mental disorders. In the post injury models, prior anxiety or depression were both strong risk factors for new onset anxiety and new onset depression. This is in line with established knowledge that anxiety and depression are both re-occurring conditions and risk factors for each other.(9) Overall, workers with complicated medical histories including somatic co-morbidity, mental co-morbidity, or prior work injury, were more likely to have anxiety and depression.   4. Work factors   87 For both men and women, health occupations were significantly associated with prevalent anxiety and depression in the year before injury. An American study of the general non-injured working population also found an elevated prevalence of depression among workers from the health care sector,(168) and an Australian study found that workers in health care had higher risk of mental health claims than workers from other industries.(74) High levels of emotional labour and greater use of health services (resulting in higher likelihood of diagnosis) and mental health awareness by workers in health occupations are suggested mechanisms.(74,168) For women, business, finance, and administration occupations were also associated with prevalent anxiety and depression in the year before injury, and management occupations were associated with depression only. There is very little literature on the prevalence of anxiety and depression across occupations by gender to compare this finding. However, some differential effects of occupational psychosocial risk factors and risk of depression by gender have been documented. For example, Wang et al. found that effort-reward imbalance at work was associated with depression in women but not men.(169) Further research is needed to determine if gender differences in exposure to psychosocial risk factors, or gender differences in the mental health effects of psychosocial risk factors account for women’s higher prevalence of anxiety and depression disorders in these occupations. There were no clear trends in the distribution of new onset anxiety and depression by occupation in the post-injury models due to large confidence intervals. For men, a positive dose response relationship was observed between increasing firm size and prevalent anxiety and depression in the year before injury and for new onset disorders in the three months following injury. This relationship was not observed in the women’s pre-injury or post injury adjusted model. In the current study, firm size was likely a surrogate measure for work environment and working conditions and this variable may have captured additional effects not captured by the occupation measures. However, further research is needed to better understand the relationship between the firm size variable and mental health, and gender differences in this relationship.   88 4.3.5 Strengths and limitations The major strengths of this study include the use of a population-based sample that represents all compensated lost-time work injury for spine or upper limb strain or sprain in BC from 2000 to 2013, as well as the use of longitudinal and comprehensive health data from the pre and post injury time periods. Despite the strengths of this study, there are some potential limitations. First, in the Canadian general population, anxiety and depression disorders are under-diagnosed and under-treated.(83,111,170–172) This is also true of injured worker populations.(63) Workers with an anxiety or depression disorder that did not receive a physician diagnosis would be classified as ‘none’ in our study. This could bias the prevalence estimates downward and possibly bias measures of association between risk factors and prevalent or incident anxiety or depression. However, the extent of Type II error classification bias in the current study is likely less than it would have been, had anxiety and depression disorders been identified using self-reported diagnoses that are less sensitive than methods using health claims data.(64) Second, workers may be under medical surveillance after work injury, especially prior to return to work. Medical surveillance or a greater number of primary care visits due to the work injury, may lead to an increase in the proportion of anxiety and depression disorders that receive a diagnosis and subsequent treatment (rather than going undiagnosed and untreated) during the year after injury relative to the year before injury. This could explain the 2% increase in the prevalence of anxiety and depression in the year after injury compared to the year before injury. Third, it is possible that some new onset post-injury anxiety and depression cases may have been missed in the post-injury time period as workers may become more reliant on the workers’ compensation system versus the public health care system for mental health services. The health claims data provided by the Ministry of Health contained physician visits (including those to a psychiatrist or general practitioner for mental health) and dispensed anxiolytics and anti-depressants insured by WorkSafeBC, but visits to a psychologist or councilor insured by WorkSafeBC were not included in the case definitions. Fourth, due to the high number of odds ratios output from the regression analyses, there is a high probability of Type 1 error among the odds ratios. However, many consistent patterns regarding the effects of certain risk factors were observed across the case groups (anxiety only, depression only, and co-morbid anxiety and depression), across the men’s and women’s models, across the pre-injury and post-injury models, and across levels of the independent variables (e.g. increasing income deciles) – indicating meaningful findings   89 beyond that expected due to random chance alone, many of which are in line with established literature and add face validity to the findings. Lastly, we were not able to examine possible exacerbations of existing anxiety and depression disorders by the injury or the associated work disability with the current administrative data. 4.3.6 Implications  The current findings suggest that workers with an anxiety or depression disorder are a large enough subpopulation of the workers with lost-time upper limb or spine strain or sprain work injury to warrant attention at a policy level. In addition to new onset conditions attributable to the injury, pre-existing anxiety and depression disorders are also important considerations for understanding injured workers’ mental health. The majority of anxiety and depression disorders present in the year before injury will persist into the year after injury, and thus have the potential to affect return to work processes. Health care professionals and policy makers should consider that workers with an anxiety or depression disorder in addition to lost-time upper limb or spine strain or sprain work injury are more likely to be women, middle aged, have a lower income, and to have a more complicated clinical profile including a greater number of somatic comorbidities. Given this, the affordability of mental health care and its integration with other health care delivery in a workers’ compensation context are important considerations. Collectively these descriptive analyses provide a detailed portrait of anxiety and depression disorders among workers with lost-time upper limb or spine strain or sprain work injury for consideration by health care, insurance, and human resources professionals who interact with injured workers through various parts of the social safety net.   90 Chapter 5: Impacts of pre-existing and new onset anxiety and depression on sustained return to work (Research Objective 2) 5.1 Background Anxiety and depression disorders frequently co-occur with physical disorders.(162,173) When co-morbid with one or more physical disorders, they are associated with an increased number of physical symptoms, exacerbation of physical symptoms, prolonged and worsened clinical course for the physical disorder, lower functional ability, and greater role impairment.(134,173–175) Much is known about the clinical and functional effects of anxiety and depression disorders; however, the impact of these common mental disorders on return to work after musculoskeletal injury is less clear. Since 2001, six review articles were identified that included anxiety and/or depression in the analyses of risk factors for return to work outcomes after musculoskeletal injury (Table 11). Conclusions from these articles indicate either that anxiety and depression are not associated with return to work outcomes or that there is insufficient evidence to make a conclusion (Table 11). A comprehensive and systematic review by Iles et al. found moderate evidence that anxiety, and strong evidence that depression, was not associated with return to work outcomes.(176) For most of the reviews, an inadequate number of studies meeting the inclusion criteria was an issue, leading to conclusions of no association based on a limited number of studies (sometimes as low as one) (Table 11). Two recent, prospective cohort studies of Canadian workers with musculoskeletal injury published after the aforementioned systematic reviews also present inconsistent evidence. Depression symptoms were associated with poor return to work in a sample of 332 injured workers in Ontario (anxiety was not examined),(63) while neither anxiety nor depression were associated with return to work in a sample of 62 injured workers in Quebec.(177)  91   Table 11: Summary of review articles that examined the role of anxiety and depression as risk factors for return to work outcomes after musculoskeletal injury    Anxiety  Depression Author Injury type Conclusion #a  Conclusion #a Shaw  (2001) (178)  Acute low back pain Not assessed   No association 1 Steenstra  (2005) (179) Acute low back pain No evidence of association 2  Inconclusive  3 Kuijer  (2006) (180) Chronic low back pain No evidence of association 1  Inconclusive  2 Iles  (2008) (176) Acute low back pain No evidence of association  7  No evidence of association  13 Kent  (2008) (181) Acute low back pain Inconclusiveb  3  No evidence of association 7 Clay  (2010) (182) Acute musculoskeletal injury Not assessed   No available studies 0c a The number of studies included in the review article that assessed the given risk factor (anxiety or depression). bInconclusive: No clear conclusions made by the authors, usually due to conflicting or insufficient evidence. c No available studies: while a search of the literature was performed, no relevant studies meeting the review criteria were found Evidence from the psychiatric and disability literature, beyond study samples for musculoskeletal injury, suggests that anxiety and depression disorders are risk factors for work disability. A 2016 review article found that depression is a risk factor, while anxiety is not a risk-factor, for poor return to work, following a work injury or illness (not specific to musculoskeletal injuries).(183) Psychiatric co-morbidity (including anxiety, depression, and substance abuse disorders) has also been associated with non return to work after herniated disc surgery, and unemployment 12 to 18 months after occupational asthma diagnosis.(184,185) In the general population, anxiety and depression disorders are each independently associated with increased sickness absence from work (174,175), work impairment,(186)  and unemployment,(187,188)  even after accounting for other co-morbid mental and physical disorders, and confounders.  Gender differences in disability due to anxiety and depression disorders is a growing area of research.(189) Among Canadian employees, the rate of compensated short-term disability leave for non-occupational depression is higher for women than for men, but men are less likely to return to work following such a leave.(190) Outside of the injured worker literature, findings based on small clinical samples have suggested that the effects of anxiety and depression disorders on disability are greater for women than for men, but recent population-based epidemiological evidence reports the opposite.(189) The gender specific effects of anxiety and   92 depression disorders on return to work after physical injury are not well understood as prior studies have adjusted or matched for gender, rather than using stratification or interaction/effect modification methods. The impacts of anxiety and depression disorders on work disability could be greater for men compared to women, due to a lower frequency of help seeking behavior among men, and/or a differential impact of work disability for masculine identities with higher personal and sometimes societal valuation placed on income production and ability to perform heavy labour among men.(189) For the latter, work interference by anxiety and depression leading to work disability could pose a greater threat to masculine identities than feminine ones, and the effects of this could result in greater feelings of loss or inadequacy among men that further extenuate the condition of work disability. Conversely, while the gendered division of paid and unpaid labour has continued to equalize in more recent years, with men becoming more involved in unpaid labour and women becoming more involved in paid labour, as of 2015, women in Canada were still spending a greater number of daily hours on work activities than men when both paid and unpaid work were considered.(191) A greater number of hours spent on work by women could slow down recovery and deplete women’s emotional, physical, or motivational resources necessary to cope with work re-entry. Each of these gendered characteristics, behaviours, and social roles could increase vulnerability to the psychological challenges of work disability and physical injury or interfere with recovery and return to work capacity, especially when combined with anxiety or depression.  Given the positive associations of anxiety and depression with pain, functional disability, and work disability from the broader psychiatric and disability literature, deductive reasoning would suggest that that anxiety and depression are risk factors for poor return to work after musculoskeletal injury. However, there is insufficient evidence to directly support this hypothesis and reviews of existing evidence suggest no association. Prior studies are limited by small study samples, short follow up time, and a lack of gender analyses. A secondary consequence of small study samples in addition to low power is that gender analyses using statistical inference is often not feasible. Further, a number of prior studies have examined low back pain, but other common work injuries such as upper limb strains or sprains that are associated with longer disability durations have received relatively less attention. Research in this area is needed to inform return to work policy and programming for physical work injury, in   93 particular the inclusion and use of mental health services, and the importance of gender-based strategies for mental health service delivery by workers’ compensation systems.  The objective of the current study was to address existing gaps and limitations in the literature by conducting a large population-based cohort study for injured workers in the Canadian province of BC. Using linked administrative databases, the effects of anxiety and depression disorders (alone and co-morbid), as diagnosed by a physician, on the probability of sustained return to work following lost-time upper limb or spine strain or sprain work injury was examined. These relationships were examined using multivariable models adjusted for known and potential confounders and using up to one year of follow up time. Gender differences in the strength and direction of these associations were also examined. 5.2 Methods A detailed description of the methods is provided in Chapter 3, but a summary of key methods is reiterated below. 5.2.1 Study sample The study sample consisted of accepted lost-time claims for upper limb or spine strain/sprain (as indicated by the primary diagnosis in the claim file) from 2009 to 2013, for workers between 19 to 64 years of age at the time of injury, registered in the provincial health services plan for at least 275 days (9 months) in both the year before and the year after injury (n=96,870). The study period (2009 to 2013) was chosen based on the availability of the detailed return to work data necessary to derive the outcome variable. To increase the probability that claims in the reference group did not have anxiety or depression, 7,967 (8.2%) claims with an anxiety or depression health care event that did not meet the case definitions (e.g. one single diagnosis from a physician) in the year before injury were excluded. From the remaining 88,903 claims, 4,038 (4.5%) were excluded due to missing data for the study variables. The final study sample consisted of 84,865 claims. 5.2.2 Study variables 1. Outcome variable Sustained return to work was defined as the number of days from injury to the day where the worker returned to non-modified or full job duties and hours (based on their pre-injury job duties   94 and hours), and had no further wage loss benefits or modified work days associated with the claim (Figure 7). Two years of follow up data was used to determine sustained return to non-modified work, prior to censoring at 365 calendar days. By censoring, up to 365 total disability days of observation following injury could be included in the analytic models for each claim, but a full 740 day window was taken into consideration to identify sustained return to non-modified work (Figure 7). In this way, it was ensured that all workers classified as reaching sustained return to non-modified work during the first 365 days following injury truly had no time loss recurrence events after their date of sustained return to work for at least a full year.    Figure 7: Visual description of the sustained return to non-modified work outcome variablea using four hypothetical return to work trajectories a Time to sustained return to non modified work: Worker A - 30 days; worker B - 100 days; worker C - 600 days censored at 365 days; worker D - 740 days censored at 365 days. 2. Primary explanatory variable Based on health care events in the year prior to injury, workers were classified as having anxiety only, depression only, co-morbid anxiety and depression, or none (no anxiety and no depression) in the year prior to injury. Like in the previous analyses, workers were considered to have anxiety or depression during the time period of interest (in this case the year before injury), if 0 days Injury 365 days 1 year 740 days 2 years Worker A Worker B Worker C Worker D 30 days 100 days 600 days Off work or modified work Non-modified work   95 they had at least one index event meeting the corresponding case definition criteria (Table 2 and Table 3 in Chapter 3) during the time period.  3. Potential confounders  The following variables were included as potential confounders of the relationship between anxiety/depression and sustained return to non-modified work: 1) sociodemographic: gender, age, income, dependents in the home, and location, 2) injury: body part, incident type, secondary diagnosis on the claim, 3) clinical: somatic co-morbidity score, other mental co-morbidity (not including anxiety or depression), and prior claims, 4) work: firm size, shift type, and occupation. These variables are described in detail in section 4.2. 5.2.3 Analyses All analyses were conducted using SAS 9.4.(139) 1. Impact of pre-existing anxiety and depression prevalent in the year prior to injury The baseline sociodemographic, injury, clinical, and work characteristics of the study sample were examined using frequencies and proportions. Unadjusted and adjusted Cox models (192) stratified by gender were used to examine the impact of anxiety and depression disorders from the year before injury on the probability of sustained return to non-modified work by calculating hazard ratios (HR) with 95% confidence intervals. The proportionality assumption was tested for the anxiety and depression variable using Kaplan-Meier curves and by testing for interaction with the log of the time variable (number of days since injury) in the Cox models. No evidence of non-proportionality was found for the anxiety or depression variable. The effect of each confounder was analyzed by adding the potential confounder independently to the unadjusted models. All potential confounders were included in the final adjusted models. As the somatic co-morbidity index score variable was both a strong confounder and a composite measure, possible residual confounding due to somatic co-morbidity was investigated by adding in dummy variables for each ICD-9 diagnostic category (mental disorders excepted for this analyses) one by one to the fully adjusted Cox proportional hazard models. For each ICD-9 diagnostic category, a value of ‘1’ was assigned to the dummy variable if the worker had at least one physician visit or hospitalization with a diagnosis from that category in the year before injury.     96 1 a) Additive effects While multiplicative Cox models are the most common tool for analyzing time to event data with censoring in observational epidemiology, additive effect measures are more easily interpreted and policy relevant. To address this, the impacts of pre-existing anxiety and depression disorders prevalent in the year before injury were examined on the additive scale using direct-adjusted survival curves and life table methods as described by Zhang et al.(193) Using this method, unadjusted and adjusted median times to sustained return to non-modified work and interquartile ranges were calculated for each level of the primary explanatory variable (anxiety and depression in the year before injury) for both men and women. Censoring the sustained return to non-modified work variable at 365 days did not affect the median or interquartile range estimates, as these values were all less than 365. 1 b) Effect modification by gender  The effect size, direction, and statistical significance of effect modification of the primary relationship (impact of pre-existing anxiety and depression on sustained return to non-modified work) by gender were examined on the multiplicative and additive scales. For the multiplicative scale, a non-stratified Cox model was constructed for the entire study sample with all potential confounders (including gender), and an interaction term for gender with the primary explanatory variable (anxiety and depression in the year before injury), included. For the additive scale, the relative excess risk due to interaction (RERI) was calculated using methods by Li and Chambless (2007).(194)  2. Impact of new onset anxiety and depression during the return to work period The impacts of new onset anxiety and/or depression occurring during the return to work period on the probability of sustained return to non-modified work were also examined. To do this, Cox models stratified by gender, adjusted for all confounders were replicated with a study sample restricted to the ‘none’ group without anxiety or depression prior to injury. For the exposure variable, a time varying measure of anxiety and depression post injury was used (none, anxiety only, depression only, and co-morbid anxiety and depression). The first relevant index event occurring after injury but before sustained return to non-modified work was used to determine the timing of anxiety and or depression onset. Once a worker was classified as having anxiety only, depression only, or co-morbid anxiety and depression they remained in this classification   97 until the end of follow-up. For descriptive purposes, the median time to case onset for workers who developed a new anxiety or depression episode during the return to work period was calculated. 5.3 Results 5.3.1 Distribution of the study sample The baseline characteristics of the study sample were similar to those described in the previous chapter (Table 5), and thus are not re-presented here.  5.3.2 Distribution of sustained return to non-modified work by anxiety and depression status For men, workers with no anxiety or depression (i.e. the none group) were more likely to reach sustained return to non-modified work within 365 days (81.0%) than workers with anxiety only (75.9%), depression only (78.7%), and co-morbid anxiety and depression (77.4%) (Table 12). For women, minimal differences were observed between the groups (none 85.8%; anxiety only 84.6%; depression only 85.0%; anxiety and depression 83.4%).  Table 12: Final study sample of lost-time upper limb or spine strain or sprain work injury claims in BC by anxiety and depression case status in the year before injury and sustained return to non-modified work within 365 days (yes/no), 2009 to 2013  Men (N=48,915)  Women (N=35,950) Case status  Total n (%) SRTWa n %  Total n (%) SRTWa n % None 41,827 (85.5) 33,873 (81.0)  24,677 (68.6) 21,172 (85.8) Anxiety only  1,970 (4.0)  1,495 (75.9)  3,100 (8.6) 2,623 (84.6) Depression only 1,483 (3.0) 1,167 (78.7)  2,025 (5.6) 1,722 (85.0) Anxiety & depression  3,635 (7.4) 2,812 (77.4)   6,148 (17.1) 5,129 (83.4) a Proportion of the total group that reached sustained return to non-modified work (SRTW) within 365 days 5.3.3 Impact of pre-existing anxiety and depression prevalent in the year prior to injury  1. Unadjusted and adjusted Cox models In the unadjusted Cox models, compared to workers in the none group, workers with anxiety only (Men HR: 0.85 (95% CI: 0.80, 0.89); Women HR: 0.94 (95% CI: 0.90, 0.98)), depression only (Men HR: 0.91 (0.86, 0.96); Women HR: 0.96 (0.92, 1.01), or co-morbid anxiety and depression (Men HR: 0.89 (0.86 to 0.93); Women HR: 0.93 (0.90 to 0.96)) in the year prior to injury had lower probability of sustained return to non-modified work (Table 13). Similar but attenuated associations were observed in the adjusted Cox models with all potential confounders,   98 although the 95% confidence intervals around the estimates now included ‘1’ for anxiety only in the men’s model, and for anxiety only and depression only in the women’s model.     99 Table 13: Unadjusted and adjusteda results for Cox regression models, probability of sustained return to non-modified work after lost-time upper limb and spine strain or sprain work injury in BC from 2009 to 2013  Men (N=48,915)    Women (N=35,950)  Unadjusted  HR (95% CI)  Adjusted HR (95% CI)   Unadjusted HR (95% CI)  Adjusted HR (95% CI)  Case status (based on the year before injury, time invariant)      None 1 1  1 1      Anxiety only 0.85 (0.80-0.89) 0.88 (0.84-0.93)  0.94 (0.90-0.98) 0.96 (0.92-1.00)      Depression only 0.91 (0.86-0.96) 0.94 (0.89-1.00)  0.96 (0.92-1.01) 0.98 (0.93-1.03)      Anxiety and depression 0.89 (0.86-0.93) 0.94 (0.90-0.97)  0.93 (0.90-0.96) 0.95 (0.92-0.98) Socio-demographic       Age group (years)           19 to 24 1 1  1 1      25 to 29 0.97 (0.93-1.02) 0.95 (0.91-0.99)  0.96 (0.91-1.01) 0.89 (0.84-0.94)      30 to 39 0.89 (0.86-0.92) 0.86 (0.83-0.89)  0.87 (0.83-0.91) 0.82 (0.78-0.87)      40 to 49 0.83 (0.80-0.86) 0.82 (0.78-0.85)  0.83 (0.80-0.87) 0.79 (0.76-0.83)      50 to 59 0.79 (0.76-0.81) 0.80 (0.76-0.83)  0.82 (0.79-0.86) 0.78 (0.74-0.82)      60 to 64 0.70 (0.66-0.74) 0.74 (0.70-0.78)  0.84 (0.79-0.90) 0.79 (0.74-0.85) Income quintile           1: lowest 1 1  1 1      2 1.03 (0.99-1.06) 1.02 (0.98-1.06)  1.11 (1.08-1.15) 1.10 (1.06-1.13)      3 1.07 (1.04-1.11) 1.07 (1.03-1.11)  1.19 (1.15-1.23) 1.17 (1.13-1.21)      4 1.05 (1.02-1.09) 1.09 (1.05-1.13)  1.15 (1.10-1.19) 1.12 (1.08-1.17)      5: highest 1.04 (1.00-1.07) 1.09 (1.05-1.13)  1.28 (1.23-1.33) 1.17 (1.12-1.23) Dependents           0 1 1  1 1      1 or more 1.04 (1.02-1.06) 1.03 (1.00-1.05)  0.97 (0.94-0.99) 0.97 (0.94-1.00) Location           Urban 1 1  1 1      Rural 0.88 (0.85-0.91) 0.93 (0.90-0.97)  0.96 (0.93-1.00) 0.98 (0.95-1.02) Injury      Injured body part           Sacroiliac 1 1  1 1      Back/neck 0.94 (0.92-0.97) 0.96 (0.94-0.99)  0.94 (0.91-0.98) 0.95 (0.91-0.98)      Upper limb 0.69 (0.67-0.72) 0.70 (0.68-0.72)  0.78 (0.75-0.81) 0.76 (0.73-0.79) Incident type           Exertion 1 1  1 1      Traumatic 1.11 (1.02-1.19) 1.08 (1.00-1.17)  1.12 (1.06-1.18) 1.11 (1.05-1.17)      Fall/slip/trip 0.86 (0.84-0.89) 0.98 (0.95-1.01)  1.01 (0.98-1.04) 1.11 (1.07-1.15)      Contact object 1.05 (1.00-1.10) 1.19 (1.14-1.25)  1.08 (1.02-1.13) 1.20 (1.14-1.26)      Transportation 0.85 (0.82-0.89) 0.94 (0.90-0.98)  1.01 (0.95-1.08) 1.09 (1.02-1.17)      Bodily motion 0.95 (0.93-0.97) 0.98 (0.96-1.01)  1.05 (1.02-1.08) 1.01 (0.98-1.04) Secondary diagnosis on the claim           No 1 1  1 1      Yes (not anx or dep) 0.81 (0.78-0.83) 0.81 (0.79-0.83)  0.86 (0.83-0.88) 0.81 (0.79-0.84)     100  Men (N=48,915)     Women (N=35,950)  Unadjusted  HR (95% CI)  Adjusted  HR (95% CI)   Unadjusted  HR (95% CI)  Adjusted  HR (95% CI)  Clinical      Somatic co-morbidity           0 1 1  1 1      1 1.00 (0.96-1.03) 0.99 (0.96-1.03)  0.97 (0.91-1.03) 0.96 (0.90-1.02)      2 0.96 (0.92-0.99) 0.95 (0.92-0.99)  0.95 (0.89-1.00) 0.94 (0.88-0.99)      3 0.90 (0.87-0.93) 0.90 (0.87-0.94)  0.93 (0.88-0.99) 0.92 (0.87-0.97)      4 0.88 (0.85-0.92) 0.90 (0.86-0.93)  0.90 (0.84-0.95) 0.89 (0.84-0.95)      5 or more 0.82 (0.79-0.85) 0.85 (0.81-0.88)  0.84 (0.79-0.89) 0.85 (0.80-0.90) Mental co-morbidity (other than anxiety or depression)           0 1 1  1 1      1 or more 0.86 (0.82-0.90) 0.91 (0.86-0.96)  0.95 (0.90-1.01) 0.99 (0.94-1.05) Prior claims           0 1 1  1 1      1 0.82 (0.79-0.85) 0.85 (0.81-0.88)  0.84 (0.79-0.89) 0.85 (0.80-0.90)      2 0.86 (0.82-0.90) 0.91 (0.86-0.96)  0.95 (0.90-1.01) 0.99 (0.94-1.05)      3 or more 0.99 (0.96-1.01) 0.99 (0.96-1.01)  0.97 (0.94-0.99) 0.96 (0.93-0.99) Work      Firm size           30 or less 1 1  1 1      31 to 150 1.17 (1.14-1.20) 1.13 (1.10-1.16)  1.20 (1.15-1.25) 1.19 (1.14-1.24)      151 to 1000 1.19 (1.16-1.22) 1.15 (1.12-1.18)  1.19 (1.14-1.24) 1.18 (1.14-1.23)      1001 to 10000 1.28 (1.25-1.32) 1.18 (1.15-1.22)  1.22 (1.17-1.26) 1.21 (1.16-1.26)      10,001 or more 1.53 (1.44-1.63) 1.28 (1.19-1.37)  1.37 (1.32-1.43) 1.31 (1.25-1.36) Shift type           Fixed 1 1  1 1      Other 0.89 (0.88-0.91) 0.88 (0.86-0.90)  1 (0.98-1.03) 0.94 (0.92-0.97) Occupation           Sales and services 1 1  1 1      Art, culture, recreation 0.94 (0.85-1.05) 0.99 (0.89-1.11)  1.13 (1.03-1.25) 1.04 (0.94-1.15)      Business, finance, admin. 