UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Delayed recovery and chronic disability in patients with whiplash-associated disorders Dufton, John Andrew 2003

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

Item Metadata

Download

Media
831-ubc_2004-0009.pdf [ 5.45MB ]
Metadata
JSON: 831-1.0091251.json
JSON-LD: 831-1.0091251-ld.json
RDF/XML (Pretty): 831-1.0091251-rdf.xml
RDF/JSON: 831-1.0091251-rdf.json
Turtle: 831-1.0091251-turtle.txt
N-Triples: 831-1.0091251-rdf-ntriples.txt
Original Record: 831-1.0091251-source.json
Full Text
831-1.0091251-fulltext.txt
Citation
831-1.0091251.ris

Full Text

Delayed Recovery and Chronic Disability in Patients with Whiplash-Associated Disorders By John Andrew Dufton B.Sc , Queen's University, 1995 D.C., Canadian Memorial Chiropractic College, 1999 A thesis submitted in partial fulfillment of the requirements for the degree of Master, of Science in The Faculty of Graduate Studies (Department of Health Care and Epidemiology) We accept this thesis as conforming to the required standard The University of British Columbia August, 2003 © John Andrew Dufton, 2003 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t of the requirements f o r an advanced degree at the U n i v e r s i t y of B r i t i s h Columbia, I agree that the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r reference and study. I f u r t h e r agree that permission f o r extensive copying of t h i s t h e s i s f o r s c h o l a r l y purposes may be granted by the head of my department or by h i s or her r e p r e s e n t a t i v e s . I t i s understood that copying or p u b l i c a t i o n of t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be allowed without my w r i t t e n permission.. Department of Health Care and Epidemiology The U n i v e r s i t y of B r i t i s h Columbia Vancouver, Canada Delayed Recovery and Chronic Disability in Patients with Whiplash-Associated Disorders. John Dufton, Masters of Science in Health Care and Epidemiology, Faculty of Graduate Studies, University of British Columbia, 2003. Narrative Abstract Whiplash is a common injury after a motor vehicle collision resulting in significant pain and disability for those injured. The prognosis of these disorders is highly variable and difficult to predict and evidence suggests that both medical and external non-injury related factors are important in determining recovery. This study is an extensive exploratory analysis investigating the association between a number of personal, clinical, and non-injury related factors and delayed clinical improvement after soft-tissue injuries sustained in a motor vehicle collision. Data were collected shortly after injury ensuring each patient was enrolled in the study at a similar point of recovery, and the outcome was measured with a valid and reliable disability questionnaire. The source of the data was the clinical database from a national network of 48 Canadian physiotherapy and rehabilitation facilities. A cohort of 2185 adult patients from this database was assembled for analysis. Multivariate logistic regression analysis revealed eight predictors associated with delayed recovery as measured by a minimal clinical improvement: 1) older age, 2) female gender, 3) increasing lagtime between injury date and presentation for treatment, 4) initial pain location, 5) province of injury, 6) higher initial pain intensity, 7) lawyer retention, and 8) work status (currently working). The variable measuring increasing initial pain intensity interacted with both the lawyer retention and the work status variables. A model i i predicting early improvement was developed and validated. A secondary cross-sectional analysis of the acute and chronic whiplash population suggests that significant differences between these groups are apparent at 3 months and that the acute patient population should be analyzed separately from the chronic patient population. Researchers and clinicians in all jurisdictions should be cognizant of the potential for non-injury related factors to delay recovery, and aware of the interaction between the initial intensity of a patient's pain and other covariates when confirming these results. iii Table of Contents Abstract i i Table of Contents iv List of Tables vii List of Figures i x Chapter 1 Introduction 1 1.1 Study Purpose 3 1.2 Specific Objectives 3 1.3 Thesis Overview 4 Chapter 2 Literature Review 6 2.1 Introduction 6 2.2 Literature Search and Selection of Articles. 7 2.3 Assessment of Methodological Quality 8 2.4 Results of Literature Review 9 2.4.1 Classification of Prognosis Studies 11 2.4.2 Recovery From Acute Whiplash 14 2.4.3 Prognostic Factors for the Recovery of Acute Whiplash Injuries 16 2.4.3.1 Sociodemographic Factors 16 2.4.3.2 Postcollision Signs and Symptoms 17 2.4.3.3 General Health Before Injury 20 2.4.3.4 Legal and Compensation Factors 21 2.4.3.5 Crash-Related Factors 21 2.5 Summary 21 Chapter 3 Background of CBI Materials and Methods 25 3.1 Introduction 25 3.2 Patient Referral Process 25 3.3 Canadian Back Institute Methods 26 3.3.1 Clinical Database 27 3.3.1.1 Spinal Assessment Form 28 3.3.1.2 Patient Questionnaire 28 3.3.1.3 Spinal Discharge Form 30 3.3.2 Data Entry and Collection 30 3.3.3 Treatment 31 3.4 Database Limitations 32 3.5 Summary 33 Chapter 4 Study Design, Materials and Methods 34 4.1 Overall Study Design 34 4.2 Inclusion Criteria 34 4.3 Final Data Set 36 i v 4.4 Data Source 36 4.4.1 Missing Data 38 4.4.2 Imputation 39 4.5 Research Variables 40 4.5.1 Outcome Variable 40 4.5.2 Explanatory Variables 43 4.5.2.1 Personal Characteristics 43 4.5.2.2 Clinical Variables 45 4.5.2.3Treatment Variables 48 4.6 A Comparison of Acute and Chronic Whiplash Associated Disorders 50 4.6.1 Additional Research Variables 50 4.7 Ethics 51 Chapter 5 Statistical Analysis Methods 53 5.1 Inception Cohort 53 5.1.1 Descriptive Statistics - 53 5.1.2 Logistic Regression Modeling Strategy 53 5.1.2.1 Univariable Analysis 53 5.1.2.2 Multivariate Analysis 54 5.2 Model Validation 55 5.4 A Comparison of Acute and Chronic Whiplash Associated Disorders 57 5.5 Missing Values 57 Chapter 6 Results 59 6.1 Cohort Characteristics 59 6.1.1 Personal Characteristics 59 6.1.2 Clinical Characteristics 59 6.1.3 Treatment Related Characteristics 59 6.1.4 Outcome Variable 60 6.2 Univariable and Multivariate (Within Category) Logistic Regression 62 6.2.1 Personal Characteristics 62 6.2.2 Clinical Characteristics 63 6.2.3 Treatment Related Characteristics 64 6.3 Multivariate (All Variable Categories) Logistic Regression 65 6.4 Model Selection and Validation 66 6.5 Final Multivariate Study Model 68 6.5.1 Sensitivity and Specificity of Study Model 69 6.6 Missing Values 70 6.6.1 Inception cohort - Study Population and Excluded Records 70 6.6.2 Missing Covariate Values 71 6.7 Comparison Study of Acute and Chronic Whiplash Associated Disorders 72 6.7.1 Comparison Cohort Characteristics 72 6.7.1.1 Personal Characteristics 73 6.7.1.2 Clinical Characteristics 74 6.7.1.3 Treatment Related Characteristics 74 6.7.1.4 Outcome Variable 76 v 6.7.2 Early Chronic Pain 76 Chapter 7 Discussion 78 7.1 Introduction 78 7.2 Prognostic Variables from Multivariate Logistic Regression 78 7.3 Thesis Strengths and Limitations 88 7.3.1 Strengths • 88 7.3.2 Limitations. 93 7.4 Conclusion 95 References 98 Appendices 105 Appendix A: The Quebec Task Force Classification of Whiplash- 106 Associated Disorders Appendix B: Criteria for the Appraisal of the Methodological Quality of 107 Cohort Studies Appendix C: Canadian Back Institute Spinal Assessment Form - page 1. 108 Appendix D: Canadian Back Institute Questionnaire (CBIQ). 110 Appendix E: Alternate approaches of imputation for missing data values 111 Appendix F: Physical examination conduction tests for each anatomical 112 neurological area Appendix G: Waddell's Non-organic Signs 115 Appendix H : Common techniques of internal validation for logistic regression 116 Appendix I: Table comparing the study population (in the comparison study 117 of acute vs. chronic injury) and those that were excluded due to not responding to either the entry or exit patient questionnaire Appendix J: Comparison of the outcome, personal, clinical and treatment 118 related characteristics in acute, early chronic, and chronic patient populations after a motor vehicle collision. Appendix K : Provincial Variation in the Management of Whiplash Associated 122 Disorders. vi: List of Tables Table 2.1 - Number of articles retrieved based on whiplash and prognosis 10 text word search and their combination from a MEDLINE and C I N A H L January 2000-March 2003 Table 2.2 - Studies excluded based on critical appraisal review criteria 10 Table 2.3 - Classification of the prognostic studies that met the critical 11 appraisal criteria for inclusion in the literature review. Studies classified based on the target population and strength of the evidence Table 2.4 - Methodological quality of cohort studies on the prognosis 12 of acute whiplash Table 2.5 - Summary of design characteristic of prognostic studies of acute 13 whiplash injuries Table 2.6 - Summary of Prognosis of Acute Whiplash Injuries by Target 15 population Table 2.7 - Summary of findings of the two previously published 16 systematic reviews of the whiplash literature Table 2.8 - Summary of prognostic factors for recovery from acute 18 whiplash injuries Table 2.9 - Established and potential prognostic factors in the recovery 22 from an acute whiplash-associated disorder reported in recent and previous literature and systematic reviews Table 4.1 - Subject inclusion criteria 34 Table 6.1 - Baseline characteristics of 2185 patients with acute presentation 60 after a motor vehicle collision Table 6.2 - Crude and adjusted odds ratios for the association of the 63 variables in the personal characteristic category with a negative outcome (< 10% change in questionnaire score) Table 6. 3 - Crude and adjusted odds ratios for the association of the 64 patient's clinical characteristics with a negative outcome (< 10% change in questionnaire score) Table 6.4 - Crude and adjusted odds ratios for the association of the 65 treatment related characteristics with a negative outcome (< 10% change in V l l questionnaire score) Table 6.5 - Adjusted odds ratios for the association of the significant 66 variables from the three category specific models with a negative outcome (< 10% change in questionnaire score) Table 6.6 - Defining characteristics and area under the ROC curve for the 67 competing models Table 6.7- Partitions for the Hosmer and Lemeshow test for the model 68 chosen by validation Table 6.8 - Adjusted odds ratios for the association of the variables in the 69 final study model and negative outcome (< 10% change in questionnaire score) Table 6.9 - Adjusted odds rations for the association of lawyer retention 69 and work status and negative outcome stratified by initial pain intensity Table 6.10- Sensitivity and specificity of the final study model with 70 varying cut points Table 6.11 - Missing Data - Characteristics associated with completion 71 of the initial pain intensity variable Table 6.12 - A comparison of personal characteristics in individuals 73 injured in a motor vehicle collision Table 6.13 - A comparison of clinical and treatment related characteristics 75 in individuals injured in motor vehicle collisions Table 6.14- Comparison of outcomes in individuals injured in a motor 76 vehicle collision (acute vs. chronic populations) viii List of Figures Figure 3.1- Flow chart of patients and data collection procedures through 27 the Canadian Back Institute(CBI) system Figure 4.1 - Final working data set of individuals injured in a motor 37 vehicle collision attending a Canadian Back Institute (CBI) clinic in Canada between 1998-2001 Figure 5.1- Overview of the logistic regression model building strategy 55 Chapter 1 Introduction Whiplash is defined as an acceleration-deceleration mechanism of energy transfer to the neck, usually occurring during a motor vehicle collision (1). The direction of the impact is usually rear-end or side impact but other mechanisms of collision also generate similar forces. Injuries to the neck or other areas of the body are a common occurrence after such a trauma. The clinical manifestations of these injuries have been termed whiplash-associated disorders (WAD) and the list of reported signs and symptoms includes: neck pain, headache, facial pain, temporo-mandibular joint pain, dysphagia, visual disturbances, vertigo, concentration difficulties, interscapular pain, and upper and lower extremity numbness and pain (1). Low back pain is also commonly reported (2, 3). The incidence of whiplash claims has proven to be highly variable in different provinces across Canada. For example, incidence rates have varied from 70 claims per 100,000 persons in Quebec (1) to 900 claims per 100,000 in British Columbia (4). In Saskatchewan, the incidence rate ranged from 302 claims per 100,000 persons under a no-fault insurance system to 700 claims per 100,000 under a tort system (2). Insurance policy is thought to be a major determinant for these observed differences among the Canadian provinces; however, both Quebec and Saskatchewan operate under a no-fault system (where pain and suffering are not compensated) suggesting other important policy and/or cultural influences are at work. The presence and recovery of chronic pain and disability after a motor vehicle collision similarly demonstrates considerable variability. Indeed, depending on the insurance 1 jurisdiction as few as 2% and as many as 28% of those involved in a crash are still compensated one year after their collision (2, 5). Overall, the natural history of the condition is generally thought to be favourable (1) but recent research has highlighted the unpredictable course of recovery (6). The management of whiplash has proven to be challenging for clinicians, and insurers alike. Accumulating evidence suggests that WADs are not only a medical issue but also a condition that is influenced by external non-injury related factors. This increasingly substantial body of research has created considerable controversy, and few health care issues have received more emotionally laden attention. Although physical trauma likely has a role in the expression of disability secondary to whiplash, it is becoming increasingly clear that compensation systems and legal factors also contribute in many cases (2, 6-8). Based on the scientific evidence to date, the Quebec Task Force on Whiplash-Associated Disorders (QTF) (1), and more recently Cote and his colleagues (6) have systematically reviewed the whiplash literature and made a number of recommendations regarding future research. Included in its research agenda, the QTF called for the standardization of data collection, assessment, and treatment procedures, as well as the use of valid and reliable outcome measures, to aid the effective investigation of prognostic factors for the recovery of whiplash (1). Cote et al (6) commented that there is a lack of rigorous investigation of prognostic factors in the primary care settings (physician, 2 physiotherapist, chiropractor, dentist), and that a further understanding of the legal, compensation and cultural forces on recovery is a priority. CBI Health (CBI), a national network of physiotherapy and rehabilitation service providers in Canada, offered the unique opportunity of access to their clinical database to the candidate for the study of delayed recovery and chronic disability among whiplash patients. The advantages of using this database were its large size, broad scope of personal, clinical and treatment-related variables, standardized data collection and treatment procedures, as well as, the diverse geographic distribution of the patients. CBI has an established position providing rehabilitation services for patients with W A D across Canada, as such both the acute and more controversial chronic patient populations seek treatment at their clinics, and are available for study. 1.1 Study Purpose The overall purpose of this study was to investigate which prognostic factors are predictive of a poor response to treatment in the early stages of a whiplash-associated disorder (WAD), and from an extensive exploratory analysis to suggest important risk factors for confirmatory analysis. 1.2 Specific Objectives: • To update previous systematic reviews of the literature on the prognosis of acute whiplash-associated disorders (WAD) and to synthesize the current literature and evaluate the consistency of reported prognostic factors. 3 • To investigate which personal, clinical, and treatment related factors are prognostic markers of delayed early recovery in a cohort of patients presenting for treatment at the secondary care physiotherapy setting. • To investigate in an exploratory manner the effect of legal and compensation issues on the recovery from an acute whiplash injury. • To develop and validate a prognostic model that identifies patients who are likely to demonstrate meaningful improvements early in the course of their recovery in the clinical setting. • To describe extensively the differences between the acute and chronic whiplash populations. 1.3 Thesis Overview Chapter 2 provides a review of the medical literature concerning the prognosis of acute whiplash. This chapter is an update to previous systematic reviews of the whiplash literature and a synthesis of both the recent and previous literature is provided. Chapter 3 provides background information regarding the clinical database of CBI Health that was used in this study. The data collection procedures are reviewed. In addition, the referral process, assessment procedures, treatment and discharge procedures for patients attending CBI are described. Chapter 4 describes the study design, materials and methods including: inclusion criteria, the working dataset, methods for handling missing data, the outcome and explanatory 4 variables and ethics. The primary study and a secondary analysis comparing acute and chronic W A D populations are described. Chapter 5 describes the statistical analysis methods. A description of the logistic regression modeling strategy, model validation, and methods for choosing a final multivariate model are provided. Statistical methods for the secondary analysis comparing the acute and chronic W A D populations are also described. Chapter 6 presents descriptive statistics for the study cohort of patients with acute whiplash injuries. Logistic regression modeling and validation results are presented. A secondary descriptive analysis of patients with chronic whiplash injuries is provided. Chapter 7 contains a discussion of the significant prognostic variables, study strengths and limitations and conclusions. 5 Chapter 2 Literature Review 2.1 Introduction There have been two previous systematic reviews of the whiplash literature. The Quebec Task Force (QTF) on Whiplash-Associated Disorders (WAD) provided a "best evidence synthesis" of the pre-1995 literature (1). The QTF commented that the overall quality of evidence available for evidence-based recommendations was poor; nevertheless, they concluded that the prognosis for whiplash-associated disorders was generally favourable (1). The QTF also developed a classification system of whiplash-associated disorders (Appendix A) that was designed to categorize patients according to the severity of their signs and symptoms post-injury. The system was developed to facilitate the evaluation of research and aid in clinical patient management (1). A n update of the QTF review was published in 2001 by Cote et al (6). Cote and his colleagues observed that the prognosis for whiplash injuries varies considerably depending on the insurance compensation system and the source population considered (6). They further concluded that older age, female gender, baseline neck pain intensity, baseline headache intensity, and radicular signs and symptoms were associated with delayed recovery (6). Cote et al additionally proposed a new conceptual framework for future reviews and for the classification of the whiplash prognosis literature. Using their framework, each study is categorized by the methodological quality (internal validity), the target population and generalizability (external validity), and the strength of the evidence (three mutually exclusive categories) (6). They identified four common source populations in the literature: population-based, insurance-based, hospital-based emergency department, and 6 primary care (physician, physiotherapist, chiropractor) cohorts (6), and suggested that each study be considered relative to the other studies with a common source population. Finally, studies are categorized based on the strength of evidence (exploratory or confirmatory). Three phases of study design (phase I-III) are considered. Phase I studies are exploratory or descriptive studies that generate hypotheses (using univariate analysis) regarding the association of a number of potential prognostic factors and a whiplash recovery outcome (6). Phase II studies are extensive exploratory research that focuses on sets of prognostic factors (using multivariate analysis, thereby controlling for a number of variables simultaneously) and/or attempts to determine which factors are of most prognostic importance (6). Phase III studies are large confirmatory models based on pre-stated hypotheses that investigate the prognostic importance of a particular exposure (6). Phase III studies provide detailed information regarding the independence, strength, and direction of a prognostic factor's association with a whiplash recovery related outcome. For the present research project, an update of the more recent literature on the prognosis of acute whiplash injuries was conducted. 