0.93 (0.88-0.98) 0.89 (0.85-0.94)  1.10 (1.05-1.16) 1.07 (1.02-1.13)      Health 1.21 (1.15-1.28) 1.22 (1.15-1.29)  1.17 (1.14-1.20) 1.08 (1.04-1.12)      Management 0.84 (0.77-0.91) 0.90 (0.83-0.98)  1.00 (0.92-1.08) 0.96 (0.88-1.03)      Natural/applied sciences 0.83 (0.78-0.88) 0.84 (0.79-0.90)  0.92 (0.80-1.05) 0.84 (0.74-0.96)      Primary industry 0.81 (0.76-0.86) 0.84 (0.79-0.89)  0.79 (0.70-0.89) 0.79 (0.70-0.89)      Processing, utilities 0.85 (0.82-0.89) 0.83 (0.80-0.87)  0.85 (0.80-0.91) 0.83 (0.78-0.89)      Social science, govern. 1.29 (1.19-1.40) 1.34 (1.23-1.45)  1.29 (1.23-1.35) 1.22 (1.17-1.28)      Trades, transport 0.84 (0.81-0.86) 0.84 (0.82-0.87)  0.90 (0.85-0.95) 0.87 (0.83-0.92) a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, occupation     101 2. Unadjusted and direct adjusted survival curves The unadjusted median time to sustained return to non-modified work was 35 days for men in the none group, and 52, 43, and 45 days for men with anxiety only, depression only, and co-morbid anxiety and depression. Stated differently, compared to men in the none group, men with anxiety only, depression only, and co-morbid anxiety and depression took 17, 8, and 12 days longer, respectively, to reach sustained return to non-modified work (Table 14). The same pattern for the unadjusted median time to sustained return to non-modified work by baseline exposure status was observed for women, but the unadjusted absolute differences between the none and the case groups was less pronounced (None 40 days, anxiety only 46 days, depression only 42 days, co-morbid anxiety and depression 44 days). Women with anxiety only, depression only, and co-morbid anxiety and depression took 6, 2, and 4 days longer to reach sustained return to non-modified work than women in the none group, respectively. Similar results were found in the stratified survival models with direct adjustment for potential confounders but the effects of anxiety and depression were attenuated. After adjustment for potential confounders, men in the anxiety only, depression only, and co-morbid anxiety and depression took 14, 7, and 7 days longer to reach sustained return to non-modified work than men in the none group, respectively; and these same corresponding values for women were 5, 2, and 3 days longer (Table 14) Table 14: Unadjusted and direct adjusteda results for survival curves by anxiety and depression status in the year before injury, time (days) to sustained return to non-modified work after lost-time upper limb or spine strain or sprain work injury in BC from 2009 to 2013  Men (N=48,915)  Women (N=35,950) Case status Unadjusted Median (IQR) Adjusted Median (IQR)  Unadjusted Median (IQR) Adjusted Median (IQR) None 35 (10, 148) 34 (10, 153)   40 (10, 109) 40 (10, 111) Anxiety only  52 (12, 313) 48 (12, 240)  46 (11, 125) 45 (11, 123) Depression only 43 (12, 201) 41 (11, 177)   42 (11, 115) 42 (10, 114) Anxiety & depression  45 (11, 237)   41 (10, 206)  44 (11, 125) 43 (11, 122) a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, occupation  3. Effect modification by gender In the non-stratified Cox adjusted model, the co-efficient estimates for the interaction terms between gender and the primary explanatory variable (anxiety and depression in the year before injury) were all greater than 0, but only the positive co-efficient estimate for the interaction between anxiety only and women was significant at a 95% level of confidence (Table 15).   102 Interpretation of this in conjunction with the co-efficients for the gender and anxiety and depression variables (the non interaction terms) (Table 15) demonstrates that, on a multiplicative scale, anxiety only (compared to none) was associated with a lower probability of sustained return to work for both women and men, but that the strength of this relationship was significantly greater for men than for women (i.e. men experienced a lower probability of return to work attributable to anxiety than did women, even though anxiety was also associated with a lower probability of return to work among women). Similar but non-significant interactions were observed between depression and gender, and comorbid anxiety and depression with gender on a multiplicative scale.  Similar results were found on the additive scale with the RERI indicating that the negative effects of anxiety only on the probability of sustained return to non-modified work were significantly greater for men than for women, and the negative effects of depression only and co-morbid anxiety and depression were non-significantly greater for men than for women (Table 15). Table 15: Effect modification of the relationship between anxiety and depression (based on the year before injury) and sustained return to non-modified work by gender on the multiplicative and additive scales controlling for potential confoundersa, lost-time upper limb or spine strain or sprain claims in BC from 2009 to 2013  Multiplicative scale Additive scale  Co-efficient  (b) estimate 95% CI p-Values RERI  estimate 95% CI p-Values Anx only  -0.13 -0.18 to -0.08 <0.01    Dep only  -0.07 -0.12 to -0.01 0.03    Anx & Dep  -0.07 -0.11 to -0.03 <0.01    Women  -0.02 -0.05 to -0.00 0.04    Anx only*women 0.09 0.02 to 0.15 <0.01 0.08 0.02 to 0.14 <0.01 Dep only*women 0.05 -0.03 to 0.13 0.20 0.01 -0.03 to 0.06 0.54 Anx & dep*women 0.02 -0.03 to 0.07 0.37 0.04 -0.03 to 0.11 0.27 a Adjusted for age group, gender, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, occupation  4. Associations of potential confounders with sustained return to non-modified work  Baseline characteristics associated with increased probability of sustained return to non-modified work in both the men’s and women’s adjusted Cox models included: higher wage, incidents that were traumatic in nature or involved contact with an object (compared to over exertion   103 incidents), increasing firm size; and occupations in health, or in social science, government or education (compared to sales and service) (Table 13). Characteristics associated with lower probability of sustained return to non-modified work in both the men’s and women’s adjusted models included: older age groups, rural location, strains or sprain to the back/neck or upper limb (compared to the sacroiliac region), a secondary diagnosis on the claim, increasing levels of somatic co-morbidity, prior claims, a non-fixed shift type (compared to a fixed shift type), and occupations in natural and applied sciences, primary industry, processing, manufacturing, or utilities, or trades, transport or equipment operation (compared to sales and service) (Table 13).  Having a dependent was associated with increased probability of sustained return to non-modified work for men but decreased probability for women (Table 13). Compared to exertion incidents, transportation incidents were associated with lower probability of sustained return to non-modified work for men, and higher probability for women. Falls, slips, and trips were associated with increased probability of sustained return to non-modified work for women, but no association was observed for this incident type for men.  Having a mental co-morbidity other than anxiety or depression was associated with lower probability of sustained return to non-modified work for men, but had no effect for women. Business, finance and administrative occupations were associated with lower probability of sustained return to non-modified to work for men, but higher probability for women.  5. Confounding analysis  The addition of individual potential confounders to the unadjusted Cox models stratified by gender was associated with only minor changes in the hazard ratios for the anxiety and depression variables on the return to work outcome (<10%) (data not shown). The strongest confounder in the both the men’s and women’s models was the somatic co-morbidity index score, although the addition of this variable to the men’s and women’s unadjusted models only caused the hazard ratios for the anxiety and depression variable to change in value by 2% to 4%. 5.3.4 Impact of new onset anxiety and depression occurring during the return to work period Among men and women who developed new onset anxiety or depression during the return to work period (i.e. after injury but before sustained return to non-modified work), the median time   104 to the first anxiety or depression health service event was shorter for women (Anxiety: 77 days, depression: 99 days, anxiety and depression: 67 days) than for men (Anxiety: 78 days, depression: 104 days, anxiety and depression: 97 days) (Table 16). In terms of the impacts of these new onset disorders on return to work, for both men and women, new onset anxiety only (Men HR: 0.82, 95% CI: 0.72 to 0.95; Women HR: 0.89, 95% CI: 0.79 to 1.01,), depression only (Men HR: 0.69, 95% CI: 0.56 to 0.86; Women HR: 0.72, 95% CI: 0.58 to 0.90,), and co-morbid anxiety and depression (Men HR: 0.63, 95% CI: 0.54 to 0.74; Women HR: 0.65, 95% CI: 0.57 to 0.74) were associated with lower probability of sustained return to non-modified work, in Cox models adjusted for confounders.  Descriptively, there was some support for effect modification by gender on the relationship between new onset anxiety and depression disorders on the probability of sustained return to non-modified work, although the statistical significance of this was not tested and the differences between men and women were not striking. Consistent with the analysis for pre-existing anxiety and depression conditions, new onset anxiety and depression disorders were associated with lower probability of sustained return to non-modified work for both men and women, but the effect sizes of these relationships were descriptively greater for men. This is indicated by the lower adjusted hazard ratios (further away from 1 and below 1) for men compared to women (Table 16).    105 Table 16: Frequency and timing (days) of new onset anxiety and/or depression disorders and impact on sustained return to non-modified work (SRTW) on the additive scale after lost-time upper limb or spine strain or sprain work injury in BC from 2009 to 2013  Men (N=41,827)  Women (N=24,677) New onset anxiety or depressiona N (%) Median (IQR) days to disorder onset Adjusted  HR (95% CI)  to SRTWb  N (%) Median (IQR) days to disorder onset Adjusted  HR (95% CI)  to SRTWb None 40,474 (96.8)  1  23,479 (95.1)  1 Anxiety only  522 (  1.3) 78 (26, 180) 0.82 (0.72-0.95)  502 (  2.0) 77 (30, 164) 0.89 (0.79-1.01) Depression only 272 (  0.7) 104 (38, 155) 0.69 (0.56-0.86)  199 (  0.8) 99 (36, 192) 0.72 (0.58-0.90) Anxiety & depression  559 (  1.3) 97 (38, 193) 0.63 (0.54-0.74)  497 (  2.0) 67 (22, 145) 0.65 (0.57-0.74) a Only new onset disorders occurring after injury but before sustained return to non-modified work were considered bAdjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, occupation    106  5.4 Discussion 5.4.1 Summary of main findings For men, pre-existing anxiety only, depression only, and co-morbid anxiety and depression were associated with lower probability of sustained return to non-modified work after lost-time upper limb or spine strain or sprain work injury, although the 95% confidence interval for depression only included ‘1’. Similar directions of association were found for women, although the effect size for depression was small, and the 95% confidence intervals for anxiety only and depression only both included ‘1’. Among men, the effect sizes for the relationships between pre-existing disorders and the probability of sustained return to work were all descriptively greater than corresponding estimates obtained from the women’s sample. This indicates that gender is a potential modifier of these relationships. Further evidence of modification by gender was found for the relationship between pre-existing anxiety and sustained return to work in tests of statistical significance. For pre-existing disorders, anxiety only was associated with the longest time to sustained return to non-modified work in both men and women (an excess of 14 and 5 calendar days respectively, compared to the none groups). For new onset conditions, anxiety only, depression only, and comorbid anxiety and depression were all associated with decreased probability of sustained return to work for both men and women; and the effects remained descriptively stronger for men than for women, although no tests of statistical significance for effect modification by gender was used for new onset disorders. Lastly, the decreased probability of sustained return to work associated with new onset disorders was descriptively larger than for pre-existing ones.   5.4.2 Anxiety and depression and return to work The finding that anxiety and depression disorders are associated with decreased probability of sustained return to work is consistent with what would be expected based on the associations of these disorders with work disability in the broader psychiatric and disability literature. Despite this consistency, potential mechanisms of these relationships may differ between these populations due to the unique contexts within which they occur (e.g. a workers’ compensation context versus the general potential labour force including unemployed individuals), although there are likely many shared mechanisms as well. Potential mechanisms for musculoskeletal work injury populations are discussed in the following paragraph. It is also important to note that   107 the current findings differ from many studies specific to musculoskeletal injury and return to work. Prior studies on return to work after musculoskeletal injury may have been underpowered due to small sample size, and may have underestimated the impacts of anxiety and depression on return to work due to short term follow up or use of a return to work outcome such as first return to work that does not capture recurring episodes of work disability within a claim. Other possible reasons for inconsistent findings across studies include differences in anxiety and depression measurement and study sample selection criteria. Use of self-reported anxiety and depression symptoms during the return to work time-period, and exclusion of workers with an anxiety or depression diagnoses prior to injury, may result in a higher representation of workers with temporary situational anxiety or depression symptoms in the anxiety and depression case groups of other studies.  It is possible that temporary or situational symptoms that resolve themselves without treatment, may not significantly impact return to work, as observed in prior studies. Possible mechanisms of the anxiety-return to work, and depression-return to work relationships for workers with musculoskeletal work injury may be: 1) the amplification of pain, 2) activity disruption or a general loss of interest in social, occupational, and recreational activities that promote well-being and likely also return to work processes, 3) resistance or non-adherence to therapeutic treatment of the musculoskeletal injury, 4) self-assessed inability to perform work tasks effectively due to symptoms of fatigue, stress, fear, or anxiety, especially in the context of returning to a work environment that caused musculoskeletal injury, 5) difficulty navigating social interactions involved in the return to work process,(53,55–58) possibly due to mental health stigma or a negative response to mental symptoms by others,(63) and 6) exacerbation of mental symptoms by procedural aspects of the workers’ compensation experience such as adjudication decisions and medical assessments.  5.4.3 Co-morbid anxiety and depression versus anxiety only or depression only Evidence from primary care research suggests that co-morbid anxiety and depression is associated with a higher severity of illness, lower quality of life and greater impairment in work and psychosocial functioning compared to depression only and anxiety only.(195) Based on the aforementioned literature, it was hypothesized a priori that co-morbid anxiety and depression would have a greater negative impact on return to work than anxiety only or depression only. The findings on the impacts of new onset anxiety and depression on return to work support this   108 hypothesis as the largest effect size was observed for the comorbid anxiety and depression group in both the men’s and women’s models. However, findings on the impacts of pre-existing anxiety and depression on return to work did not indicate that co-morbid anxiety and depression have more severe impacts than anxiety alone or depression alone. Diagnostic coding practices in BC may have resulted in an underestimate of the impacts of co-morbid anxiety and depression on return to work in the current study, especially relative to anxiety alone or depression alone. The comorbid anxiety and depression group had a higher frequency of code 50b than the anxiety alone group and the depression alone group (Table 28). According to Bilsker and colleagues, the BC specific diagnostic code 50b for anxiety/depression is generally reserved for more minor episodes;(129) although secondary sources to verify or refute this statement are not readily available. Inclusion of code 50b in the anxiety and depression case definitions may have led the co-morbid anxiety and depression case group to contain a relatively high number of workers with minor episodes, in comparison to the depression only and anxiety only groups.  5.4.4 Gender differences The finding that the negative influence of pre-existing anxiety on return to work is significantly greater for men than for women is similar to recent epidemiological findings in the general population that men experience more role, social, and cognitive disability due to anxiety and depression than women.(80,81) The current finding for anxiety may be due to gender differences in the timing of the first anxiety diagnosis, the type of treatment received after the first diagnosis, occupational performance expectations, access to social support, emotional disclosure, and graduated return to work opportunities. In Canada, odds of mental health service use for anxiety in the past year is 1.6 times higher for women with anxiety than for men with anxiety.(83) Treatment for anxiety may buffer the potential negative effects of this disorder on return to work, as well as buffer the negative mental health effects of a significant life event such as being injured and off work/on disability benefits. Findings from the current study also suggest that men may seek medical care at a later period in the clinical course of a mental disorder. Gender differences in the timing of the first diagnosis among workers who developed a new onset disorder during the return to work time period were observed in the current study for depression alone (5 days earlier for women) and comorbid anxiety and depression (30 days earlier for women), but not for anxiety alone (1 day earlier for women). As proposed by Scott and Collings:   109 1) men’s high occupational performance expectations in combination with a high load of psychological distress (feelings of fear or worry), may lead men with anxiety to self-assess themselves as unable to work; 2) men may be less likely to disclose negative emotions related to an episode of anxiety that in turn may create feelings of isolation or less opportunities for mental health support; and 3) men may have smaller social support networks than women that may also lead to less opportunities for mental health support during a significant health/life event, although the evidence for this latter point is mixed.(189) Lastly, evidence suggests that men with work injury in BC have fewer graduated return to work options than women.(196) A longer detachment from the workplace (due to an absence of graduated return to work and thus an absence of the therapeutic effects of work) may facilitate the bi-directional relationships of anxiety with work disability, resulting in longer time to sustained return to work.  In the background section of this chapter, a greater number of total hours spent on a combination of paid and unpaid work was provided as a possible reason for why women might experience greater difficulty in return to work compared to men, especially when an anxiety or depression disorder is present. The current study is unable to rule in or out any of these potential mechanisms. However, the findings do suggest that, at a population level, potential mechanisms to explain greater mental health related disability among men are prominent. 5.4.5 Strengths and limitations Strengths of the current study include the use of a longitudinal population-based dataset, measures of anxiety and depression from before and after injury based on health care events, a longer follow up period than most prior relevant studies, minimal loss to follow up, and use of a return to work outcome that accounts for recurring instances of lost-time. There are some potential limitations regarding the measurement of anxiety and depression.  First, workers with an anxiety or depression disorder could be misclassified as ‘none’ due to under treatment and under diagnosis of these disorders, or use of private mental health services (not captured in the health claims data) as opposed to public ones. This would have resulted in an underestimation of the main effects observed in the current study. To minimize this bias, workers with an anxiety or depression health care event not sufficient to meet the case definitions were excluded (i.e. the ‘case-like’ group from Chapter 3). Building on this issue, differential   110 misclassification of workers with depression or anxiety by gender, could have contributed to the findings for gender modification. In particular, men with mild or depression or anxiety may have been less likely to receive a diagnosis or treatment than women with mild depression or anxiety. However, the impacts of this issue are estimated to be small, especially due to the exclusion of the ‘case-like’ group in the current analyses – as this group likely represents a high proportion of mild or transient cases.  Based on Canadian data from the early 2000’s, men with depression were approximately 3% to 5% less likely to have a mental health service contact with a physician (general practitioner or specialist) than women with depression.(197,198) This gender difference disappeared when only specialist mental health service contacts are considered.(197,198) This could suggest that in the current study, men with injuries occurring at the beginning of the study period (early 2000’s) who also had mild or moderate depression may have been more likely (~3%) to be misclassified as ‘none’ than similar women, but that no gender differences in misclassification would be expected for men and women with more severe depression. However, there remains the possibility that general practitioners may be more likely to refer men to specialty mental health care than women, even when symptoms are similar. Data from the early 2000s also suggests that men with anxiety disorders were less likely to seek mental health service contact with a physician than women during this time period, although the gender differences in use of specialty care were non-significant.(83)  Lastly, data from 2014 suggests that in more recent years men and women with a self-reported anxiety or depression disorder are equally as likely to seek treatment.(199) This suggests that, if differential misclassification of anxiety and depression by gender is driving the findings regarding gender modification, the effects of this would be more prominent during the earlier study period (early 2000s) than the later period (late 2000s, early 2010s). This remains a topic of investigation for future analyses. The presence of subclinical or short duration anxiety or depression symptoms in the ‘none’ group during the post injury time period is also not fully characterized. Other studies indicate that even after excluding workers with pre-existing depression from prior to injury (defined as any self-reported diagnosis of depression in the year prior to injury), 40% to 50% of the remaining workers have high levels of depression symptoms following lost-time musculoskeletal injury. (36,37,140,200,201) Based on this, it would be   111 expected that a notable proportion of the ‘none’ group experienced high levels of depression or anxiety symptoms during follow up, even if the criteria for a disorder (based on symptom severity and of sufficient duration for a physician visit and a diagnosis) was not met. The inclusion of workers with subclinical or short duration symptoms in the ‘none’ group in the current study could have contributed to smaller effect sizes than had these workers been excluded from the study. Similarly, their inclusion in the case groups would have led to smaller effect sizes if their symptoms were not associated with longer disability durations. 5.4.6 Implications Workers’ compensation benefits and programs intended to improve return to work after lost-time upper limb or spine strain or sprain work injury should take pre-existing anxiety and depression disorders into consideration, in addition to new onset disorders attributable to the injury; and, gender specific strategies may be warranted to optimize return-to-work outcomes. Based on these findings, further investigation of how anxiety and depression affect return to work processes and transitions, during the time course from injury to sustained return to non-modified work, and the identification of gender specific strategies to address this, is warranted.    112 Chapter 6: Impacts of pre-existing anxiety and depression on return to work and lost-time recurrence events (Research Objective 3) 6.1 Background Sustained return to non-modified job duties is a common return to work goal for workers with lost-time injury,(33) especially for injuries like low back pain typically characterized by return to work within the first 30 days of disability.(29) Following injury, progression towards a return to work goal such as this is a dynamic and complex process often characterized by multiple events, transitions, and phases.(33,202) According to Young et al (2005), return to work phases include being off work (lost-time), work re-entry, retention (remaining at work after re-entry), and advancement.(33)  Recurrence of lost workdays after work re-entry is common for lost-time musculoskeletal injuries. In the year following initial return to work for low back pain, approximately 20% of workers in Quebec and 25% of workers in BC experience additional time off work due to the injury.(203,204) Episodes of lost-time recurrence after initial return to work for low back pain are also longer and more costly than first episodes, and as a result, recurrences contribute disproportionately to the total economic and disability burden of low back pain.(205,206)   There has been much investigation of risk factors for poor return to work outcomes, but not for lost-time recurrences specifically. Further, few studies delineate between factors that impact probability of return to work (from lost-time) and factors that impact risk of lost-time recurrence (after initial return to work), even though the factors that influence these two distinct events may vary. In regards to mental health and lost-time recurrence after lost-time musculoskeletal injury, Bultmann et al. (2007) found a positive association between high levels of depression symptoms at one month post injury and the probability of having experienced a lost-time recurrence in the one month prior.(207) While this study was longitudinal, participants were assessed at 1-month post injury and at 6 months follow-up allowing for the possibility of lost-time recurrence to precede depression. A prospective Canadian cohort study in Ontario found a positive prospective association between the development of self-reported post-injury depression symptoms and lost-time recurrence.(36,63) However, Marras et al. (2007) found that psychological symptoms were not associated with recurrence of lost work days for low back pain.(208)    113 Intermediate return to work events such as work reentry or recurrence of lost-time can be used to assess progression to a sustained return to work goal.(33) A better understanding of the factors that affect these intermediate events can provide a rich context to inform the timing and nature of interventions intended to improve the return to work process. While high levels of depression symptoms during the post-injury period have been positively associated with lost-time recurrence, no study has examined clinically verified disorders already present or recently present at the time of injury. Rather, the focus has been on new onset depression symptoms that develop post-injury. Further, few studies have examined if anxiety is a risk factor for lost-time recurrence. This is relevant as pre-existing anxiety and depression disorders are both common and could potentially represent early prognostic factors for poor return to work and recurrence events. Further evidence is needed to demonstrate that anxiety and depression disorders prospectively impact the risk of lost-time recurrence, to inform the development of interventions based on modifiable risk factors that precede recurrence events.  