2.2 Literature Search and Selection of Articles. Two electronic databases were searched: M E D L I N E (1966 to March 2003), and CINAHL (1982 to March 2003). Articles published before January 1, 2000 were excluded because they had been previously reviewed (1, 6). The search was based on the keywords: (whiplash injuries, OR neck injuries, OR neck pain), and (predict, OR prognosis, OR prognostic, OR risk factor). A combination of these two searches using 7 the Boolean " A N D " operator was applied to limit the search. A l l searches were limited to the English language. An initial screen of the titles and abstracts was conducted to ensure that all articles were studies of acute whiplash injuries, and not opinion letters or narrative reviews. If the primary subject matter was unclear from the abstract the full article was evaluated to ensure the content concerned acute whiplash injuries. After the initial screen, the remaining articles were further evaluated based on criteria established from previous systematic reviews of the whiplash literature (1,6). Specifically studies were included for further review i f the following criteria were met: 1) cohort or case-control studies on the prognosis of acute whiplash (duration less than 3 months), 2) systematic reviews on whiplash-associated disorders, and 4) publication after January 1, 2000. Studies that included injuries not associated with a motor vehicle collision, patients with severe injuries (such as fracture or dislocation), patients younger than 18 years of age, or less than 20 subjects, as well as opinion papers, were not included. A n assessment of the methodological quality was completed for the included articles. 2.3 Assessment of Methodological Quality The candidate reviewed the articles that met the inclusion criteria. A subset of the critical appraisal criteria presented in Appendix B was used to assess each article for scientific admissibility. These criteria have been used by other authors in the systematic review of the whiplash literature to determine "fatal flaws" in methodology and to identify information and selection biases (6). Specifically each article was evaluated in terms of the following: 1) source population identified in terms of time, place and sampling frame, 2) inclusion and exclusion criteria were adequately described and appropriate, 3) the zero time (or start of follow-up) was identified, 4) the prognostic factors were measured in an 8 appropriate manner, 5) the outcome was adequately defined and appropriately measured, and 6) the overall participation rate was reported to be at least 60% or an analysis of factors associated with participation was conducted. For each of these criteria the study was graded in a yes/no fashion. If any criteria were graded no for a particular study then it was excluded from further review. The articles that met the evaluation criteria were classified according to a new conceptual framework (study population and phase of study) suggested by Cote et al (6) and were evaluated according to a list of criteria described in Appendix B. These criteria are similar to those used by others in the systematic review of observational studies (6, 9). The following characteristics of each study were further scrutinized: source population, sample size, prognostic factors, follow-up time, primary outcome measure, and risk estimates. The results were qualitatively synthesized and no attempts at statistical pooling were made. The results were judged consistent i f at least two studies, or i f 75% of the studies reported similar results. When making a judgment about a particular prognostic factor the two previous systematic reviews were also considered in conjunction with the more recent literature. For example, i f a predictor from one of the previous reviews was considered inconsistent due to limited study, and this same prognostic factor was evaluated in the current literature the results from both the previous review and current literature will be considered to judge the importance of the prognostic factor. 2.4 Results of Literature Review Table 2.1 displays the results of the text word search. 9 Table 2.1 - Number of articles retrieved based on whiplash and prognosis text word search and their combination from a MEDLINE and CINAHL January 2000-March 2003 Database Whiplash papers Prognosis papers Combined (whiplash and prognosis papers) M E D L I N E C I N A H L Text words: 1287 478 Whiplash injuries, neck injuries, neck pain 65397 3955 Predict, prognosis, prognostic, risk factor 108 21 The Boolean "AND" operator applied to the whiplash and prognosis searches After excluding duplicate articles the combined search from M E D L I N E and CINAHL produced a total of 115 articles. The review of the titles and abstracts of these articles resulted in the further exclusion of: 73 articles with subject matter other than whiplash, 16 opinion letters or narrative reviews, 9 articles other than prognostic studies of acute whiplash, and two articles in a previous review. The full text of the remaining 15 articles was retrieved. An additional 5 articles that initially had unclear subject matter (based on the abstract), were excluded after the full text review as articles with subject matter other than whiplash. Ten articles were considered for evaluation. Four articles were further rejected because they did not meet one or more of the critical appraisal criteria (Table 2.2). Table 2.2 - Studies excluded based on critical appraisal review criteria •Review Criteria* First Author 1 2 3 4 5 6 Ovadia(lO) N N Y Y N N Khan (11) Y N N Y Y Y Miettinen(12) Y N N N Y Y Soderlund(13) Y Y N Y Y Y ""Evaluation criteria 1) source population identified, 2) inclusion and exclusion criteria described and appropriate, 3) the zero time was identified, 4) the prognostic factors were measured in an appropriate manner, 5) the outcome was adequately described and appropriately measured, and 6) the overall participation rate was reported to be at least 60% or analysis of participation was included. N=No, Y=Yes. Studies in table evaluated using criteria based on Cote et al (6). 10 2.4.1 Classification of Prognosis Studies Of the 6 papers included in the review, four were cohort studies composed of patients from hospital emergency departments (3, 14-16); however, the study by Sterner et al also included patients from the primary care setting of general practitioners, and two of the studies reported on the same cohort of patients (Table 2.3). General population cohorts were used in two studies (17, 18) (Table 2.3). Table 2.3 - Classification of the prognostic studies that met the critical appraisal criteria for inclusion in the literature review. Studies classified based on the target population and strength of the evidence (phase of investigation) PHASES OF INVESTIGATION TARGET POPULATION Hospital Emergency Department Primary Care Insurance General Population Phase III - Explanatory study (hypothesis testing) Cote et al (3) Phase II - Exploratory study (multi-variable models) Hartlingetal (3, 13), Sterner et al (15), Kyhlback et al (14) Suissa et al (17) Phase I - Descriptive Study (univariable statistics) Five studies (3, 14-16, 18) used multivariate regression analysis to investigate a set of predictor variables and thus were phase II - exploratory studies. Only one study was a phase III confirmatory or explanatory study (17). In this study Cote et al tested the pre-stated hypothesis that investigated whether neck pain intensity, physical functioning and depressive symptomatology were associated with time-to-claim closure. A summary of the methodological quality and the design characteristics of each included study are presented in Table 2.4 and Table 2.5 respectively. 11 Table 2.4 - Methodological quality of cohort studies on the prognosis of acute whiplash Evaluation Criteria* First Author 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Population-based Cote (17) Y Y Y Y Y Y n/a Y Y Y Y Y Y Y Y Y Y Y Y n/a Suissa(18) Y Y Y Y Y n/a n/a Y Y Y n/a Y Y S Y Y Y Y S n/a Hospital-based emergency department Harding (14) Y Y Y Y Y n/a N S S Y n/a Y S S Y Y Y Y S N Harding (3) Y Y Y Y S n/a N Y S Y n/a Y S S S Y S Y S N Kyhlback Y S Y Y Y n/a N S s Y n/a Y S s Y Y Y Y Y N (15) Sterner (16) Y Y Y Y Y n/a N Y s Y n/a N S Y N Y Y Y S N •Evaluation Criteria: 1) research question and objective are clearly defined, 2) source population is identified, 3) inclusion and exclusion criteria are described and appropriate 4) the overall participation rate was reported to be at least 60% or analysis of participation was included, 5) follow-up is reported, explained and reasonable, 6) loss to follow-up is equal in both groups, 7) sample size is pre-planned based on the objective of the study, 8) statistical analysis is appropriate for the objective of the study, 9) adjustment is made for important variables, 10) zero time is identified, 11) baseline comparability of various groups is reported, 12) same data collection procedures for all members of the cohort, 13) important baseline variables are measured, valid, and reliable, 14) all aspects of a prognostic factor are measured (dose, level, duration) and done so adequately (previous, baseline, follow-up), 15) Regular follow-up are accomplished, 16) other prognostic factors are measured, 17) duration of follow-up is adequate for the objective of the study , 18) outcome is defined and measurable, 19) outcome is valid, 20) outcome assessment was blind (Appendix B). N = no, Y = yes, S = substandard, n/a = not applicable Overall the methodological quality of the studies using population-based cohorts was good (Table 2.4). Both studies had large sample sizes and studied a wide variety of prognostic factors. In their study of Quebec residents, Suissa et al did not adjust adequately for the initial whiplash pain intensity, nor did these authors report on the validity of their outcome (18). A l l methodological criteria were achieved by Cote and his colleagues (17). Overall, the methodological quality of the studies using population-based cohorts was better than studies using hospital cohorts. A number of methodological flaws were noted in the emergency department cohorts (Table 2.4). Only the study conducted by Kyhlback et al documented the validity of their outcome measure (15). One of the studies by Hartling and her colleagues (3) as well as the study by Sterner et al (16) failed to regularly follow-up on their subjects. In the study by Hartling et al, a single follow-up at 6 months provided some insight into the early recovery from a whiplash associated 12 disorder, however this was not the stated objective of this article and a longer follow-up is preferred such that overall long-term outcomes may be observed. Sterner et al, also had only a single follow-up, at 16 ± 2 months. Although the duration of follow-up time is adequate in this study, the lack of regular follow-ups could have resulted in the introduction of bias as subjects attributed unrelated or spontaneous spinal pain episodes to their previous motor vehicle collision. The sample sizes of all the hospital-based cohorts were relatively small and none calculated a priori sample size estimates. A l l of the studies used multivariate analysis and identified important prognostic factors; however, not all adjusted for previously identified factors (age, gender, baseline neck and headache intensity, radicular signs and symptoms) that are known to be associated with delayed recovery from acute whiplash (6). Table 2.5 - Summary of design characteristic of prognostic studies of acute whiplash injuries Study Author Source Case Definition Sample Follow-Up Outcome Population Size Measure Population-based Cote et al (17) Province of Whiplash, 18 yrs or 5398 670- 1215 Time-to-claim Saskatchewan, older, report injury to days closure Canada residents, insurance company July 1994-Dec. 1995 Suissa et al Province of Motor vehicle collision. 2843 12 months Time on (18) Quebec, Canada ICD-9 code 847.0, compensation 1997 report injury to insurance company Hospital-based emergency department Hartling et al Kingston, Ontario, Whiplash, 18 yrs or 353 6 months Presence of (3) Canada residents, older, first visit to WAD* Oct. 1995-Mar. emergency department 1998 Hartling et al Kingston, Ontario Whiplash, 18 yrs or 353 6 months Presence of (14) residents, Oct. older, first visit to WAD 1995-Mar. 1998 emergency department Kyhlback et al Sweden residents, Whiplash, 18 yrs or 83 12 months Symptom (15) Jan. 1997-May older, first visit to intensity, pain 1998 emergency department disability index Sterner et al Umea, Sweden Whiplash, 16-64 yrs, 356 16 months Presence of (16) residents, all persons seeking disability Jan.l997-Feb. treatment at the hospital 1998 emergency department or general practitioner *Whiplash-associated disorder 13 2.4.2 Recovery From Acute Whiplash There is little consistency in the outcomes used to evaluate recovery or the follow-up duration across studies (Table 2.5). This makes comparison in overall recovery between studies difficult. Two of the studies reviewed (17, 18) used an administrative proxy of the recovery time (time-to-claim closure (17), and time receiving compensation (18) ). Although these outcomes are thought to closely parallel the actual recovery time (2, 17) the use of the outcome time-to-claim closure is often criticized for it's administrative nature (19). For example, it is possible that some individuals continue to have significant disability despite closure of their claim. In addition, it is difficult to determine how the time-to-claim closure correlates with a patient's perception of recovery. The remainder of the studies reviewed used self-report from the patient to determine i f persistent symptoms or disability remained at follow-up. This outcome may also be difficult to interpret because individuals will have different interpretations and perceptions of recovery. Some patients may consider only the complete absence of pain to indicate improvement, while others may consider the return to their normal routines of daily living or successful readjustment to living with pain and disability to indicate meaningful improvement (20). The differences in primary outcome and the duration of follow up partly explain the considerable variation seen in the overall recovery from whiplash in different jurisdictions. However, differences in prognosis are also to be expected in different insurance jurisdictions. In general individuals from provinces (2) or countries (7) that operate in a no-fault jurisdiction have faster recovery time and suffer less chronic whiplash pain (6); however, considerable variation in recovery time also exists in areas 14 with similar insurance systems suggesting other factors are at play. For instance, both Saskatchewan and Quebec do not provide pain and suffering compensation payments; however, large differences in recovery time exist for these two jurisdictions (6). To date there is no universally accepted recovery outcome for the study of prognosis of whiplash-associated disorders. The perception of recovery will mean different things to different individuals. Similarly definitions of recovery will vary depending on the stakeholder. An insurer may consider the point at which an individual has reached a maximal medical improvement postcollision, whereas the patient or a clinician may seek the complete resolution of symptoms. Table 2.6 - Summary of Prognosis of Acute Whiplash Injuries by Target population Study Author Primary Outcome Population-based Cote et al (17) Claim duration: median time to claim closure in tort 433 days (95% CI: 403-457), median time to claim closure in no-fault 198 days (95% CI: 190-206). Suissa etal (18) Compensation at 1 year (%). 1.4% WAD I, 1.8% WAD II, 4.8 % WAD III Hospital-based Hartling et al (3) Presence of WAD at 6 months: 35.3% have symptoms consistent with WAD Hartling etal (14) Presence and grade of WAD at 6 months: 10.2% WAD 0, 31.7% WAD I, 57.5 % WADII, 0.6% WAD III Kyhlback et al (15) Visual Analogue Scale (VAS) at baseline and 1 year - mean (SD): 47.2 (21.6), 41.9 (27.2) Pain Disability Index (PDI) at baseline and 1 year-mean (SD): 25.2 (16), 21.7 (18.8) Self-Efficacy Scale (SES) at baseline and 1 year - mean (SD): 140.3 (40.6), 146.2 (48.8) Sterner etal (16) Presence of Disability at 6 months: 32% (26% symptoms affect work and leisure, but not on sick leave, 4% on sick leave, 2% modified work duties) 15 [ 2.4.3 Prognostic Factors for the Recovery of Acute Whiplash Injuries Previous systematic reviews of the whiplash literature (1,6) have considered all studies published prior to July 2000. The results of these reviews are summarized in Table 2.7. The prognostic factors for the six recent studies are summarized in Table 2.8 and considered below. Table 2.7 - Summary of findings of the two previously published systematic reviews of the whiplash literature Review Study Author Prognostic Factors Consistently Associated with Delayed Recovery Cote et al. Sociodemographic: Older age (2, 5, 21) Female gender (2, 5, 21, 22)} Postcollision Symptoms: Baseline neck pain intensity(2, 23) Baseline headache intensity(2, 23) Radicular signs and symptoms (2, 24) Consistent evidence was not found for: Work status, work activities, marital status, education level, number of dependents previous neck pain, previous headache, crash-related factors Prognostic factors from a single study associated with delayed recovery: Anxiety before collision, reduced/painful jaw movement, percentage of body in pain, concentration problems, initial health care provider, lawyer involvement, compensation system, not at fault for collision (2), anxiety, fatigue, sleep disturbances, illness/disability worry, forgetfulness, stress, cervical spine osteoarthritis, cranial nerve/brainstem disturbance, visual disturbances (24) Spitzer et al. Prognostic Factors Consistently Associated with Delayed Recovery Postcollision Signs and Symptoms Severity of initial injury (25, 26) Prognostic factor from a single study associated with delayed recovery: Older age, radicular symptoms, self-reported cognitive impairment (25), previous headache (27), musculoskeletal or neurological sign within 3 days of collision (26) 2.4.3.1 Sociodemographic Factors The two Canadian population-based studies (17, 18) controlled for the risk factors age and gender but did provide risk estimates for these variables (Table 2.8). Suissa et al. 16 (18) reported in the text that both female gender and older age were associated with delayed recovery. In addition, both of these cohorts have been evaluated previously (2, 5), and older age and female gender were found to be significantly associated with delayed recovery in the previous studies (2, 5). The authors investigating patients from emergency departments (3, 14-16) reported risk estimates for the sociodemographic factors: age, gender, education, and work status (Table 2.7). Hartling et al. (3) and Kyhlback et al. (15) found that older age was associated with the presence of future symptoms; however, Sterner et al. (16) did not find such an association. Female gender was found to be associated with delayed recovery in two of the three cohorts (15, 16). Using a univariate analysis, Hartling et al. additionally reported that missed time from work and modified work and leisure activities were associated with delayed recovery (3). Only one study investigated the effect of education level. Sterner et al. (16) found that individuals with less than a university education were more likely to have persistent disability at follow-up. There is consistent evidence in the recent literature and from previous studies that older age and female gender are associated with delayed recovery. 2.4.3.2 Postcollision Signs and Symptoms In a population-based cohort from Saskatchewan, using an explanatory model to test the hypothesis that initial neck pain intensity, depressive symptomatology and physical functioning after a motor vehicle collision were associated with delayed recovery, Cote et al found that all three of these factors were significant (17) (Table 2.8). In the other population-based cohort from Quebec, Suissa et al (18) reported that neck pain on 17 Table 2.8 - Summary of prognostic factors for recovery from acute whiplash injuries Study Author Significant Prognostic Factors - Risk Estimate (95% CI) Outcome Measure Population-based Cote et al (17) Initial neck pain intensity 0-34 days HRR=1.11 (0.97-1.27) 35-115 days HRR=0.84 (0.77-0.91) 116-235 days HRR=0.76 (0.69-0.84) 236-358 days HRR=0.81 (0.74-0.90) >358 days HRR=0.87 (0.83-0.92) Functional status HRR=1.17 (1.12-1.23) Depressive Symptomatology HRR=0.63 (0.51-0.77) Time-to-claim closure Suissa et al (18) Neck pain on palpation HRR=0.85 (0.76-0.96) Muscle pain HRR=0.85 (0.74-0.97) Arm numbness or pain HRR=0.64 (0.55-0.76) Shoulder numbness or pain HRR=0.83 (0.71-0.97) Headache postcollision HRR=0.82 (0.73-0.92) QTF classification WAD II HRR=0.82 (0.75-0.89) WAD III HRR=0.61 (0.51-0.73) Nonsignificant tenderness on palpation, decreased neck mobility, neck pain on mobilization, muscle spasm, muscle stiffness, radiation of pain or numbness to back or chest, dizziness, loss of consciousness, visual disturbances, anxiety or insomnia Time on compensation Hospital-based Hartling et al (3) Age group* 18-30 OR=1.00 31-50 OR=1.51 (0.86-2.65) 51-70 OR=3.78 (1.84-7.75) Number of initial symptoms 0-2 OR=1.00 3 OR=2.05 (0.64-6.61) 4 OR=2.71 (0.91-8.07) 5 OR=6.71 (2.39-18.81) 6 OR=5.87 (2.01-17.16) 7 OR=9.87 (3.24-30.11) 8 OR=17.81 (5.80-54.64) 9 OR=22.67 (5.21-98.72) Upper back pain OR=2.91 (1.65-5.12) Arm numbness or weakness OR=2.18 (1.22-3.87) Vision disturbances OR=1.96 (1.00-3.86) Nonsignificant: nausea and/or vomiting, neck stiffness, headaches, low back pain, shoulder pain, neck pain Presence of WAD Hartling et al (14) QTF classification of WAD WAD I OR=0.78 (0.32-1.88) WAD II OR= 1.87 (0.69-5.07) Modified WAD II a OR= 1.17 (0.49-2.77) Modified WAD II b OR=3.10 (1.18-8.19) Presence of WAD Kyhlback et al (15) Self-efficacy scale, QTF classification, and female gender significantly associated with VAS. Self-efficacy scale, older age, and gender associated with PDI. Symptom intensity (VAS), Pain Disability Index (PDI) Sterner et al (16) Female Gender OR=2.02 (1.13-3.63) Previous neck pain OR=3.17(1.34-7.46) QTF classification of WAD WAD II & III OR=2.03 (1.08-3.88) Educational level (lower) OR=2.08 (1.09-3.98) Nonsignificant: age, accident type, previous headache, previous back complaints Presence of disability •Similar risk estimate for age in the study by Hartling et al (3) reported for two different models. The results from model 1 displayed in this table. HRR=hazard rate ration, OR=odds ratio 18 palpation, muscle pain, headache, and radicular signs and symptoms were associated with delayed recovery. In addition this study found that the Quebec Task Force (QTF) classification of whiplash-associated disorders (WAD) was an important prognostic factor. Specifically, that the more severe grades of W A D took longer to recover (18). In a hospital based cohort from Kingston, Ontario, Hartling et al. (3, 14) found that the number of initial symptoms, as well as the specific symptoms: upper back pain, radicular signs and symptoms, and visual disturbances were associate with the presence of W A D at 6 months. They also reported that a modified version of the QTF classification of W A D (WAD II was subdivided into two categories based on the presence abnormal cervical range of motion) was predictive of delayed recovery (14) (Table 2.8). In addition, using univariate analysis these authors reported that: initial pain intensity, depression, anxiety, concentration difficulty, fatigue, and sleep disturbances were associated with delayed recovery (3). They did not report findings for these variables using multivariate analysis. The other two hospital-based cohorts (15, 16) both found the QTF classification of W A D to be an important prognostic factor. Kyhlback et al. also reported that an individual's self-efficacy (defined as a measure of a patient's confidence in performing daily activities despite pain) was associated with pain and disability at 1-year postcollision (15). On the basis of four studies, there is consistent evidence that the QTF classification of W A D is associated with delayed recovery (14-16, 18). Two recent studies and the previous whiplash literature provide consistent evidence that initial neck pain intensity (1, 3, 6, 17), and radicular signs and symptoms (3, 6, 18) are associated with delayed 19 recovery. There is also consistent evidence that depressive symptomatology is associated with delayed recovery (3, 17), but this prognostic factor was not adequately described or analyzed in one of the studies (3), and therefore it is not possible to comment on the consistency of this variable. Similarly, only one study (17) investigated an instrument to measure physical functioning (physical functioning scale of the SF-36), number of initial symptoms (3), and self-efficacy (15); therefore, it is not possible to comment on the consistency of these factors. 2.4.3.3 General Health Before Injury Two of the hospital-based cohorts examined the importance of general health prior to the collision (Table 2.8). In univariate analysis, Hartling et al reported that previous neck pain was associated with delayed recovery (3). This group did not report the results of a multivariate analysis for this covariate. Similarly, Sterner et al found that previous neck pain was associated with prolonged disability; however, previous back complaints and headaches were not. Two additional studies that investigated the association of previous self-reported neck pain with delayed recovery reported conflicting results (6). Based on the findings of three studies (3, 8, 16) previous neck pain appears to be an important prognostic factor in the recovery of acute whiplash; however, it should be noted that in two of these studies (3, 8) statistical adjustment for other factors was not performed. 2.4.3.4 Legal and Compensation Factors The Saskatchewan insurance system changed from a tort-compensation system (where claimants received payments for pain and suffering) to a no-fault system (which included no such pain and suffering payments) during the study period of the Cote et al study. 20 Although it was not the primary objective of this group to investigate the influence of the insurance system, they did find that the median time to claims closure was reduced by 235 days in the no-fault insurance system (17). The effect of the insurance system on recovery for this cohort of patients has been previously reported (2). No other studies investigated the association of insurance system or other legal or compensation issues with a delay in recovery; therefore, it is not possible to comment on the consistency of the evidence for these factors. 2.4.3.5 Crash-Related Factors Only one study investigated crash-related factors. Hartling et al did not find a significant association for any of the following factors: position in vehicle, road conditions, type of road (location), preparation for crash, head position, direction of collision, seatbelt use, head movements after collision, type of transmission, size of vehicle involved, posted speed limit, or brake light in rear window (3). Similarly, previous reviews of the literature (6) have not found consistent results for the association of crash-related factors with delayed recovery. 2.5 Summary The quality of evidence to date remains mainly exploratory, and there is still considerable inconsistency in the reported association of various prognostic factors and recovery from W A D (Table 2.9). 21 Table 2.9 - Established and potential prognostic factors in the recovery from an acute whiplash-associated disorder reported in recent and previous literature and systematic reviews Prognostic factor Consistent Limited Inconsistent evidence* evidence** evidence Sociodemographic Older age (2, 5, 15, 16, 18,21) X Female gender (2, 3, 5, 15, 18, 21) X Work status (2, 3, 5, 21) X Work activities (2, 3, 24) X Education level (2, 16) X Marital status (3, 5) X Number of dependants (2, 5) X Postcollision signs and symptoms / clinical Initial pain intensity (1,2, 14, 17, 23) X Increasing grade of injury (QTF classification) (14- X 16, 18) Radicular signs and symptoms (3, 17, 18, 24) X Depressive symtomatology (3, 17) X Physical functioning (17) X Anxiety / stress / worry (3, 24) X Self-efficacy (15) X Fatigue / sleep disturbance (3, 24) X Forgetfulness / concentration difficulties (2, 3) X Cranial nerve/brainstem disturbance (24) X Cervical spine osteroarthritis (24) X Reduced/painful jaw movements (2) X Number of symptoms / percentage of body in pain X (2, 3) Visual disturbances (3, 18, 24) X Medical history before injury Previous neck pain (3, 8, 16) X Previous back pain (16) X Previous headache (2, 16, 24) X Anxiety before collision (2) X Crash-related Seat belt use (2, 3, 5,21,23) X Head position (2, 3, 23, 24) X In stationary car during collision (23, 24) X Position in vehicle (2, 5) X Size of vehicle collided with (3, 5) X Initial health care / treatment-related Initial health care provider (2) X Legal / Compensation Insurance jurisdiction (2) X Lawyer involvement (2) X Not at fault for collision (2) X * Evidence was considered consistent if at least two studies or 75% of all studies reported consistent results on the importance of a specific risk factor. ** Evidence was considered limited if only a single study has investigated a particular risk factor or limited statistical analysis (univariate statistics) prevented comment on the consistency of the evidence. 22 Some potentially important prognostic factors have been investigated by only a single study (for example, lawyer involvement, insurance jurisdiction, initial health care provider, and physical functioning). In other circumstances confirmation of the potential importance of a particular factor was limited by a univariate statistical analysis. For example, both Cote et al and Hartling et al reported that postcollision depression resulted in delayed recovery; however, the study by Hartling and her colleagues lacked statistical adjustment for other factors when investigating the effect of depression (although this was not the primary objective of this study). Evidence was considered limited when only a single study was available or the statistical analysis was descriptive in nature, the evidence was considered limited. Furthermore, differences in the definition of other prognostic factors across studies, such as psychological variables related to anxiety, stress, and worry, limited the qualitative synthesis of these particular factors. Thus, further exploratory and explanatory work is necessary for confirmation of these (often controversial) findings. In future research particular attention should be made to the clear definition of prognostic factors and the use of multivariate statistical analysis. Based on the most recent literature and previous studies, there is consistent evidence that female gender (6, 15, 16, 18), older age (3, 6, 15, 18), increasing grade of injury by the Quebec Task Force classification (14-16, 18), initial intensity of neck pain and headache (3, 6, 17), radicular signs and symptoms, and previous neck pain(3, 6, 16)are associated with delayed recovery (Table 2.9). The lack of consistency and limited available information for many potentially important prognostic variables has generated considerable controversy. In particular, further consideration of legal and compensation 23 factors, the choice of health care provider and other treatment related factors, and psychological variables are necessary. In contrast to the results reported from previous reviews, the recent literature usually considered the simultaneous effect of many variables (phase II studies) in favour of less sophisticated univariate analysis methods (phase I studies). However, since 2000 there has been only one study considered a phase III explanatory study. Several recent articles studied patients presenting to a hospital emergency department, yet there continues to be limited information from other primary or secondary care clinical settings (physician, physiotherapist, or chiropractor). In addition, although there were two recent population-based studies, both of these cohorts have been previously considered in the literature. The assembly of new large cohorts of whiplash subjects is necessary, particularly in primary care settings other that hospital emergency departments. 24 Chapter 3 Background of CBI Materials and Methods 3.1 Introduction This study used the clinical database from the Canadian Back Institute (CBI) Corporate Office. This database contains data from a national network of 48 Canadian physiotherapy and rehabilitation clinics. An understanding of the type of patients CBI typically provides service for, the referral process for these patients, the standard treatments they receive, the clinical course of their condition, and the data collection procedures is necessary and provides the context for the current study. Prior to a description of the clinical database an overview of how an individual injured in a motor vehicle collision would potentially become a patient at CBI is provided. That is, a description of the how an individual progresses form injury to treatment to recovery. 3.2 Patient Referral Process Typically, when a patient is injured in a motor vehicle collision in Canada the initial provider of care is a physician (2). The patient may attend an emergency department or a private office. The majority of individuals will present with soft-tissue injuries to the neck (1) or low back (2, 3). The physician will generally decide on one of three possible methods of management: 1) CBI treatment, 2) other non-physician health care professional treatment, or 3) continued management with the family physician. Of those patients referred to CBI, greater than 90% have complaints of neck or low back pain (28). Figure 3.1 shows the flow of patients from initial injury to recovery. A patient may also 25 directly seek care at CBI or with another health care professional without a previous physician visit. This occurs in a minority of the cases. Patients that attend CBI become part of the clinical database on the day of their initial assessment day. Each patient typically receives a course of therapy until the point at which either the doctor, therapist, or patient deem themselves sufficiently recovered. If recovery is delayed a further referral or discharge from CBI may follow. A cycle of specialist, physician and other health care worker visits may occur until that point where a patient feels that a maximal recovery has occurred or the patient accepts a settlement from their auto insurance provider. In all provinces, the patient's auto insurance company makes payment for CBI treatment. Claim closure for injuries sustained in motor vehicle accidents varies depending on the insurance jurisdiction, however the majority of claims are closed within 1 to 2 years (6, 29). 3.3 Canadian Back Institute Methods Figure 3.1 depicts the rehabilitative process for whiplash-associated disorders (WAD) referred to CBI. Upon referral to a CBI clinic, an appointment is made for an initial assessment. During this appointment the patient provides demographic information and completes a patient questionnaire regarding their injuries. A registered physiotherapist with specific training in spinal pain rehabilitation performs a standardized interview and physical examination. After considering historical and examination findings, a diagnosis is rendered and a treatment program designed. Data entry personnel enter the 26 demographic, initial patient questionnaire and spinal assessment form findings into the computerized clinical database. After a treatment program is complete the patient is discharged or referred elsewhere. Clinical data from all clinics within CBFs national network are downloaded to a central location, the CBI Corporate Office. Motor vehicle collision -resulting in patient seeking treatment Physician Consultation Referral to local CBI Patient Questionnaire Diagnosis Treatment Patient Questionnaire Discharge Spinal Assessment Form 1) Data entry 2) Download to Corporate Office 3) Consolidated data from all clinics Figure 3.2- Flow chart of patients and data collection procedures through the Canadian Back Institute (CBI) system 3.3.1 Clinical Database The forms for data entry into the clinical database are the Spinal Assessment Form, Spinal Discharge Form and patient questionnaire (completed twice, once on presentation to the clinic, and once at discharge). 27 3.3.1.1 Spinal Assessment Form Prior to the development of the Spinal Assessment Form, the history taking and physical examination process consisted entirely of open-ended questions and free hand writing. In an effort to standardize assessment procedures a step-by-step structured approach consisting of a series of close-ended questions was developed. This new approach captures the essential aspects of the history and physical examination findings and results in an efficient mode for data entry and potential analysis. The Spinal Assessment Form was developed by CBI's Medical Director (an orthopaedic surgeon), the senior physiotherapist, the Teaching and Development staff, and other clinicians of CBI. The Spinal Assessment Form was implemented across the network of clinics in 1992 (Appendix C). 3.3.1.2 Patient Questionnaire The questionnaire used for this study is based on a previously published instrument, the Low Back Outcome Score (LBOS) (30). This questionnaire has been modified and renamed the Canadian Back Institute Questionnaire (CBIQ) (Appendix D). The CBIQ was designed with consideration of the various potential causes of injury as well as locations of spinal pain. The questionnaire is used for patients with both neck and back pain. With the exception of the first question, the CBIQ does not ask questions regarding the specific anatomical site of the patient's pain; rather the effect of pain (in general terms) on function is the primary focus. The CBIQ includes questions regarding: 1) area 28 affected 2) cause of injury, 3) onset of pain, 4) length of time in pain, 5) household chores, 6) sport and leisure activities, 7) lawyer involvement, 8) smoking, 9) rest required, 10) doctor visits, 11) pain medication, 12) walking, 13) sitting, 14) standing, 15) lifting, 16) dressing, 17) sleeping, 18) traveling, and 19) working. The CBIQ includes differentially weighted questions to determine an overall score. In a similar scoring system to the LBOS, each response on the CBIQ is scored on a four-point scale. Pain and active daily activities (such as household chores or leisure activities) are weighted more that treatment and rest required, which in turn is weighted more than passive activities (such as sitting or standing). The CBIQ produces two outcome scores. The full questionnaire is scored out of 100, and includes all of the modifications that were made to the LBOS. The second outcome (the function score) is generated from a subset of the full questionnaire. The function score is produced by removing the questions: cause of injury, onset of pain, length of time in pain, and smoking status, thereby more closely resembling the LBOS. Assessment of validity of self-report scales is made difficult by the lack of a gold standard for which to compare results. However, in a correlation study the CBIQ was compared to the Oswestry Disability Index (ODI)(31). The Pearson correlation coefficient (r) was -0.72 (p<0.01), indicating adequate validity to measure function (based on the ODI as an external criterion)(31). In addition the CBIQ demonstrated the ability to predict prognosis and discriminate between low- and high-risk individuals for time on compensation benefits after occupational back injury (32). 29 3.3.1.3 Spinal Discharge Form The Spinal Discharge Form includes a second application of the patient questionnaire, four questions regarding patient satisfaction with treatment, and i f applicable, therapist recommendations for return to work. 3.3.2 Data Entry and Collection A data field has been created for each form response in a format that mimicked the actual paper version. Upon completion of the initial assessment and data collection forms, receptionists enter the data verbatim directly into the clinical database at each individual clinic. Upon data entry the questionnaire outcome scores are generated automatically. No coding or interpretation is required. The data is downloaded monthly to a central collection site, the CBI corporate office in Toronto. Here the data is consolidated to one database. The CBI Training and Development team train all staff on the standardization of data collection and data form completion. Registered physiotherapists at CBI complete two training courses. These courses cover the standardized approach to clinical data collection, assessment form completion, patient interview, treatment strategies, and discharge planning. The curriculum regarding data collection and form completion includes: 1) review of questions and responses, 2) overview of the salient information from the Clinical Forms User's Guide, 3) diagnosis using the CBI classification system, 4 ) role playing interviews by course participants, and 5) assessing real patients and completing the assessment form as a group. The goal of 30 these efforts is to standardize data collection, and assessment across therapists within the company. CBI regards 1994 as the operational inception year of the clinical database. 3.3.3 Treatment The Canadian Back Institute clinics provide active physiotherapy primarily for mechanical spinal pain of musculoskeletal origin. They are secondary care rehabilitation facilities that focus on pain control in acute, sub-acute and chronic ambulatory populations. Rare patients with suspected systemic disease and cases sustaining trauma sufficient to produce severe bony injury or major neurological sequelae are referred elsewhere. A l l patients follow a structured CBI protocol of active exercise. The duration of each treatment session and adjunctive physiotherapeutic techniques applied vary by the patient and are left to the discretion of the therapist. Treatment is progressed through three stages of recovery: 1) pain control - approximately ten days of treatment stressing back and neck education and exercises that produce pain reduction, 2) recovery of movement -approximately ten days of treatment emphasizing active exercises through the complete range of motion of the injured area, 3) physical conditioning- approximately four weeks of cardiovascular training via a stationary bike, stair climber or walking program, progressive strengthening of the trunk and extremity musculature using a combination of free weights and machine training, and potentially work conditioning. The number of treatment hours per day in each stage, and the total treatment time are adapted to the individual needs of the patient. Home exercises are additionally given to most patients 31 ><"> early in the treatment regime. The conclusion of the treatment program can occur at any time during its course when normal activities or return to work become possible. 