The objective of the current study was to examine the impacts of pre-existing anxiety and depression disorders prevalent in the year before injury on the probability of: 1) return to non-modified work (from lost-time) and 2) recurrence of lost-time (after initial return to non-modified work) after lost-time upper limb or spine strain or sprain work injury. These two types of return to work events were analyzed separately using multivariable models suitable for the analyses of recurring events, adjusted for known confounders using up to two years of follow up time.  6.2 Methods A detailed description of the data sources is provided in Chapter 3, but a summary of key methodological steps is reiterated here. 6.2.1 Study sample The study sample consisted of accepted lost-time claims for upper limb or spine strain/sprain (as indicated by the primary diagnosis in the claim file) from 2009 to 2013, for workers between 19 to 64 years of age at the time of injury, registered in the provincial health services plan for at least 275 days (9 months) in both the year before and the year after injury (n=96,870). The study period (2009 to 2013) was chosen based on the availability of the detailed return to work data   114 necessary to derive the outcome variables. To increase the probability that claims in the reference group did not have depression or anxiety, 7,967 (8.2%) claims with an anxiety or depression health care event that did not meet the case definitions (e.g. one single diagnosis from a physician) in the year before injury were excluded. From the remaining 88,903 claims, 4,996 (5.6%) claims belonging to multiple job-holders were excluded as return to work trajectories for these workers are complicated and can have overlapping return to work phases not suitable for analysis (e.g. worker was off work for job A but at work for job B during the same time period depending upon the nature and cause of the injury). Lastly, 5,771 (6.9%) of 83,907 claims were excluded due to missing data for the study variables, or for administrative error in the detailed return to work data. Examples of administrative error in the detailed return to work data included multiple records documenting conflicting return to work phases (e.g. off work and full return to work) for the same time period, even though the worker was a single job-holder. The final study sample included 78,136 claims. 6.2.2 Study variables 1. Outcome variables Detailed return to work calendar data obtained from the provincial workers’ compensation system in BC was used to construct two outcome variables that were analyzed separately: return to work and recurrence (i.e. lost-time recurrence associated with the injury claim). Included in each claim record were the start and end dates of every return to work event. Possible return to work events in the data included off work (short term disability), no return to work (long term disability), modified return to work (at work but with modifications to the pre-injury job duties or hours), other employment earnings (at work but with a new job, new employer, and/or new industry compared to the pre-injury job), and return to work (at work with no modifications to the pre-injury job duties or hours). For the current study, the events of off work, no return to work, and modified return to work were grouped together and collectively referred to as lost-time. Modified workdays were grouped together with off workdays because these workdays are defined as lost-time or disability days by the workers’ compensation system (i.e. events other than return to full work), and the focus of the research questions is on return to non-modified work for spine and upper limb strain and sprain injuries. Finally, the return to work and other employment earnings states were also grouped together as these both represent return to non-  115 modified work. The return to work outcome variable used for analysis was therefore defined as return to non-modified job duties after a period of registered lost-time associated with the claim.  For the second outcome measure of interest, recurrence was defined as recurrence of lost-time associated with the injury claim after a period of return to non-modified job duties. There are multiple ways to measure injury recurrence including recurrence of symptoms, medical care, or any sickness absence. Definitions of recurrence based on lost-time documented by workers’ compensation are the least sensitive.(209) Based on this, de Vet and colleagues (2002) recommend that a standardized definition of registered lost-time recurrence that includes no minimum length requirement be adapted.(210) This recommendation was followed for the definition of recurrence in the current study. However, a drawback of no minimum duration for a work disability recurrence event is that short duration recurrences may represent temporary compensated rest days that do not substantially contribute to higher claim costs or long term disability (due to their short duration), and thus are of less interest to workers’ compensation management and policy. To test if the findings were sensitive to the definition of the outcome measures, analyses were repeated for return to work events that only included seven consecutive calendar days (approximately five work days) of return to non-modified job duties, and only recurrences that included seven consecutive calendar days of lost-time (after a first return to work).  2. Primary explanatory variable Based on health care events in the year prior to injury, workers were classified as having anxiety only, depression only, co-morbid anxiety and depression, or none (no anxiety and no depression) in the year prior to injury. Like in the previous analyses, workers were considered to have anxiety or depression during the time period of interest (in this case the year before injury), if they had at least one index event meeting the corresponding case definition criteria (Table 2 and Table 3 – Chapter 3) during the time period.  3. Potential confounders The following variables were included as potential confounders: 1) sociodemographic: gender, age, income, dependents in the home, and residential location; 2) injury: body part, incident type, and secondary diagnosis on the claim; 3) clinical: somatic co-morbidity score, other mental co-  116 morbidity (not including anxiety or depression), and prior claims; and 4) work-related: firm size, shift type, and occupation. These variables are described in detail in section 4.2. 6.2.3 Analysis All analyses were stratified by gender and conducted using S.A.S. 9.4.(139) The baseline sociodemographic, injury, clinical, and work characteristics of the study sample were examined using frequencies and proportions.  The initial intention was to analyze multiple recurrence events, but this was not possible due to a low frequency of claims with two or more recurrences. Of the claims in the study sample, 2,431 (7.6%) of men and 2,627 (5.7%) of women had at least one recurrence event (Figure 8), and 272 (0.8%) of men and 279 (0.6%) of women had two or more recurrence events (data not pictured). To account for this, claims were censored after the second return to work event, or at the end of follow up (760 calendar days after the first lost-time day), whichever occurred first. A two-year follow up (760 days) was chosen based on availability of the data and prior findings that risk of a first lost-time recurrence continues beyond the first year after injury.(203) Due to censoring, only the first recurrence event and the first two return to work events (first initial return to work after the first lost-time period, and second return to work after the first recurrence event) were considered for the outcome variables. Lastly, as mentioned earlier, a sensitivity analysis was conducted to examine if the findings were sensitive to the definition of the return to work and recurrence outcome variables; in particular the threshold of time required for entry to a new state (at least one calendar day in the primary analysis and at least 7 calendar days in the sensitivity analysis).      117  Figure 8: Number of claims according to the number of return to work and recurrence events for men (left) and women (right), for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013 (% = percent of total claims in the study sample or percent of the claims at risk of the specific return to work event) To examine the associations of pre-existing anxiety and depression with the probability of return to work (including both first and second return to work events), a variation of the Cox proportional hazards model, adapted for analyses of ordered repeating event data known as the Prentice-Williams-Peterson (PWP) model was used.(211) The PWP accounts for the correlated nature of repeated events by stratifying the data based on the prior number of events during the follow up period, and then strata specific and common (pooled across strata) effect estimates are produced. While the PWP has not been widely used for the analyses of return to work data, its application to this area is supported by findings from Navarro et al. (2009), who compared statistical approaches to the analyses of recurrent work-related sickness absence.(212) Recommendations from Navarro et al. (2009) included the use of: 1) instantaneous models such as extensions of the Cox model like the PWP to analyze recurrent work-related sickness absence over count based models such as the Poisson model, as the former takes the timing of events into account while the latter does not; and 2) models with event specific baseline hazards like the PWP when event probability is dependent on the number of prior events. For analysis of the return to work outcome variable, all claims were included in the first stratum but only those with a first return to work and a first recurrence (i.e. at risk for a second return to   118 work event) were included in the second stratum. Discontinuous risk intervals were used such that a claim was only at risk of return to work while on lost-time (i.e. a claim was not at risk of return to work while at work and performing non-modified job duties). There are two versions of the PWP model and both were conducted here as they measure different aspects of time. The PWP Total Time (PWP-TT) model uses time since entry to the study as the time frame to construct the risk sets, while the PWP Gap Time (PWP-GT) model uses time since the previous event (i.e. time spent at risk of the current event) as the time frame to construct the risk sets. For the current study, study entry was defined as the first lost-time day following injury. For the recurrence outcome, as only the first event was considered, there was no need to account for the correlated nature of repeated events and traditional Cox proportional hazard models were used. Like the models described above, the models for the recurrence outcome had discontinuous risk intervals such that claims were only at risk of recurrence while on non-modified job duties and not during lost-time. Also like the models above, both gap time and total time models were conducted.  For all models, the proportional hazards assumption was tested by examining interactions between the primary explanatory variable (anxiety and depression) with the log of the time variable. Unadjusted and adjusted hazard ratios (HR) and 95% confidence intervals (CI) were reported to describe the associations of anxiety and depression with the outcomes. For the adjusted analyses, all potential confounders listed above were accounted for. 6.3 Results 6.3.1 Distribution of the study sample The baseline characteristics of the study sample were similar to those described in the previous chapter, and thus are not presented here (Table 5).  6.3.2 Description of return to work and recurrence events The majority of workers had a first return to work event (83.6% of men and 87.7% of women), and of those with a first return to work event, 6.9% of men and 8.6% of women had a lost-time recurrence event (Figure 8). Among workers with a lost-time recurrence event, the median time to first recurrence was approximately two weeks following the first lost-time day (17 days for men and 14 days for women), and the median duration of the first recurrence phase was   119 approximately six weeks for men (41 days) and five weeks for women (34 days) (Table 17). Of the workers who had a first recurrence event, the majority returned to work (i.e. 2nd return to work event; 78.6% of men and 85.6% of women). The rate of return to work following the first recurrence (i.e. second return to work) was slightly lower than the rate of return to work following the first lost-time period (i.e. first return to work) (Figure 8). Table 17: Frequency, timing and duration of first recurrence events for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013  Frequency of first recurrence Timing onseta  Median days (IQR) Durationb Median days (IQR) Men 2,627  17 (7, 47) 41 (7, 163) Women 2,431 14 (7, 31) 34 (7, 94) a Timing of the first recurrence event (days) using the first lost-time day as day 0 b Duration of the first recurrence events (days), starting at the first day of lost-time recurrence (after initial return to work) and ending at work re-entry (2nd return to work) 6.3.3 Cox regression and PWP models Overall, findings from the gap time and total time models were similar (Table 18 and Table 19) and for this reason, results from the total time models are emphasized in the following section. 1. Return to work  1 a. Men’s results For men, there was no indication that the effects of anxiety or depression on return to work varied across strata of the return to work event variable and only the common hazard ratios (pooled across strata) are described here. In the unadjusted total time model, men with anxiety only (cHR=0.87, 95% CI: 0.82 to 0.91), depression only (cHR=0.92, 95% CI=0.87 to 0.98), and comorbid anxiety and depression (cHR=0.92 , 95% CI=0.88 to 0.95) were less likely to return to work than men with no anxiety and no depression (Table 18). The 95% confidence intervals for the men’s effect estimates did not cross ‘1’, providing further support for this finding. Results from the total time adjusted model were similar (cHRanxiety only=0.90, 95% CI: 0.85 to 0.94; cHRdepression only=0.95, 95% CI= 0.89 to 1.00; cHRanxiety and depression=0.96, 95% CI: 0.92 to 0.99) (Table 19).  1 b. Women’s results In the women’s unadjusted total time model, there was evidence of effect modification by the prior events variable as the HRs for the second return to work event were larger than the   120 corresponding HRs for the first return to work event, with no overlapping 95% confidence intervals (Table 18). However, this effect was less consistent in the unadjusted and adjusted women’s gap time models, and was no longer present in the total time model after adjustment for confounders (Table 18 and Table 19). This suggests that effect modification by the prior events variable in the women’s unadjusted total time model, may have been due to time dependent confounding that is present when time since injury is used as the time scale.   In the unadjusted total time model, women with anxiety only (cHR=0.96, 95% CI: 0.92 to 1.00), and comorbid anxiety and depression (cHR=0.94 , 95% CI=0.92 to 0.97) were less likely to return to work than women with no anxiety and no depression, while no effect was observed for women with depression only (cHR=0.99, 95% CI=0.94 to 1.04) (Table 18). After adjustment, the effect estimates were all closer to ‘1’ and all of the 95% confidence intervals either included or crossed ‘1’ suggesting no strong relationship (cHRanxiety only=0.98, 95% CI: 0.94 to 1.02; cHRdepression only=1.00, 95% CI= 0.95 to 1.05; cHRanxiety and depression=0.97, 95% CI: 0.94 to 1.00) (Table 19).  2. Recurrence 2 a. Men’s results In the men’s unadjusted total time model, anxiety only (HR=1.23, 95% CI: 1.02 to 1.48) and comorbid anxiety and depression (HR=1.42, 95% CI: 1.26 to 1.61) were associated with a higher probability of recurrence, but depression only was not (HR=1.03, 95% CI: 0.83 to 1.29) (Table 18). In the men’s adjusted total time model, similar but attenuated effects were observed (HRanxiety only: 1.17, 95% CI: 0.97 to 1.41; HRdepression only: 1.01, 95% CI: 0.81 to 1.27; HRanxiety and depression=1.28, 95% CI: 1.12 to 1.46) (Table 19). 2 b. Women’s results In the women’s unadjusted total time model, anxiety (HR=1.31, 95% CI: 1.15 to 1.50) and comorbid anxiety and depression (HR=1.23, 95% CI: 1.11 to 1.36) were associated with higher probability of recurrence, and while depression only was also associated with a higher probability or recurrence, the effect size was smaller and the 95% CI was wide and included ‘1’ indicating a lower precision of the estimate (HR: 1.08, 95% CI: 0.91 to 1.29) (Table 18). In the women’s adjusted total time model, similar but attenuated effects were observed for anxiety only   121 (HRanxiety only: 1.24, 95% CI: 1.08 to 1.42) and comorbid anxiety and depression (HRanxiety and depression=1.16, 95% CI: 1.04 to 1.29) but the effect estimate for depression was very close to 1 indicating no association (HRdepression only: 1.02, 95% CI: 0.86 to 1.22). 3. Gender differences In descriptive analyses, women were more likely to return to work than men but they were also more likely to have a recurrence (Figure 8). Among women, 87.7% had a first return to work event, and of those that experienced a recurrence of work disability following their first return to work, 85.6% had a second return to work event. For men, these values were 83.6% and 78.6% respectively. Among women that had a first return to work event, 8.6% had a recurrence, whereas among men, 6.8% had a recurrence. For the relationship between anxiety alone and return to work, the men’s and women’s adjusted effect estimates and 95% confidence intervals suggest that either, 1) there is an association for men but not women, or 2) anxiety alone is associated with lower likelihood for both men and women, but the strength of this detrimental relationship is greater for men. The most notable gender difference in the analysis was that depression was non-significantly associated with lower likelihood of return to work for men (based on the direction of effect), and not associated with return to work for women.  4. Sensitivity Analysis The results from the sensitivity analysis investigating the robustness of the outcome definitions for both return to work and recurrence were similar to the primary analysis presented above, and for this reason, they are presented in the Appendix only (Table 30 and Table 31, p. 193).  122 Table 18:Unadjusted estimates of the association between anxiety and depression in the year before injury and return to work and recurrence events lasting 1 day or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013  Men  Women  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep Return to work PWP TT             1st RTW HR  0.86 (0.82-0.91) 0.93 (0.88-0.98) 0.92 (0.89-0.96)  0.93 (0.90-0.97) 0.95 (0.90-1.00) 0.91 (0.89-0.94)      2nd RTW HR 0.91 (0.74-1.12) 0.79 (0.61-1.03) 0.89 (0.77-1.03)  1.52 (1.33-1.73) 1.81 (1.52-2.15) 1.56 (1.41-1.72)      Common HR 0.87 (0.82-0.91) 0.92 (0.87-0.98) 0.92 (0.88-0.95)  0.96 (0.92-1.00) 0.99 (0.94-1.04) 0.94 (0.92-0.97) PWP GT             1st RTW HR  0.86 (0.82-0.91) 0.93 (0.88-0.98) 0.92 (0.89-0.96)  0.96 (0.92-1.00) 0.97 (0.93-1.02) 0.94 (0.91-0.97)      2nd RTW HR 0.94 (0.76-1.15) 0.81 (0.63-1.05) 0.90 (0.78-1.04)  1.05 (0.91-1.21) 1.21 (1.00-1.45) 1.04 (0.93-1.16)      Common HR 0.97 (0.83-0.91) 0.92 (0.87-0.98) 0.92 (0.89-0.95)  0.96 (0.92-1.00) 0.99 (0.94-1.04) 0.94 (0.91-0.97) Recurrence Cox TT HR 1.23 (1.02-1.48) 1.03 (0.83-1.29) 1.42 (1.25-1.61)  1.31 (1.15-1.50) 1.08 (0.91-1.29) 1.23 (1.11-1.36) Cox GT HR 1.30 (1.08-1.56) 1.10 (0.88-1.38) 1.42 (1.25-1.62)  1.29 (1.13-1.47) 1.06 (0.89-1.26) 1.21 (1.09-1.34)  Table 19: Adjusteda estimates of the association between anxiety and depression in the year before injury and return to work and recurrence events lasting 1 day or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013  Men Women  Anx only Dep only Anx & dep Anx only Dep only Anx & dep Return to work PWP TT            1st RTW HR  0.89 (0.85-0.94) 0.96 (0.90-1.02) 0.95 (0.92-0.99) 0.97 (0.93-1.02) 0.98 (0.95-1.04) 0.96 (0.93-0.99)      2nd RTW HR 0.91 (0.74-1.12) 0.79 (0.61-1.03) 0.92 (0.79-1.07) 1.04 (0.90-1.20) 1.04 (0.89-1.20) 1.07 (0.95-1.20)      Common HR 0.90 (0.85-0.94) 0.95 (0.89-1.00) 0.96 (0.92-0.99) 0.98 (0.94-1.02) 1.00 (0.95-1.05) 0.97 (0.94-1.00) PWP GT            1st RTW HR  0.89 (0.85-0.94) 0.96 (0.90-1.02) 0.95 (0.92-0.99) 0.97 (0.93-1.02) 0.98 (0.94-1.04) 0.96 (0.93-0.99)      2nd RTW HR 0.96 (0.78-1.19) 0.84 (0.65-1.10) 0.96 (0.83-1.12) 1.06 (0.92-1.23) 1.25 (1.04-1.50) 1.05 (0.94-1.18)      Common HR 0.90 (0.86-0.95) 0.95 (0.90-1.01) 0.96 (0.92-0.99) 0.98 (0.94-1.02) 1.00 (0.95-1.05) 0.97 (0.94-1.00) Recurrence Cox TT HR 1.17 (0.97-1.41) 1.01 (0.81-1.27) 1.28 (1.12-1.46) 1.24 (1.08-1.42) 1.02 (0.86-1.22) 1.16 (1.04-1.29) Cox GT HR 1.14 (0.95-1.28) 0.98 (0.78-1.23) 1.31 (1.14-1.50) 1.24 (1.09-1.42) 1.03 (0.86-1.23) 1.17 (1.05-1.30) a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, occupation TT=total time model; GT=gap time model   123 Table 20: Summary table depicting adjusteda associations of anxiety and depression in the year before injury with return to work and lost-time recurrence events lasting 1 day or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013  Men Women  Anx only Dep only Anx & dep Anx only Dep only Anx & dep Return to work PWP TT            1st RTW HR             2nd RTW HR            Common HR     1.00  PWP GT            1st RTW HR             2nd RTW HR            Common HR     1.00  Recurrence Cox TT HR       Cox GT HR        = greater likelihood of the event (i.e. adjusted HR>1) with 95% CI that does not cross one  = greater likelihood of the event (i.e. adjusted HR>1) with 95% CI that does cross one  = less likelihood of the event (i.e. adjusted HR<1) with 95% CI that does not cross one  = less likelihood of the event (i.e. adjusted HR<1) with 95% CI that does cross one a Adjusted for age group, income quintile, dependents, location, injured body part, incident type, secondary diagnosis on the claim, somatic co-morbidity index score, other mental co-morbidity (that is not anxiety or depression), prior claims, firm size, shift type, occupation TT=total time model; GT=gap time model   124 6.4 Discussion 6.4.1 Summary of main findings This study demonstrated that pre-existing anxiety and depression disorders impact both return to work and lost-time recurrence events. Among men with lost-time upper limb or spine strain or sprain work injury, anxiety, depression, and comorbid anxiety and depression, were associated with lower probability of return to non-modified job duties, although the 95% confidence interval for depression included '1’; and anxiety and comorbid anxiety and depression were associated with higher probability of lost-time recurrence, although the 95% confidence interval for anxiety included ‘1’. Among women, anxiety and comorbid anxiety and depression were associated with lower probability of return to non-modified job duties; although the 95% confidence intervals for both of these included ‘1’, and anxiety and comorbid anxiety and depression were associated with higher probability of lost-time recurrence.  6.4.2 Differences in the impacts of anxiety and depression disorders  Like in the previous chapter, the findings from the current chapter suggest that for both men and women, pre-existing anxiety alone and co-morbid anxiety and depression are larger drivers of poor return to work outcomes than pre-existing depression alone. A possible hypothesis for this finding is that symptoms that are more characteristic of anxiety than depression (e.g. fear or worry) may have a greater impact on return to work and lost-time recurrence than symptoms that are more characteristic of depression (e.g. sadness or low mood). One consideration is that return to work events may produce or exacerbate anxiety symptoms (e.g. fear or worry about the following: ability to perform work duties, exacerbation of physical injury symptoms such as pain or functional limitation, co-workers reactions or opinions, and re-exposure to the initial cause of physical injury upon return to the workplace). In contrast, return to work events may have some beneficial effects for depression through social interactions, sense of purpose or accomplishment related to one’s work duties, and increased daily activity. 6.4.3 Contributions to the literature The contributions of the current study to the literature on mental health and return to work after physical injury are outlined here. First, it adds to a body of evidence that poor mental health is associated with a more complicated return to work trajectory.(36,63,207) In particular, it adds to the literature by demonstrating that pre-existing anxiety and comorbid anxiety and depression   125 (but not depression alone) are early prognostic factors associated with increased risk of lost-time recurrence. To our knowledge, there has been little examination of the association between anxiety and lost-time recurrence, and prior examination of the association between depression and lost-time recurrence has provided limited evidence for depression’s prospective effects (or lack thereof as pre-existing depression was not associated with recurrence in this study). While it is likely that associations between anxiety and specific return to work events (like first or second return to work and lost-time recurrence) are bidirectional, evidence of a prospective association can inform interventions, particularly the timing of interventions intended to improve return to work for workers with a pre-existing disorder, and possibly also workers who develop new anxiety symptoms during the return to work process. Second, there is a mixed body of evidence regarding the impacts of anxiety on return to work after musculoskeletal injury, with several studies reporting no association.(176) The current findings indicate that use of a return to work outcome measure that adequately captures lost-time recurrence may be necessary to capture the full effects of anxiety on the overall return to work process. 6.4.4 Potential mechanisms It is hypothesized that the mechanisms through which pre-existing anxiety or comorbid anxiety and depression increase risk of lost-time recurrence are similar to the mechanisms through which anxiety and depression negatively impact return to non-modified work. As reviewed in the previous chapter, these include: 1) the amplification of pain, 2) activity disruption or loss of interest in activities that promote well-being, 3) resistance or non-adherence to therapeutic treatment of the musculoskeletal injury, 4) self-assessed inability to perform work tasks effectively, 5) difficulty navigating social interactions involved in the return to work process,(53,55–58) and 6) exacerbation of mental symptoms by procedural aspects of the workers’ compensation experience. Short-term use of compensated rest days (e.g. a single day off work registered with workers’ compensation) by workers with an anxiety disorder, or comorbid anxiety and depression, to temporarily recuperate, did not explain the finding that anxiety alone, and comorbid anxiety and depression were positively associated with risk of recurrence. In the sensitivity analysis, only recurrences lasting seven calendar days or longer were considered, and findings regarding the impacts of anxiety alone, and comorbid anxiety and depression, on the risk of recurrence did not change substantially.    126 6.4.5 Strengths and limitations A major strength of this study is detailed return to work data for each calendar day following injury that allowed for the analysis of ordered recurrent events with event specific estimates, while taking into account information on the timing of events. We chose to examine a measure of recurrence that was limited to lost-time associated with a workers’ compensation claim. This was done to provide evidence that can inform disability management in a workers’ compensation context. Based on findings by Young et al. (2013), use of workers’ compensation data to measure lost-time recurrence is concordant with self-reported registered and non-registered time off due to the injury.(204) However, it fails to capture other work disability outcomes. These include re-injury that results in a new workers’ compensation claim, decreased work performance while at non-modified job duties, indirectly related sickness absence (e.g. time off work due to exacerbation of comorbid conditions), directly related sickness absence that is registered as a sick day(s) via the workplace benefit plan rather than as a lost-time benefit via workers’ compensation, and instances where workers chose to terminate employment for reasons related to the work injury. Another limitation of the current study was the inability to collect data on changing levels of anxiety and depression symptoms throughout the study period, particularly in the control group (i.e. the none group). A study by Franche et al. (2009) that excluded workers with a pre-existing depression diagnosis, found that the prevalence of high self-reported depression symptoms at one and six months post injury for upper spine or upper limb strain or sprain was 43% and 27%.(36) This indicates that our findings likely underestimate the effects of depression on return to work and recurrence, and a similar underestimate could have occurred for anxiety.  