3.4 Database Limitations Using a clinical database for epidemiological research eliminates many of the potential difficulties associated with administrative data. In addition procedures, such as standardized assessment forms, and training courses in data collection and input, have been implemented at CBI to attempt to improve the completeness of the clinical database and reduce examiner bias. However, an awareness of some possible limitations of the clinical database is necessary. The CBI database included only the sample of patients who perceived that their injuries required treatment. Some individuals with minor injuries and disability will not seek therapy. Others may not have received a referral for physiotherapy services, and may receive treatment with another health care professional. There may be systematic differences between the various groups of patients who seek therapy elsewhere (for example with a massage therapist or chiropractor). In addition, the majority of patients in the CBI clinical database have their treatment costs paid by their auto insurance company. Some individuals whose collision resulted in minor damage to their vehicle or who have had their claims denied for other reasons may not be captured by the database; however, those who opt to pay for their treatment by themselves will be included. Nevertheless some individuals with minor injuries will be excluded. 3 2 As with any database there are limits to the quantity, primary focus, and quality of information provided. The CBI database included sociodemographic and clinical information. Limited data on psychological and workplace factors were available. Missing data resulted when certain questions on the Spinal Assessment Form were not completed. The overall response rate for the survey was good; however, due to low response rate for individual survey items, two variables of interest were not included in this study (previous spinal pain and an indicator of constant pain). 3.5 Summary The clinical database represents the reality of the management of whiplash-associated disorders in Canada. Not all patients injured in motor vehicle collisions will seek professional treatment. Those that perceive that their injuries require some sort of treatment usually initiate therapy with their physician but may additionally see multiple other health care professionals. The database includes individuals either in their first or recurrent motor vehicle collision, who have either immediate or delayed onset of symptoms, and who seek care. The patients presenting to CBI are referred by their physician, other health care professional, other individual involved with their case (for example a claims adjustor, lawyer), or by self-referral. The database consists of many variables of interest (focusing primarily on sociodemographic variables, personal characteristics and clinical information) and is collected in six provinces (British Columbia, Alberta, Saskatchewan, Ontario, Quebec and Nova Scotia) across Canada. The next chapter will focus on additional methodological issues, and a description of the study design. 33 Chapter 4 Study Design, Materials and Methods 4.1 Overall Study Design This was a retrospective inception cohort of individuals injured in a motor vehicle collision (MVC) enrolled in rehabilitation programs in 6 provinces in Canada. Data were originally collected by self-report survey and questionnaire, and physiotherapist assessment. A l l data from patients who presented to one of the Canadian Back Institute Health's (CBI) 48 network clinics, between January 1, 1998 and December 31, 2001 after a M V C were extracted for analysis. The years of the study were selected based on the availability of national, electronic data from CBI. A secondary analysis of all patients presenting after M V C regardless of inception time will be conducted to compare acute and chronic W A D . 4.2 Inclusion Criteria Table 4.1 lists the subject inclusion criteria. Table 4.1 - Subject inclusion criteria 1) Injured in a motor vehicle collision. 2) Assessed for primary complaints of neck or low back pain at any one of the 48 CBI locations across Canada. 3) Less than 91 days between reported injury date and first day of treatment at CBI 4) Completion of both the entry and exit patient questionnaires / complete follow-up information. 5) Adults aged 18 to 65 years 6) Less than 91 days between reported injury date and first day of treatment at CBI 7) No previous history of spinal surgery 8) Completed a Visual Analogue Scale (VAS) on the intensity of their current pain. Inclusion criteria were based on the following rationale. Patients injured in motor vehicle collisions were targeted because treatment of these injuries is a primary focus of CBI. Those individuals who were injured in a motor vehicle collision at work were excluded from the present study because the compensation structure for the injured worker is 34 different. It was thought that these differences would produce systematic differences in the distribution of the descriptive variables. Individuals injured in motor vehicle collisions commonly report neck, and/or low-back pain, yet whiplash injuries are typically associated with only neck pain. Therefore, the decision was made to target both locations thereby providing a more comprehensive description of whiplash-associated disorders. A study investigating the prognosis of any condition requires the construction of an inception cohort (33, 34). This ensures that each individual within the cohort are at a similar stage in the course of recovery. Only individuals with acute and sub-acute neck and back pain (defined as less than 91 days) were included (35, 36). The chronic cases were excluded, as the purpose of the model building exercise was to find predictors of delayed recovery in those individuals prior to chronicity. The inception point for this study was the start date of treatment (which had to occur within 91 days of the motor vehicle collision). Completion of the patient questionnaire on the first visit at CBI, as well as upon discharge was required to ensure that complete follow-up information was available for all subjects, and thus the primary outcome could be generated. Patients with prior spine surgery were also excluded from the study population. These patients are rare and not typical of most individuals with soft tissue injuries sustained in a motor vehicle collision. 35 The intensity of pain on presentation to CBI (as measured by a visual analogue scale) was considered to be an essential variable; therefore, this was included as one of the inclusion criteria. 4.3 Final Data Set The extraction of data from the clinical database of the 48 Canadian CBI clinic locations produced an initial list of 16,404 records. This represented all patients treated at CBI who were injured in a motor vehicle collision. Of these patients 8,945 (54.5%) presented with acute injuries as defined by less than or equal to 91 days since their injury date. Complete follow-up information was available for 3,472 (38.8%) records. This indicated that both entry and exit questionnaires had been completed. From this population a further 245 (7.1%) records were excluded because of age restrictions, and 15 records (0.5%) due to previous spinal surgery. A further 875 records (28.6%) were excluded with missing intensity of pain data as scored by the V A S , and 123 records (3.8%) with missing or inconsistent age data. A minimal number of other subjects had missing or inconsistent work status or injury dates (see Figure 4.1). In total, of the 8,945 original acute patient records, 5,473 records (61.2%) did not have complete follow-up information, while an additional 1,027 (11.5%) were excluded due to missing or inconsistent data. 4.4 Data Source This source of data for this study was the clinical database from the Canadian Back Institute (CBI) Corporate Office. A detailed description of the data collecting procedures was provided in Chapter 3. Briefly, upon presentation to one of CBFs rehabilitation clinics each patient completed a questionnaire regarding their current condition and 36 provided sociodemographic information. A standardized assessment by registered physiotherapist was performed. This examination provided a number of clinical variables of interest. After the treatment period the patient once again completed a questionnaire regarding any changes in their condition. The information from the two patient questionnaires and the spinal assessment form are downloaded to CBFs Corporate Office where the database is analyzed and maintained. Initial query 16.404 records Chronic injury (> 91 days from injury date) n = 7,459 Acute injuries f - 91 days from • injury date), not occurring at work n -8.945 Missing follow-up outcome information n = 5,473 follow-up information available n = 3,472 Other inclusion criteria not met and /or additional missing data n= 1,287 Had previous spinal surgery n= 15 Study Cohort n = 2.185 Missing Visual Analogue Scale n = 875 Aged< 18 or >65 years, n = 245 Age missing n - 123 Work status missing n = 29 Figure 4.3 - Final working dataset of individuals injured in a motor vehicle collision attending a Canadian Back Institute (CBI) clinic in Canada between 1998-2001 4.4.1 Missing Data Upon assembly of the dataset (n = 2,185), the data were examined for blank fields, and inconsistent formats (37). The frequency distribution of essential variables was examined. The inclusion criteria for the inception cohort resulted in the removal of a 37 significant number of records (see Figure 4.1). A total of 5,473 records (61.2%) were excluded because either the entry or exit questionnaire was not completed. This variable was essential for generating the primary outcome variable. A further 878 records (28.7%) with missing data on the visual analogue scale measuring intensity of spinal pain (VAS). This variable was considered an essential variable to control for in the analysis of prognostic factors. Because both the outcome and pain intensity variables were considered essential, imputation techniques were not considered and records missing these data were excluded from the analyses. To examine potential systematic differences between the respondents and non-respondents a comparative analysis was conducted for a number of these key variables. Alternate methods of analysis were conducted for additional variables of interest (based on previous findings in the literature or clinical suspicion), but with a significant number of missing values. This occurred for three variables of interest (previous episodes of spinal pain, constant versus intermittent pain, and physical demands at work). A total of 1,273 (58.3%) and 1,264 (57.8%) of the records were missing values regarding previous episodes of spinal pain, and constant versus intermittent pain respectively. The proportion of records missing data were considered to large for imputation procedures. To examine the potential association of these two excluded variables with the outcome of interest, a separate analysis using those records with completed information on these variables was conducted. A third variable (physical demands at work) had missing values for 270 records (12.4%). A method of imputation was utilized for adjusting for the smaller proportion of missing values for this variable. 38 4.4.2 Imputation There is no ideal method for dealing with missing values in a dataset. For this reason every attempt should be made during the collection of the data to ensure completeness. Nevertheless missing values for some covariates or outcomes are a reality in most epidemiological studies. A number of methods have been suggested for dealing with these missing values. Perhaps the most common approach is complete subject analysis (38). In this approach only those records with complete information on all variables are included in the analysis (38). Miettinen (39) has stated that this is the only approach that will assure that no bias has been introduced into the data. This declaration, however, assumes that the data that are missing at random. In this scenario the subset of records with complete information are equivalent to a random sample of the entire study population (38). Complete subject analysis is not without difficulty. The precision of the model will be reduced by the reduction in sample size. In addition should a number of different variables have missing values (while at the same time the rest of the record is complete) then this method can be wasteful. Alternative approaches of dealing with missing values include using an imputed variable based on: central tendency, regression estimates, indicator variable or multiple imputation estimates. These approaches are explained further in Appendix E. In simulation models there is not one particular method for handling missing data that produces vastly superior results (38). Methods based on regression estimates, the indicator method or multiple imputation, along with complete subject analysis have 39 produced similar results (38, 40). As stated previously, in datasets with a small sample size one must consider the loss of precision that results with complete subject analysis. A l l of the above methods are common in epidemiological study. Some authors have suggested that the regression and indicator methods have limitations due to the introduction of bias (40). This occurs i f the variable with the missing values is a confounder (41). If C% of the confounder is missing then C% of the confounding is potentially uncontrolled (39). The indicator method and the complete subject analysis method have the additional benefit of convenience of application. For this study the physical demands of work variable was analyzed using two separate methods. The variable was evaluated with both the complete subject analysis and indicator methods. The variable was dropped from later analyses due to lack of significant association with outcome; therefore, the completed information in the records with the missing physical demands variable was available for study without the loss of precision that would have been expected had these records been excluded. 4.5 Research Variables 4.5.1 Outcome Variable The outcome measure for this study was change in the Canadian Back Institute Questionnaire (CBIQ) score between the initial and discharge visits. The CBIQ is a 19-item survey based on the Low Back Outcome Score (LBOS) (30). Greenough and Fraser (30) concluded that their instrument (LBOS) was more comprehensive and more 40 discriminating that the Oswestry Disability Score, the Waddell Disability Rating, and the Waddell Physical Impairment Rating. The LBOS had a correlation coefficient of -0.87 (p<0.001) with the Oswestry Disability Scale (30). In multivariate regression: age, sex, compensation status, psychiatric disturbance and employment status were all significantly related to questionnaire score (30). Holt et al. (42) further examined the reliability of the LBOS and concluded that this instrument has good internal consistency (cronbach alpha coefficient=0.85), and test-retest reliability (84% agreement at 1 week). Khatri et al. recently reported the minimally clinically important difference in LBOS in a population of low back pain patients (43). Based on an external criterion of the patient's global perception or recovery (no change, a minimal improvement, good-but incomplete recovery, or complete recovery) they found that a 10% change or 7.5 points was the minimal clinically important change. As noted earlier the questionnaire in this study in a modified version of the LBOS (now called the CBIQ). Additional questions regarding how pain is affecting the ability to sit, stand and lift have been added. In addition, a question regarding contacting a lawyer has been added. This CBIQ correlates well with the Oswestry Disability Index (Pearson correlation coefficient (r) = -0.72), and has demonstrated the ability to predict prognosis (31). For this study the change in CBIQ was the outcome of interest. The CBIQ was first rescaled to remove the lawyer question from the total score. This was done because this explanatory variable was of interest for analysis. Therefore, it was not appropriate to 41 investigate this variable while it still contributed to the scoring of the outcome. The CBIQ (function outcome) is scored out of a maximum 70 points. The lawyer question is scored out of a possible 9 points. With this question removed from the rescaled CBIQ has a total possible score of 61 points. For this study the minimally clinically important change in CBIQ score was deemed to be a 10% change or 6.1 points. Minimum clinically important change in questionnaire score is a common outcome in patients with low back and neck pain (44), however it is important to consider the clinical relevance of the demonstrated change. Statistically significant changes in score may be observed (particularly when a large sample size is used) that have little clinical significance. For this reason, this study used an outcome that considered the clinical significance of the change in CBIQ. Another issue regards the meaning of the change that occurred. Each individual will have different starting scores on the CBIQ, and a change of 6.1 points will have different meaning depending on the individual (20, 45). For example, an individual with significant disability and pain on their initial presentation who achieved the minimally clinical improvement (yet is still significantly disabled), and an individual who presented with relatively minor disability and pain who fully recovered, will both be included in the positive outcome category. The nature of the outcome did not discriminate between two such individuals. However, both individuals described above are demonstrating important improvements in their condition, regardless of their initial state, and the outcome will reflect this change. 42 4.5.2 Explanatory Variables A total of 25 explanatory variables were analyzed in the inception cohort. Some of the variables had not been previously demonstrated as prognostic factors in the recovery from whiplash-associated disorders (WAD), however clinical intuition suggested they might be worthy of exploration. The variables were grouped into three clusters: personal characteristics (8 variables), clinical (9 variables) and treatment related (8 variables). 4.5.2.1 Personal Characteristics The personal characteristics that were examined consisted of: 1) age, 2) gender, 3) lagtime from injury data to treatment date, 4) lawyer involvement with case, 5) physical demands of job, 6) province where injury and treatment occurred, 7) smoking status, and 8) work status. Age. Patients provided their birth date upon presentation to the clinic. Gender. This was a categorical variable (0=male, l=female). Physical Demands. The physical demand categories were modeled after the taxonomy developed by the U.S. Department of Labour (46). Responses were coded: l=sedentary, 2=light, 3=medium, 4=heavy. Patients were asked to rate their overall job tasks. 43 Lagtime. The lagtime between injury and treatment (duration of the current episode) in days was generated by subtracting the injury date from the date of presentation to the clinic. Lawyer. Each patient in this study was asked on two occasions whether or not they have contacted a lawyer about their injury, once at the initial presentation and once at discharge. The combination of these two variables was use to determine whether a patient had legal representation at or prior to assessment at CBI, or i f they retained legal representation at some point after initial presentation by prior to discharge. The two variables were coded 0=no lawyer involvement, and l=lawyer retained. Province. Subjects presented to one of a 48 CBI Physiotherapy and Rehabilitation clinics across Canada. Based on the clinic location a province variable was created for analysis. The included provinces were coded as: l=British Columbia, 2=Alberta, 3=Saskatchewan, 4=Ontario, 5=Quebec, and 6=Nova Scotia. For analysis, British Columbia was chosen as the reference category due to the high incidence of whiplash claims in this province. The incidence of whiplash claims in British Columbia is reported to be higher than all other Western countries (4). Smoking Status. For analysis this variable was coded 0=non-smoker, l=current or former smoker 44 Work Status. On their initial visit to the clinic, patients were asked i f they were currently working. For analysis responses were coded: 0=no, l=yes. 4.5.2.2 Clinical Variables The list of clinical variables that was examined consisted of: 1) comorbid conditions, 2) headaches, 3) neurological examination results, 4) non-organic signs of pain focused behaviour, 5) number of symptoms on initial presentation, 6) pain radiation to the arm or leg, 7) primary pain location, 8) thoracic (mid back) pain, and 9) Visual Analogue Scale of pain intensity. Comorbidity. Subjects were asked about their past medical history. Specifically they were queried regarding the presence of coronary artery disease, hypertension, rheumatoid arthritis, diabetes mellitus, malignancy, chronic obstructive pulmonary disease, or other comorbid medical condition. For analysis, comorbidity was recorded as a binary variable: 0=none, l=any comorbid condition(s). Headache. The presence of headaches was recorded as a binary variable: 0=no, l=yes. Number of Symptoms. A number of symptoms are common to patients with W A D . For analysis, a variable was created that totaled each of these symptoms. For example, each patient was asked about the presence of neck pain, mid back pain, low back pain, upper extremity pain, lower extremity pain, and headache. Depending on the number of body areas where pain was present, each individual received a score from 0 to 6. 45 Neurological Examination - Bicep Reflex, Anterior Deltoid, Bicep Power, Tricep Reflex, Extensor Digitorum Longus, Tricep Power, Knee Reflex, Ankle Dorsiflexion, Ankle Reflex, Ankle Plantarflexion, and Gluteus Maximus. A physiotherapist's physical examination of the cervical and lumbar spine was performed to rule out any disease states that may require direct attention or referral to a specialist. The clinical database contained the physiotherapists' clinical interpretation of whether these tests were positive or negative. The number of neurological tests performed depends on clinical findings and therapist discretion; therefore, not every patient will get every test. In the dataset this makes it difficult to identify cases where a test was done (but did not need to be performed) from those who were not tested but should have been. To overcome this issue, the conduction tests were categorized into summary variables representing their specific anatomic implications in terms of spinal level of likely neurological involvement: C56 (bicep reflex, anterior deltoid, bicep power), C78 (tricep reflex, extensor digitorum longus, tricep power), L4 (knee reflex, ankle dorsiflexion) and SI (ankle reflex, ankle plantarflexion, gluteus maximus). Appendix F presents a summary of which combination of conduction test results comprised positive or negative neurological signs. For analysis this variable was coded with a binary variable (0=normal neurological conduction, l=abnormal neurological conduction) Non-organic signs. Using physical examination as a means of identifying illness behaviour in back pain patients has become increasingly popular since the introduction of 46 Waddell's non-organic signs in 1980 (47). Waddell et al. identified five non-organic signs, each consisting of one or two tests. The tests assess a patient's pain behaviour in response to certain maneuvers (Appendix E). These tests have also been used for patients with a presenting complaint of neck pain (48). A high number of positive non-organic signs in patients with either neck or back pain indicates pain behaviour. For analysis results were recoded into a binary variable: 0=low (0,1,or2 out of 5), l=high (3,4, or 5 out of 5). A patient with three or more positive non-organic signs was defined as having a clinically significant pattern of non-mechanical, pain focused behaviour (47). Pain Location. Patients recorded their primary complaint. This categorical variable was coded l=neck, 2=back, and 3=neck & back. Pain Radiation. The location of the most distal symptom was recorded. This variable was recoded as a binary variable. For subjects with a complaint of neck pain, 0=no radiation to the upper limb, l=arm, forearm, and/or hand radiation. For subjects with a complaint of back pain, 0=no radiation to the lower limb, l=thigh, calf, and/or foot radiation. Thoracic (mid-back) pain. The presence of thoracic spinal pain was recorded during the initial assessment. This variable was recorded as a binary variable: 0=no, l=yes. Visual Analogue Scale (VAS). Subjects were asked to rate the intensity of their back or neck pain based on the score on a visual analogue scale. The visual analogue scale, using 47 a 0 to 10 scale where 0 represents no pain and 10 represents worst possible pain, is widely used for measuring pain and has been shown to be a reproducible method of measuring pain (49, 50). 4.5.2.3 Treatment Variables The list of treatment variables that was examined consisted of: 1) bed rest, 2) duration of treatment, and number of rehabilitation sessions, 3) concurrent and previous treatment, 4) health care seeking behaviour, and 5) radiographs and other investigations. Bed Rest. Previous literature suggests that prescribing bed rest may delay recovery from episodes of low back and neck pain (51, 52). This variable was coded separately from the remainder of the previous treatments (0=no bed rest prescribed, l=bed rest previously prescribed). Duration of the Treatment Program. A continuous variable; the number of calendar days from the initial presentation to the clinic until discharge. In addition, the number of actual treatment sessions was investigated. Regardless of entry time into the study cohort, data was collected until the patient's discharge date; therefore, the follow up time was not truncated by the end of the study period. Previous and Concurrent Treatment. Although the initial provider of services was not known for this cohort, patients were asked about their previous investigations and treatment, as well as concurrent treatment. Specifically each patient was asked i f 48 myelography/computed tomography(CT)/magnetic resonance imaging(MRI), bonescan, x-ray, blood work, electrical studies, or no investigations had been previously performed Patients were also asked i f either previous or concurrent manipulation/mobilzation, modalities, active exercise, massage, bed rest, other treatment, or no other treatment had been performed. For analysis the x-ray variable was coded 0=no x-rays taken, l=x-rays taken. The other investigations variable was coded 0=no other investigations performed, l=myelography/CT/MRI, bone scan, blood work, or electrical studies were performed. Two variables were created for the previous investigations (x-rays and other investigations) because in a typical clinical setting, x-rays and the other investigations would not be performed during the same clinical encounter. It was not known i f previous treatment was performed by a single or multiple providers. For analysis a binary variable was created (0=no previous treatment/, l=any previous treatment. It was not know i f concurrent treatment was performed by a single of multiple providers. For analysis concurrent treatment was a binary variable (0=no concurrent treatment, l=any concurrent treatment). Health Care Seeking Behaviour. A variable capturing the health care seeking behaviour of each patient was created. This was a sum of the x-ray, other investigations, previous 49 treatment, and concurrent treatment variables. A score of 0 to 4 was possible for this variable. 4.6 A Comparison of Acute and Chronic Whiplash Associated Disorders Using a cohort of all patients who completed both entry and exit questionnaires (n=6,195) a secondary descriptive comparison of acute and chronic W A D was conducted. Further exclusions of this dataset were made based on the injury occurring at work (n=156), age restrictions (n=343), and injury date or work status missing (n=l 15). This resulted in a working data set of 5581 records. This comparison used an available-case analysis approach. This method included all acute and chronic cases with observed data for a particular variable, for any one specific analysis (but the same number of cases may not be present for other analyses) (53). This method assumes that the missing data are missing at random; that is, that the information available does not differ in any systematic way from the missing information (53). Three groups were compared in a cross-sectional manner: those with an acute presentation (defined as presentation less than or equal to 91 days since injury), early chronic presentation (defined as greater than 91 days but less than 6 months since injury), and chronic presentation (greater than 6 months since injury date). 4.6.1 Additional Research Variables In addition to the outcome variable and explanatory variables described earlier, additional variables were included for descriptive analysis. 50 Raw CBIQ entry and exit scores, and CBIQ change score. In addition to the primary outcome based on a minimally clinically important change, a comparison of the average change in CBIQ, as well as, the raw sores on presentation and discharge was conducted. Work Modifications. Patients were asked about any modification in the hours of work or duties performed at work since their injury. The hours at work variable was coded 0=pre-episode hours, l=reduced hours. The modified duties variable was coded 0=pre-episode duties, l=modified duties. For analysis a binary variable of work modifications was coded 0 if hours at work and duties at work were the same as pre-episode, and 1 i f either the hours at work or the duties at work had been altered. Constancy. Patients were asked if their pain was constant or intermittent. Constancy of pain was recorded into a binary variable: (^intermittent or no pain, l=constant pain. Previous Episodes. Patients were asked i f they had any previous episodes of back or neck pain. A binary variable was created for previous back pain episodes and previous neck pain episodes (0=no, l=yes). Previous Spinal Surgery. A binary variable (0=no, l=yes). 4.7 Ethics In 1990, the International Epidemiology Association (54) concluded that: 51 It is not feasible to obtain the consent, informed or otherwise, of all individuals whose records have become part of a large database such as a nationwide system of linked records, or the archival records of a general hospital. In these and similar situations, informed consent to use such records for epidemiological study may reasonable be delegated to an ethics review committee. Consent for access of the clinical database of CBI Health, was received by the candidate from the Medical Director and Research Department Data Steward. It was not feasible to contact the expected 16 000 individuals whose records would be included in the database. To protect the anonymity of the patients within the clinical database the CBI Research Department (Data Stewart, Lynda Wilson) provided a dataset with the unique identifiers removed from each file. In addition CBI clinics are located only in major Canadian centers. This ensured the protection of confidentiality of personal information and records. In accordance to these guidelines, the Clinical Research Ethics Board at the University of British Columbia approved the ethical issues concerned with this thesis project. Having documented the materials and methods, the next chapter will focus on the statistical analysis required for the dataset. 52 Chapter 5 Statistical Analysis Methods 5.1 Inception Cohort 5.1.1 Descriptive Statistics Preliminary analysis involved descriptive statistics of the assembled study cohort for each variable included in this portion of analysis. Means and percentages were calculated for continuous and categorical study variables to describe the study population. Study population characteristics were described for those with a negative and positive outcome. Additional cross tabulations and correlation coefficients of study variables were computed to investigate possible confounding and to facilitate multivariate model construction. 5.1.2 Logistic Regression Modeling Strategy For this study, logistic regression modeling was used to evaluate associations between explanatory variables and the dichotomous minimally clinically important change in Canadian Back Institute Questionnaire (CBIQ) outcome. The explanatory variables were grouped into three categories: personal characteristics, clinical, and treatment related variables. The three stage logistic regression modeling strategy employed (55) is described below. Figure 5.1 displays an overview of the model building strategy. 5.1.2.1 Univariable Analysis Within each category of variables, univariable logistic regression analysis was performed and crude odds ratios and 95% confidence intervals were generated to investigate the association of each explanatory variable with the dichotomous outcome variable. At this 53 stage the rigid application of a strict 5% significance level for choosing variables was not used. Instead variables with regression coefficient (beta) values for which the P value was less than or equal to 0.10 on the Wald X2 statistic were retained for a multivariate within category model. For categorical variables with more that two categories, i f one or more levels were selected the models were fit using all levels. Due to collinearity, the covariates relating to, number of symptoms, and health seeking behaviour, were not entered into a multivariate model with the two composite covariates that included these items as part of the composite score. A l l variables in the three category specific models with beta values that maintained a P value of less that or equal to 0.10 were included in a multi-category model. 5.1.2.2 Multivariate Analysis The multivariate models were generated by using all significant variables (P value < 0.10) from the three previous category specific models. The provisional multivariate model included variables with coefficient values that demonstrated a significance level of 0.05 or less on the W a l d X statistic. In addition, the a priori decision was made to adjust for age, gender and initial pain intensity. Adjusted odds ratios and 95% confidence intervals were generated for the variables of the model. Bivariate interaction terms were investigated for potential inclusion into the final model. Interaction terms with a P value less than or equal to 0.05 on the Wald X2 statistic were considered for inclusion in a model. The variables involved in the interaction were also retained in these models. Once a provisional model was constructed using the above strategies, a number of variables that had been previously excluded due to a lack of 54 significance were reconsidered. A model validation process considered any gains in performance that could be acquired with the inclusion of a specific variable. This model validation process was used to determine the choice of the final most appropriate model. Figure 5.1- Overview of the logistic regression model building strategy Step 1 Step 2 Comparison Inception Study: — Cohort: acute and acute chronic patients patients only Step 3 Personal characteristics Step 4 Step 5 Clinical variables Treatment variables Re-fit with all significant variables from three category specific models, and interaction terms Validation of various models 5.2 Model Validation In a general sense, model validation provides information that can help one determine i f a model indeed does what it intends to do (that the predictions are accurate for the purpose). Model validation usually involves a method of comparison to determine how well the individual predicted probabilities agree with the actual observed responses (56). A number of validation methods are available to predict the accuracy of a model; however, this essential step in the model building process is seldom reported in the literature (57, 58). When validation is performed, forms of internal validation are commonly used. These validation procedures derive estimates based on the same dataset that was used to build the model. In contrast external validation uses an external source of data, such as data from another centre (perhaps collected by other researchers) to validate the model (59). Temporal validation uses subsequent data from the same centre for the purposes of validation. This method is no different than splitting the data into two datasets seen at different time periods (56). Common techniques for internal validation 55 include: simple sample statistics, split-sample cross validation, and leave-one-out cross validation (Appendix H). Single sample statistics such as, Akaike's Information Constant (AIC) or Schwarz Bayesian Criterion (SC) may over estimate performance because the same data that was used to build the model is also used for validation (60, 61). Split-sample and the various form of cross validation are methods for obtaining nearly unbiased internal assessments of accuracy (62). Cross validation has the additional advantage of not significantly reducing the sample size of the training sample. For this study the models that were constructed were validated with leave-one-out cross validation. Cross validation was used as a method for final model section based on the predictive ability of the model in conjunction with other non-statistical criteria. The discriminative accuracy (the model's performance in discriminating outcome) of the competing models was compared by generating Receiver Operator Characteristic (ROC) curves for each model. An ROC curve was constructed using the predicted probabilities generated from the leave-one-out cross validation for each competing provisional model. The model with the largest area under the ROC curve, and thus the best predictive accuracy, was chosen as the final model. The calibration of probability predictions was also analyzed. A comparison of the predicted and the observed responses in deciles of risk was conducted for the final model using the Hosmer and Lemenshow goodness-of-fit statistic (55). 56 5.4 A Comparison of Acute and Chronic Whiplash Associated Disorders To illustrate the differences between acute and chronic W A D populations, a secondary descriptive comparison was conducted using the entire cohort of patients who completed both entry and exit questionnaires. Patients presenting with acute injury (defined as 91 days or less since their injury date), were compared to patients with early chronic presentations (greater than 91 days but less than 6 months since their injury date), and to patients with chronic injuries (greater than or equal to 6 months since injury date). An available case analysis approach allowed for an examination of most variables in the database using the information on the cases that provided the information specific to a variable while disregarding those that did not provide information on the same variable. These assessments will be valid i f the patients with missing values represent a simple random sample from the cohort (53). Descriptive statistics (means and percentages) were used to describe the distribution of the various study variables and treatment outcomes using the chi-square test for dichotomous variables and the one-way analysis of variance (ANOVA) test for continuous variables. The results were adjusted for multiple comparisons using the Bonferroni correction. For categorical variables, i f the expected frequency count was less than 5 in one of the groups being compared the Fisher's exact test was used. 5.5 Missing Values Methods to assess the presence of self-selection bias were computed for both the inception cohort and the comparison study. The inclusion of only those patients who 57 completed both the entry and exit questionnaire and the V A S of pain intensity resulted in the exclusion of a significant number of records. Descriptive statistics to compare the baseline characteristics of participants and non-participants for key variables were assessed. In addition, two variables of interest were excluded from the inception cohort (previous episodes of spinal pain, and presence of constant pain) due to low response rates (> 50 % missing). Separate univariable and multivariate logistic regression models were constructed to investigate the potential influence of these two covariates on the primary outcome (10% change in questionnaire score), among patients with completed data. A l l analyses were conducted with SAS System for Windows version 8.2 and S-Plus 2000 for Windows. 58 Chapter 6 Results 6.1 Cohort Characteristics 6.1.1 Personal Characteristics A total of 2185 patients who sustained injuries in a motor vehicle collision (MVC) were included in the inception cohort. The mean age of the cohort population was 35.6 years. Fifty-five percent (55.3%) were female, 53.4% were off work due to their M V C , and 38.3% retained a lawyer at some point prior to their discharge from CBI. The mean time between the injury date and the presentation to the clinic was 31.7 days. The majority of the patients were from Ontario (37.6%), British Columbia (29.1%), or Alberta (23.3%). Table 6.1 includes the remaining frequency distributions of the personal, clinical, and treatment related characteristics. 6.1.2 Clinical Characteristics Almost thirty percent (29.7%) of the patients had a primary complaint of neck pain, 55.9% had complaints of both neck and low back pain, while the remainder (14.4%) reported only low back pain. It was common for other areas of the body to be affected. 28.1 % reported headaches, while 43.9% reported pain in the mid-back. The average initial pain intensity rating was 6.0 (Range 0-10, SD=2.0) (Table 6.1). 6.1.3 Treatment Related Characteristics Twenty-two percent (22.4%) of the patients received previous treatment prior to presentation to CBI, and 9.9% received concurrent treatment from another health care provider. The average duration of the treatment program at CBI was 72.7 days, and 59 during this time the patients received an average of 22.9 rehabilitation/treatment sessions (SD=16.5) (Table 6.1). 