6.4.6 Implications Current findings suggest that interventions intended to reduce the negative impacts of pre-existing anxiety and depression disorders on the overall return to work process after lost-time upper limb or spine strain or sprain work injury should begin in the initial lost-time phase to facilitate initial return to work and then continue once back at non-modified work to prevent recurrence. Such interventions could include ongoing social support from managers and co-workers, pain monitoring and management by the treating clinician, mental health treatment   127 including cognitive behavioral therapy, and therapeutic treatment for the physical injury (e.g. physiotherapy).  Another consideration to improve return to non-modified work for workers with a pre-existing anxiety disorder is use of graduated return to work in the form of modified job duties, or a graduated increase in the number of working hours. There is growing evidence that graduated return to work is effective for injured workers in general.(213,214) While its effectiveness among injured workers with pre-existing anxiety has not been examined, it could be particularly effective in this population as it may address some of the mechanisms linking anxiety to increased work disability. In particular, graduated return to work may decrease stress or worries about return to non-modified work, and increase perception that return to non-modified work is possible and sustainable. Likewise, inclusion of positive social interactions as a part of graduated return to work could lessen anxiety symptoms and in turn facilitate return to non-modified work. Recent findings from BC report low rates of modified return to work for lost-time musculoskeletal work injury, especially for workers in small firms, suggesting room for improvement in terms of supports and services to increase use of modified work.(196)  While several potential interventions are discussed here, evidence to support their use, especially for workers with anxiety and depression disorders in addition to lost-time upper limb or spine strain or sprain work injury is lacking. Future research should examine the effectiveness of these suggested interventions and the duration for which they should continue after return to non-modified work.     128 Chapter 7: Impacts of physician mental health services on return to work for workers with pre-existing anxiety (Research Objective 4) 7.1 Background In the general population, anxiety disorders are associated with increased risk of work disability including sickness absence from work (174), work impairment,(186) and unemployment(188). Pharmacotherapy and psychological counseling treatments have been demonstrated to reduce anxiety symptoms and improve role functioning in anxiety patients, although there is more evidence for the former than the latter.(117) In regards to work functioning, anxiety outpatients who were enrolled in a psychological counseling program demonstrated long-term improvements in work ability, functional capacity, and occupational role performance.(215,216) Research on the effects of pharmacotherapy and psychological counseling treatments on work disability outcomes for workers with anxiety is limited, and there is little to no examination of this topic among workers with musculoskeletal work injury who remain attached to the labour market but are at risk of long-term work disability. Pharmacotherapy and psychological counseling treatments known to stabilize anxiety symptoms (117) may better prepare workers with anxiety to cope with, and respond positively to, difficult life events or experiences, like musculoskeletal work injury. The low rates of anxiety treatment in Canada suggest that this is a potential policy area with room for improvement.(83–87)  The primary purpose of the current study was to explore associations between pharmacotherapy and psychological counseling treatments with return to work after lost-time upper limb or spine strain or sprain work injury for workers with a pre-existing anxiety disorder. As there is little research in this area, especially for injured worker populations, the current study utilized quasi-experimental epidemiological approaches and linked administrative databases from the Canadian province of BC to provide preliminary evidence for this important health and social issue. Findings from such analyses are intended to inform future intervention studies, and to provide evidence to inform policy and practice as the field evolves. Due to the unique study sample, the findings will be informative for clinicians who treat anxiety disorders in worker populations, especially workers with, or at risk of, lost-time upper limb or spine strain or sprain work injury; as well as workers’ compensation systems interested in addressing anxiety disorder co-morbidity   129 among workers with lost-time upper limb or spine strain or sprain work injury; and employers interested in supporting employees with anxiety disorders and helping these employees to decrease their risk of work disability. Anxiety disorders are the focus of the current chapter (as opposed to anxiety and depression disorders as in previous chapters) due to earlier evidence that the negative impacts of pre-existing anxiety on return to work are greater than that of pre-existing depression (Chapter 5: section 5.3.3 and Chapter 6: section 6.3.3). 7.1.1 Overview of methodological approach Use of administrative databases to examine healthcare practices and their corresponding outcomes is an increasingly common practice. Population-level research of this nature is complementary to smaller clinical studies that have limited external generalizability. Challenges of observational research using administrative databases include non-random treatment assignment and unmeasured confounding leading to a biased estimate of the relationship of interest. When treatment is not assigned randomly, confounding due to indication can lead to comparison groups that are imbalanced across levels of the confounding factors. For example, among patients with mental disorders, more severe mental symptoms are associated with increased treatment contact and intensity.(217) More severe mental symptoms may also negatively impact return to work outcomes (Figure 9). Thus, in an observational study of workers with physical injury and anxiety, failure to account for the potential confounding effects of anxiety severity could lead to a spurious or biased estimate of the association between mental health services and return to work. Bias due to measured confounders can be accounted for through analytic methods such as adjustment, matching, stratification, or propensity scores; however, such methods are not adequate to remove bias by unmeasured confounders.    130       Figure 9: Analytic framework for research question on the association between treatment for anxiety and return to work outcomes after musculoskeletal injury Instrumental variables (IVs) are one means of addressing bias due to unmeasured confounders when randomization is not feasible. They can be thought of as a tool that mimics random treatment allocation or as a natural or quasi-experiment.(218) Important assumptions of an instrumental variable analysis include that the instrument is: 1) associated with treatment, 2) associated with the outcome only through treatment (i.e. the only mechanism linking the instrumental variable to the outcome is the treatment), and 3) not associated with unmeasured confounders.  Physician treatment preferences are a common instrumental variable used to estimate treatment effects for health services when confounding due to indication and unmeasured confounders are an issue.(218) In this method, individual patient’s treatment is estimated based on the physician’s treatment of other patients. In order for this to be effective, adequate variation in treatment preferences across physicians must exist beyond those due to differences in physician’s patient populations.(218) In the area of mental health, instrumental variables including physician treatment preferences have effectively estimated the treatment effects of 1) guideline concordant Instrumental variables Treatment preferences of worker A’s general practitioner – based on treatment of other anxiety patients  Treatment variables Physician mental health services received by worker A  Outcome variable Time to return to work (paid disability days) for worker A Unmeasured confounder Worker A anxiety disorder severity Measured confounders Age  Gender Income  Occupation Somatic co-morbidity Firm size Location (urban/rural) Dependents  Depression Other mental co-morbidities Injury year   131 or evidenced-based care for depression on future clinical status and employment (219,220), 2) type of anti-depressant prescription (tricyclics versus selective serotonin reuptake inhibitors) on risk of hospital admission for self-harm or death by suicide (221), 3) psychosocial assessment of patients presenting to the emergency room for self-harm on risk of repeat self-harm (222), and 4) anti-depressant use on risk of hip fracture.(223)  In the BC public health care system, physicians are the de facto health service providers for mental disorders, as public funding for mental health services from other service providers is extremely limited. Physicians are also the only service providers of pharmacotherapy for mental disorders. Workers may seek non-pharmaceutical based mental health services (e.g. counseling) from other service providers (e.g. psychologist) outside of the public health system, but these are paid for out-of-pocket or through a workplace benefit plan. Despite alternative treatment options in the private health systems, publically funded physicians remain the most commonly sought out mental health treatment provider used by approximately 85% of anxiety patients who seek treatment in Canada.(224) Thus, physician services, are the predominant public policy lever in the BC public health system (as it exists currently) for ensuring adequate mental health care for working populations, especially those at risk of work disability. For this reason, and the availability of information on physician services in the administrative data, instrumental variables based on physician treatment preferences were developed to estimate the probability of pharmacotherapy and psychological counseling treatment for workers with a pre-existing anxiety disorder at the time of musculoskeletal work injury. 7.2 Methods 7.2.1 Study sample The study sample consisted of accepted lost-time claims from 2000 to 2013 for upper limb or spine strain or sprain work injury (as indicated by the primary diagnosis in the claim file), belonging to workers that were between 19 to 64 years of age at the time of injury, registered in the provincial health services plan for at least 275 days (9 months) in both the year before and the year after injury (n=294,913) (Figure 10). A longer study period was chosen for the current analysis compared to Chapters 5 and 6 to increase the sample size. Next, the study sample was restricted to workers with prevalent anxiety in the year prior to their injury using similar methods   132 as Puyat and colleagues (2017)(Figure 11).(128) Diagnoses of anxiety (ICD-9 code 300, 308, 309; and ICD-10 code F341, F4, and F68) in the year prior to injury were identified from the health claims data for workers with claims meeting the above criteria. Next, the earliest recorded anxiety diagnosis during the year prior to injury was used as an index date, and health claims data from prior to the index date was reviewed to identify other anxiety diagnoses. A criterion of at least two outpatient diagnosis or at least one inpatient diagnoses of anxiety in a 12-month period was used to increase the likelihood of a true case. If the earliest diagnosis did not meet this criterion, then the next earliest diagnosis was used as an index date and so on in a reiterative manner until the criterion was met or no diagnoses remained. A total of 276,535 (93.8%) claims were excluded for workers not meeting the anxiety criteria. For the 18,378 remaining claims, the general practitioner that provided the greatest number of services (up to one per day) to the worker during the year prior to injury was identified. In the event that two or more general practitioners were tied for providing the greatest number of services to a given worker, the general practitioner with a preference for a higher number of mental health related visits per anxiety patient was selected for the analyses. Claims assigned to a general practitioner for whom no other anxiety patients could be identified during the study period (2000 to 2013) were excluded (n=1,557, 8.5% excluded, n=16,821 remaining). Next, 379 claims were excluded for missing data (2.3% excluded, n=16,442 claims remaining). In the event that more than one claim belonging to a single worker met all of the above criteria, only the claim with the earliest injury date (within the timespan of the study period) was included (n=2,274, 13.8% excluded). The final study sample consisted of 14,168 claims.      133  Figure 10. Construction of the study sample consisting of workers with lost-time upper limb or spine strain or sprain work injury and a prevalent anxiety disorder in the year before injury in BC from 2000 to 2013   317,512 lost time claims met the age and injury criteria, 2000 to 2013 22,599 (7.1%) claims with insufficient medical registration period excluded 294,193 remaining 18,378 remaining 276,535 (93.8%) claims with no prevalent anxiety in the year before injury excluded 16,821 remaining 1,557 (8.5%) claims assigned to general practitioner with no other anxiety patients in the research extract excluded 379 (2.3%) claims with missing data excluded 16,442 remaining Final study sample 14,168 2,274 (13.8%) claims for workers with earlier occurring claims already selected into the study sample excluded   134   Figure 11: Timeline for identifying prevalent anxiety in the year prior to injury and receipt of mental health services among workers with lost-time upper limb or spine strain or sprain work injury in BC, Canada A. Time frame used to capture anxiety index diagnosis (one year before injury) B. Time frame to capture secondary anxiety diagnoses to determine if the case definition for prevalent anxiety was met (one year before index diagnosis) C. Time frame used to capture mental health services from a physician for workers identified as having prevalent anxiety in the year before injury (one year after index diagnosis) *Defined as the earliest occurring anxiety diagnosis in the year prior to injury 7.2.2 Study variables The selection of the study variables was informed by the analytic framework (Figure 9). 1. Outcome variable As the detailed return to work data was not available prior to 2009, sustained returned to non-modified work was not a feasible outcome measure for the study period 2000 to 2013. Given this, the return to work outcome variable for the current study was defined as the cumulative number of disability days paid that were associated with the claim in the first two years following injury. Disability days paid includes both lost-time days and modified workdays associated with the injury on the claim, where wage-loss benefits were paid by workers’ compensation to the worker. 2. Minimal adequate treatment, pharmacotherapy, and psychological counseling (treatment variables) For each worker in the study sample, anxiety treatments received in 12 months following the index diagnoses were identified and three binary treatment variables were constructed: 1) receipt of minimal adequate treatment (yes/no), 2) receipt of pharmacotherapy (at least one anti-Injury 0 -365  A Time in days C -305  Index diagnosis* B  +60 -670   135 depressant or anxiolytic dispensing event - yes/no), and 3) receipt of counseling (at least one counseling event with a physician – yes/no). Based on clinical practice guidelines for anxiety, and studies on treatment efficacy and adequacy for common mental disorders, minimal adequate treatment was defined as 1) at least one anxiolytic or anti-depressant dispensing event and at least two mental health related physician visits (up to one per day) to allow for patient follow up and monitoring; or 2) at least four psychological counseling events with a physician (up to one per day) in the 12 months following the index date diagnosis. (83,87,111,117) For psychological counseling, preliminary descriptive analyses (Table 27, Table 28, and Table 29 in the Appendix, p. 191 to 192), and other research on physician treatment of common mental disorders in BC, suggests that four sessions is a reasonable cut point for typical treatment observed in administrative health data, despite it being less than the 12 recommended sessions in treatment guidelines.(111) Further, during the study period, up to four psychological counseling sessions per patient (worker) per year was easily billable by physicians under the public health insurance plan whereas coverage for a greater number of sessions was limited and uncommon. 3. Physician treatment preferences (instrumental variables) Instrumental variables were developed to account for confounding of the relationships between the treatment variables and the outcome (return to work) by severity of anxiety symptoms. General practitioners are responsible for the majority of mental health primary care in BC, especially for common disorders like anxiety and depression. Psychiatrists also provide mental health care in this province, however outpatient access to a psychiatrist is dependent on general practitioner referral. Thus, access to any outpatient primary mental health care service (from a psychiatrist or from a general practitioner) is dependent on general practitioner treatment practices. For this reason, both services from the assigned general practitioner and those from psychiatrists for the same patient were included in the measures of general practitioner treatment preference. To assess how the decision to include mental health services from a psychiatrist might have affected the analysis, a sensitivity analysis was conducted where anxiety patients who received mental health services from a psychiatrist in the year following their index diagnosis were excluded from the calculation of physician treatment preferences (the instrumental variables). Workers with anxiety who were treated by a psychiatrist in the year   136 following their index diagnosis were also excluded from the study sample used for this sensitivity analysis. For each general practitioner assigned to a claim in the study sample, other anxiety patients seen by that practitioner during the study period (2000 to 2013) were identified from the Medical Services Plan file within the research extract, and the same criterion for identifying prevalent anxiety was used as described during construction of the study sample.(128) This file includes historical physician billing data and diagnoses (back to 1991) for all BC workers who had a workers’ compensation claim for upper body work injury between 2000 and 2013 (n=415,436). Thus, in addition to workers from the study sample, BC residents with a history of work-injury from outside the study sample were used to determine general practitioner treatment preferences.  Next, using the larger anxiety patient sample, measures of general practitioner treatment preference were calculated for each claim on an individual basis. This was done to ensure that the worker from the claim of interest in the main analyses was not included in the instrumental variable measure. General practitioner treatment preferences were calculated as the proportion (continuous variables) of anxiety patients for a given general practitioner that received: 1) minimal adequate treatment (as defined above), 2) anxiolytics or anti-depressants (at least one dispensing event in the 12 months following the index diagnosis), and 3) counseling (at least one event in the 12 months following the index diagnosis). 4. Potential confounders Potential confounders of the relationships between the treatment variables and return to work were defined as in the previous chapters, but were restricted to those relevant to the current research question and the current analytic framework (Figure 9). Potential confounders for the current research question included age, gender, income, occupation, somatic co-morbidity index score, firm size, location (urban/rural), dependents under age 18, co-morbid depression, other mental co-morbidities (other than anxiety or depression), and injury year. Due to low cell size (i.e. less than 40 men or less than 40 women in a given level of the variable) and for model simplicity, the occupation variable was reduced from 10 levels to 6. Workers that received a diagnosis of depression (ICD-9 codes 311 or 296 or ICD-10 code F3 [excluding F341]) from a physician in the 24 months preceding their injury were considered to have comorbid depression.   137 Injury year was also added as a confounder to account for policy changes in coverage for mental health services under the medical services plan. 7.2.3 Analysis 1. Assessment of the relationship between anxiety treatment and return to work  1 a. Instrumental variable analysis For the first stage of the instrumental variable analysis, binary (yes/no) measures of treatment (minimal adequate treatment, pharmacotherapy, and counseling) were each individually regressed on the corresponding instrumental variables (physician treatment preference) and measured potential confounders (included as independent variables) using logistic regression models with individual claims as the unit of analysis. Fitted values were retained for the second stage of analyses and used as the instrumental variables. For the second stage, associations between the instrumental variables and the natural log of disability days were estimated using linear regression models adjusted for all measured potential confounders. The outcome (disability days) variable was log transformed to achieve normality. Co-efficient estimates and their 95% confidence intervals, as well as the implied percent effect were reported for the instrumental variables in the second stage models.  The implied percent effect is the percent change in the dependent variable associated with a one-unit change in X. It can be approximated from the corresponding co-efficient from the log-linear model, however, direct approximations will be biased downward, and the extent of this downward bias increases with increasing values of the co-efficient.(225) To reduce this downward bias, the following adjustment recommended by Palmer (2011) was made: percent implied effect = exp (?̂̂?) − 1 with  𝛽 ̂ representing the estimate of the co-efficient.(225) The instrumental variable analysis was conducted using the ivreg command and the sandwich estimate for variance from the AER package in the statistical package R.(226)  1 b. Conventional regression analysis Observational associations between the actual treatments and the log of the disability days were also calculated using unadjusted and adjusted linear regression models (i.e. conventional regression analysis). While this is not a necessary part of an instrumental variable analysis, it was conducted for comparison reasons and to assess the degree of confounding in a conventional regression analysis due to unmeasured variables.    138       2. Assessment of instrumental variable assumptions It is not possible to directly test the assumption that the instrumental variables have no unaccounted association with unmeasured confounders. However, when measured confounders are balanced across levels of the instrumental variable, the assumption of balance is more likely to carry over to unmeasured confounders.(227) To identify potentially imbalanced measured confounders, the prevalence of measured confounders was examined across levels of actual treatments, and the prevalence difference (prevalence treated – prevalence untreated) was calculated. Next, to assess if the instrumental variables improved confounder balance, the prevalence of measured confounders was examined across the lower and upper quartiles of the instrumental variable (data not shown), and the prevalence difference (prevalence treated – prevalence untreated) was calculated. A lower prevalence difference across quartiles of the instrumental variable compared to levels of actual treatment is indicative of improved confounder imbalance.  To descriptively assess the assumption that the instrumental variables were associated with treatment, the frequency of actual treatment was examined across quartiles of the corresponding instrumental variable. Next, the coefficient and 95% confidence interval for each instrumental variable in the first stage of the instrumental variable analysis was examined and the partial F and partial R2 values were reported.(228) The partial F-test tests the null hypothesis that the co-efficient for the instrumental variable is 0. As a rule of thumb, a partial F-value of less than 10 indicates a weak instrumental variable.(229) The partial R2 value describes the amount of variance in the model explained by the instrumental variable. Higher R2 values are more desirable, but small R2 values have also been reported for good instrumental variables.(227) All analyses were stratified by gender based on earlier findings that the impacts of anxiety on time to sustained return to non-modified work are greater for men than for women. Analyses were conducted using S.A.S. 9.4 (in addition to the ivreg package in R as mentioned earlier).(139)   139 7.3 Results The full study sample consisted of 5,810 men and 8,358 women. There were 3,158 unique general practitioners assigned to at least one or more claims (n=14,168), with a mean number of 10 (standard deviation 9.36) anxiety patients (other than the worker from the claim of interest) per general practitioner. The overall proportion of claims that received minimal adequate treatment (i.e. at least one pharmaceutical event and two mental health visits or at least four counseling events), at least one counseling event, and at least one prescription dispensing event in the year following the index diagnosis was 56.6%, 57.5%, and 63.7% for men, and 59.0%, 61.1%, and 67.0% for women respectively. The study sample restricted to workers who did not receive services from a psychiatrist consisted of 4,922 men and 7,542 women.  The characteristics of the full study sample by treatment status and gender are given in Table 21. For both men and women, the mean number of disability days was greater among the treatment groups than the non-treatment groups (Table 21). Among men, minimal adequate treatment, counseling, and prescription services were associated with an excess of approximately 6 (54.3 days versus 47.9 days), 5 (53.9 days versus 48.2 days), and 10 (55.1 days versus 45.1 days) days respectively. Among women, these same values were 2 (53.5 days versus 51.9 days), 3 (54.1 days versus 50.8 days), and 5 days (54.4 days versus 49.5 days).   7.3.1 Assessment of independence of the instrumental variables with unmeasured confounders Study variables were well balanced across levels of the treatment groups with the exception of the other mental diagnosis (anxiety and depression excluded) and depression diagnosis variables (Table 21). Imbalance of these two aforementioned variables was less pronounced when examined across the highest and lowest quartiles of the instrumental variables (Table 22). This indicates that the instrumental variables improved balance of the other mental diagnoses and depression diagnoses variables. However, some imbalance remained for the depression variable across quartiles of the instrumental variables for minimally adequate treatment (prevalence differences: men=4.7 and women=4.5) and prescription services (prevalence differences: men=6.1 and women=7.6) (Table 22).    140 Table 21: Distribution of study variables across levels of actual treatment for the full study sample of workers with pre-existing anxiety in addition to lost-time upper limb and spine strain or sprain in BC from 2000 to 2013  Men Women  Minimally adequate treatment Counselling Prescription Minimally adequate treatment Counselling Prescription  No  Yes No Yes No  Yes No Yes No  Yes No Yes Disability days                  Mean 47.9 54.3 48.2 53.9 45.1 55.1 51.9 53.5 50.8 54.1 49.5 54.4      Median 15.0 16.0 13.0 17.0 14.0 17.0 20.0 21.0 19.0 21.0 18.0 21.0 Socio-demographic  Percent Distribution (%) Age group (years)                  19 to 24 9.4 7.3 8.2 8.2 9.7 7.4 7.1 6.9 6.1 7.6 7.3 6.8      25 to 29 10.9 11.2 11.0 11.1 11.3 11.0 9.0 9.4 8.9 9.5 9.9 8.9      30 to 39 27.6 29.1 28.2 28.6 27.8 28.8 25.1 25.3 23.9 26.0 25.5 25.0      40 to 49 30.4 31.7 31.2 31.1 30.7 31.4 33.1 34.1 35.1 32.8 32.1 34.4      50 to 59 17.9 18.0 18.1 17.8 17.5 18.2 22.6 21.9 23.1 21.6 22.2 22.2      60 to 64 3.7 2.8 3.3 3.1 3.1 3.2 3.1 2.5 3.0 2.5 3.0 2.6 Income quintile                  1: lowest 15.5 17.4 15.6 17.3 15.8 17.0 28.2 33.2 29.3 32.3 28.3 32.6      2 16.9 18.0 17.8 17.3 17.3 17.7 23.2 23.3 24.2 22.7 22.7 23.5      3 17.8 19.0 18.2 18.7 18.6 18.4 24.6 22.6 24.4 22.9 24.2 23.1      4 25.5 22.7 24.5 23.4 24.5 23.5 14.7 11.6 13.7 12.4 14.9 11.9      5: highest 24.4 22.9 23.9 23.3 23.7 23.4 9.2 9.3 8.4 9.7 9.8 8.9 Dependents                  0 62.8 67.3 64.2 66.2 63.0 66.7 53.4 60.4 56.6 58.1 52.5 60.0      1 or more 37.2 32.7 35.8 33.8 37.0 33.3 46.6 39.6 43.4 41.9 47.5 40.0 Location                  Urban 7.7 7.6 7.8 7.6 7.0 8.1 9.2 9.7 9.2 9.6 8.6 9.9      Rural 92.3 92.4 92.2 92.4 93.0 91.9 90.8 90.3 90.8 90.4 91.4 90.1 Clinical             Somatic co-morbidity 0 4.6 3.7 4.7 3.7 4.5 3.8 1.7 1.4 2.1 1.1 1.7      0 0 12.5 10.5 12.4 10.6 12.8 10.6 7.3 5.8 7.3 5.9 7.7      1 0 18.9 17.1 18.6 17.4 19.3 17.1 14.4 12.2 14.1 12.5 14.6      2 0 22.3 21.1 20.8 22.2 22.4 21.2 19.1 17.6 18.8 17.8 19.4      3 0 17.1 17.9 17.4 17.7 16.7 18.1 19.3 18.6 19.3 18.6 18.7      4 0 24.6 29.6 26.1 28.4 24.2 29.3 38.2 44.4 38.3 44.1 37.8      5 or more 0 4.6 3.7 4.7 3.7 4.5 3.8 1.7 1.4 2.1 1.1 1.7 Mental co-morbiditya                  0 89.1 76.3 87.4 77.8 86.3 79.4 95.2 85.4 93.2 87.0 94.0 87.1      1 or more 10.9 23.7 12.6 22.2 13.7 20.6 4.8 14.6 6.8 13.0 6.0 12.9 Depression diagnosis                  No 74.3 37.3 63.4 46.0 72.6 42.4 74.4 35.6 61.0 45.4 74.1 40.3      Yes 25.7 62.7 36.6 54.0 27.4 57.6 25.6 64.4 39.0 54.6 25.9 59.7      141  Men Women  Minimally adequate treatment Counselling Prescription Minimally adequate treatment Counselling Prescription  No  Yes No Yes No  Yes No Yes No  Yes No Yes Work  Percent Distribution (%) Firm size                  30 or less 28.8 28.5 29.6 27.9 27.3 29.4 15.2 17.0 17.0 15.8 14.4 17.2      31 to 150 25.2 26.4 25.7 26.1 24.9 26.5 19.9 21.0 20.8 20.4 19.8 20.9      151 to 1000 25.1 23.3 24.1 24.0 25.2 23.4 21.8 20.4 20.5 21.2 22.5 20.2      1001 to 10000 18.2 19.1 18.1 19.1 19.4 18.3 26.5 24.7 25.0 25.7 26.2 25.1      10,001 or more 2.7 2.7 2.6 2.9 3.2 2.5 16.7 16.8 16.7 16.8 17.0 16.6 Occupation                  Management/admin 6.4 6.1 6.5 6.0 6.7 5.9 9.4 9.4 9.5 9.4 9.3 9.5      Health 4.9 5.0 5.5 4.6 4.9 5.0 30.7 29.0 29.7 29.7 31.0 29.0      Industry/utilities 15.9 14.8 15.7 15.1 15.7 15.1 5.0 4.8 5.8 4.4 4.9 4.9      Culture/services 16.6 18.7 16.9 18.4 18.0 17.6 40.9 40.8 40.7 40.9 41.0 40.8      Science/government 3.2 4.3 3.4 4.2 3.7 3.9 8.5 9.7 8.3 9.8 8.6 9.5      Trades/transport 53.0 51.1 52.1 51.7 50.9 52.5 5.5 6.3 6.2 5.8 5.2 6.3 Injury year                  2000 7.2 6.7 5.7 7.8 7.6 6.5 5.9 6.5 4.9 7.1 6.1 6.3      2001 8.0 7.6 6.8 8.5 8.2 7.5 6.9 6.4 5.5 7.4 7.3 6.3      2002 7.1 6.4 6.0 7.2 7.0 6.5 6.5 5.6 5.3 6.4 6.1 5.9      2003 6.3 6.6 6.0 6.9 6.3 6.6 6.7 6.3 6.1 6.7 6.6 6.4      2004 7.1 6.8 7.1 6.8 7.1 6.9 6.4 6.1 6.2 6.3 6.6 6.1      2005 8.0 7.0 8.4 6.7 8.0 7.1 7.4 6.7 7.6 6.6 7.4 6.8      2006 7.3 7.3 8.2 6.6 7.4 7.2 7.2 7.7 8.0 7.2 7.5 7.5      2007 8.8 7.8 8.7 7.8 8.5 8.1 7.4 8.0 8.3 7.4 7.3 8.0      2008 8.7 7.8 9.4 7.4 8.4 8.1 8.7 8.5 9.7 7.8 8.6 8.5      2009 5.9 6.5 6.2 6.2 5.5 6.6 6.7 7.0 6.9 6.9 7.1 6.8      2010 5.9 6.5 6.1 6.3 6.3 6.2 7.2 7.5 7.6 7.2 6.9 7.6      2011 6.1 7.4 6.9 6.8 6.0 7.3 7.1 7.3 7.6 7.0 6.9 7.4      2012 6.7 7.2 7.0 6.9 6.4 7.3 8.3 8.3 8.7 8.0 7.7 8.6      2013 6.9 8.5 7.5 8.0 7.2 8.1 7.6 8.1 7.7 8.0 8.0 7.8 a Mental diagnosis other than depression or anxiety *Variables with poor balance are marked in bold    142 Table 22: Comparison of the prevalence differencea across levels of actual treatment to the prevalence difference across the highest and lowest quartiles of the instrumental variables (IV)  Men Women  Minimally adequate treatment Counselling Prescription Minimally adequate treatment Counselling Prescription  Actual  IV Actual  IV Actual  IV Actual  IV Actual  IV Actual  IV Socio-demographic              Age group (years)                  19 to 24 -2.2 1.1 0.0 0.9 -2.3 3.1 -0.1 2.1 1.5 1.3 -0.5 2.2      25 to 29 0.2 -0.3 0.1 0.1 -0.3 0.6 0.4 0.3 0.6 1.0 -1.0 0.7      30 to 39 1.5 3.2 0.4 -2.2 1.1 3.4 0.2 -1.8 2.1 -1.7 -0.5 1.0      40 to 49 1.2 -1.0 -0.1 -1.5 0.7 -3.1 1.0 -2.3 -2.3 -2.1 2.4 -1.0      50 to 59 0.2 -2.5 -0.3 2.0 0.7 -2.6 -0.7 1.6 -1.5 0.9 0.0 -1.6      60 to 64 -1.0 -0.5 -0.2 0.8 0.1 -1.3 -0.7 0.1 -0.5 0.6 -0.4 -1.2 Income quintile                  1: lowest 1.9 -1.3 1.7 -0.6 1.2 2.3 5.0 0.6 3.0 -2.8 4.3 3.4      2 1.2 2.4 -0.4 -1.2 0.4 0.4 0.0 0.6 -1.4 1.2 0.8 1.2      3 1.2 -2.7 0.5 -2.1 -0.2 -2.2 -2.0 -0.1 -1.5 0.2 -1.1 -1.5      4 -2.8 -1.1 -1.1 2.2 -1.0 -1.2 -3.1 -0.5 -1.3 -0.1 -3.1 -1.5      5: highest -1.6 2.7 -0.7 1.7 -0.3 0.7 0.1 -0.6 1.3 1.5 -0.9 -1.6 Dependents                  0 4.6 4.5 2.0 2.8 3.7 5.3 7.0 3.2 1.5 1.9 7.4 3.0      1 or more -4.6 -4.5 -2.0 -2.8 -3.7 -5.3 -7.0 -3.2 -1.5 -1.9 -7.4 -3.0 Location                  Urban -0.1 3.3 -0.2 2.6 1.1 3.9 0.5 1.0 0.4 1.8 1.3 2.9      Rural 0.1 -3.3 0.2 -2.6 -1.1 -3.9 -0.5 -1.0 -0.4 -1.8 -1.3 -2.9 Clinical             Somatic co-morbidity                  0 -0.9 0.7 -1.0 0.2 -0.7 0.6 -0.3 -0.5 -1.0 -0.8 -0.3 0.3      1 -2.0 0.6 -1.8 -0.7 -2.2 1.2 -1.4 0.5 -1.4 0.4 -1.9 0.7      2 -1.8 1.6 -1.2 1.6 -2.2 0.7 -2.2 0.8 -1.7 0.8 -2.3 0.4      3 -1.2 0.6 1.4 2.7 -1.2 0.7 -1.5 -1.3 -0.9 0.9 -1.8 -1.4      4 0.8 0.2 0.3 1.7 1.3 -0.7 -0.7 0.3 -0.7 1.8 0.3 -1.0      5 or more 5.1 -3.6 2.3 -5.6 5.1 -2.6 6.2 0.3 5.7 -3.1 6.0 1.0 .    143  Men Women  Minimally adequate treatment Counselling Prescription Minimally adequate treatment Counselling Prescription  Actual  IV Actual  IV Actual  IV Actual  IV Actual  IV Actual  IV Clinical             Mental co-morbiditya                  0 -12.7 -3.0 -9.6 -0.5 -6.9 -3.8 -9.8 -0.6 -6.3 1.2 -6.9 -1.4      1 or more 12.7 3.0 9.6 0.5 6.9 3.8 9.8 0.6 6.3 -1.2 6.9 1.4 Depression diagnosis                  No -37.0 -4.7 -17.4 0.3 -30.2 -6.1 -38.8 -5.5 -15.6 -0.9 -33.8 -7.6      Yes 37.0 4.7 17.4 -0.3 30.2 6.1 38.8 5.5 15.6 0.9 33.8 7.6 Work Firm size                  30 or less -0.2 -2.8 -1.8 1.0 2.1 1.7 1.8 -0.6 -1.1 0.8 2.7 2.2      31 to 150 1.2 1.4 0.4 -0.3 1.5 -0.8 1.1 -0.1 -0.3 0.4 1.1 -0.2      151 to 1000 -1.8 1.6 0.0 -1.8 -1.9 0.5 -1.4 -0.5 0.7 -1.8 -2.3 -1.5      1001 to 10000 0.9 0.0 1.1 0.5 -1.1 -1.5 -1.7 -1.8 0.8 -1.2 -1.1 -1.3      10,001 or more 0.0 -0.2 0.3 0.6 -0.7 0.0 0.2 3.0 0.0 1.7 -0.4 0.9 Occupation                  Management/admin -0.3 -0.1 -0.4 1.8 -0.8 -0.5 0.0 -1.4 -0.1 0.0 0.2 -0.3      Health 0.1 0.7 -0.9 0.2 0.1 0.6 -1.7 0.4 0.0 -1.0 -2.0 -1.3      Industry/utilities -1.1 0.1 -0.6 -2.6 -0.6 0.5 -0.2 0.4 -1.4 -2.2 0.0 0.9     Culture/services 2.0 0.9 1.6 0.5 -0.4 -2.0 0.0 1.2 0.3 3.4 -0.2 -0.2      Science/government 1.1 -1.0 0.8 1.3 0.2 -0.1 1.1 0.8 1.5 1.5 0.9 1.0      Trades/transport -1.9 -0.6 -0.4 -1.2 1.5 1.6 0.8 -1.3 -0.4 -1.7 1.1 -0.2      144   Men Women  Minimally adequate treatment Counselling Prescription Minimally adequate treatment Counselling Prescription  Actual  IV Actual  IV Actual  IV Actual  IV Actual  IV Actual  IV Injury year                  2000 -0.6 -0.7 2.0 0.3 -1.1 -2.0 0.6 -0.2 2.3 0.3 0.2 -0.7      2001 -0.3 -0.1 1.8 0.7 -0.7 -1.5 -0.5 -3.8 1.9 -1.0 -0.9 -2.9      2002 -0.6 -1.5 1.2 0.7 -0.5 -2.2 -1.0 -0.9 1.1 1.2 -0.3 -1.6      2003 0.2 -1.4 0.9 0.3 0.3 -0.9 -0.4 -1.8 0.6 0.3 -0.2 -2.2      2004 -0.4 -2.4 -0.3 -0.1 -0.2 -1.4 -0.3 -2.5 0.1 -0.1 -0.5 -1.5      2005 -1.0 0.6 -1.7 1.2 -0.9 -0.8 -0.6 -1.6 -1.0 -0.3 -0.5 -1.9      2006 -0.1 -0.8 -1.6 -0.9 -0.2 0.1 0.5 -0.8 -0.8 0.9 0.0 -0.1      2007 -1.1 -0.3 -0.9 -0.8 -0.5 0.5 0.6 0.2 -0.9 -0.6 0.7 -1.0      2008 -0.9 1.3 -2.0 -1.8 -0.3 1.6 -0.2 1.8 -1.9 0.6 -0.1 0.6      2009 0.6 1.0 0.0 0.7 1.1 -0.2 0.3 1.3 -0.1 -0.9 -0.3 0.6      2010 0.6 0.7 0.2 -0.3 -0.2 1.6 0.3 2.5 -0.3 -0.2 0.7 3.2      2011 1.4 1.7 -0.1 1.2 1.3 1.3 0.1 2.0 -0.6 -1.1 0.4 2.5      2012 0.5 0.0 -0.1 -1.3 0.9 1.9 0.0 2.3 -0.7 -0.2 0.8 3.4      2013 1.7 2.0 0.5 0.1 0.9 1.9 0.6 1.3 0.3 1.0 -0.1 1.4 a PD for actual treatment= prevalence difference = prevalence treated – prevalence untreated; PD for instrumental variable= prevalence highest quartile (Q4) – prevalence lowest quartile (Q1) *A lower prevalence difference indicates better balance. Prevalence differences of more than 10% are bolded     145 7.3.2 Assessment of instrumental variable strength (First stage regression) For all treatment types, there was a positive association between actual treatment and physician treatment preference (Table 23 and Table 24). Based on the conventional criterion of instrumental variable strength (Partial F statistic <10 = weak), the instrumental variable for minimal adequate treatment was weak for men but not women (F values from the full and restricted study samples respectively: men=1.7 and 10.8, women=29.7 and 50.2), while the instrumental variables for prescription and counseling services were strong for both men and women (Table 24). The instrumental variable for counseling (F values from the full and restricted study samples respectively: men=259.9 and 607.7, women=499.4 and 912.3) was the strongest followed by the instrumental variable for prescription services (F values from the full and restricted study samples respectively: men=34.5 and 118.8, women=110.0 and 175.5). In the full study sample, after adjustment for confounders, physician treatment preference explained 4.5% of the variation in counseling for men and 6.0% for women. Also in the full study sample, little of the variation in prescription services (men=0.6%, women=1.5%) and minimal adequate treatment (men= 0.0%, women=0.4%) was explained by physician treatment preference. The strength of the instrumental variables improved when the instruments were limited to services from a general practitioner and workers treated by a psychiatrist were excluded from the study sample. This is reflected in the larger partial F-values (despite the smaller sample size), odds ratios, and partial R2 values for the restricted study sample in comparison to the corresponding values obtained from the full study sample (Table 24). Overall, the instrumental variables were stronger for women than for men.  Table 23: Unadjusted proportion of workers that received actual treatment by quartiles of the instrumental variables  Men Women  Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Full study sample Minimal adequate treatment 55.2 55.2 55.9 60.1 53.6 58.3 59.8 64.4 Counseling 44.1 52.1 62.7 70.9 44.3 56.3 66.8 75.3 Prescription 56.9 62.2 65.7 69.3 54.5 66.6 72.3 73.0 Workers treated by a psychiatrist excluded Minimal adequate treatment 46.7 45.7 47.2 54.0 44.9 52.7 52.7 59.4 Counseling 26.1 35.7 51.7 70.0 28.9 47.6 61.9 74.5 Prescription 46.5 60.3 63.1 68.3 49.4 63.8 67.6 72.9 Q1= lowest quartile (contains workers whose general practitioner was least likely to provide treatment based on the proportion of other anxiety patients to receive treatment); Q2 = second lowest quartile; Q3= second highest quartile; Q4 = highest quartile (contains workers whose general practitioner was most likely to provide treatment based on the proportion of other anxiety patients to receive treatment)   146   Table 24: Characteristics of the first-stage instrumental variable (IV) regression models   Unadjusted OR  (IV->T) Adjusted ORa  (IV->T) Partial F-valueab Partial r2 valueac Full study sample (nmen=5,810, nwomen=8,358) Minimal adequate treatment Men 1.26 (0.98 to 1.61) 1.20 (0.91 to 1.58) 1.7 0.0002 Women 1.87 (1.52 to 2.30) 1.89 (1.50 to 2.38) 29.7 0.0038 Counseling Men 4.96 (4.05 to 6.08) 5.44 (4.40 to 6.73) 259.9 0.0446 Women 6.24 (5.26 to 7.40) 6.07 (5.84 to 8.32) 499.4 0.0602 Prescription Men 2.34 (1.82 to 3.00) 2.22 (1.70 to 2.90) 34.5 0.0062 Women 3.50 (2.82 to 4.34) 3.40 (2.70 to 4.28) 110.0 0.0148 Workers treated by a psychiatrist excluded (nmen=4,922 nwomen=7,452) Minimal adequate treatment Men 2.02 (1.39 to 2.94) 1.97 (1.31 to 2.96) 10.8 0.0022 Women 3.71 (2.73 to 5.03) 3.26 (2.36 to 4.49) 50.2 0.0067 Counseling Men 18.18 (14.22 to 23.24) 19.49 (15.13 to 25.10) 606.7 0.1183 Women 17.57 (14.40 to 21.43) 18.72 (15.27 to 22.95) 912.3 0.1178 Prescription Men 7.13 (5.10 to 9.97) 6.96  (4.88 to 9.93) 118.8 0.0254 Women 8.29 (6.29 to 10.93) 7.02 (5.23 to 9.42) 175.7 0.0262 a Adjusted for all study covariates b A partial F-value of less than 10 indicates a week instrument. The F-value is a function of both the strength of the instrument and the sample size. c A higher partial r2 value indicates a stronger instrument.  7.3.3 Estimation of treatment effect  1. Instrumental variable regression analyses (second stage regression) Several of the 95% confidence intervals for the instrumental variable coefficients were wide and crossed 0, indicating low precision for the instrumental variable effect estimates (Table 25). In this section, the direction of association as indicated by the instrumental variable coefficients is emphasized, to highlight trends observed across this section of the findings. 1 a) Minimal adequate treatment In the full study sample, minimally adequate treatment trended towards no association with disability days for men (?̂?=0.01, 95% CI: -0.22 to 0.23) and less disability days for women (?̂?=-0.13, 95% CI: -0.32 to 0.06). In the restricted study sample, minimally adequate treatment was associated with less disability days for both men (?̂?=-0.16, 95% CI: -0.43 to 0.10) and women (?̂?=-0.20, 95% CI: -0.40 to 0.01). Based on the restricted study sample, minimally adequate treatment was associated with a 14.8% and 18.1% reduction in disability days for men and women respectively.              147  1 b) Counseling In the full study sample, counseling was associated with less disability days for both men (?̂̂?=-0.32, 95% CI: -0.62 to -0.02) and women (?̂̂?=-0.14, 95% CI: -0.38 to 0.10). Similar results were found in the restricted study sample that excluded workers treated by a psychiatrist. Based on the restricted study sample, counseling was associated with a 19.7% and 17.3% (estimates from adjusted model 2) reduction in disability days for men and women respectively. 1 c) Prescription services In the full study sample, prescription services were associated with more disability days for men (?̂̂? =0.06, 95% CI: -0.21 to 0.34) and less disability days for women (?̂̂? =-0.16, 95% CI: -0.38 to 0.06). In the restricted study sample, prescription services were associated with less disability days for both men (?̂?=-0.07, 95% CI: -0.34 to 0.21) and women (?̂?=-0.21, 95% CI: -0.44 to 0.01). For men, the effect size for prescription services was small in both analyses. Based on the restricted study sample, prescription services were associated with a 5.8% and 18.1% (estimates from adjusted model 2) reduction in disability days for men and women respectively. 2. Conventional regression analyses (for comparison purposes) The instrumental variable analyses yielded wide confidence intervals that were consistently wider than comparable confidence intervals from the conventional analysis, highlighting a common draw back of the instrumental variable approach - a loss of precision compared to conventional regression (Table 26).  In conventional analysis, all effect estimates were positive (i.e. each treatment variable was associated with more disability days) (Table 26). Compared to the conventional analyses, the instrumental variable analyses demonstrated a change in the direction of effect (from positive to negative), with two exceptions. Among men in the full study sample, minimally adequate treatment and prescription services were positively associated with more disability days in the instrumental variable analyses. This may be because the instrumental variables used to obtain these effect estimates were weak (partial F-values of 1.67 and 34.5 respectively) (Table 24) and may not have been strong enough to remove all confounding. In the restricted study sample, these same instrumental variables for men were stronger (partial F-values of 10.8 and 50.2)   148 (Table 24); and as such, minimally adequate treatment and prescription services were associated with less disability days for men in the restricted study sample. Collectively, this suggests that findings from the conventional analyses were driven by confounding due to anxiety symptom severity, and that strong instrumental variables can reduce these confounding effects. Table 25: Instrumental variable two stage least squares linear regression estimates of the impacts of minimal adequate treatment, counseling, and prescription services on disability days (log-transformed)   Adjusted Model 1a Adjusted Model 2b   Co-efficient  (95% CI) Implied % effect  Co-efficient  (95% CI) Implied % effect Full study sample Minimal adequate  treatment Men 0.01 (-0.22 to 0.23) 1.0   Women -0.13 (-0.32 to 0.06) -12.2   Counseling Men -0.32 (-0.62 to -0.02) -27.4 -0.46 (-0.79 to -0.13) -36.9 Women -0.14 (-0.38 to 0.10) -13.1 -0.08 (-0.34 to 0.19) -7.7 Prescription Men 0.06 (-0.21 to 0.34) 6.2 0.27 (-0.04 to 0.58) 31 Women -0.16  (-0.38 to 0.06) -14.8 -0.14 (-0.37 to 0.10) -13.1 Workers treated by a psychiatrist excluded Minimal adequate  treatment Men -0.16 (-0.43 to 0.10) -14.8   Women -0.20 (-0.40 to 0.01) -18.1   Counseling Men -0.22 (-0.46 to 0.02) -19.7 -0.22 (-0.46 to 0.02) -19.7 Women -0.21 (-0.41 to 0.00) -18.9 -0.19 (-0.40 to 0.02) -17.3 Prescription Men -0.07 (-0.34 to 0.21) -6.8 -0.06 (-0.34 to 0.22) -5.8 Women -0.21 (-0.44 to 0.01) -18.9 -0.20 (-0.43 to 0.02) -18.1 a Adjusted model 1 = adjusted for all potential confounders including age, income, occupation, somatic co-morbidity index score, firm size, location (urban/rural), dependents under age 18, co-morbid depression, other mental co-morbidities (other than anxiety or depression), and injury year b Adjusted model 2 (applicable to counseling and prescription services only)= adjusted for all potential confounders listed above and the other treatment modality (counseling or prescription services).  *The coefficients can be interpreted as the number of excess disability days (log transformed) associated with treatment. Implied % effect can be interpreted as the percent change in disability days associated with a 1-unit change in X (i.e. no treatment versus treatment).    149 Table 26: Conventional linear regression estimates of the impacts of minimal adequate treatment, counseling, and prescription services on disability days (log-transformed)   Unadjusted Adjusted Model 1a Adjusted Model 2b   Co-efficient  (95% CI) Implied % effect  Co-efficient  (95% CI) Implied % effect  Co-efficient  (95% CI) Implied % effect Full study sample Minimal adequate  treatment Men 0.09 (0.01 to 0.17) 9.4 0.07 (-0.02 to 0.15) 7.3   Women 0.03 (-0.04 to 0.09) 3.0 0.03 (-0.04 to 0.10) 3.0   Counseling Men 0.13 (0.05 to 0.21) 13.9 0.13 (0.05 to 0.21) 13.9 0.11 (0.03 to 0.20) 11.6 Women 0.03 (-0.04 to 0.10) 3.0 0.04 (-0.03 to 0.11) 4.1 0.04 (-0.03 to 0.11) 4.1 Prescription Men 0.16 (0.08 to 0.25) 17.4 0.12 (0.04 to 0.21) 12.7 0.11 (0.03 to 0.20) 11.6 Women 0.11 (0.04 to 0.18) 11.6 0.10 (0.03 to 0.17) 10.5 0.10 (0.03 to 0.17) 10.5 Workers treated by a psychiatrist excluded Minimal adequate  treatment Men 0.06 (-0.02 to 0.15) 6.2 0.04 (-0.05 to 0.12) 4.1   Women 0.02 (-0.05 to 0.09) 2.0 0.01 (-0.06 to 0.08) 1.0   Counseling Men 0.15 (0.06 to 0.24) 16.2 0.14 (0.05 to 0.22) 15.0 0.13 (0.05 to 0.22) 13.9 Women 0.01 (-0.07 to 0.08) 1.0 0.01 (-0.06 to 0.08) 1.0 0.01 (-0.06 to 0.08) 1.0 Prescription Men 0.12 (0.03 to 0.21) 12.7 0.09 (0.00 to 0.17) 9.4 0.08 (-0.01 to 0.17) 8.3 Women 0.09 (0.02 to 0.17) 9.4 0.08 (0.00 to 0.15) 8.3 0.08 (0.00 to 0.15) 8.3 a Adjusted model 1 = adjusted for all potential confounders including age, income, occupation, somatic co-morbidity index score, firm size, location (urban/rural), dependents under age 18, co-morbid depression, other mental co-morbidities (other than anxiety or depression), and injury year b Adjusted model 2 (applicable to counseling and prescription services only)= adjusted for all potential confounders listed above and the other treatment modality (counseling or prescription services).  *The coefficients can be interpreted as the number of excess disability days (log transformed) associated with treatment. Implied % effect can be interpreted as the percent change in disability days associated with a 1-unit change in X (i.e. no treatment versus treatment).   150 7.4 Discussion 7.4.1 Summary of main findings and assessment of the instrumental variables Using a quasi-experimental study design and instrumental variables based on physician treatment preference, this study examined the effects of minimal adequate treatment, counseling and pharmaceutical treatment on return to work for workers with anxiety in addition to lost-time upper limb or spine strain or sprain work injury. Counseling was associated with less disability days for both women and men, and minimal adequate and pharmaceutical treatment were associated with less disability days for women. For men, there was inconsistent evidence regarding the associations of minimal adequate and pharmaceutical treatment with disability days, possibly due to weak instrumental variables, or in the case of pharmaceutical treatment, possible absence of a true effect for men (as indicated by the small effect sizes).  Several of the 95% confidence intervals for the treatment effect estimates obtained using the instrumental variables crossed 0. While this does not imply that the point estimates are invalid, it does affect the level of certainty with which conclusions can be drawn.(230) This is a common drawback of instrumental variable analyses. Compared to conventional regression, instrumental variable analyses generally require a much larger sample size to obtain the same degree of precision, especially when instrument strength is low or confounding is strong.(230) Sample sizes for pharmaco-epidemiological studies using physician treatment preference as an instrumental variable for treatment are commonly in the 100,000 range. Despite low precision for some effect estimates in the current study, the findings consistently demonstrated that when the instrumental variables were sufficiently strong, treatments were associated with less wage loss days. The instrumental variables were consistently stronger for woman than for men. It is possible that physicians have different treatment preferences for women with anxiety than for men with anxiety. It is also likely that of the anxiety patients used to estimate physician treatment preference, there were more women than men, due to the higher prevalence of anxiety among women; and this could account for why the instrumental variables were stronger for the women in the study sample. During the analyses, a research decision was made to include both women and men in the calculation of physician treatment preference to increase the number of patients that these measures were based on. Future research should examine the impact of this decision   151 on the results, as well as adjustments to the definitions of the instrumental variables to improve the strength of these tools for men.  Point estimates obtained using the prescription and minimal adequate treatment instrumental variables may be biased upward due to a possible violation of the assumption of no unaccounted association between the instrumental variable and unmeasured confounders for both men and women. Compared to actual treatment, these instrumental variables improved the balance of important confounders like comorbid depression and other mental comorbidities, however some imbalance remained for these variables. This increases the plausibility that the instrumental variables for prescription services and minimal adequate treatment could be related to unmeasured confounders such as the severity of anxiety symptoms. In contrast, the instrumental variable for counseling sufficiently balanced the comorbid depression and other mental comorbidities variables for both men and women. The instrumental variable for counseling was also several times stronger than the instrumental variables for minimal adequate treatment and prescription services.  Collectively, the testing of the instrumental variable assumptions performed here demonstrated that the counseling and prescription instrumental variables were strong predictors of actual treatment values and able to improve balance of important confounders for both women and men. However, the counseling instrumental variable performed better than the prescription services instrumental variable in both of these domains. The instrumental variable for minimal adequate treatment was weak for men but strong for women, and while this instrumental variable improved the balance of important confounders for both men and women, like the prescription services instrumental variable, it did not perform as well in this domain as the counseling instrumental variable. This indicates the potential utility of these instrumental methods, especially for counseling, for future research on outcomes of physician mental health services.  7.4.2 Potential mechanisms linking primary care mental health services and return to work  There is a small body of evidence that high quality treatment of anxiety in a primary care context can improve work outcomes.(216,231,232) Ways that counseling might improve return to work after physical injury for workers with pre-existing anxiety include decreased situational   152 catastrophizing and all-or-none thinking, reduced worry and fear of failure, reduced avoidance of potentially difficult or embarrassing situations (e.g. returning to work while still dealing with some degree of physical injury that may require work modifications), improved self-esteem or confidence, increased belief in ability to return to work, and increased resilience to stressful situations. There is some evidence that short-term cognitive behavioral therapy is an effective anxiety relapse-prevention strategy and that treatment response is long lasting even after treatment cessation.(233)  Point estimates suggested that pharmaceutical treatment might improve return to work for women, but there was no clear evidence of a similar association for men. Pharmaceutical treatment might improve return to work for workers with anxiety through either pain reduction or an attenuation of the anxiety response.(234) However, pharmaceutical treatment could also negatively impact return to work through unwanted negative side effects (e.g. decreased cognitive functioning, fatigue, blurred vision). In addition, evidence indicates that attenuation of an anxiety response by medication only works as long as medication is continued, and once discontinued, symptom relapse is common.(233) As a wide time frame was used to identify dispensing events (including the one year prior to injury), the effects of pharmaceutical treatment on return to work may have been diluted by discontinued treatment or non-adherence. Greater non-adherence among men compared to women could explain the absence of a suggested treatment effect in men in the current study. However, findings from the Canadian Community Health Survey suggest that men have better adherence to psychotropic medications than women.(235) Likewise, there is little evidence to suggest that anxiety pharmacotherapy is more effective for women, although there is a lack of research on gender differences to treatment response.(236) Other possible explanations for an absent pharmaceutical treatment effect in men but not women include: gender differences in the relative frequency of anxiety disorder subtypes (e.g. unlike most anxiety disorders, social anxiety shows no gender differences in prevalence),(237) differences in physician treatment preferences for pharmacotherapy based on patient gender, and unaccounted for residual confounding in the instrumental variable analyses for pharmaceutical treatment by anxiety symptom severity, particularly in the men’s sample. This last point is supported by the following: a greater imbalance of other mental co-morbidities (not including anxiety and depression) across quartiles of the instrumental variable in the men’s   153 sample than the women’s, the strength of the instrumental variable for pharmaceutical services was weaker for men than for women, and in the conventional regression analysis, the number of additional disability days associated with pharmaceutical treatment was greater for men than for women, suggesting that the relationship between pharmaceutical treatment and disability days was more strongly confounded by anxiety symptom severity in the men’s sample, than the women’s.  7.4.3 Strengths and limitations This study implemented an instrumental variable quasi-experimental research design to overcome challenges commonly encountered in observational research to generate evidence on an important health and social issue where there previously was none. This is a strength compared to conventional regression analyses unable to account for bias due to unmeasured confounding. Despite this, certain limitations of the study should be noted.  First, as mentioned earlier, the wide standard errors for the point estimates in the instrumental variable analyses compared to the conventional regression analyses imply that the instrumental variable analyses in the current study were limited by small sample size. This demonstrates that instrumental variable analyses may have limited usefulness for studies of this nature with sample sizes similar to or smaller than the current sample. Second, mental health services accessed through private health systems and some mental health services accessed through workers’ compensation were not available for analyses. This limited the scope of the study to publically insured mental health services, instead of the full range of mental health services available to workers. While this is a limitation, physicians are the most commonly utilized mental health service providers by people with anxiety,(224) and the predominant public policy lever for improving mental health treatment in the current BC public health system. Further, the lack of data on private services was considered non-differential in the current study, and would have exerted a conservative bias on the estimates. Third, as an observational study, the counseling and pharmaceutical treatment measures were limited to current physician treatment preferences, and some physicians may not deliver these services in a guideline concordant manner. For example, in British Columbia, General Practitioners receive relatively little training in counseling programs for anxiety disorders   154 compared to psychiatrists and psychologists. It was not possible to assess the type or quality of counseling services identified in the claims data, and there remains the possibility that some of these services fell short of evidence-based standards. As another example, the measure of pharmaceutical services was binary (yes/no) and did not distinguish between prescriptions for evidenced-based medications at an adequate dosage versus prescriptions not meeting these criteria. The instrumental variable analyses conducted here estimate the impacts of current general practitioner treatment practices on return to work some of which may include low quality or non-evidence based counseling or pharmaceutical services. Thus, the current study may underestimate the potential impacts that these services could have if delivered in an evidence-based manner, and in the case of counseling services, delivered by clinicians with specialized training.  