6.1.4 Outcome Variable In total 1574 of the 2185 patients (72.0%) demonstrated at least a minimally clinically important improvement (10% change or 6.1 points) as measured by the Canadian Back Institute Questionnaire (CBIQ) between initial visit and discharge. The mean change in raw score on the CBIQ for the entire cohort was 13.5 (SD=11.0, median=13). For those defined with a positive outcome (10% improvement in questionnaire score) the mean change in CBIQ was 18.3 (SD=8.6, median=17.0). For those with a negative outcome the mean change in CBIQ was 0.9 (SD=5.0, median=2.0). Table 6.1 - Baseline characteristics of 2185 patients with acute presentation after a motor vehicle collision Variable Total Population (n = 2185) Positive Outcome (£10% change in questionnaire score) (n=1574) Negative Outcome (<10% change in questionnaire score) (n=611) Personal Characteristics Age Mean age - yr (SD) Gender - no. (%) Male Female Lagtime between injury date and clinic presentation Mean duration - days (SD) Median Work status due to collision - no. (%) Off work Still working Physical demands of work - no. (%) Sedentary Light Medium Heavy Missing 35.6(10.6) 977 (44.7) 1208 (55.3) 31.7 (23.7) 25.0 1166 (53.4) 1019(46.6) 363 (16.6) 370(16.9) 514(23.5) 668 (30.6) 270(12.4) 35.3 (10.5) 731 (46.4) 843 (53.6) 30.3 (23.2) 23.0 904 (57.4) 670 (42.6) 253 (16.1) 256 (16.3) 373 (23.7) 514(32.7) 178(11.3) 36.4(10.8) 246 (40.3) 365 (59.7) 35.4 (24.7) 30.0 262 (42.9) 349 (57.1) 110(18.0) 114(18.7) 141 (23.1) 154 (25.2) 92(15.1) 60 Table 6.1 - Continued baseline characteristics of 2185 patients with acute presentation after a motor vehicle collision. —— Retained a lawyer on or before first visit to clinic - no. (%) Yes 672(30.8) 477 (30.3) 195 (31.9) No 1513 (69.2) 1097(69.7) 416(68.1) Retained a lawyer at some point prior to discharge - no. (%) Yes 837 (38.3) 561 (35.6) 276 (45.2) No 1348 (61.7) 1013 (64.4) 335 (54.8) Smoking Status - no. (%) Current/Former 778 (35.6) 588(37.4) 190(31.1) Non-smoker 1407(64.4) 986 (62.6) 421 (68.9) Province - no. (%) British Columbia 635 (29.1) 438(27.8) 197(32.2) Alberta 510(23.3) 335 (21.3) 175 (28.6) Saskatchewan 72(3.3) 50(3.2) 22(3.6) Ontario 822(37.6) 624(39.6) 198(32.4) Quebec 31(1.4) 25(1.6) 6(1.0) Nova Scotia 115(5.3) 102(6.5) 13(2.1) Clinical Characteristics Location of Pain- no. (%) Neck Neck & Back Back Other Area Affected no. (%) Headache Mid-back Upper extremity Lower extremity Number of Symptoms Mean#(SD) Median Neurological Tests - no. (%) Positive Negative Non-organic Signs - no. (%) 3+ Comorbid Conditions - no. (%) Yes No Initial Pain Intensity (range 0-10) Mean intensity - VAS (SD) Median Treatment Related Characteristics Duration of Treatment Program Mean duration -days (SD) Median Number of Treatment Sessions Mean number - # (SD) Median Previous Investigations - no. (%) X-rays Other 649 (29.7) 1222 (55.9) 314(14.4) 613 (28.1) 960 (43.9) 335 (15.3) 274(12.5) 2.6 (1.3) 2.0 33 (1.5) 2152 (98.5) 127 (5.8) 256 (11.7) 1929 (88.3) 6.0 (2.0) 6.0 72.7 (64.6) 56.0 22.9(16.5) 20.0 594 (27.2) 45 (2.1) 480(30.5) 850 (54.0) 244(15.5) 441 (28.0) 696 (44.2) 249(15.8) 192(12.2) 2.5 (1.3) 2.0 23(1.5) 1551 (98.5) 93 (5.9) 173 (11.0) 1401 (89.0) 6.1 (2.0) 6.0 72.6 (65.3) 55.0 22.9(16.9) 20.0 438 (27.7) 33 (2.1) 169 (27.7) 372 (60.9) 70(11.5) 172 (28.2) 264 (43.2) 86(14.1) 82(13.4) 2.6 (1.2) 2.0 10(5.6) 601 (94.4) 34 (5.6) 83 (13.6)) 528 (86.4) 5.9 (2.0) 6.0 72.8 (62.8) 58.0 22.9(15.4) 20.0 156 (25.5) 12(2.0) 61 Table 6.1 - Continued baseline characteristics of 2185 patients with acute presentation after a motor vehicle collision. Previous Treatment - no. (%) Yes 490 (22.4) 332 (21.1) 158 (25.9) No 1695 (77.6) 1242 (78.9) 453 (74.1) Concurrent Treatment - no. (%) Yes 217(9.9) 152(9.7) 65 (10.6) No 1968 (90.1) 1422(90.3) 546 (89.4) Health Care Seeking Mean # (SD) 1.6 (0.9) \.6 (0.9) 1.6 (0.9) Median 1.0 1.0 1.0 6.2 Univariable and Multivariate (Within Category) Logistic Regression Table 2, 3, and 4 display the univariable logistic regression coefficient estimates and associated statistics for the explanatory variables by each category of variable (personal, clinical, and treatment related). In addition, the odds ratios (OR) adjusted for each of the other significant variables within the variable category are presented. 6.2.1 Personal Characteristics The univariable analysis revealed that personal characteristics: older age, female gender, work status (currently working), lawyer retention prior to discharge, province of M V C , less physically demanding work, and lagtime were associated with a negative outcome. A l l of these variables except for the physical demands at work remained significant while controlling for each other (Table 6.2). The age variable was rescaled to reflect a 10-year increase in age. Similarly, the lagtime (between injury and presentation to the clinic) variable was rescaled to a month lagtime. 62 Table 6.2 - Crude and adjusted odds ratios for the association of the variables in the personal characteristic category with a negative outcome (< 10% change in questionnaire score) UNIVARIABLE ANALYSIS MULTIVARIATE ANALYSIS Personal Characteristics Crude OR (95% CI) p-value Adjusted OR (95% CI) p-value Age (per decade) 1.10 (1.06-1.20) 0.037 1.14(1.04-1.25) <0.0001 Gender Male (ref.) Female 1.29 (1.06-1.56) 0.009 1.29 (1.06-1.58) 0.0123 Lagtime between injury and presentation to clinic (per month) 1.30 (1.16-1.46) <0.0001 1.19(1.05-1.35) 0.0058 Work status Off work (ref.) Currently working 1.80(1.49-2.71) <0.0001 1.84(1.49-2.28) <0.0001 Physical demands sedentary (ref.) light medium heavy missing 1.02 (0.75-1.40) 0.87 (0.65-1.17) 0.69 (0.52-0.92) 1.19(0.85-1.67) 0.8814 0.3543 0.0111 0.3142 1.07 (0.71-1.48) 1.02 (0.75-1.39) 0.85 (0.62-1.17) 1.19(0.84-1.71) 0.6953 0.9128 0.3193 0.3299 Layer retention - late No (ref) Yes 1.49 (1.23-1.80) O.0001 1.57 (1.28-1.93) <0.0001 Province British Columbia (ref) Alberta Saskatchewan Ontario Quebec Nova Scotia 1.16(0.91-1.49) 0.98 (0.58-1.66) 0.71 (0.56-0.90) 0.53 (0.22-1.32) 0.28 (0.12-0.52) 0.2376 0.9351 0.0032 0.1746 <0.0001 0.88 (0.67-1.54) 0.90 (0.52-1.56) 0.69 (0.53-0.89 0.63 (0.25-1.58) 0.26 (0.14-0.49) 0.3424 0.7144 0.0048 0.3230 O.0001 Lawyer retention - early No (ref.) Yes 1.08 (0.89-1.32) 0.4643 Not included in multivariate model N A Smoking Status No (ref.) Yes 1.17 (0.84-1.63) 0.3439 Not included in multivariate model N A NA= not applicable because variable not included in multivariate model 6.2.2 Clinical Characteristics The clinical variables pain location (pain located simultaneously in the low back and neck), and the presence of comorbid medical conditions were associated with a negative outcome, both at the univariate and multivariate level (Table 6.3). 63 Table 6. 3 - Crude and adjusted odds ratios for the association of the patient's clinical characteristics UNIVARIABLE ANALYSIS MULTIVARIATE ANALYSIS Clinical Characteristics Crude OR (95% CI) p value Adjusted OR (95% CI) p value Pain Location Neck (ref.) Neck & Back Back 1.24(1.05-1.54) 0.82 (0.59-1.12) 0.0319 0.18 1.24(1.00-1.53) 0.82 (0.59-1.22) 0.0334 0.2663 Comorbid Conditions No (ref.) Yes 1.27(0.96-1.69) 0.0913 1.26 (0.95-1.67) 0.1071 Headache No (ref.) Yes 1.01 (0.82-1.24) 0.9505 Not included in multivariate model NA Mid-back No(ref.) Yes 0.96 (0.78-1.16) 0.6696 Not included in multivariate model NA Pain Radiation to Extremity No (ref.) Yes 1.02 (0.82-1.27) 0.8724 Not included in multivariate model NA Number of Symptoms 1.03 (0.96-1.11) 0.3651 Not included in multivariate model NA Neurological Testing Negative (ref.) Positive 1.12 (0.53-2.37) 0.7630 Not included in multivariate model NA Non-Organic Signs (3+) No (ref) Yes 0.94 (0.63-1.41) 0.6985 Not included in multivariate model NA Initial Pain Intensity 0.97(0.93-1.01) 0.1733- Not included in multivariate model NA NA= not applicable because variable not included in multivariate model 6.2.3 Treatment Related Characteristics A n association between those patients who had received previous treatment and a negative outcome was observed (Table 6.4). 64 Table 6.4 - Crude and adjusted odds ratios for the association of the treatment related characteristics with a negative outcome (< 10% change in questionnaire score) UN I VARIABLE ANALYSIS MULTIVARIATE ANALYSIS Treatment Related Characteristics Crude OR (95% CI) p value Adjusted OR (95% CI) p value Previous Treatment No (ref.) Yes 1.31 (1.05-1.62) 0.0167 1.31 (1.05-1.62) 0.0167 Bed rest No (ref.) Yes 0.52 (0.06-4.42) 0.5459 Not included in multivariate model Duration of treatment (days) 1.00(1.00-1.00) 0.9429 Not included in multivariate model NA Treatment Sessions (no.) 1.00 (0.99-1.00) 0.9820 Not included in multivariate model NA Previous x-rays No (ref.) Yes 0.89(0.72-1.10) 0.2793 Not included in multivariate model NA Other investigations No (ref.) Yes 0.94 (0.48-1.83) 0.8469 Not included in multivariate model NA Concurrent Treatment No (ref.) Yes 1.11 (0.82-1.51) 0.4911 Not included in multivariate model NA Health Seeking Behaviour 1.04(0.94-1.16) 0.4297 Not included in multivariate model NA NA= not applicable because variable not included in multivariate model Based on these results the following variables from all categories were included in a multivariate model: age, gender, work status, lawyer retention prior to discharge, province, lagtime, pain location, comorbid medical conditions, and previous treatment. 6.3 Multivariate (All Variable Categories) Logistic Regression At the more stringent 5% significance level, and while controlling for variables from all categories, the following characteristics were associated with a negative outcome: older age, female gender, work status (currently working), lawyer retention prior to discharge, province of M V C , increasing lagtime between injury date and treatment, and pain 65 location. The parameter estimations and associated statistics are displayed in Table 6.5 This model was additionally adjusted for the initial pain intensity. Table 6.5 - Adjusted odds ratios for the association of the significant variables from the three Variable Parameter Estimate SE OR (95% CI) p value Age (per decade) 0.159 0.00498 1.17 (1.06-1.29) O.0014 Gender Male (ref) Female 0.2180 0.1094 1.24(1.04-1.54) 0.0462 Work status Off work (ref) Currently working 0.7136 0.1143 2.04(1.63-2.55) O.0001 Lagtime between injury and presentation to clinic (per month) 0.059 0.0023 1.19 (1.04-1.37) 0.0114 Lawyer Prior to Discharge No (ref) Yes 0.4638 0.1162 1.59 (1.26-2.00) O.0001 Pain Location Back (ref) Neck & Back Neck 0.3983 0.2390 0.1662 0.1803 1.49 (1.08-2.01) 1.27 (0.89-1.81) 0.0165 0.1849 Province British Columbia (ref) Alberta Saskatchewan Ontario Quebec Nova Scotia -0.0959 0.0736 -0.2482 -0.2843 -1.1201 0.1546 0.2988 0.1382 0.4755 0.3189 0.91 (0.67-1.23) 1.08 (0.89-1.81) 0.78 (0.60-1.02) 0.75 (0.30-1.91) 0.33 (0.18-0.61) 0.5350 0.8053 0.0725 0.5500 0.0004 Initial Pain Intensity -0.0127 0.0266 0.99 (0.94-1.04) 0.6760 6.4 Model Selection and Validation A total of 17 different models containing interaction terms and other covariates were compared using cross validation techniques. The areas under the Receiver Operator Characteristic (ROC) curves for these models and the defining characteristics are displayed in Table 6.6. These provisional models were generated, to see i f the inclusion of interaction terms or a variable that was previously eliminated due to a lack of significance (during the univariable or multivariate analysis) could improve the predictive ability of the final model. 66 Table 6.6 - Defining characteristics and area under the ROC curve for the competing models Competing Models Model Characteristics Area under ROC curve Model 1 Significant variables adjusted for pain intensity - age, gender, work status, lawyer, lagtime, pain location, intensity, province 0.632 Model 2 Model 1 with smoking status 0.634 Model 3 Model 1 with comorbid variable 0.631 Model 4 Model 1 with duration of treatment variable 0.630 Model 5 Model 1 with non-organic signs 0.627 Model 6 Model 1 with smoking status, comorbid, and duration variable 0.631 Model 7 Model 1 with second order interaction terms for province variable 0.615 Model 8 Model 1 with second order interaction terms for intensity variable 0.634 Model 9 Model 1 with second order interaction terms for age variable 0.628 Model 10 Model 1 with second order interaction terms of smoking status variable 0.632 Model 11 Model 1 with second order interaction terms for lawyer variable 0.630 Model 12 Model 1 with second order interaction terms for gender variable 0.629 Model 13 Model 1 with significant interaction terms: intensity*lawyer, intensity*work status 0.636 Model 14 Model 13 with smoke status 0.640 Model 15 Model with all confounding variables and significant interaction terms 0.605 Model 16 Parsimonious model including: age, gender, work status, lawyer, intensity 0.614 Model 17 Parsimonious model including: age, gender, province, intensity 0.591 Many of the models demonstrated similar areas under the ROC curve. The manipulation of a few similar variables produced the models with the best discriminative ability. Generally, models containing the variable smoking status, and/or the interaction terms of initial pain intensity*lawyer and initial pain intensity*work status performed better than models without these terms. Not surprisingly, including a number of non-significant interaction terms or erroneous variables resulted in a reduction in the performance of the model. A number of simple models containing fewer variables (Model 16 and 17 in Table 6) resulted in less discrimination by the model. Model 14 containing the significant interaction terms and the smoking status variable was chosen as the best model based on having the largest area under the ROC curve. The overall agreement between the observed and predicted outcomes at deciles of risk in this model was assessed with the Hosmer-Lemenshow test statistic (55). The test statistic was not significant (p=0.24) indicating that the logistic model adequately describes the risk throughout all of the risk groups. Table 6.7 displays the observed and predicted responses for both negative and positive outcome for the 10 risk subgroups. 67 Table 6.7- Partitions for the Hosmer and Lemeshow test for the model chosen by validation Negative Outcome Positive Outcome Risk Total Observed Expected Observed Expected Group 1 218 15 23.02 203 194.98 2 218 43 34.79 175 183.21 3 220 39 43.08 181 176.92 4 218 47 49.66 171 168.34 5 220 68 56.77 152 163.23 6 218 59 62.05 159 155.95 7 220 72 68.98 148 151.02 8 218 79 75.90 139 142.10 9 218 81 86.12 137 131.88 10 215 106 108.60 109 106.40 6.5 Final Multivariate Study Model In the final model: older age, female gender, increased lagtime between injury and presentation to the clinic, and pain location (simultaneous back and neck pain) were associated with a negative outcome. Those patients whose motor vehicle collision occurred in Nova Scotia or Ontario when compared to British Columbia (reference category) were less likely to have a negative outcome (Table 6.8). In addition, interaction terms between initial pain intensity and the retention of a lawyer, and initial pain intensity and work status were significant. The adjusted odds ratios for the variables lawyer retention and work status at three levels of pain intensity (mild, moderate, and severe) are displayed in Table 6.9. The effect of lawyer retention on outcome was stronger for those individuals with less intense pain. Similarly the effect of work status on outcome diminished as the initial pain intensity increased. The smoking status variable was not significantly associated with the outcome; however, the inclusion of this variable led to better predictive ability by the logistic model, and thus it was also included. 68 Table 6.8 - Adjusted odds ratios for the association of the variables in the final study model Variable Adjusted Odds Ratio (95% CI) p value Age (per decade) 1.14(1.04-1.25) 0.0038 Gender - Female 1.303 (1.067-1.593) 0.0096 Lagtime (per month) 1.20(1.06-1.37) 0.0049 Pain Location Neck & Back vs. Back Neck vs. Back 1.423 (1.046-1.934) 1.208 (0.866-1.686) 0.0246 0.2663 Province Alberta vs. British Columbia Saskatchewan vs. British Columbia Ontario vs. British Columbia Quebec vs. British Columbia Nova Scotia vs. British Columbia 0.926 (0.704-1.217) 1.012(0.584-1.755) 0.725 (0.559-0.939) 0.707 (0.280-1.786) 0.279 (0.150-0.517) 0.5806 0.9664 0.0147 0.4627 O.0001 Smoking status (current of former smoker) 0.812(0.658-1.002) 0.0534 Intensity*lawyer retention prior to discharge See Table6.9 0.0020 Intensity*work status (currently working) See Table 6.9 0.0460 Table 6.9 - Adjusted odds rations for the association of lawyer retention and work status and negative outcome stratified by initial pain intensity Variable - Stratification Odds Ratio (95% CI) Lawyer retention stratified by pain intensity. Mild intensity (VAS 0-4) Moderate intensity (VAS 5-7) Severe intensity (VAS 8-10) 2.17(1.34-3.57) 1.57(1.17-2.11) 1.26 (0.66-2.41) Work Status (currently working) Mild intensity (VAS 0-4) Moderate intensity (VAS 5-7) Severe intensity (VAS 8-10) 3.50(2.14-5.73) 1.9.1 (1.43-2.54) 1.03 (0.50-2.13) 6.5.1 Sensitivity and Specificity of Study Model The area under the ROC curve for this final study model was 0.64. The sensitivity, specificity, and the number of false positives and false negatives varied depending on the choice of cut-point used to classify a predicted outcome as positive or negative based on the subjects predicted probability. A number of different cut-points and the corresponding sensitivity, specificity, false positives, and false negatives were calculated (Table 6.10). Using a cut-point of 0.24 the sensitivity of the final study model to predict those patients with a negative outcome was 0.74, and the specificity was 0.45. This cut-69 point is optimal i f the objective is minimizing the total number of false positives and false negatives (63). Table 6.10 - Sensitivity and Specificity of the final study model with varying cut points . Cut-Point Sensitivity Specificity False False (%) (%) positives negatives (%) (%) 0.180 94.7 17.0 83.0 5.3 0.200 88.5 26.2 73.8 11.5 0.220 81.4 35.6 64.4 18.6 0.240 74.1 44.9 55.1 25.9 0.260 65.5 53.4 46.6 34.5 6.6 Missing Values 6.6.1 Inception cohort - Differences between study population and excluded records A total of 3,060 patients were eligible for inclusion in the cohort study to investigate prognostic factors associated with delayed recovery and disability among acute W A D patients. Of these, 875 patients were excluded because they did not complete a Visual Analogue Scale of their initial pain intensity. A comparison of patients with and without the V A S data are displayed in Table 6.11. Those with V A S data were identified by the following characteristics: more likely to have hired a lawyer, less likely to have taken time off work due to their injuries, less likely to have comorbid medical conditions, and less likely to have had previous treatment. The completion of the V A S scale also varied depending on the province of origin. Quebec, Alberta and Saskatchewan had participation rates of 91.2%, 89.0%>, and 89.0% respectively, whereas those injured in Nova Scotia, Ontario and British Columbia participated at rates of 78.2%, 68%, and 62.5% respectively. This resulted in differences in the geographic makeup of the study population and the subjects that were excluded. There was not a significant difference in the overall outcome between the two groups or in the following study variables: age, 70 gender, lagtime, smoking status, or duration of treatment. The differences between individuals who responded to both the entry and exit questionnaires and those who did not are displayed in Table 1 of Appendix I. Those that completed both questionnaires were less likely to smoke or have comorbid medical conditions, and the duration of their treatment programs was longer by ten days; however, most of the observed differences were not considered clinically significant. Table 6.11 - Missing Data - Characteristics associated with completion of initial pain intensity variable (VAS) Variable Study Missing Difference SE of p value Population VAS between groups difference (n=2185) (n=875) (95% CI) Positive Outcome (>10% change in questionnaire score) - % 72.0 73.6 -1.6 (-5.1, 1.8) 0.0177 NS Age - mean years (SD) 35.6(10.6) 35.8(10.1) 0.3 (-1.1,0.58) 0.4252 NS Gender - % female 55.3 57.9 -2.6 (-6.5, 1.2) 0.0198 NS Lagtime - mean days (SD) 31.7(23.8) 30.5 (22.8) 1.2 (-0.7, 3.0) 0.9429 NS Duration of treatment - mean days 72.7 (64.6) 76.6 (70.6) -3.9 (-9.2, 1.3) 2.6650 NS (SD) Work status - % off work 53.4 63.9 -10.5 (-14.3,-6.7) 0.0194 O.0001 Lawyer Retention % yes 30.7 27.0 3.7 (0.2, 7.2) 0.0179 0.0036 Smoking Status - % current/former 35.6 34.5 1.1 (-2.6,4.8) 0.0191 NS Comorbid Medical Conditions - % 5.5 11.7 -6.2 (-8.5, -3.9) 0.0119 <0.0001 Previous Treatment - % yes 9.1 22.4 -13.3 (-16.3,-10.3) 0.0154 <0.0001 Non-organic signs - % with 3+ 3.7 5.8 -2.1 (-3.8, -0.3) 0.0089 0.0152 Geographic Composition - % of group composed of British Columbia 29.1 43.5 -14.4 (-18.2,-10.6) 0.0193 <0.0001 Alberta 23.3 7.2 16.1 (13.6, 18.6) 0.0126 <0.0001 Saskatchewan 3.3 1.0 2.3(1.3,3.3) 0.0051 0.0004 Ontario 37.6 44.2 -6.6 (-10.5,-2.7) 0.0197 0.0007 Quebec 1.4 0.3 1.1 (0.5, 1.7) 0.0031 0.01 Nova Scotia 5.3 3.7 1.6 (0.0,3.1) 0.0080 NS 6.6.2 Missing Covariate Values Two variables (previous spinal pain episodes and the constant vs. intermittent pain variable) were excluded from the model building exercise due to low response rates (> 50% missing values). Separate univariable and multivariate (adjusted for other predictors found to be associated with outcome) logistic regression models were constructed to investigate the potential influence of these two covariates on the primary outcome among 71 patients with completed data. Among the 912 individuals who provided information regarding previous episodes of spinal pain those who reported previous episodes were less likely to have a negative outcome (p=0.0084). The adjusted odds ratio (95% CI) for this association was 0.653 (0.475-0.897). A significant association between the constant vs. intermittent pain variable was not observed in the 921 individuals who provided this information (adjusted OR (95% CI) = 1.145(0.784-1.674), p=0.48). 6.7 Comparison Study of Acute and Chronic Whiplash Associated Disorders During the study period a total of 6195 patients who sustained injuries in a motor vehicle collision were identified in the CBI clinical database. Of theses, 5581 of these patients met the inclusion criteria for a comparison study of the acute and chronic patient populations. 