Fourth, the study sample was restricted to workers with anxiety who accessed physician or hospital services associated with a diagnosis of anxiety at least twice in the 24 months before injury (with at least one of these events occurring in the 12 months before injury). Thus, findings may not be generalizable to workers with anxiety who do not access physician services, or workers with sub-threshold anxiety.  Fifth, physician’s mental health treatment preferences could be related to other treatment preferences (e.g. treatment of somatic health conditions) or physician characteristics that also affect return to work. While this would be a violation of the assumption that the instrumental variable affects the outcome only through the treatment of interest (and not other treatments), it still has meaningful policy implications as it implies that interventions to improve physician treatment practices may improve return to work and reduce work disability among workers with anxiety at a population level. Whether or not other physician treatment practices (not related to mental health services) or physician characteristics influenced the current findings could be examined by replicating the analyses in the future using a negative control population (e.g. workers with physical work injury but no history of mental disorders). 7.4.4 Recommendations for improvements to the instrumental variable methodology  Future research should investigate further improvements to the instrumental variable methods explored here. Namely, increased strength of the instrumental variables should increase the   155 precision of the point estimates for the main relationships of interest – allowing firmer conclusions to be drawn. As the current study was exploratory in nature, with the intention to assess the suitability of an instrumental variable approach to the research question, study sample, and health claims data, a full examination of alternative instrumental variable definitions and methods were not explored. Possible improvements to the instrumental variable methodology to be explored by future research include: 1) exclusion of workers with depression or other mental co-morbidities from the study sample as well as the calculation of the instrumental variable measures. In the current study, we explored the effect of excluding workers treated by a psychiatrist from the study sample and calculation of the instrumental variables. This increased the strength of the instrumental variables, likely by decreasing the amount of confounding due to anxiety symptom severity. 2) Calculation of separate instrumental variable measures for the men’s study sample (based on other anxiety patients who are men) and the women’s study sample (based on other anxiety patients who are women). While regression models from the current study were stratified by gender, the instrumental variables were calculated using anxiety patients who were men and who were women. This was done to increase the number of anxiety patients available for each physician and to obtain more reliable measures of physician treatment preference. The effects of stratifying and calculating separate instrumental variable measures by age and socio-economic status should also be explored. 3) The use of varying time frames to calculate physician treatment preference (e.g patients treated in the last 10 years, versus the last 5 years, versus the last year, versus the most recent patient only). In the current study, physician treatment preference was based on all anxiety patients seen during the study period (2000 to 2013) (once again to increase the number of patients per physician). 4) Less specific case definitions for workers with anxiety. In the current study, only workers with at least 2 diagnoses of anxiety within 365 days of each other were considered to have anxiety. This criterion was used for both selection into the study sample, and the identification of other anxiety patients to calculate physician treatment preference. Alternative case definitions, such as one-anxiety diagnoses should be examined. This may i) make the study more externally generalizable to milder or less chronic anxiety cases (although it may also increase cofounding due to anxiety symptom severity), ii) increase the sample size, iii) increase the representation of general practitioners who   156 tend not to monitor their anxiety patients through follow up visits in the study. This latter point might increase the variability in physician treatment preference and result in a stronger instrumental variable. 5) More refined measures of the instrumental variables beyond ‘did the worker receive the service (yes/no)’ should be examined. This may not be possible for counseling given the lack of detail in the data and current counseling practices but should be possible for pharmaceutical services. For example, instrumental variables that capture whether or not a physician’s prescribing preferences are concordant with evidence based recommendations for drug type and dosage should be explored. And lastly, 6) extension of these methods to workers with musculoskeletal work injury and pre-existing depression should be examined. 7.4.5 Conclusion These findings confirm that, in the absence of good measures for anxiety symptom severity, estimates of associations between anxiety treatments and work disability will be biased when using conventional regression analysis. Based on the findings for the current study, an alternative pseudo-experimental approach using instrumental variables based on physician treatment preference for anxiety is proposed as a means to address the limitations of conventional regression. The findings indicate that this approach can be used to remove confounding due to unmeasured anxiety severity, but that the effectiveness of the instrumental variables can vary by treatment type (e.g. counseling versus pharmaceutical treatment) and worker characteristics (e.g. men versus women). Future research is needed to improve the effectiveness of these instruments, and to explore the instrumental variable assumptions in greater depth. More research on this promising method will increase the certainty with which conclusions can be drawn regarding associations between anxiety treatments and work disability outcomes. The preliminary results provided here, which suggest that anxiety treatments are associated with less work disability after musculoskeletal strain or sprain work injury, have face validity and are an impetus for further research of these associations in other study populations, as well the use of non-observational study designs such as controlled trials.     157 Chapter 8: Discussion This dissertation adds to the body of knowledge on mental health and musculoskeletal work injury by addressing gaps in the literature and methodological limitations of prior studies. Using cohorts of workers with lost-time upper limb or spine strain or sprain work injury, the research chapters examined the descriptive epidemiology of anxiety and depression disorders; the impacts of pre-existing and new onset anxiety and depression disorders on the probability of sustained return to non-modified work; the impacts of pre-existing anxiety and depression disorders on the probability of return to non-modified work (from lost-time), and recurrence of lost-time (after initial return to work); and the impacts of physician mental health services on disability days for workers with pre-existing anxiety. Gender analyses were also conducted for each research chapter to examine if any of the above measures of disease frequency or effect estimates varied by gender, thus extending knowledge of gendered experiences of mental health and work disability in a physical work injury context.  In this final chapter, a summary of the results and contributions to the literature is provided followed by a summary of the overall strengths and limitations of the thesis, and lastly implications of the research findings and recommendations for future research are discussed. 8.1 Summary of results and contributions 1. Descriptive epidemiology of anxiety and depression disorders Findings from Chapter 4 demonstrated that pre-existing anxiety and depression disorders are common among injured workers, as well as being chronic and recurring disorders. In the study sample, 13.2% of men and 29.8% of women had one or both disorders in the year before injury, and more often than not, pre-existing disorders from the year before injury persisted or recurred in the year following injury. Findings also demonstrated that lost-time upper limb or spine strain or sprain work injury was not a significant driver of medically recognized new onset anxiety or depression disorders, as the prevalence of anxiety and depression disorders in the year following injury was only slightly elevated (~1 to 2%) compared to the year before injury.  Chapter 4 also examined factors associated with the presence of anxiety and depression disorders in the study sample. Factors associated with an anxiety or depression disorder or both in the year prior to injury for both men and women included older age (especially middle age), lower   158 income, a greater number of physical comorbidities, mental disorders other than anxiety and depression, and prior workers’ compensation claims. Factors unique to women associated with an increased probability of prevalent anxiety and/or depression included having a dependent in the home under the age of 19 years (for anxiety only); living in a rural area (for depression only and for comorbid anxiety and depression); and working in occupations in business, finance, and administration (for anxiety only, depression only, and comorbid anxiety and depression). Factors unique to men associated with an increased probability of prevalent anxiety and/or depression included having a multi-site work injury such as a concussion concurrent with lost-time upper limb or spine strain or sprain work injury (anxiety only, and comorbid anxiety and depression), and working in a larger firm (anxiety only, depression only, and comorbid anxiety and depression). Unique to men, living in a rural area was associated with a decreased prevalence of anxiety only, and comorbid anxiety and depression.   This is the first population-level study of anxiety and depression in workers with physical lost-time work injury that uses clinical diagnoses from longitudinal health claims data in Canada, and one of only a few such studies internationally. Most other studies on this topic are based on self-reported anxiety and depression symptoms or diagnoses collected in the post-injury time period.(35–37) In the few studies that measured anxiety or depression using clinical structured interviews or clinically verified diagnoses from health claims data, measurement was limited to the post-injury time period, and these studies were not population based.(10–12,131) As new onset anxiety or depression symptoms or disorders during the post-injury time period has been the focus of most research on injured worker populations, a contribution of the current research is identification and characterization of a separate group of workers affected by pre-existing anxiety and depression disorders, who are often excluded from research on this topic through study sample selection criteria.(36,63) Lastly, there has been a limited examination of factors associated with anxiety and depression disorders in injured worker populations, especially for some of the work-related factors included in this study. 2. Impact of anxiety and depression on time to sustained return to non-modified work  Findings from Chapter 5 demonstrated that both recent anxiety and depression disorders from before injury, as well as new onset anxiety and depression disorders arising during the return to work process, were associated with a lower probability of sustained return to non-modified work.   159 When only recent pre-existing disorders were taken into consideration, anxiety alone had the largest negative impact on return to work, and the effect size for this relationship was significantly greater for men than for woman. Among men, those with recent anxiety took 14 extra calendar days to reach sustained return to non-modified work compared to similar men with no recent anxiety or depression. Among women, this difference was five calendar days. Given that the adjusted median time to sustained return to non-modified work in men and women with no anxiety and no depression was 34 and 40 calendar days respectively, anxiety disorders increased the time to return to work by 41.2% for men and 12.5% for women. These results add to an existing body of literature that has shown mixed results to-date, possibly due to prior study limitations that include small sample size, short follow up time, and the measurement of mental health symptoms in the post injury time period only and by self-report. Further, authors of systematic reviews on this topic have had difficulty drawing conclusions due to a limited number of high-quality studies (Table 11). The current study is a contribution to the literature as it addresses several of the aforementioned limitations through a longitudinal study design, population level data, and clinical diagnoses from before and after injury. The current findings are also in line with the broader mental health and disability literature beyond that of musculoskeletal work disability, that suggests that anxiety and depression disorders are associated with increased work disability.(174,175,186–188) The findings regarding gender are also relatively novel as there has been limited examination of gender differences in work disability associated with common mental disorders, especially for workers with physical work injury. In particular, the current findings suggest that certain aspects of masculine gender interact with anxiety symptoms to lower the probability of return to work relative to feminine gender and relative to those without anxiety symptoms. 3. Impact of recent anxiety and depression disorders on return to work and lost-time recurrence events Findings from the Chapter 6 demonstrated that pre-existing anxiety and depression disorders are associated with lower probability of return to work and higher probability of lost-time recurrence. Among men, anxiety, depression, and comorbid anxiety and depression were associated with a lower probability of return to non-modified work; and anxiety only and   160 comorbid anxiety and depression were associated with a higher probability of lost-time recurrence. Among women, anxiety only and comorbid anxiety and depression were associated with a lower probability of return to non-modified work; and anxiety only, and comorbid anxiety and depression were associated with a higher probability of lost-time recurrence.  This study compliments an existing body of evidence that depression symptoms are associated with poor return to work trajectories such as those with no return to work or late return to work events, or those with both return to work and lost-time recurrence event.(63,207) However, this study offers a more nuanced investigation of these relationships than prior studies in this area as it investigated the impacts of anxiety alone and comorbid anxiety and depression, as opposed to just depression alone on return to work, as well as lost-time recurrence.(63,207)  Further, the current study measured return to work status (at non-modified work versus lost-time) on a daily basis for up to two years, as opposed to other studies with larger measurement intervals.(63,207)  This study can inform the timing of interventions intended to reduce the negative impacts of pre-existing anxiety and depression on return to work within critical recovery windows, although more research is needed to determine the exact timing, duration and mode of intervention for workers with pre-existing disorders. The Readiness for Return to Work Model recommends ongoing depression specific support even after initial return to work for workers with both musculoskeletal injury and depression,(34) although empirical evidence for this recommendation is lacking. The current study suggests that anxiety may have a greater role in lost-time recurrence than depression for both women and men, and that anxiety specific supports that begin during lost-time and continue after initial return to work may also be warranted. Further, depression was also associated with lower probability of return to non-modified work in men (but not women), suggesting that depression specific supports may be particularly important during the lost-time. Further, such depression specific supports should be sensitive to depression-related challenges and issues that commonly occur among men with work injury during the return to work process. 4. Impact of physician mental health services on disability days for workers with anxiety In Chapter 7, an exploratory analysis used instrumental variables, based on physician treatment preferences to estimate the effect of physician mental health services received by individuals with a pre-existing anxiety disorder on return to work outcomes after lost-time upper limb or spine strain or sprain work injury. Results suggest that physician-counseling services were   161 associated with improved return to work for both men and women, and that pharmaceutical treatment was associated with improved return to work for women, although further investigation and refinement of the instrumental variable methods are warranted. There is a limited body of literature suggesting that for anxiety outpatients, anxiety treatment can improve work ability and reduce sickness absence days.(215,216) The current study adds to this existing body of evidence by examining associations between anxiety treatment and work disability in a specific population of workers with lost-time upper limb or spine strain or sprain work injury and a pre-existing anxiety disorder from prior to injury. The findings also demonstrate the potential of instrumental variable methods to address research questions in this area, when conventional regression methods are affected by bias due to unmeasured or poorly measured confounders. 8.2 Methodological limitations and considerations The strengths and limitations of each study/analyses have already been discussed in depth in the respective discussion sections of each research chapter. Methodological limitations and considerations of the dissertation as a body of work are discussed here.  While use of administrative health claims data to identify diagnosed medical conditions is a strength, use of this data also has its limits. A limitation already discussed in previous chapters includes low sensitivity for mental health conditions, as sub-clinical symptoms and untreated disorders are not likely to be captured in the case groups. This could lead to an underestimation of prevalence and also an underestimation of the true impacts of anxiety and depression on return to work.  Another drawback of the administrative data is that physician billing practices only code diagnoses using the first three digits of the ICD-9 codes in BC, with the first three digits representing umbrella definitions of anxiety and depression. As a consequence, examination of specific types of anxiety and depression disorders was not possible. The current findings suggest that anyone with an umbrella diagnosis for anxiety or depression may benefit from some early supports after lost-time upper limb or spine strain or sprain work injury. However, should these results have been driven by specific subtypes of anxiety or depression but not others, the main effects could have been underestimated, and the provision of support to all workers meeting the   162 umbrella definitions could be an inefficient use of resources. More specific measures of symptom severity and disorder subtype could be used to identify high-risk groups and develop more targeted interventions. More specific measures of anxiety and depression may be achievable using clinical or small samples. However, an advantage of the population based studies conducted here is the opportunity to identify overall patterns and relationships in the population and pick up on relationships that are robust to variability across the injured worker population of BC. Some potential confounders were not readily available in the administrative data records and this may have resulted in residual confounding. This includes anxiety and depression symptom severity and individual level psychosocial work factors. The implications of unmeasured symptom severity have already been discussed in the previous paragraph and in Chapter 7 and are not reiterated here. Psychosocial work factors have been associated with depression in the literature, although the causal direction of this relationship is not well established.(238) Psychosocial work factors have also been associated with poor return to work.(239) There could be some residual confounding in the analyses for Chapters 5 and 6 due to unmeasured psychosocial work factors, although the analyses did adjust for occupation, a measure that captures some psychosocial aspects of work. Other measures not readily available in the data known to play a role in the return to work process include measures of pain, physical injury severity (e.g. size of the physical lesion), and psychological factors. This is not necessary a limitation of the dissertation per se as these variables were not hypothesized as potential confounders of the relationships of interest with the anxiety and depression explanatory variables (Figure 2). However, having no readily available measures for these variables precluded an investigation of their role in the relationships of anxiety and depression disorders with return to work. To create a more homogenous study sample, only accepted claims with a primary ICD-9 code for strain or sprain of the spine or upper limb were selected for inclusion. This criterion excluded other common but more severe occupational soft tissue injuries to these body areas such as a partial tear of the rotator cuff or intervertebral disc displacement. As a consequence, the findings may not be generalizable to injuries thought of as more severe from a biomedical standpoint. It is   163 possible that the role of anxiety and depression in return to work, and the timing of anxiety depression onset, varies according to injury factors such as severity indicative of tears and displacements. Lastly, according to the study’s conceptual framework (Figure 2), workers’ compensation and healthcare system factors, as well as the overall societal context, may have contributed to the current findings. Thus, external generalizability may be limited accepted injury claims in jurisdictions that are similar to BC in terms of these aforementioned ecological factors.   8.3 Implications and considerations for future research, policy, and practice  The implications and considerations of the findings for future research, policy, and practice are discussed here. While suggestions are made for policy and practice, it is important to note that more research is needed to confirm the effectiveness of these suggestions.  Collectively, these results suggest that workers with recent or current anxiety or depression episodes at the time of lost-time upper limb or spine strain or sprain work injury are a large enough subgroup of the total injured worker population to warrant attention at the policy level, as they represent approximately 1 in 10 men and 3 in 10 women. The mental health needs of these workers and appropriate intervention efforts may differ from workers with no recent anxiety or depression episode at the time of injury but who are at risk of new onset symptoms post-injury. Policy makers should consider that workers with an anxiety or depression disorder, in addition to lost-time upper limb or spine strain or sprain work injury are more likely to be women, middle aged, have a lower income, and to have a more complicated clinical profile including a greater number of somatic comorbidities. Given this, affordability of anxiety and depression-related health care and its integration with other health care delivery are important considerations.  Findings indicate that both pre-existing and new-onset disorders are associated with lower probability of sustained return to work, suggesting that both pre-existing and new onset disorders should be considered in the claim management process. Currently in BC, case managers may consider mental disorders that arise from or are exacerbated by a work-related injury or illness, for inclusion on a claim. Once a mental disorder is included on a claim, the worker becomes eligible for compensated mental health care services intended to treat the disorder. Pre-existing mental disorders that are not exacerbated by a work-related injury are not eligible for inclusion, and methods to ascertain if mental symptoms were caused or exacerbated by a work-injury are   164 limited. As a consequence of this, many injured workers with a co-morbid anxiety or depression disorder, likely rely on public or private health care services for mental health care. Possible drawbacks of this include: under treatment, unaffordability of mental health care especially for evidence-based counseling programs and pharmaceutical treatment, and a lack of return-to-work focused mental health treatment. Even if not considered eligible for compensated mental health care, workers with pre-existing anxiety and depression disorders may benefit from consideration of their mental health needs as a part of the claim management process.  Another consideration is that improvements to mental health care in the public health system and better employee benefits for mental health services; as well as decreased mental stigma and increased mental health literacy in the general and working populations, might help to ensure that workers with anxiety and depression disorders receive appropriate mental health treatment. This in turn, might help mitigate any negative impacts of anxiety and depression disorders on work outcomes following physical work injury, especially for pre-existing anxiety and depression disorders that ideally should already be nested within a mental treatment plan or program from prior to injury. Another important finding was that pre-existing anxiety disorders were associated with an increased risk of lost-time recurrence after initial return to non-modified work. This suggests that continued consideration of injured workers’ mental health needs, even after an initial return to work, may be beneficial.  The findings also offer important considerations for the role of gender in the inter-relationships between mental health, lost-time upper limb or spine strain or sprain, and return to work. First, women were more likely to have an anxiety or depression disorder than men. However, the detrimental impacts of anxiety and depression disorders on the probability of sustained return to non-modified work were greater for men than for women, especially for pre-existing anxiety. These two findings suggest that policy makers should consider gender sensitive approaches to address mental health in injured worker populations, although more research is needed to inform what these approaches might be. Two main findings from this dissertation offer important considerations for how return to work is defined by future research in this area. First, anxiety and depression disorders were associated   165 with lower probability of sustained return to non-modified work after lost-time upper limb or spine strain or sprain work injury. This is in contrast to several other studies that report no association (Table 11). It is difficult to know the exact reason(s) for this discrepancy in findings across studies, but these could include comparatively longer follow up time in the current study (two years), and use of a return to work measure that captures lost-time recurrence. In comparison, some studies use a follow up time as short as one month or only capture the first return to work event, and rarely are differences in modified and non-modified return to work captured. Return to work measurement for future research in this area should include longer duration of follow up and possible recurrence events to capture the total effects of anxiety and depression disorders on full return to work trajectories. A second finding that pre-existing anxiety disorders were associated with increased risk of lost-time recurrence after initial return to work reinforces this research recommendation.  Administrative data offers valuable resource-efficient opportunities to examine health and other social issues at a population level. However, the potential of administrative data for health and social research could be improved by increasing the depth and accuracy of the information available in the data. For example, the current study used a proxy measure for the number of dependents in the home based on the number of dependents registered on each worker’s medical service plan. While this proxy measure is better than no measure, it fails to capture dependents registered under the plan of another caregiver/legal guardian. Other variables for which administrative data exist, that could be added to the registry files, include marital/common law status, highest level of education obtained, and employment status. Likewise, for the WorkSafeBC files, the addition of information on individual level work exposures and injury severity that is commonly collected as a part of the return to work process could be beneficial for research of this nature, as would more complete information on the number and type of diagnoses accepted on a claim. Lastly, inclusion of symptom severity measures for both mental and physical health conditions to the Ministry of Health Data, as well as use of more specific ICD-9 codes for the physician billing data would be beneficial. However, physicians are not likely to provide this information if not required for billing purposes, as is currently the case.   166 8.4 Future research recommendations Findings from these epidemiological studies provide an impetus for further research in this area, as follows. 1. Descriptive epidemiology Findings from Chapter 4 suggested that the injury was not a major stimulus of new onset, clinically recognized anxiety and depression disorders. However, it remains unclear if the injury was a major stimulus for the exacerbation of pre-existing anxiety and depression disorders, including a new episode or exacerbated symptoms post-injury not picked-up in the billing data This last question is of relevance to adjudication decisions by workers’ compensation systems regarding the inclusion of a secondary condition on the claim. Often, as in BC, inclusion of secondary conditions is limited to those that were caused or exacerbated by the injury. Findings from Chapter 4 also indicate that the prevalence of pre-existing anxiety and depression disorders among workers with lost-time upper limb or spine strain or sprain is high. However, this finding is not conclusive as there was no non-injured comparison group. Estimates of anxiety and depression disorder prevalence for the general BC population stratified by age and gender based on health claims data recently became available.(142,240) However, the case definitions differed from the ones used here. If comparable case definitions are used, future research could generate better estimates for determining if pre-existing anxiety and depression disorders are elevated among injured workers, while controlling for age and gender. Being able to further refine these population-based estimates to the labour force would be ideal for comparison purposes, but working status is not readily collected as part of health claims within the public health care system in BC or Canada. Comparison research of this nature (comparing findings among the injured worker population with the overall labour force), might increase the uptake of the current findings by workers’ compensation systems, and encourage their consideration of pre-existing mental disorders in the return to work process. 2. Gender and work disability due to mental disorders While this study found some evidence that gender may moderate the association between anxiety and sustained return to work, more research is needed to better understand gender differences in work disability due to mental disorders. In particular, what contextual influences might lead to greater work disability among men with an anxiety disorder compared to similar women? A   167 better understanding of these contextual influences can inform conceptual frameworks on mental health, gender and work disability that can, in turn, inform research and practice in the area of return to work and other work disability contexts. 3. Identification of mediators and moderators of the relationships between anxiety and depression disorders and return to work outcomes Chapter 5 established that anxiety and depression disorders are associated with lower probability of sustained return to work. However, beyond the role of gender as a modifier, other moderators and mediators of these relationships were not examined. Findings indicate that pain amplification, and resistance to therapeutic treatment among workers with depression are plausible mediators, (58,241) but empirical evidence for self-efficacy and stigma as mediators is more limited. A better understanding of the mediators involved in these relationships and their relative contribution to return to work outcomes could help inform interventions to help workers with an anxiety or depression disorder return to work after physical injury. Future research might also investigate if socio-economic status moderates the associations of anxiety and depression with return to work, as well as differences in these associations by gender. In Canada, low socio-economic status is associated with higher risk of both anxiety and depression,(40,47) and a higher probability of acceptability barriers to mental care.(242) Workers with low socio-economic status may also have greater difficulty negotiating work accommodation due to lower social capital in the workplace, and when combined with stigma around mental illness, this could result in negative compounding effects on return to work. Lastly, workers with low socio-economic status might have less access to employer paid benefits for prescription anxiety or depression treatments and private counseling due to non standard working relationships, and this could contribute to poor return to work outcomes among low socio-economic status workers with an anxiety or depression disorder. 4. Mental health needs assessment Another potential area of research to inform interventions is assessment of injured workers’ mental health needs. This could include future questions on: what are the gaps in mental treatment among workers with a mental disorder and physical work injury? What sort of mental health services are workers willing to use if available, and what are the barriers and facilitators   168 that influence treatment initiation and adherence in a return to work context? How can return to work policy and procedures be customized to better meet the individual needs of workers with a mental disorder? What are the work accommodation needs of workers with a mental disorder in addition to physical work injury and how do these vary from the needs of workers with no mental disorder and just physical injury? Are there additional challenges in negotiating accommodation for injured workers with a comorbid mental disorder and if so, what are these? What aspects of the return to work process are associated with decreased or increased anxiety or depression symptoms? Lastly, how do answers to each of the aforementioned questions vary for women and men, or by levels of socio-economic status? 5. Recurrence  The current findings on the associations between pre-existing anxiety and depression disorders and lost-time recurrence within the context of a single claim could be expanded using the same datasets to include other policy relevant recurrence events such as a new work injury (not associated with the original claim from an administrative standpoint), or recurrence of symptoms associated with the injury as indicated by healthcare utilization. Other potential recurrence-related outcomes not captured by the current data for consideration by future research include voluntarily leaving a job or the labour market due to ongoing injury related-symptoms. Further examination of this topic will provide a more comprehensive picture of the long-term work disability and quality of life impacts of anxiety and depression disorders.  6. Identification of effective interventions and cost-effectiveness analyses Lastly, more research is needed to identify policies and procedures to promote improved return to work for workers with a mental disorder in addition to physical injury. In conjunction to this, cost-effectiveness analyses will be necessary to inform policy makers’ decision-making regarding the uptake of such interventions. Specific to the current work, further refinement of the instrumental variable measures introduced in Chapter 7 is necessary to better predict use of counseling and pharmaceutical treatment for workers with anxiety, in order to provide unbiased estimates of the associations between these anxiety treatments and return to work outcomes for workers with lost-time upper limb or spine strain or sprain work injury, especially men.    169 8.5 Summary The purpose of this research was to explore the patterns and return to work impacts of anxiety and depression disorders among workers with lost-time upper limb or spine strain or sprain work injury. Findings suggest that workers with an anxiety or depression disorder prior to injury comprise a significant proportion of the injured worker population, especially for women. Both pre-existing and new onset anxiety and depression disorders were associated with a lower probability of sustained return to work after lost-time upper limb or spine strain or sprain work injury, and pre-existing anxiety was associated with a higher probability of lost-time recurrence after initial return to work. A general trend observed across the findings was that the negative impacts of pre-existing anxiety disorders on return to work appeared to be greater than the negative impacts of pre-existing depression. Some gender differences were evident in these results. First, the effect size of the association between pre-existing anxiety and sustained return to work was greater for men than for women, and pre-existing depression was associated with sustained return to work for men but not women. Lastly, the findings suggest that further research on the effects of treatment for anxiety and depression disorders, or other mental health interventions, on return to work outcomes for workers with a pre-existing anxiety or depression disorder in addition to lost-time upper limb or spine strain or sprain work injury, is warranted. These findings offer important and novel contributions to our understanding of anxiety and depression disorders including their work disability impacts in a lost-time musculoskeletal work injury context, as well as contextual information to inform future research and the management of these disorders.    170 References  1.  James SL, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and Injuries for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392:1789–858.  2.  Punnett L, Prüss-Üstün A, Nelson DI, Fingerhuf MA, Leigh J, Tak SW, et al. Estimating the global burden of low back pain attributable to combined occupational exposures. Am J Ind Med. 2005;48(6):459–69.  3.  WorksafeBC. WorkSafeBC Statistics 2017. Richmond, BC; 2018.  4.  Rush AJ, Polatin P, Gatchel RJ. Depression and chronic low back pain: establishing priorities in treatment. Spine (Phila Pa 1976). 2000 Oct 15;25(20):2566–71.  5.  Viana MC, Lim CCW, Garcia Pereira F, Aguilar-Gaxiola S, Alonso J, Bruffaerts R, et al. Previous mental disorders and subsequent onset of chronic back or neck pain: findings From 19 countries. J Pain. 2018;19(1):99–110.  6.  Waghorn G, Chant D, Lloyd C. Labor force activity among Australians with musculoskeletal disorders comorbid with depression and anxiety disorders. J Occup Rehabil. 2006 Jun;16(2):241–52.  7.  Koehoorn M, McLeod C, Fan J, Hogg-Johnson S, Amick B. What a difference gender makes.... duration of work disability claims. Annual Symposion on Environmental, Occupational, and Public Health. Semiahmoo, Washington. January 9-10th, 2014.  8.  Macpherson RA, Koehoorn M, Fan J, Quirke W, Amick BC, Kraut A, et al. Do differences in work disability duration between men and women vary by province in Canada? J Occup Rehabil. 2019;29(3):560–8.  9.  American Psychiatric Association. Diagnostic and statistical manual of nental disorders. 5th ed. Arlington, VA: American Psychiatric Publishing; 2013. 1–947 p.  10.  Dersh J, Gatchel RJ, Polatin P, Mayer T. Prevalence of psychiatric disorders in patients with chronic work-related musculoskeletal pain disability. J Occup Env Med. 2002;44(5):459–68.  11.  Dersh J, Gatchel RJ, Mayer T, Polatin P, Temple OR. Prevalence of psychiatric disorders in patients with chronic disabling occupational spinal disorders. Spine (Phila Pa 1976).   171 2006;31(10):1156–62.  12.  Holmes ACN, Donnell MLO, Williamson O, Hogg M, Arnold C. Persistent disability is a risk factor for late-onset mental disorder after serious injury. Aust N Z J Psychiatry. 2014;(Published online).  13.  McRae L, O’Donnell S, Loukine L, Rancourt N, Pelletier C. Report summary – mood and anxiety disorders in Canada, 2016. Heal Promot Chronic Dis Prev Canada. 2016;36(12):314–5.  14.  Kent PM, Keating JL. The epidemiology of low back pain in primary care. Chiropr Osteopat. 2005;13:13.  15.  Hoy D, Bain C, Williams G, March L, Brooks P, Blyth F, et al. A systematic review of the global prevalence of low back pain. Arthritis Rheum. 2012;64(6):2028–37.  16.  Public Health Agency of Canada (PHAC). The economic burden of illness in Canada, 2010 [Internet]. Ottawa, Canada; 2018 [cited 2019 Jul 11]. p. 1–58. Available from: https://www.canada.ca/content/dam/phac-aspc/documents/services/publications/science-research/economic-burden-illness-canada-2010/economic-burden-illness-canada-2010.pdf 17.  March L, Smith EUR, Hoy DG, Cross MJ, Sanchez-Riera L, Blyth F, et al. Burden of disability due to musculoskeletal (MSK) disorders. Best Pract Res Clin Rheumatol. 2014;28(3):353–66.  18.  WorkSafeBC. Statistics 2016. Richmond, BC; 2016.  19.  Waters TR. National efforts to identify research issues related to prevention of work-related musculoskeletal disorders. J Electromyogr Kinesiol. 2004;14(1):7–12.  20.  Costa BR, Vieira ER, da Costa BR, Vieira ER. Risk factors for work-related musculoskeletal disorders: a systematic review of recent longitudinal studies. Am J Ind Med. 2010;53(3):285–323.  21.  National Research Council. The dynamics of disability: measuring and monitoring disability for social security programs. Wunderlich G, Rice D, Amado N, editors. Wahington, DC: The National Academies Press; 2002. 1–347 p.  22.  Nagi S. Some conceptual issues in disability and rehabilitation. In: Sussman M, editor. Sociology and Rehabilitation. Wahington, DC: American Sociological Association; 1965.  23.  Verbrugge LM, Jette AM. The disablement process. Soc Sci Med. 1994;38(1):1–14.  24.  Schultz IZ, Crook J, Fraser K, Joy PW. Models of diagnosis and rehabilitation in   172 musculoskeletal pain-related occupational disability. J Occup Rehabil. 2000;10(4):271–93.  25.  Schultz IZ, Stowell AW, Feuerstein M, Gatchel RJ. Models of return to work for musculoskeletal disorders. J Occup Rehabil. 2007;17(2):327–52.  26.  Pincus T, Williams A. Models and measurements of depression in chronic pain. J Psychosom Res. 1999;47(3):211–9.  27.  Schultz IZ, Crook JM, Berkowitz J, Meloche GR, Milner R, Zuberbier OA. Biopsychosocial multivariate predictive model of occupational low back disability. Spine (Phila Pa 1976). 2002;27(23):2720–5.  28.  Spitzer WO, Leblanc FE, Dupuis M. Scientific approach to the assessment and management of activity-related spinal disorders: a monograph for clinicians. Report of the Quebec Task Force for Spinal Disorders. Spine (Phila Pa 1976). 1987;12(7 Supplement 1):S9–55.  29.  Loisel P, Durand P, Abenhaim L, Gosselin L, Simard R, Turcotte J, et al. Management of occupational back pain: the Sherbrooke model. Results of a pilot and feasibility study. Occup Environ Med. 1994;51(9):597–602.  30.  Krause N, Ragland D. Occupational disability due to low back pain: a new interdisciplinary classification based on a phase model of disability. Spine (Phila Pa 1976). 1994;19(9):1011–20.  31.  Loisel P, Durand M, Berthelette D, Vézina N, Baril R, Gagnon D, et al. New paradigm for the management of occupational back pain. Dis Manag Heal Outcomes. 2001;9(7):351–60.  32.  Loisel P, Buchbinder R, Hazard R, Keller R, Scheel I, Van Tulder M, et al. Prevention of work disability due to musculoskeletal disorders: the challenge of implementing evidence. J Occup Rehabil. 2005;15(4):507–24.  33.  Young AE, Roessler RT, Wasiak R, McPherson KM, Van Poppel MNM, Anema JR. A developmental conceptualization of return to work. J Occup Rehabil. 2005;15(4):557–68.  34.  Franche R-L, Krause N. Readiness for return to work following injury or illness: conceptualizing the interpersonal impact of health care, workplace, and insurance factors. J Occup Rehabil. 2002 Dec;12(4):233–56.  35.  Wenzel HG, Haug TT, Mykletun A, Dahl AA. A population study of anxiety and   173 depression among persons who report whiplash traumas. J Psychosom Res. 2002;53(3):831–5.  36.  Franche R-L, Carnide N, Hogg-Johnson S, Cote P, Breslin FC, Bultmann U, et al. Course, diagnosis, and treatment of depressive symptomatology in workers following a workplace injury: a prospective cohort study. Can J Psychiatry. 2009;54(8):534–45.  37.  O’Hagan FT, Ballantyne PJ, Vienneau P. Mental health status of Ontario injured workers with permanent impairments. Can J Public Heal. 2012;103(August):303–8.  38.  Ballenger JC, Davidson JRT, Lecrubier Y, Stein DJ, Wittchen H. Consensus statement on generalized anxiety disorder from the International Consensus Group on Depression and Anxiety. J Clin Pscyhiatry. 2001;62(suppl 11):53–8.  39.  Bandelow B, Michaelis S. Epidemiology of anxiety disorders in the 21st century. Dialogues Clin Neurosci. 2015;17(3):327–35.  40.  Watterson RA, Williams JVA, Lavorato DH, Patten SB. Descriptive epidemiology of generalized anxiety disorder in Canada. Can J Psychiatry. 2017;62(1):24–9.  41.  The Government of Canada. The human face of mental health and mental illness in Canada 2006. Ottawa, Canada; 2006.  42.  Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62:593–602.  43.  Blanco C, Ph D, Rubio J, Wall M, Ph D, Wang S, et al. Risk factors for anxiety disorders: common and specific effects in a national sample. Depress Anxiety. 2014;31(9):756–64.  44.  Asmundson GJG, Katz J. Understanding the co-occurrence of anxiety disorders and chronic pain: state-of-the-art. Depress Anxiety. 2009;26:888–901.  45.  Hartikainen S, Lönnroos E, Louhivuori K. Medication as a risk factor for falls: critical systematic review. Journals Gerontol Ser A. 2007;62(10):1172–81.  46.  Coryell W, Akiskal HS, Leon AC, Winokur G, Maser JD, Mueller TI, et al. The time course of nonchronic major depressive disorder. Arch Gen Psychiatry. 1994;51(5):405–10.  47.  Patten SB, Jian LW, Williams JVA, Currie S, Beck CA, Maxwell CJ, et al. Descriptive epidemiology of major depression in Canada. Can J Psychiatry. 2006;51(2):84–90.  48.  Kroenke K, Wu J, Bair MJ, Krebs EE, Damush TM, Tu W. Reciprocal relationship   174 between pain and depression: a 12-month longitudinal analysis in primary care. J Pain. 2011;12(9):964–73.  49.  Covic T, Adamson B, Spencer D, Howe G. A biopsychosocial model of pain and depression in rheumatoid arthritis: a 12-month longitudinal study. Rheumatology. 2003;42(11):1287–94.  50.  Bair MJ, Robinson RL, Katon W, Kroenke K. Depression and pain comorbidity. Arch Intern Med. 2003;163(20):2433.  51.  Kendler KS, Karkowski LM, Prescott CA. Causal relationship between stressful life events and the onset of major depression. Psychiatry Interpers Biol Process. 1999;156:837–41.  52.  Tiesman H, Peek-Asa C, Whitten P, Sprince N, Stromquist A, Zwerling C. Depressive symptoms as a risk factor for unintentional injury: a cohort study in a rural county. Inj Prev. 2006;12(3):172–7.  53.  Han C, Pae CU. Pain and depression: a neurobiological perspective of their relationship. Psychiatry Investig. 2015;12(1):1–8.  54.  Symreng I, Fishman SM. Pain: clinical updates. Int Assoc Study pain. 2004;XII(7):1–6.  55.  Theunissen M, Peters ML, Bruce J, Gramke H-F, Marcus MA. Preoperative anxiety and catastrophizing. Clin J Pain. 2012;28(9):819–41.  56.  Peolsson M, Gerdle B. Coping in patients with chronic whiplash-associated disorders: a descriptive study. J Rehabil Med. 2004;36(1):28–35.  57.  Sullivan MJL, Simmonds M, Velly A. Pain, depression, disability and rehabilitation outcomes. Revised version. Montreal Quebec: Institut de recherche Robert-Sauvé en santé et en sécurité du travail (IRSST); 2011.  58.  Slepian P, Bernier E, Scott W, Niederstrasser NG, Wideman T, Sullivan M. Changes in pain catastrophizing following physical therapy for musculoskeletal injury: the influence of depressive and post-traumatic stress symptoms. J Occup Rehabil. 2014 Mar;24(1):22–31.  59.  Cole DC, Mondloch M V., Hogg-Johnson S. Listening to injured workers: how recovery expectations predict outcomes - a prospective study. CMAJ. 2002;166(6):749–54.  60.  Petrie KJ, Weinman J, Sharpe N, Buckley J. Role of patients’ view of their illness in predicting return to work and functioning after myocardial infarction: longitudinal study.   175 BMJ. 1996;312(7040):1191–4.  61.  Loukine L, O’Donnell S, Goldner EM, McRae L, Allen H, O’Donnell S, et al. Health status, activity limitations, work-related restrictions and level of disability among Canadians with mood and/or anxiety disorders. Heal Promot chronic Dis Prev Canada  Res policy Pract. 2016;36(12):289–301.  62.  Wang J, Patten S, Currie S, Sareen J, Schmitz N. Perceived needs for and use of workplace accommodations by individuals with a depressive and/or anxiety disorder. J Occup Environ Med. 2011;53(11):1268–72.  63.  Carnide N, Franche R-LL, Hogg-Johnson S, Côté P, Breslin FC, Severin CN, et al. Course of depressive symptoms following a workplace injury: a 12-month follow-up update. J Occup Rehabil. 2015;26(2):204–15.  64.  O’Donnell S, Vanderloo S, McRae L, Onysko J, Patten SB, Pelletier L. Comparison of the estimated prevalence of mood and/or anxiety disorders in Canada between self-report and administrative data. Epidemiol Psychiatr Sci. 2016;25(04):360–9.  65.  Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for measuring depression. Arch Gen Psychiatry. 1961;4:561–71.  66.  Beck AT, Epstein N, Brown G, Steer RA. An inventory for measuring clinical anxiety: psychometric properties. J Consult Clin Psychol. 1988;56:893–7.  67.  Hamilton M. The assessment of anxiety states by rating. Br J Med Psychol. 1959;32(1):50–5.  68.  Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23:56–62.  69.  Zigmod AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand. 1983;67(6):361–79.  70.  Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1(3):385–401.  71.  Canadian Institutes of Health Research. How to integrate sex and gender into research [Internet]. 2019 [cited 2019 Apr 29]. Available from: http://www.cihr-irsc.gc.ca/e/50836.html 72.  Messing K, Punnett ÃL, Bond M, Alexanderson K, Pyle J, Zahm S, et al. Be the fairest of them all: challenges and recommendations for the treatment of gender in occupational   176 health research. Am J Ind Med. 2003;43:618–29.  73.  Islam SS, Velilla AM, Doyle EJ, Ducatman AM. Gender differences in work-related injury/illness: analysis of workers compensation claims. Am J Ind Med. 2001;39(1):84–91.  74.  Berecki-Gisolf J, Smith PM, Collie A, Mcclure RJ. Gender differences in occupational injury incidence. Am J Ind Med. 2015;58(3):299–307.  75.  Steel Z, Marnane C, Iranpour C, Chey T, Jackson JW, Patel V, et al. The global prevalence of common mental disorders: a systematic review and meta-analysis 1980-2013. Int J Epidemiol. 2014;43(2):476–93.  76.  Riecher-Rössler A. Sex and gender differences in mental disorders. The Lancet Psychiatry. 2017;4(1):8–9.  77.  Kuehner C. Why is depression more common among women than among men? The Lancet Psychiatry. 2017;4(2):146–58.  78.  Stock SR, Nicolakakis N. Gender differences in duration of work absence for non-traumatic work-related musculoskeletal disorders. Occup Environ Med. 2016;73(Suppl 1):A188.  79.  Lederer V, Rivard M, Mechakra-Tahiri SD. Gender differences in personal and work-related determinants of return-to-work following long-term disability: a 5-year cohort study. J Occup Rehabil. 2012 Dec;22(4):522–31.  80.  Scott K, Collings S. Gender and the association between mental disorders and disability. J Affect Disord. 2010;125(1–3):207–12.  81.  Derdikman-Eiron R, Indredavik MS, Bakken IJ, Bratberg GH, Hjemdal O, Colton M. Gender differences in psychosocial functioning of adolescents with symptoms of anxiety and depression: longitudinal findings from the Nord-Trøndelag Health Study. Soc Psychiatry Psychiatr Epidemiol. 2012;47(11):1855–63.  82.  Thornicroft G, Patel V. Including mental health among the new sustainable development goals. BMJ. 2014;349(August):8–10.  83.  Roberge P, Fournier L, Duhoux A, Nguyen CT, Smolders M. Mental health service use and treatment adequacy for anxiety disorders in Canada. Soc Psychiatry Psychiatr Epidemiol. 2011;46(4):321–30.  84.  Kosteniuk J, Morgan D, D’Arcy C. Treatment and follow-up of anxiety and depression in   177 clinical-scenario patients: survey of Saskatchewan family physicians. Can Fam Physician. 2012;58(3):152–8.  85.  Talbot F, Clark DA, Yuzda WS, Charron A, McDonald T. ‘Gatekeepers’ perspective on treatment access for anxiety and depression: a survey of New Brunswick family physicians. Can Psychol. 2014;55(2):75–9.  86.  Roberge P, Hudon C, Pavilanis A, Beaulieu MC, Benoit A, Brouillet H, et al. A qualitative study of perceived needs and factors associated with the quality of care for common mental disorders in patients with chronic diseases: the perspective of primary care clinicians and patients. BMC Fam Pract. 2016;17(1):1–15.  87.  Roberge P, Normand-Lauzière F, Raymond I, Luc M, Tanguay-Bernard MM, Duhoux A, et al. Generalized anxiety disorder in primary care: mental health services use and treatment adequacy. BMC Fam Pract. 2015;16(1).  88.  Canada Health Act. RSC 1985, c C-6 [Internet]. [cited 2019 Dec 14]. Available from: http://canlii.ca/t/532qv 89.  Workers Compensation Act, RSBC 1996, c 492 [Internet]. [cited 2019 Dec 14]. Available from: ttp://canlii.ca/t/542t0 90.  Hurley J, Pasic D, Lavis JN, Mustard C, Culyer AJ, Gnam W. Parallel Lines Do Intersect: Interactions between the Workers ’ Compensation and Provincial Publicly Financed Healthcare Systems in Canada Quand les lignes parallèles se croisent: interaction entre les commissions des accidents du travail et les systèmes. Healthc Policy. 2008;3(4):100–12.  91.  Alamgir H, Koehoorn M, Ostry A, Tompa E, Demers PA. How many work-related injuries requiring hospitalization in British Columbia are claimed for workers’ compensation? Am J Ind Med. 2006;49(6):443–51.  92.  Alamgir H, Siow S, Yu S, Ngan K, Guzman J. Compensation patterns for healthcare workers in British Columbia, Canada. Occup Environ Med. 2009;66(6):381–7.  93.  Shannon HS, Lowe GS. How many injured workers do not file claims for workers’ compensation benefits? Am J Ind Med. 2002;42(6):467–73.  94.  Lippel K. Compensation for musculoskeletal disorders in Quebec: Systemic discrimination against women workers? Int J Heal Serv. 2003;33(2):253–81.  95.  Association of Workers’ Compensation Boards of Canada. Detailed key statistical measures report 2000 to 2013 [database online] [Internet]. [cited 2017 Nov 14]. Available   178 from: http://awcbc.org/?page_id=14 96.  Population Data BC. About PopData [Internet]. 2016 [cited 2018 Jan 19]. Available from: https://www.popdata.bc.ca/aboutus 97.  WorkSafeBC [creator]. WorkSafeBC Claims and Firm Level Files. Population Data BC [publisher]. Data Extract. WorkSafeBC (2016). http://www.popdata.bc.ca/data. 2016.  98.  BC Ministry of Health [creator]. Consolidation File (MSP Registration & Premium Billing). Population Data BC [publisher]. Data Extract. MOH (2016). http://www.popdata.bc.ca/data. 2017.  99.  BC Ministry of Health [creator]. Medical Services Plan (MSP) Payment Information File. Population Data BC [publisher]. Data Extract. MOH(2016). http://www.popdata.bc.ca/data. 2017.  100.  Canadian Institute for Health Information [creator]. Discharge Abstract Database (Hospital Separations). Population Data BC [publisher]. Data Extract. MOH (2016). http://www.popdata.bc.ca/data. 2017.  101.  BC Ministry of Health [creator]. PharmaNet. BC Ministry of Health [publisher]. Data Extract. Data Stewardship Committee (2016). http://www.popdata.bc.ca/data. 2017.  102.  World Health Organization. Anatomical Therapeutic Chemical (ATC) Classification [Internet]. WHO Collaborating Centre for Drug Statistics Methodology. 2011 [cited 2018 Jan 9]. Available from: https://www.whocc.no/atc/structure_and_principles/ 103.  Government of Canada - Health Canada. Read Me File - Drug Product Database (DPD) Data Extract - Health Canada [Internet]. 2017. Available from: https://www.canada.ca/en/health-canada/services/drugs-health-products/drug-products/drug-product-database/read-file-drug-product-database-data-extract.html 104.  Chamberlayne R, Green B, Barer ML, Hertzman C, Lawrence WJ, Sheps SB. Creating a population-based linked health database: a new resource for health services research. Can J Public Heal. 1998;89(4):270–3.  105.  Population Data BC. Data linkage [Internet]. 2016 [cited 2018 Jan 22]. Available from: https://www.popdata.bc.ca/datalinkage 106.  Byrne N, Regan C, Howard L. Administrative registers in psychiatric research: a systematic review of validity studies. Acta Psychiatr Scand. 2005;112(6):409–14.  107.  Marrie RA, Fisk JD, Stadnyk KJ, Tremlett H, Wolfson C, Warren S, et al. Performance of   179 administrative case definitions for comorbidity in multiple sclerosis in Manitoba and Nova Scotia. Chronic Dis Inj Can. 2014;34(2–3):145–53.  