6.7.1 Comparison Cohort Characteristics Of the entire group, 3916 (64.8%) reported a positive outcome, after a mean treatment duration of 73.19 days (SD=63.42, median=57.0 days). During this time period each patient on average attended 22.6 rehabilitation sessions (SD=16.1, median=20). A comparison of outcomes as well as the personal, clinical and treatment related characteristics between the acute (< 91 days since injury date) and chronic (> 6 months since injury date) patient populations revealed several significant differences. t 72 6.7.1.1 Personal Characteristics Table 6.12 provides a summary of the personal characteristics of the acute and chronic populations. Those individuals with chronic pain and disability after a motor vehicle collision were more likely to be female, and to have retained a lawyer. Chronic patients were more likely to have returned to work; however, less physically demanding work was more common in this population. The distribution of chronic patients across the provinces differed from the distribution of acute patients across the provinces. A larger proportion of the chronic population was composed of patients from British Columbia, Alberta, Saskatchewan and Quebec, when compared to the proportion of patients from these provinces in the acute population. The median lagtime between injury and presentation to the clinic for each of the provinces was: British Columbia (88 days), Alberta (136 days), Saskatchewan (138 days), Ontario (21 days), Quebec (148 days), and Nova Scotia (28 days). Table 6.12 - A comparison of personal characteristics in individuals injured in a motor vehicle collision Personal Characteristics Acute Chronic Difference SE of p value (N = 3075) (N = 1548) Acute-Chronic (95% CI) Difference Age no. of respondents 3075 1548 Mean age - yr(SD) 35.7(10.6) 36.2(10.9) -0.5 (-1.5,0.2) 0.3335 0.2876 Female Sex - no./total no. 0.0002 (%) 1719/3075 (55.9) 954/ 1548 (61.6) -5.7 (-8.7, -2.7) 0.0153 Lagtime no. of respondents 3053 1548 Mean no. of days (SD) 31.4 (23.5) 434.0(321.6) -403 (-414, -391) 5.8503 N/A Median 25 348 Off work because of collision no. / total no. (%) 1749/3075 (56.9) 506/ 1548 (32.7) 24.2 (21.2, 27.1) 0.0149 <0.0001 Physical Demands of Work -no./total no (%) Sedentary 447/3075(14.5) 239/ 1548(15.4) -0.9 (-3.0, 1.3) 0.0111 0.4152 Light 480/3075 (15.6) 282/ 1548 (18.2) -2.6 (-4.9, -0.3) 0.0118 0.0024 Medium 682 / 3075 (22.2) 305/ 1548 (19.7) 2.5 (0.0-0.5) 0.0126 0.0525 Heavy 875/3075 (28.5) 377/ 1548 (24.4) 4.1 (1.4, 6.8) 0.0136 0.0031 Missing 591 /3075 (19.2) 345/1548 (22.3) -3.1 (-5.6, -0.6) 0.0127 0.0143 Modified work duties because of collision. no. / total no (%) 466/ 1310*(35.6) 279/ 1024*(27.2) 8.4(4.6, 12.2) 0.0191 <0.0001 73 Table 6.12 - Continued comparison of personal characteristics in individuals injured in a motor vehicle collision. Retained a lawyer on or before first visit to clinic no./total no. (%) 913/3075 (29.7) 943/ 1548 (60.9) -31.2 (-34.1,-28.3) 0.0149 <0.0001 Retained a lawyer at some point prior to discharge no./total no. (%) 1127/3075 (36.7) 935/ 1548 (60.4) -23.7 (-26.7,-20.1) 0.0151 <0.0001 Smoking Status - current / former - no./total no. (%) 1054/ 3072 *(34.3) 561 / 1547*(36.3) 2.0 (-4.9, 0.9) 0.0149 0.1888 Province - no./total no. (%) British Columbia 1020/3075 (33.2) 573/ 1548 (37.0) -3.8 (-6.7, -0.9) 0.0149 0.0094 Alberta 575 /3075 (18.7) 563/ 1548 (36.4) -17.7 (-20.5,-14.9) 0.0141 <0.0001 Saskatchewan 81 /3075 (2.6) 96/ 1548 (6.2) -3.6 (-4.9, -2.3) 0.0068 <0.0001 Ontario 1216/3075 (39.5) 259/ 1548 (16.7) 22.8 (20.2, 25.3) 0.0129 . 0.0001 Quebec 43 /3075 (1.1) 36/ 1548 (2.3) -1.2 (-2.3,-0.3) 0.0042 <0.0001 Nova Scotia 149/3075 (4.9) 21 / 1548 (1.4) 3.5 (2.4, 4.4) 0.0050 <0.0001 *Number of respondents for each patient population that answered the survey item 6.7.1.2 Clinical Characteristics The acute population reported more symptoms on presentation than the chronic population (Table 6.13). Specifically, additional complaints such as simultaneous neck and low back pain, and mid-back pain were more common. The chronic population; however, reported more low back and lower extremity syndromes, and demonstrated pain focused behaviour as determined by Waddel's non-organic signs more often (47). Comorbid medical conditions were also more common in the chronic population. There was a small difference in the initial pain intensity between the two populations. 6.7.1.3 Treatment Related Characteristics Not surprisingly, the chronic population was more likely to have received previous treatment and investigations prior to their presentation to CBI (Table 6.13). Regarding their rehabilitation programs at CBI, there was no difference in the overall duration of the treatment programs between the two populations; however, the acute population received more treatment sessions. 74 Table 6.13 - A comparison of clinical and treatment related characteristics in individuals injured in motor vehicle collisions Clinical Characteristics Acute Presentation Chronic Difference SEfor p value (N = 3075) (N = 1548) Acute-Chronic (95% CI) Difference Location of Pain - no./total no. (%) Neck 916/3075 (29.8) 399/ 1548 (25.8) 4.0(1.3,6.7) 0.0138 0.0043 Neck & Back 1764/3075 (57.4) 819/ 1548 (52.9) 4.5 (1.4, 7.5) 0.1555 0.0040 Back 395/3075 (12.8) 330/ 1548 (21.3) -8.5 (-10.9, -6.1) 0.0120 <0.0001 Areas Affected - no/total no. (%) Headache 772/3075 (25.1) 382/ 1548 (24.7) 0.4 (-0.2, 3.0) 0.0135 0.7506 Mid-Back 1227/3075 (39.9) 544/ 1548 (35.1) 4.8(1.9, 7.7) 0.0150 0.0017 Extremity Pain - no./total no. (%) Upper Limb 422/3075 (13.7) 196/ 1548(12.7) 1.0 (-1.1,3.0) 0.0105 0.3166 Lower Limb 331 /3075 (10.8) 280/ 1548 (18.1) -7.3 (-9.5, -5.1) 0.0113 <0.0001 Number of Symptoms Mean (SD) 2.47(1.24) 2.43 (1.31) 0.04 (-0.04, 0.11) 0.0394 0.1161 Neurological Testing - no./total no. (% Positive) 39/3075(1.3) 25/ 1548 (1.6) -0.3 (-1.0, 0.4) 0.0038 0.3410 Non-Organic Signs - no./total no. (%) Any 344/3075 (11.2) 234/ 1548(15.1) -3.9 (-6.0,-1.2) 0.0107 <0.0001 3+ 160/3075 (5.2) 95/ 1548 (6.1) -0.9 (-0.2, 0.5) 0.0073 0.1891 Previous Episodes - no./total no. (%) Neck Pain 683/910(75.1) 398 / 527 (75.5) -0.4 (-5.0,4.2) 0.0236 0.8434 Back Pain 471 / 771 (61.2) 333/504(66.1) -4.9 (-10.3, 0.4) 0.0274 0.0715 Previous Spinal Surgery - no. /total no. (%) Cervical 3/3075 (0.1) 5 / 1548(0.4) -0.3 (-0.6, 0.0) 0.0017 0.0347 Lumbar 12/3075 (0.4) 15/ 1548(1.0) -0.6 (-1.1,-0.6) 0.0028 0.0148 Comorbid Conditions no./total no. (%) - Yes 309/3075 (10.1) 196 / 1548(12.7) -2.6 (-4.6, -0.6) 0.0100 0.0072 Presence of Constant Pain (Yes) no./total no. (%) 918/ 1172 (78.3) 490 / 642 (76.3) 2.0 (-0.6, 4.5) 0.0131 0.3276 Intensity of Pain Intensity Mean VAS (SD) 6.0 (2.0) 5.8(2.1) 0.2 (0.06, 0.34) 0.0729 0.0224 Treatment Characteristics Bed rest previously prescribed -no./total no. % 6 / 3075 (0.2) 13/ 1548 (0.8) -0.6 (-1.1,-0.1) 0.0024 0.0012 Duration or treatment program Mean days (SD) 73.8 (66.3) 73.6 (60.0) 0.2 (-3.7, 4.1) 2.0069 0.3891 Number of treatment sessions Mean # (SD) 23.7(17.1) 21.0(14.5) 2.6(1.8,3.8) 0.5066 <0.0001 Treatment Costs ($) Mean (SD) 2097(1779) 1724(1601) 373 (268, 478) 53.657 O.0001 Previous Investigations -no../total no. (%) X-rays 737 / 3075 (24.0) 572/ 1548 (37.0) -13 (-15.8,-10.2) 0.0145 <0.0001 Other 58/3075 (1.9) 159/ 1548(10.3) -8.4 (-10, -6.8) 0.0081 <0.0001 Previous Treatment - no./ total no. (%) Yes 579/3075 (18.8) 774/ 1548 (50.0) -31.2 (-34, -28) 0.0145 <0.0001 Concurrent Treatment - no./total no. (%) Yes 251 /3075 (8.2) 248/ 1548(15.6) -7.4 (-9.5, -5.3) 0.0105 <0.0001 Health care seeking - no./total no. (%) Mean (SD) 1.5 (0.8) 2.1 (1.2) -0.6 (-0.7, -0.5) 0.0304 <0.0001 75 6.7.1.4 Outcome Variable Seventy-two percent (72.3%) of the patients who presented to the clinic in an acute state demonstrated a positive outcome, in comparison to 51.6% of the chronic patients (Table 6.14). Those with a chronic whiplash injury were not only less likely to demonstrate a minimal clinically important improvement in disability, but also demonstrated less overall improvement on the CBIQ. Table 6.14- Comparison of outcomes in individuals injured in a motor vehicle collision (acute vs. chronic populations) Outcome Acute Chronic Difference SE for p value 9^1 days ^6mos Acute-Chronic Difference (N = 3075) (N = 1548) (95% CI) Change in CBIQ - no./total no. - % with positive outcome 2224 / 3075 (72.3) 806/1548 20.2 (17.3,23.1) 0.0150 O.0001 (> 10% change) (52.1) Mean Change in CBIQ Mean (SD) 13.8(11.1) 6.8 (8.4) 7.0 (6.4, 7.7) 0.321 O.0001 Median 13.0 6.0 CBIQ Raw Score at Presentation* 29.3 (10.5) 32.2 (9.0) -2.9,(-3.5,-2.3) 0.311 <0.0001 Mean (SDO 29.0 33.0 Median CBIQ Raw Score at Discharge* Mean (SD) 43.1 (12.2) 39.0(10.2) 4.1 (3.4,4.8) 0.361 O.0001 Median 43.0 39.0 * Raw CBIQ scores out of 61. A higher score indicates less disability. 6.7.2 Early Chronic Pain An additional comparison group was composed of individuals who presented after 91 days but prior to 6 months. A number of differences between the acute population and this early chronic group were evident (see Appendix J, tables 1-4). Most of the differences echoed the findings in the above comparison with the chronic population, although at an earlier time. The early chronic group were more likely to have returned to work, to have retained a lawyer, to have demonstrated pain focused behaviour, to have low back syndromes and to have received previous treatment and investigations. Some trends were apparent in the data with the addition of the early chronic comparison group. 76 In particular the proportion of individuals who had a positive outcome, who were female and who retained a lawyer increased from acute through early chronic to chronic presentation. 77 Chapter 7 Discussion 7.1 Introduction The identification of predictors of early recovery after a motor vehicle collision is an important step in understanding and potentially reducing the burden of illness created by whiplash-associated disorders (WAD). This study has found a number of predictors for poor recovery in the secondary care (physiotherapy) setting, to be considered for further confirmatory evaluation. While adjusting for other covariates, older age, female gender, increasing lagtime between injury date and presentation for treatment, initial pain location, and province of injury were associated with a poor outcome (<10% improvement in disability by questionnaire score over the treatment period). In addition the interaction between work status and initial pain intensity, as well as, lawyer retention and initial pain intensity was found to be associated with a negative outcome, and as such, must be considered for future studies. 7.2 Prognostic Variables from Multivariate Logistic Regression Age. The odds of a poor outcome following whiplash-associated disorder (WAD) increased with each year increase in age. Older individuals were more likely to demonstrate minimal early improvement in comparison to their younger counterparts. This predictive ability of age has been demonstrated in previous studies (1,6) investigating whiplash recovery outcomes in both population- and hospital-based cohorts. For example, Cassidy and his colleagues reported that in the province of Saskatchewan there was approximately a 10% decrease in the time-to-claim closure for each decade increase in age (2). Biologically, older individuals may take longer to heal after a 78 musculoskeletal injury and the related prolonged inactivity (64, 65), or have less ability to compensate for the functional deficiencies that resulted from the motor vehicle collision (MVC). Potentially, this could have resulted in the slower improvements in the early stages of recovery observed in this study. Some have theorized that degenerative changes (24) in older individuals may be related to delayed recovery. Gender. Several studies, investigating different source populations, have reported the importance of gender as a prognostic factor for recovery (2, 6, 16). This study observed a similar finding. It is unclear why women have a delayed recovery from W A D . Some authors have suggested that women are more vulnerable to musculoskeletal injury due to smaller muscle mass in the neck or smaller cross-sectional area of the muscle fibers (16, 65). Alternatively, some authors have highlighted that women and men experience pain differently (66). Keogh and Herdenfeldt have suggested that women report more pain experiences and more negative responses to pain when compared to men, and demonstrate lower thresholds and tolerance to a variety of noxious stimuli (66). In addition, men and women use different coping strategies when under stress (66). A number of these biologically and psychologically based theories could have explained the gender differences in this study. Lagtime. The current study found that patients who waited the longest to seek treatment at a CBI clinic following their injury were less likely to demonstrate further meaningful improvement during the follow up period. Each additional month of lagtime was associated with an approximately 20% increase in the odds of a poor outcome. Several 79 sources (1, 52, 67) support an early, active approach to treatment for W A D . Interventions such as early exercise and advice to return to normal daily activities have been shown to shorten recovery time (67-69). Individuals who presented late to a CBI clinic may not have received the preceding interventions in a timely fashion post injury, resulting in a poor recovery outcome. Alternatively, whiplash-associated disorders are thought to have a favourable natural history in that they tend to resolve with time (1). As such, one might consider that patients with the longest lagtime reap the benefits of the additional healing window. However, at some point after a motor vehicle collision (MVC) the behavioural and psychological consequences of having a W A D become established. This may have occurred in those patients that delayed in their presentation to CBI and/or received a non-active therapeutic approach elsewhere. Initial Pain Location. In this study injuries to the neck, with or without low back pain, were associated with a poor outcome. Several studies have examined the initial postcollision symptoms and their effect on recovery. Depending on the study, various symptoms, such as neck pain on palpation (18), headache (70), upper back pain (3), low back pain (71), radicular symptoms (6), and number of total symptoms (3) have been associated with delayed recovery or poor outcome. Generally those individuals with more bodily areas affected would likely demonstrate increased disability. It is reasonable to expect that these individuals have a more serious condition and may take longer to heal. In this regard, those patients in the current study with simultaneous neck and back pain were more likely to have a poor outcome. The neck could be considered a more vulnerable area for injury to occur, and this could explain the association of neck pain 80 and lack of meaningful improvement in comparison to those individuals with low back pain alone. The current study did not find a significant association with headache, mid-back pain and referred extremity pain with the outcome. There is inconsistent evidence in the literature regarding the potential effect of postcollision headache or mid back pain symptoms on recovery; however, radicular signs and symptoms are an established prognostic factor. It is unclear why the referral of pain to the extremity and/or positive neurological testing (radicular symptoms) was not associated with a negative outcome in the current study (although a trend of this nature was apparent during univariable analysis). The CBI focuses on the conservative treatment of mechanical spinal pain of musculoskeletal origin. The physiotherapists are specifically trained in the management of referred extremity pain, and early efforts are made to demystify these symptoms with the patient. Perhaps this approach of patient education and active treatment for these symptoms resulted in the timely alleviation of the anxiety and disability that radicular symptoms may have created in other cohorts. Province. A higher percentage of patients from Ontario and Nova Scotia demonstrated a positive outcome in comparison to the reference category, British Columbia. It is unclear what effect the province of residence had on determining clinical improvement to a standardized treatment program at the Canadian Back Institute (CBI). The Quebec Task Force on Whiplash-Associated Disorders (QTF) suggested that the insurance jurisdiction where an injury occurred could alter prognosis (1). This effect was subsequently confirmed in the province of Saskatchewan when the introduction of no-fault legislation that eliminated pain and suffering compensation resulted in faster recovery (2). A similar 81 effect was seen in the Australian state of Victoria, when legislation limiting compensation was introduced (7). In addition, in countries where there is little public notion of chronic symptoms and disability associated with motor vehicle collisions and litigation is limited, the condition follows a more favourable course (8, 72, 73). Previous reports (2, 6) have also highlighted that a large variation in recovery time exists even within jurisdiction with common insurance systems. For example in the provinces of Quebec and Saskatchewan, there is a wide variation in the median recovery time, despite the fact that both provinces use no-fault insurance systems (2, 5). In addition to the compensation system of the province, other policies of the insurance jurisdiction or cultural factors may be important. In the current study, British Columbia, Alberta, Ontario, and Nova Scotia operated under tort insurance systems, where pain and suffering settlements are provided. Alternatively, the provinces of Saskatchewan, and Quebec operated under no-fault insurance systems, where pain and suffering settlements are not provided. Although it is believed that recovery is faster in no-fault jurisdictions, a detailed comparison between those provinces with such insurance systems was complicated by other observed differences between the provinces. For example, in the current study, variation across the provinces existed for the lagtime between injury and presentation to the clinic, the duration and number of treatment sessions, and the cost of the treatment program (see Table 1, Appendix K). Likely, social forces and insurance policy are influencing these differences. It is impossible to determine from this data how the complex interaction of factors determined by insurance policy may have altered the clinical course and prognosis of those injured in a M V C ; however, this study found the 82 province of the compensation system to be an important predictor or recovery, and worthy of further, more detailed investigation in future studies. Work status. In this study, those individuals who took time off work were more likely to demonstrate a clinically important improvement in disability questionnaire score. While there are conflicting results regarding this variable in the literature, the findings of this study are not consistent with other Canadian reports (2, 5). This is likely due to the differences in outcome measures between the three cohorts. The previous studies used a primary outcome of recovery time (time-to-claims closure, or time on compensation); however, in the present study a minimum clinically important change in recovery by questionnaire score was used as the primary outcome. As such, the time-to-claim closure may have been less in the individuals who remained at work. An examination of the disability scores at discharge revealed that those individuals who remained at work had higher overall function scores (less disability) than individuals who were off work. In addition, the off work group were more disabled on presentation (mean CBIQ score at presentation 25.2 as compared to 35.8 for the working population) and had more physically demanding jobs (49.4 % self-reported heavy labour as compared to 17.9% heavy labour in the working population). The off work individuals in this cohort may have represented a sub-group of patients who benefited from a break from their heavy work duties while remaining active and receiving education about their injuries in the structured environment of CBI. In addition, as these off work individuals were more disabled on presentation than their working counterparts more room for subsequent improvement as measured by the CBIQ was possible. 83 Lawyer Retention. This study found a similar association between lawyer retention and poor outcome that has been reported previously (2). After a M V C , individuals who have sustained injuries, have taken time off work, had a disruption of their normal daily activities, or are concerned about their recovery, may seek compensation for their injuries. In this sense it is not surprising that legal representation was associated with poor outcome. However, this study controlled for these factors (daily activities, work status, and severity of the initial complaints), and lawyer retention remained significantly associated with poor outcome. Some authors have argued that poor recovery is inevitable when one must repeatedly prove their illness (7, 74). Having a lawyer involved with a case presumes that an individual is seeking compensation for their predicament. In this context an individual will need to prove that the magnitude of their pain, suffering, and disability are grounds for compensation, often in an adversarial environment, in which others will be denying the existence of serious injury. In this regard, the process of seeking compensation and hiring a lawyer may cause an individual to become focused on their symptoms in comparison to one not involved in litigation. This could lead not only to the poor treatment outcome demonstrated in our study, but also the delayed recovery and increased costs associated with lawyer retention reported by others. Initial Pain Intensity. Several studies have reported that the intensity of the presenting complaints is an important prognostic factor (1,6, 17). The current study did not observe such an association in multivariate analysis. Using a change in questionnaire score as the primary outcome in this study likely resulted in partial control for the intensity of initial 84 symptoms. As the CBIQ asks the individual to rate how their pain has affected the various activities of daily life, it reasons that the intensity of the initial pain would also influence these same activities. Indeed a significant correlation (r = -0.38, p<0.0001) exists between the initial pain intensity and initial CBIQ raw score. Therefore, the intensity of initial pain is at least partially controlled for due to the nature of the outcome. In addition, linear regression modeling revealed that the initial pain intensity was significantly associated with the raw CBIQ exit score while controlling for age and gender (p<0.0001). It is important to also note that those individuals with more intense complaints on presentation will have lower initial scores on the CBIQ (i.e. higher disability), thus these individuals have more room for improvement while attending treatment at CBI, and therefore have a better chance of obtaining a minimum clinical improvement. A trend of this sort is apparent in the data. Individuals with mild pain (VAS 0-4) had on average a raw change in CBIQ score of 12.8 points (SD = 10.8), whereas individuals with moderate (VAS 5-7) and severe (VAS 8-10) intensity pain had average changes of 14.0 (SD = 11.0) and 13.9 (SD =11.3) respectively. Finally, while the individuals with initially more intense pain did have more room for improvement, their overall disability was still greater at discharge. For instance, the average raw CBIQ exit score for those with the most intense pain was 36.8 (SD=12.1), while those with the mild initial pain on average scored 46.2 (SD=11.1) on discharge. In summary, this study did not find a significant association with the intensity of initial pain and poor recovery during multivariate analysis; however, it appears that this is due to the nature of the outcome. 