108.  Kisely S, Lin E, Gilbert C, Smith M, Campbell L-A, Vasiliadis H-M. Use of administrative data for the surveillance of mood and anxiety disorders. Aust N Z J Psychiatry. 2009;43(12):1118–25.  109.  Marrie RA, Yu BN, Leung S, Elliott L, Caetano P, Warren S, et al. The utility of administrative data for surveillance of comorbidity in multiple sclerosis: a validation study. Neuroepidemiology. 2013;40(2):85–92.  110.  Chronic Disease Information Working Group. BC chronic disease and selected procedure case definitions [Internet]. British Columbia Ministry of Health. 2015 [cited 2018 Jul 16]. p. 45. Available from: https://www2.gov.bc.ca/assets/gov/health/conducting-health-research/data-access/chronic-disease-registries-case-definitions.pdf 111.  Puyat JH, Kazanjian A, Goldner EM, Wong H. How often do individuals with major depression receive minimally adequate treatment? A population-based, data linkage study. Can J Psychiatry. 2016;1–11.  112.  Puyat JH, Marhin WW, Etches D, Wilson R, Martin RE, Kaur SK, et al. Estimating the prevalence of depression from EMRs. Can Fam Physician. 2013;59:2013.  113.  Marrie RA, Fisk JD, Yu BN, Leung S, Elliott L, Caetano P, et al. Mental comorbidity and multiple sclerosis: validating administrative data to support population-based surveillance. BMC Neurol. 2013;13(Cidi):16.  114.  Marrie RA, Walker JR, Graff LA, Lix LM, Bolton JM, Nugent Z, et al. Performance of administrative case definitions for depression and anxiety in inflammatory bowel disease. J Psychosom Res. 2016;89:107–13.  115.  West SL, Richter A, Melfi CA, McNutt M, Nennstiel ME, Mauskopf JA. Assessing the Saskatchewan database for outcomes research studies of depression and its treatment. J Clin Epidemiol. 2000;53(8):823–31.  116.  Williamson T, Green ME, Birtwhistle R, Khan S, Garies S, Wong ST, et al. Validating the 8 CPCSSN case definitions for chronic disease surveillance in a primary care database of electronic health records. Ann Fam Med. 2014;12(4):367–72.  117.  Katzman MA, Bleau P, Blier P, Chokka P, Kjernisted K, Van Ameringen M, et al. Canadian clinical practice guidelines for the management of anxiety, posttraumatic stress   180 and obsessive-compulsive disorders. BMC Psychiatry. 2014;14 Suppl 1(Suppl 1):S1.  118.  Lam RW, McIntosh D, Wang J, Enns MW, Kolivakis T, Michalak EE, et al. Canadian Network for Mood and Anxiety Treatments (CANMAT) 2016 clinical guidelines for the management of adults with major depressive disorder: Section 1. Disease burden and principles of care. Can J Psychiatry. 2016;61(9):510–23.  119.  Jones W. Coding of depression by physicians in B.C. Vancouver British Columbia: Mental Health Evaluation and Community Consultation Unit UBC & Centre for Health Evaluation and Outcome Sciences; 2002.  120.  Schaffer A, Cairney J, Cheung A, Veldhuizen S, Levitt A. Community survey of bipolar disorder in Canada: lifetime prevalence and illness characteristics. Can J Psychiatry. 2006;51(1):1–16.  121.  Kisely S, Lin E, Lesage A, Gilbert C, Smith M, Campbell LA, et al. Use of administrative data for the surveillance of mental disorders in 5 provinces. Can J Psychiatry. 2009;54(8):571–5.  122.  Wong J, Motulsky A, Abrahamowicz M, Eguale T, Buckeridge DL, Tamblyn R. Off-label indications for antidepressants in primary care: descriptive study of prescriptions from an indication based electronic prescribing system. BMJ. 2017;356:j603.  123.  O’Brien PL, Cummings N, Mark TL. Off-label prescribing of psychotropic medication, 2005–2013: an examination of potential influences. Psychiatr Serv. 2017;68(6):549–58.  124.  Furukawa TA, Streiner D, Young LT, Kinoshita Y. Antidepressants plus benzodiazepines for major depression. Cochrane Database Syst Rev. 2001;3:CD001026.  125.  Studney DR, Hakstian AR. A comparison of medical record with billing diagnostic information associated with ambulatory medical care. Am J Public Heal. 1981;71(2):145–9.  126.  Solberg LI, Engebretson KI, Sperl-Hillen JM, Hroscikoski MC, O’Connor PJ. Are claims data accurate enough to identify patients for performance measures or quality improvement? The case of diabetes, heart disease, and depression. Am J Med Qual. 2006;21(4):238–45.  127.  Mahar AL. The impact of severe psychiatric illness on a cancer diagnosis, treatment, and survival. Department of Public Health Sciences, Queen’s University (Doctoral dissertation); 2017.    181 128.  Puyat JH, Kazanjian A, Wong H, Goldner E, Sc MH, Ph D, et al. Comorbid chronic general health conditions and depression care: a population-based analysis. Psychiatr Serv. 2017;68(9):907–15.  129.  Bilsker D, Gloldner EM, Jones W. Health service patterns indicate potential benefit of supported self-management for depression in primary care. Can J Psychiatry. 2007;52(2):86–95.  130.  Doktorchik C, Patten S, Eastwood C, Peng M, Chen G, Beck CA, et al. Validation of a case definition for depression in administrative data against primary chart data as a reference standard. BMC Psychiatry. 2019;19(1):1–8.  131.  Dersh J, Mayer T, Theodore BR, Polatin P, Gatchel RJ. Do psychiatric disorders first appear preinjury or postinjury in chronic disabling occupational spinal disorders? Spine (Phila Pa 1976). 2007;32(9):1045–51.  132.  Lippel K. Workers describe the effect of the workers’ compensation process on their health: A Québec study. Int J Law Psychiatry. 2007;30(4–5):427–43.  133.  Dembe  a E. The social consequences of occupational injuries and illnesses. Am J Ind Med. 2001;40(4):403–17.  134.  Sareen J, Cox BJ, Clara I, Asmundson GJG. The relationship between anxiety disorders and physical disorders in the U.S. National Comorbidity Survey. Depress Anxiety. 2005;21(4):193–202.  135.  United Nations Economic Commission for Europe (UNECE). Developing gender statistics: a practical tool. Geneva, Switzerland; 2010.  136.  Cameron CM, Purdue DM, Kliewer E V, McClure RJ. Differences in prevalence of pre-existing morbidity between injured and non-injured populations. Bull World Health Organ. 2005;83(5):345–52.  137.  Gabbe BJ, Harrison JE, Lyons RA, Edwards ER, Cameron PA. Comparison of measures of comorbidity for predicting disability 12-months post-injury. BMC Health Serv Res. 2013;13(1):30.  138.  Burnham KP, Anderson DR. Multimodel inference understanding AIC and BIC in model selection. Sociol Methods Res. 2004;33(2):261–304.  139.  SAS 9.4 Software | SAS [Internet]. [cited 2014 Dec 9]. Available from: http://www.sas.com/en_us/software/sas9.html   182 140.  Casey R, Ballantyne PJ. Diagnosed chronic health conditions among injured workers with permanent impairments and the general population. J Occup Environ Med. 2017;59(5):486–96.  141.  McGee RE, Thompson NJ. Unemployment and depression among dmerging adults in 12 states, behavioral risk factor surveillance system, 2010. Prev Chronic Dis. 2015;12:140451.  142.  Jones W, Tallon C. Review of Phase 1 Findings from the Provincial Working Group: Needs-based Planning of Mental Health and Substance Use Services in British Columbia. Vancouver, BC; 2018.  143.  Peen J, Ra S, At B, The DJ. The current status of urban-rural differences in psychiatric disorders. 2010;(1):84–93.  144.  Weich S, Twigg L, Lewis G. Rural/non-rural differences in rates of common mental disorders in Britain: prospective multilevel cohort study. Br J Psychiatry. 2006;188:51–7.  145.  Breslau J, Marshall GN, Pincus HA, Brown RA. Are mental disorders more common in urban than rural areas of the United States? J Psychiatr Res. 2014;56:50–5.  146.  Leipert BD, Reutter L. Women’s health in northern British Columbia: the role of geography and gender. Can J Rural Med. 2005;10(4):241–53.  147.  Kazyak E. Midwest or lesbian? Gender, rurality, and sexuality. Gend Soc. 2012;26(6):825–48.  148.  Evenson RJ, Simon RW. Clarifying the relationship between parenthood and depression. J Health Soc Behav. 2005;46(4):341–58.  149.  Umberson D, Pudrovska T, Reczek C. Parenthood, childlessness, and well-being: a life course perspective. J Marriage Fam. 2010;72(2):612–29.  150.  Scott KM, Wells JE, Angermeyer M, Brugha TS, Bromet E, Demyttenaere K, et al. Gender and the relationship between marital status and first onset of mood, anxiety and substance use disorders. Psychol Med. 2010;40(9):1495–505.  151.  Simon RW. Revisiting the relationships among gender, marital status, and mental health. Am J Sociol. 2002;107(4):1065–96.  152.  Lamb KA, Lee GR, Demaris A. Union formation and depression: selection and relationship effects. Natl Counc Fam Relations. 2003;65(4):953–62.  153.  Grav S, Hellzèn O, Romild U, Stordal E. Association between social support and   183 depression in the general population: The HUNT study, a cross-sectional survey. J Clin Nurs. 2012;21(1–2):111–20.  154.  Nomaguchi KM, Milkie MA, Bianchi SM. Time strains and psychological well-being do dual-earner mothers and fathers differ? J Fam Issues. 2005;26(6):756–92.  155.  Rhudy JL, Meagher MW. Fear and anxiety: divergent effects on human pain thresholds. Pain. 2000;84(1):65–75.  156.  Wong WS, Lam HM, Chow YF, Chen PP, Lim HS, Wong S, et al. The effects of anxiety sensitivity, pain hypervigilance, and pain catastrophizing on quality of life outcomes of patients with chronic pain: a preliminary, cross-sectional analysis. Qual Life Res. 2014;23(8):2333–41.  157.  Taylor S, Koch WJ. Anxiety disorders due to motor vehicule accidents: nature and treatment. Clin Psychol Rev. 1995;15(8):721–38.  158.  Choi ES, Jung HS, Kim SH, Park H. The influence of workplace violence on work-related anxiety and depression experience among Korean employees. J Korean Acad Nurs. 2010;40(5):650–61.  159.  Mayou R, Bryant B, Ehlers A. Prediction of psychological outcomes one year after a motor vehicle accident. Am J Psychiatry. 2001;158(8):1231–8.  160.  Fitzharris M, Fildes B, Charlton J. Anxiety, acute- and post-traumatic stress symptoms following involvement in traffic crashes. 50th Annu Proc Assoc Adv Automot Med. 206AD;50:297–315.  161.  Chapman DP, Perry GS, Strine TW. The vital link between chronic disease and depressive disorders. Prev Chronic Dis. 2005;2(1):A14.  162.  Scott KM, Bruffaerts R, Tsang A, Ormel J, Alonso J, Angermeyer MC, et al. Depression-anxiety relationships with chronic physical conditions: results from the World Mental Health surveys. J Affect Disord. 2007;103(1–3):113–20.  163.  Wells KB, Golding JM, Burnam MA. Affective, substance use, and anxiety disorders in persons with arthritis, diabetes, heart disease, high blood pressure, or chronic lung conditions. Gen Hosp Psychiatry. 1989;11(5):320–7.  164.  Kroenke K, Jackson JL. Depressive and anxiety disorders in patients presenting with physical complaints: clinical predictors and outcome. Am J Med. 1997;103:339–47.  165.  Kessler RC, Chiu WT, Demler O, Walters EE. Prevalence, severity, and comorbidity of   184 12-Month DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62(6):617–27.  166.  Grant BF, Stinson FS, Dawson DA, Choi P, Dufour MC, Compton W, et al. Prevalence and co-occurrence of substance use disorders and independent mood and anxiety disorders. Arch Gen Psychiatry. 2004;61:807–16.  167.  Lipscomb HJ, Cameron W, Silverstein B. Incident and recurrent back injuries among union carpenters. Occup Environ Med. 2008;65(12):827–34.  168.  Wulsin L, Alterman T, Timothy Bushnell P, Li J, Shen R, Bushnell PT, et al. Prevalence rates for depression by industry: a claims database analysis. Soc Psychiatr Psychiatr Epidemiol. 2014;49(11):1805–21.  169.  Wang J, Smailes E, Sareen J, Schmitz N, Fick G, Patten S. Three job-related stress models and depression: a population-based study. Soc Psychiatry Psychiatr Epidemiol. 2012;47(2):185–93.  170.  Urbanoski K, Inglis D, Veldhuizen S. Service use and unmet needs for substance use and mental disorders in Canada. Can J Psychiatry. 2017;62(8):551–9.  171.  Vermani M, Marcus M, Katzman MA. Rates of detection of mood and anxiety disorders in primary aare. Prim Care Companion CNS Disord. 2011;13(2):e1–10.  172.  Pelletier L, O’Donnell S, Dykxhoorn J, McRae L, Patten SB. Under-diagnosis of mood disorders in Canada. Epidemiol Psychiatr Sci. 2017;26(4):414–23.  173.  Scott KM. Co-morbidity of mental and physical conditions. Second Edi. Vol. 4, International Encyclopedia of the Social & Behavioral Sciences: Second Edition. Elsevier; 2015. 390–393 p.  174.  Stein MB, Roy-byrne PP, Craske MG, Bystritsky A, Sullivan G, Pyne JM, et al. Functional impact and health utility of anxiety disorders in primary care outpatients. Med Care. 2005;43(12):1164–70.  175.  Stein MB, Cox BJ, Afifi TO, Belik S-L, Sareen J. Does co-morbid depressive illness magnify the impact of chronic physical illness? a population-based perspective. Psychol Med. 2006 May;36(5):587–96.  176.  Iles RA, Davidson M, Taylor NF. Psychosocial predictors of failure to return to work in non-chronic non-specific low back pain: a systematic review. Occup Env Med. 2008;65:507–17.    185 177.  Laisné F, Lecomte C, Corbière M. Biopsychosocial determinants of work outcomes of workers with occupational injuries receiving compensation: a prospective study. Work. 2013 Jan;44(2):117–32.  178.  Shaw WS, Pransky G, Fitzgerald TE. Early prognosis for low back disability: intervention strategies for health care providers. Disabil Rehabil. 2001;23(18):815–28.  179.  Steenstra IA, Verbeek JH, Heymans MW, Bongers PM. Prognostic factors for duration of sick leave in patients sick listed with acute low back pain: a systematic review of the literature. Occup Environ Med. 2005;62(12):851–60.  180.  Kuijer W, Groothoff JW, Brouwer S, Geertzen JHB, Dijkstra PU. Prediction of sickness absence in patients with chronic low back pain: a systematic review. J Occup Rehabil. 2006;16(3):430–58.  181.  Kent PM, Keating JL. Can we predict poor recovery from recent-onset nonspecific low back pain? A systematic review. Man Ther. 2008;13(1):12–28.  182.  Clay FJ, Newstead S V., McClure RJ. A systematic review of early prognostic factors for return to work following acute orthopaedic trauma. Injury. 2010;41(8):787–803.  183.  Cancelliere C, Donovan J, Stochkendahl MJ, Biscardi M, Ammendolia C, Myburgh C, et al. Factors affecting return to work after injury or illness: best evidence synthesis of systematic reviews. Chiropr Man Therap. 2016;24(1):32.  184.  Zieger M, Luppa M, Meisel HJ, Günther L, Winkler D, Toussaint R, et al. The impact of psychiatric comorbidity on the return to work in patients undergoing herniated disc surgery. J Occup Rehabil. 2011;21(1):54–65.  185.  Lavoie KL, Favreau H, Paine NJ, Lemière C, Joseph M, Gagnon-Chauvin A, et al. Prospective impact of psychiatric disorders on employment status and health care use in patients investigated for occupational asthma. J Occup Environ Med. 2016;58(12):1196–201.  186.  Lim D, Sanderson K, Andrews G. Lost productivity among full-time workers with mental disorders. J Ment Health Policy Econ. 2000;3(3):139–46.  187.  Jefferis BJ, Nazareth I, Marston L, Moreno-Kustner B, Bellón J ángel, Svab I, et al. Associations between unemployment and major depressive disorder: evidence from an international, prospective study (the Predict Cohort). Soc Sci Med. 2011;73(11):1627–34.  188.  Waghorn G, Chant D, White P, Whiteford H. Disability, employment and work   186 performance among people with ICD-10 anxiety disorders. Aust New Zeal J Psychiatry. 2005;39(1/2):55-66 12p.  189.  Scott KM. Sex differences in the disability associated with mental disorders. Curr Opin Psychiatry. 2011;24(4):331–5.  190.  Dewa CS, Goering P, Lin E, Paterson M. Depression-related short-term disability in an employed population. JOEM. 2002;44(7):628–33.  191.  Moyser M, Burlock A. Time use: Total work burden, unpaid work, and leisure. Ottawa: Statistics Canada; 2018.  192.  Cox DR. Regression models and life-tables. J R Stat Soc. 1972;34(2):187–220.  193.  Zhang X, Loberiza FR, Klein JP, Zhang MJ. A SAS macro for estimation of direct adjusted survival curves based on a stratified Cox regression model. Comput Methods Programs Biomed. 2007;88(2):95–101.  194.  Li R, Chambless L. Test for additive interaction in proportional hazards models. Ann Epidemiol. 2007;17(3):227–36.  195.  Hirschfeld RMA. The comorbidity of major depression and anxiety disorders: recognition and management in primary care. Prim Care Companion J Clin Psychiatry. 2001;3(6):244–54.  196.  Maas E, Koehoorn M, Mcleod CB. Descriptive epidemiology of gradual return to work for workers with a work-acquired musculoskeletal disorder in British Columbia, Canada. J Occup Rehabil. 2019;(Publish ahead of print).  197.  Gagné S, Vasiliadis H, Préville M. Gender differences in general and specialty outpatient mental health service use for depression. BMC Psychiatry. 2014;14(135):1–11.  198.  Smith KLW, Matheson FI, Moineddin R, Dunn JR, Lu H, Cairney J, et al. Gender differences in mental health service utilization among respondents reporting depression in a national health survey. Health (Irvine Calif). 2013;5(10):1561–71.  199.  O’Donnell S, Syoufi M, Jones W, Bennett K. Use of medication and psychological counselling among Canadians with mood and/or anxiety disorders. Heal Promot Chronic Dis Prev Canada. 2017;37(5):160–71.  200.  Carroll LJ, Cassidy JD, Côté P. Frequency, timing, and course of depressive symptomatology after whiplash. Spine (Phila Pa 1976). 2006;31(16):E551–6.  201.  Park SK. Exploration of the prevalence and correlates of depression among South Korean   187 workers with injuries. Work. 2011;39:345–51.  202.  Pransky G, Gatchel R, Linton SJ, Loisel P. Improving return to work research. J Occup Rehabil. 2005;15(4):453–7.  203.  Abenhaim L, Suissa S, Rossignol M. Risk of recurrence of occupational back pain over three year follow up. Br J Ind Med. 1988;45(12):829–33.  204.  Young AE, Wasiak R, Gross DP. Recurrence of work-related low back pain and disability: association between self-report and workers’ compensation data. Spine (Phila Pa 1976). 2013;38(26):2279–86.  205.  Wasiak R, Kim J, Pransky G. Work disability and costs caused by recurrence of low back pain: longer and more costly than in first episodes. Spine (Phila Pa 1976). 2006;31(2):219–25.  206.  Ricci JA, Stewart WF, Chee E, Leotta C, Foley K, Hochberg MC. Back pain exacerbations and lost productive time costs in United States workers. Spine (Phila Pa 1976). 2006;31(26):3052–60.  207.  Bültmann U, Franche R-L, Hogg-johnson S, Côté P, Lee H, Severin C, et al. Health status, work limitations, and return-to-work trajectories in injured workers with musculoskeletal disorders. Qual Life Res. 2007;16(7):1167–78.  208.  Marras WS, Ferguson SA, Burr D, Schabo P, Maronitis A. Low back pain recurrence in occupational environments. Spine (Phila Pa 1976). 2007;32(21):2387–97.  209.  Wasiak R, Pransky GS, Webster BS. Methodological challenges in studying recurrence of low back pain. J Occup Rehabil. 2003;13(1):21–31.  210.  de Vet HCW, Heymans MW, Dunn KM, Pope DP, Beek AJ Van Der, Macfarlane GJ, et al. Episodes of low back pain: a proposal for uniform definitions to be used in research. Spine (Phila Pa 1976). 2002;27(21):2409–16.  211.  Amorin LD, Cai J. Modelling recurrent events: a tutorial for analysis in epidemiology. Int J Epidemiol. 2015;44(1):324–33.  212.  Navarro A, Reis RJ, Martin M. Some alternatives in the statistical analysis of sickness absence. Am J Ind Med. 2009;52(10):811–6.  213.  Van Vilsteren M, Van Oostrom SH, De Vet HCW, Franche RL, Boot CRL, Anema JR. Workplace interventions to prevent work disability in workers on sick leave. Cochrane Database Syst Rev. 2015;(10. Art. No.: CD006955):1–92.    188 214.  Anema JR, Steenstra IA, Bongers PM, de Vet HCW, Knol DL, Loisel P, et al. Multidisciplinary rehabilitation for subacute low back pain: graded activity or workplace intervention or both? a randomized controlled trial. Spine (Phila Pa 1976). 2007;32(3):291–8.  215.  Knekt P, Lindfors O, Laaksonen MA, Renlund C, Haaramo P, Härkänen T, et al. Quasi-experimental study on the effectiveness of psychoanalysis, long-term and short-term psychotherapy on psychiatric symptoms, work ability and functional capacity during a 5-year follow-up. J Affect Disord. 2011;132(1–2):37–47.  216.  Linden M, Zubrägel D, Bär T. Occupational functioning, sickness absence and medication utilization before and after cognitive-behaviour therapy for generalized anxiety disorders. Clin Psychol Psychother. 2011;18(3):218–24.  217.  Ten Have M, Nuyen J, Beekman A, de Graaf R. Common mental disorder severity and its association with treatment contact and treatment intensity for mental health problems. Psychol Med. 2013;43(10):2203–13.  218.  Brookhart MA, Schneeweiss S. Preference-based instrumental variable methods for the estimation of treatment effects: assessing validity and interpreting results. Int J Biostat. 2007;3(1):1–19.  219.  Miranda J, Schoenbaum M, Sherbourne C, Duan N, Wells K. Effects of primary care depression treatment on minority patients’ clinical status and employment. Arch Gen Psychiatry. 2004;61(8):827–34.  220.  Schoenbaum M, Unutzer J, McCaffrey D, Duan N, Sherbourne C, Wells KB. The effects of primary care depression treatment on patients’ clinical status and employment. Health Serv Res. 2002;37(5):1145–58.  221.  Davies NM, Gunnell D, Thomas KH, Metcalfe C, Windmeijer F, Martin RM. Physicians’ prescribing preferences were a potential instrument for patients’ actual prescriptions of antidepressants. J Clin Epidemiol. 2013;66(12):1386–96.  222.  Carroll R, Metcalfe C, Steeg S, Davies NM, Cooper J, Kapur N, et al. Psychosocial assessment of self-harm patients and risk of repeat presentation: an instrumental variable analysis using time of hospital presentation. PLoS One. 2016;11(2):1–13.  223.  Uddin MJ, Groenwold RHH, de Boer A, Gardarsdottir H, Martin E, Candore G, et al. Instrumental variables analysis using multiple databases: an example of antidepressant use   189 and risk of hip fracture. Pharmacoepidemiol Drug Saf. 2016;25(Suppl. 1):122–31.  224.  Chapdelaine A, Carrier JD, Fournier L, Duhoux A, Roberge P. Treatment adequacy for social anxiety disorder in primary care patients. PLoS One. 2018;13(11):1–15.  225.  Palmer C. Interpretation of B in log-linear models [Internet]. UC Berkley Teaching Resources. 2011 [cited 2019 Aug 28]. p. 1–2. Available from: https://faculty.haas.berkeley.edu/palmer/beta_in_log-linear_regression.pdf 226.  R Topics Documented: Package ‘AER’ [Internet]. 2018 [cited 2019 May 30]. Available from: https://cran.r-project.org/web/packages/AER/AER.pdf 227.  Ertefaie A, Small DS, Flory JH, Hennessy S. A tutorial on the use of instrumental variables in pharmacoepidemiology. Pharmacoepidemiol Drug Saf. 2017;26:357–67.  228.  Zhang D. A coefficient of determination for generalized linear models. Am Stat. 2017;71(4):310–6.  229.  Staiger D, Stock JH. Instrumental variables regression with weak instruments. Econometrica. 1997;65(3):557.  230.  Boef AGC, Dekkers OM, Vandenbroucke JP, Le Cessie S. Sample size importantly limits the usefulness of instrumental variable methods, depending on instrument strength and level of confounding. J Clin Epidemiol. 2014;67(11):1258–64.  231.  Gérard A, Liard F, Crochard A, Goni S, Millet B. Disability in patients consulting for anxiety or mood disorders in primary care: response to antidepressant treatment. Neuropsychiatr Dis Treat. 2012;8:605–14.  232.  Rollman BL, Belnap BH, Mazumdar S, Houck PR, Zhu F, Gardner W, et al. A randomized trial to improve the quality of treatment for panic and generalized anxiety disorders in primary care. Arch Gen Psychiatry. 2005;62:1332–41.  233.  Otto MW, Smits JAJ, Reese HE. Combined psychotherapy and pharmacotherapy for mood and anxiety disorders in adults: review and analysis. Clin Psychol Sci Pract. 2005;12(1):72–86.  234.  Fishbain D. Evidence-based data on pain relief with antidepressants. Ann Med. 2000;32(5):305–16.  235.  Bulloch AGM, Patten SB. Non-adherence with psychotropic medications in the general population. Soc Psychiatry Psychiatr Epidemiol. 2010;45(1):47–56.  236.  Kinrys G, Wygant LE. Anxiety disorders in women: does gender matter to treatment? Rev   190 Bras Psiquiatr. 2005;27(Supl II):43–50.  237.  McLean CP, Asnaani A, Litz BT, Hofmann SG. Gender differences in anxiety disorders: Prevalence, course of illness, comorbidity and burden of illness. J Psychiatr Res. 2011;45(8):1027–35.  238.  Wang J. Work stress as a risk factor for major depressive episode(s). Psychol Med. 2005;35(6):865–71.  239.  Krause N, Frank JW, Dasinger LK, Sullivan TJ, Sinclair SJ. Determinants of duration of disability and return to work after work related injury and illness: challenges for future research. Am J Ind Med. 2001;40(4):464–84.  240.  BC Centre for Disease Control. Chronic disease dashboard [Internet]. [cited 2019 Aug 11]. Available from: http://www.bccdc.ca/health-professionals/data-reports/chronic-disease-dashboard 241.  Corbière M, Sullivan MJL, Stanish WD, Adams H. Pain and depression in injured workers and their return to work: A longitudinal study. Can J Behav Sci / Rev Can des Sci du Comport. 2007;39(1):23–31.  242.  Steele L, Dewa C, Lee K. Socioeconomic status and self-reported barriers to mental health service use. Can J Psychiatry. 2007;52(3):201–6.    191 Appendix  Table 27: Median number of each type of mental health service event in the year before injury by those that had at least one   Table 28: The proportion of claims with at least one of each type of mental health service event in the year before injury by anxiety and depression case status (based on the primary case definitions) in the year before injury MH service event  Case status in the year before injury Case-like (n=25,591) Pure anx (n=16,237) Pure dep (n=11,187) Anx & dep (n=29,800) Anx diag (physician visit) 21.5 72.7 1.7 38.8 Dep diag (physician visit) 9.2 2.8 82.4 52.4 50b (physician visit) 13.3 11.9 26.5 50.4 Antidepressant  32.2 23.6 70 76.9 Anxiolytic 28.3 37.7 11.7 29.3 Anx diag (hospital visit) 0 0.3 0 1.1 Dep diag (hospital visit) 0 0 0.7 1.4 MH service Total number events Median IQR Max Anx diag (physician visit) 58,515 1 1, 2 98 Dep diag (physician visit) 77,841 2 1, 3 99 50b (physician visit) 37,191 1 1, 2 48 Antidepressant  218,690 3 1, 6 499 Anxiolytic 63,453 1 1, 3 132 Anx diag (hospital visit) 694 2 1, 2 10 Dep diag (hospital visit) 933 1 1, 2 22   192  Table 29: Median number of each type of mental health service event in the year before injury by assigned case status (based on the primary case definitions) in the year before injury MH service Case status Case-like (n=25,591) Pure anx (n=16,237) Pure dep (n=11,187) Anx & dep (n=29,800) Median IQR Max Median IQR Max Median IQR Max Median IQR Max Anx diag (physician visit) 0 0, 0 1 1 0, 2 52 0 0, 0 1 0 0, 1 98 Dep diag (physician visit) 0 0, 0 1 0 0, 0 1 1 1, 3 75 1 0, 2 99 50b (physician visit) 0 0, 0 1 0 0, 0 1 0 0, 0 1 1 0, 1 48 Antidepressant  0 0, 1 159 0 0, 0 56 2 0, 5 289 3 1, 6 499 Anxiolytic 0 0, 1 63 0 0, 1 61 0 0, 0 37 0 0, 1 132 Anx diag (hospital visit) 0 0, 0 0 0 0, 0 2 0 0, 0 0 0 0, 0 10 Dep diag (hospital visit) 0 0, 0 0 0 0, 0 0 0 0, 0 17 0 0, 0 22      193 Table 30: Unadjusted estimates of the association between anxiety and depression in the year before injury and the return to work and recurrence events lasting 7 days or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013  Men  Women  Anx only Dep only Anx & dep  Anx only Dep only Anx & dep Return to work PWP TT             1st RTW HR  0.86 (0.81-0.90) 0.93 (0.88-0.99) 0.90 (0.87-0.94)  0.95 (0.91-0.99) 0.97 (0.92-1.02) 0.93 (0.90-0.96)      2nd RTW HR 0.80 (0.58-1.10) 0.62 (0.42-0.91) 0.86 (0.69-1.08)  1.13 (0.92-1.39) 1.62 (1.23-2.14) 1.23 (1.05-1.44)      Common HR 0.86 (0.81-0.90) 0.92 (0.87-0.98) 0.90 (0.87-0.94)  0.95 (0.91-0.99) 0.98 (0.93-1.03) 0.94 (0.91-0.97) PWP GT             1st RTW HR  0.86 (0.81-0.90) 0.93 (0.88-0.99) 0.90 (0.87-0.94)  0.95 (0.91-0.99) 0.97 (0.92-1.02) 0.93 (0.90-0.96)      2nd RTW HR 0.88 (0.64-1.21) 0.70 (0.47-1.02) 0.87 (0.69-1.08)  1.03 (0.82-1.28) 1.34 (1.00-1.79) 1.04 (0.87-1.25)      Common HR 0.86 (0.82-0.90) 0.93 (0.87-0.98) 0.90 (0.87-0.94)  0.95 (0.91-0.99) 0.98 (0.93-1.03) 0.94 (0.91-0.97) Recurrence Cox TT HR 1.32 (1.01-1.73) 1.26 (0.93-1.71) 1.32 (1.01-1.73)  1.34 (1.09-1.65) 1.07 (0.82-1.40) 1.17 (1.00-1.38) Cox GT HR 1.25 (1.96-1.63) 1.20 (0.89-1.63) 1.44 (1.20-1.74)  1.31 (1.06-1.60) 1.05 (0.80-1.37) 1.16 (0.98-1.36)  Table 31: Adjusted estimates of the association between anxiety and depression in the year before injury and the return to work and recurrence events lasting 7 days or longer for lost-time upper limb and spine strain or sprain claims in BC from 2009 to 2013  Men Women  Anx only Dep only Anx & dep Anx only Dep only Anx & dep Return to work PWP TT            1st RTW HR  0.88 (0.84-0.93) 0.96 (0.91-1.02) 0.94 (0.90-0.98) 0.97 (0.93-1.01) 0.99 (0.94-1.04) 0.96 (0.93-0.99)      2nd RTW HR 0.80 (0.58-1.10) 0.67 (0.45-0.99) 0.88 (0.69-1.12) 0.95 (0.75-1.20) 1.54 (1.13-2.09) 1.06 (0.88-1.29)      Common HR 0.89 (0.84-0.93) 0.95 (0.90-1.01) 0.94 (0.90-0.98) 0.97 (0.93-1.01) 1.00 (0.95-1.05) 0.96 (0.93-0.99) PWP GT            1st RTW HR  0.88 (0.84-0.93) 0.96 (0.91-1.02) 0.94 (0.90-0.98) 0.97 (0.93-1.01) 0.99 (0.94-1.04) 0.96 (0.93-1.01)      2nd RTW HR 0.90 (0.65-1.24) 0.77 (0.52-1.14) 0.93 (0.73-1.18) 1.02 (0.80-1.28) 1.42 (1.05-1.93) 1.04 (0.86-1.26)      Common HR 0.89 (0.84-0.94) 0.96 (0.90-1.01) 0.94 (0.90-0.98) 0.97 (0.93-1.01) 0.99 (0.95-1.05) 0.96 (0.93-0.99) Recurrence Cox TT HR 1.16 (0.88-1.52) 1.14 (0.84-1.54) 1.26 (1.04-1.53) 1.27 (1.03-1.56) 1.00 (0.77-1.31) 1.10 (0.93-1.30) Cox GT HR 1.12 (0.85-1.46) 1.10 (0.81-1.50) 1.27 (1.04-1.54) 1.26 (1.02-1.55) 1.01 (0.77-1.33) 1.11 (0.94-1.31)   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0387452/manifest

Comment

Related Items