85 Interaction Terms. The inclusion of the intensity variable revealed two significant interaction terms. The initial pain intensity (VAS) variable was involved in an interaction with both the lawyer retention and work status variables. As the intensity of pain increased, the strength of the association between lawyer retention and work status and delayed recovery diminished. No previous whiplash studies have thoroughly reported the impact of interactions among covariates. Individuals with mild (VAS 0-4) and moderate (VAS 5-7) intensity pain at initial presentation, this study found that the retention of a lawyer was associated with a lack of meaningful improvement. A similar effect has been described in individuals recovering from a closed head injury (75). In this study Binder and his colleagues reported that the impact of financial incentives on disability and maintenance of symptoms was most pronounced in individuals with minor injuries. Individuals with minor disability and pain may report less improvement, as any medical improvement may impact future compensation that they may receive through litigation. When initial disability and pain are more intense, there is more room for clinical improvement, perhaps without the perception that future compensation will be affected. It is not clear why the strength of the association between work status and a clinically important improvement increased as the initial pain intensity decreased. This is likely related to the precision of the CBIQ to measure change in individuals with minimal disability (i.e. mild intensity pain and remained at work). In comparison, a patient who was off work and likely received additional lost wages compensation might have 86 perceived and reported more disability even at low levels of initial pain intensity. In this manner these patients will have more room for improvement in comparison to a working counterpart, and therefore more likely to have demonstrated and reported a clinical improvement. The prognostic model developed using the variables discussed above (and the smoking status variable) had a sensitivity of 0.74 and a specificity of 0.45. Although this model produced the best predictions of the various models compared, the strict application of this model as a diagnostic tool for use in clinical practice is not recommended. A significant number of patients would be falsely labeled as "destined for chronicity" (false positive rate = 55.1%), and a number of patients who eventually develop chronic symptoms would be incorrectly told that rapid recovery was likely (false negative rate = 25.9%). Neither of these scenarios is desirable. A patient falsely labeled as chronic may receive more intense therapy than they need, and some intensive rehabilitation programs may reinforce sick role behaviour (32). When recovery is uncomplicated the forecasting of a poor outcome may have a negative impact on recovery. Alternatively, overly optimistic predictions of outcome can create disappointment should recovery be delayed. False negatives may be denied necessary rehabilitation and suffer additional pain and disability because they were deemed to be at low risk for developing chronic symptoms (32). In a qualitative sense the developed prognostic model alerts clinicians of a number of potential factors associated with a poor clinical outcome; however, the utility of the model for identification of patients at risk for delayed recovery is limited. 87 7.3 Thesis Strengths and Limitations 7.3.1 Strengths Inception Cohort. A n essential component of prognostic research is the assembly of an inception cohort (33, 34). Acute patients at a similar stage of recovery are needed such that the outcome of interest (early recovery or lack of early recovery) has not already occurred (1, 76). In a clinical setting of secondary care, and where the primary outcome of interest is a change in questionnaire score, an appropriate 'zero time' is the treatment start date. The longer the allowable lagtime between injury and treatment the greater the chance of enrolling patients with behavioural and psychological characteristics associated with a chronic pain state. Therefore, this study included only those patients whose lagtime between injury date and treatment start date was under 91 days. This period is prior to the definition of chronic W A D at 6 months post injury highlighted by the Quebec Task Force on Whiplash-Associated Disorders (QTF) (1). The 91-day period will allow for the accumulation of some sub-acute cases; however, with a median lagtime of 25 days (mean=32 days) most patients were still in an acute state. Ninety percent (90%) of the sample had lagtimes of less than 69 days. Having patients enrolled at a similar stage of disease helped to reduce the introduction of bias due to the issues of cohort assembly, and this variable was also controlled for in the analysis. Outcome Variable. Common outcome measures used in previous prognosis studies of acute whiplash have been the time-to-claim closure or the self-report of the presence of whiplash symptoms (6). Although, time-to-claim closure has been shown to be a good indicator of recovery from W A D (2, 17), it continues to be criticized by some authors 88 (19). Time-to-claims closure is an administrative proxy of recovery; therefore, it is possible that some individuals continue to have significant disability despite closure of their claim. Alternatively, self-report whiplash related symptoms are often used in studies; however, in addition to the recall bias introduced when asking one to remember past symptoms, previous studies rarely use a standardized instrument designed to measure pain and/or disability. In addition, the meaning of recovery will vary depending on the patient. Individuals will not only consider the resolution of their symptoms, but also the readjustment and adaptation of daily activities to work around the condition (20). In this regard, Beaton and her colleagues concluded that two individuals may place entirely different meanings on the concept of recovery, such that some individuals may not actually demonstrate a change in disability or function (20). This poses problems for the interpretation of recovery in studies that ask questions regarding the improvement in symptoms and/or improvement in the whiplash-associated disorder. In the current study, the primary outcome was a change in disability questionnaire score between entry to and exit from a CBI clinic. A clinically important change in score indicated that a patient had demonstrated at least a minimal improvement in their condition. The data was collected in a standardized manner using a valid and reliable instrument. By using a patient centered outcome measure this study has investigated a number of risk factors for delayed recovery, while at the same time addressing the criticism that proxy markers of recovery such as time-to-claim closure may not indicate 89 actual improvement in the patient. In addition, the current study controlled for the effect described by Beaton et al above by using a change in questionnaire score as the primary outcome. Regardless of the internal meaning an individual placed on "getting better", a change in questionnaire score would reflect meaningful improvements perceived by the patient. One potential limitation of using a change in questionnaire score as the primary outcome is the ability of the questionnaire to measure change in individuals with minimal disability (high initial function scores on the CBIQ). In this study this effect was particularly apparent in patients who presented with initial disability that was more than two standard deviations below the cohort's mean initial disability. These 69 highly functional people (3% of the cohort) demonstrated minimal improvements in CBIQ score (on average 2 points improvement) and therefore would not have been captured as a positive outcome, despite minimal overall disability. Overall, a total of 30 individuals (approximately 1% of the cohort) presented with such minimal initial disability that the primary outcome of minimal clinically important improvement was not measurable (i.e. out of the possible 61 points on the CBIQ these individuals scored greater than 55 points at initial presentation thereby making positive outcome impossible). It is unlikely these 30 highly functional individuals biased the results in anyway because there was so few of them; however it should be recognized that their inclusion with the other patients with a negative outcome might slightly diminish the risk estimates (odds ratios) for the various covariates of interest. Related to this concept is the fact that minimally important difference in change of questionnaire score may vary depending on the baseline score. 90 This effect has been reported with the use of other condition specific and generic questionnaires (45, 77). For example, Hagg and his colleagues reported smaller minimally important clinical differences in the lower end of the Roland-Morris questionnaire for low back pain (and vice versa at the upper end of the scale) (45). Finally, when using a change in questionnaire score as an outcome measure, the assumption is made the traits being measured remain stable over time. In the current study, the assumption was made that the concepts of pain and disability remained constant between the two administrations of the CBIQ. The relatively short time between completion of the initial and discharge questionnaires strengthen this assumption. Ultimately, the goal of treatment at CBI is to improve the quality of life as perceived by the patient. This goal holds true regardless of the baseline score of disability. For this reason the change in questionnaire score is necessary to capture improvements along the initial disability continuum. The current study is one of only a few studies to use a valid and reliable instrument to capture a whiplash recovery related outcome. Treatment Protocol. The initial treatment that a patient receives after a motor vehicle collision appears to be an important prognostic factor (2). The recommended treatment for an acute patient is an active approach that encourages a return to normal activities as soon as possible (1). In addition, the QTF, as part of their research agenda, has recommended the standardization of assessment procedures such that critical baseline information is collected (1). Assessment procedures, data collection, and treatment at 91 CBI are standardized across the company treatment centres. Therefore the same information was collected and an active treatment approach was applied for all patients over several Canadian Provinces. In accordance with previous clinical based research (78) and recommendations for the design of a prognostic factor study (79) there was little variation in CBI treatment. This reduced the possibility of confounding by treatment regimen. Multiple Insurance Jurisdictions. The insurance system where an individual has their claim managed is an important predictor of recovery (2). In addition, large variations in incidence and recovery time for W A D exist for different geographic regions (even when similar insurance systems exist) (1,6). This study used patient data from six different provinces in Canada. Quebec and Saskatchewan operated under a no-fault system during the study period, while British Columbia, Alberta, Ontario, and Nova Scotia had various forms of a tort system. In addition, three of the provinces (B.C., Sask., and Que.) had sole insurance providers, while the other three provinces (Alta., Ont, N.S.) had multiple insurance companies operating in the jurisdiction. Although the influence of the differences in insurance system was not specifically investigated, this study did show that recovery varied by province, and that while controlling for the province a number of other significant risk factors emerged in both provinces with tort and no-fault systems. Unfortunately the sample of patients from Que. and Sask. was of limited size. 92 7.3.2 Limitations Patient Selection. One must consider the selection of patients in this study when interpreting the results. Patients must have sought treatment at CBI in order to be included in the database. Individuals with mild forms of W A D may not seek treatment, and would be excluded from this study. In addition, CBI is dependent on physician referral for patient acquisition. Physicians do not refer all of their patients to this clinic system; therefore, certain types of whiplash injury may gravitate towards other types of treatment. Likely, this selection process eliminated the mild forms of W A D . However, this referral process would also eliminate severe injuries such as fractures, dislocations or head injuries that could bias the results. The effect of these referral patterns would be to increase patient homogeneity thereby narrowing predictor variable distribution, resulting in a bias towards the null (80). Despite this, the pain severity and functional status, the patient population of CBI was found to be similar to other rehabilitation providers in a pilot project by the Institute of Work & Health (81). In addition, the assembly of an inception cohort attempted to include individuals at a similar stage of recovery. Unmeasured Prognostic Factors. This study investigated the associations of a large variety of personal, clinical, and treatment related factors and clinically important recovery. However, other prognostic factors that may predict poor outcome, such as depressive symptomatology (17), and initial health care provider (2) were not collected. In addition, due to missing values, two variables of interest (previous back or neck pain, and constant vs. intermittent pain) had to be excluded. It is possible that these or other unmeasured factors influenced the described associations. 93 Missing Values. Despite the overall large sample size of this cohort, 28.6% of the subjects were excluded because they did not complete a Visual Analogue Scale regarding their initial pain intensity. Overall there was no significant difference in outcome between the non-participants (73.6 % positive outcome) when compared to the participants (72.0%) positive outcome). In addition, when comparing outcome in the participants and non-participants, an indicator variable of participation in the cohort (i.e. those individuals who completed the VAS) was not associated with outcome, while controlling for other important predictors. Finally, important factors associated with participation (i.e. work status and lawyer retention) were adjusted for in the cohort analysis. Comorbid medical conditions and previous treatment were not included in the final model despite noted differences between those in the study cohort and those excluded, in favour of a more parsimonious model. Even with the exclusion of a significant reduction of a number of records, this study represents one of the largest cohorts assembled investigating the recovery of acute whiplash. Duration of Follow-Up. Ideally the follow-up questionnaires would have been completed at identical times for each patient. However, the nature of clinical practice results in a variation in the duration that each individual receives treatment (and thus the timing of discharge and completion of the exit questionnaire). The treatment approach at CBI is standardized, however the point at which the doctor, therapist, or patient deem themselves sufficiently recovered will vary. A patient who improves rapidly will have less follow-up time because they will be discharged earlier. Likewise a patient who is 94 slow to improve will have a longer duration of treatment. Other factors such as insurance settlement or policy, and patient characteristics may also influence the duration of therapy. The median duration of treatment in the study cohort was 56 days. The duration of follow-up was not associated (p=0.94) with a clinically important improvement in the CBIQ (the dichotomous primary outcome) or the overall change in CBIQ (continuous) score (p=0.44); however, the duration of follow-up was mildly correlated (Pearson correlation coefficient) with the raw initial CBIQ score (r — -0.09) and raw discharge CBIQ score (r = -0.07). Those patients who reported higher levels of disability on presentation or discharge were followed up for a slightly longer duration. The magnitude of this difference in follow-up was relatively small, as approximately 97 % of the cohort had median follow-up times within 28 days of each other. This study would have been additionally strengthened by a longer duration of follow-up. Nevertheless, the follow-up time was suitable to monitor early clinical improvement. 7.4 Conclusion This study has provided valuable insights into the early stages of recovery from whiplash injuries of varying severity, across multiple insurance jurisdictions. This extensive exploratory work investigating the association of a number of personal, clinical and treatment related variables and early improvement of pain and disability after motor vehicle collision in the secondary care physiotherapy setting, has indicated variables that require particular attention in subsequent research. Older age, female gender, increasing lagtime between injury date and presentation for treatment, initial pain location, and province of injury were associated with minimal early improvements in response to 95 treatment. Lawyer retention and work status while interacting with the initial pain intensity were also associated with less improvement. This study sought to further investigate risk factors for delayed recover in the secondary care setting (an area with minimal previous research) and to explore how legal and compensation issues might influence recovery. Both of these objectives have been accomplished. The current study and one previous study(2) now provide consistent evidence that legal and compensation issues are important. In addition, lawyer retention was an important factor regardless of the specific policies within the insurance systems of six Canadian provinces. Initial pain intensity has been previously described as an important prognostic factor in the recovery from whiplash (1,6). However future researchers should consider the interaction that this variable has with the retention of legal services and work status, and consider other potential interactions that might occur based on the intensity of initial complaints. In addition, this study has provided a description of the chronic whiplash population as defined by the Quebec Task Force (QTF) and an early chronic group. A comparison of these two groups with the acute whiplash population was noteworthy because it demonstrated many significant differences between the populations that occurred prior to the six month definition of chronic pain suggested by the QTF (1). The information provided by this study will also be useful for clinicians in the secondary care setting who are likely to treat a similar patient population. Ideally clinicians should be able to identify those patients that might require special attention and perhaps offer more intensive or alternate treatment. Although some uncertainty will always exist when 96 determining the prognosis for a particular patient this study has provided a list of factors that will help a clinician identify a patient who wil l not make a meaningful early recovery. Clinicians must pay particular attention to non-clinical factors, as these external factors and personal characteristics were found to be particularly important, and are not the usual focus of clinical practice. 97 References 1. Spitzer WO, Skovron M L , Salmi LR, et al. Scientific monograph of the Quebec Task Force on Whiplash-Associated Disorders: redefining "whiplash" and its management. Spine. 1995;20(8 Suppl):lS-73S. 2. Cassidy JD, Carroll LJ , Cote P, Lemstra M , Berglund A , Nygren A. Effect of eliminating compensation for pain and suffering on the outcome of insurance claims for whiplash injury. NEnglJMed. 2000;342(16):1179-86. 3. Hartling L, Pickett W, Brison RJ. Derivation of a clinical decision rule for whiplash associated disorders among individuals involved in rear-end collisions. Accid AnalPrev. 2002;34(4):531-9. 4. Allen MJ . Whiplash claims and costs in British Columbia. BCMJ. 2002;44:241-42. 5. Harder S, Veilleux M , Suissa S. The effect of socio-demographic and crash-related factors on the prognosis of whiplash. J Clin Epidemiol. 1998;51(5):377-84. 6. Cote P, Cassidy JD, Carroll L, Frank JW, Bombardier C. A systematic review of the prognosis of acute whiplash and a new conceptual framework to synthesize the literature. Spine. 2001;26(19):E445-58. 7. Ferrari R, Russell AS. Epidemiology of whiplash: an international dilemma. Ann Rheum Dis. 1999;58(l):l-5. 8. Schrader H , Obelieniene D, Bovim G, et al. Natural evolution of late whiplash syndrome outside the medicolegal context. Lancet. 1996;347(9010): 1207-11. 9. Hoogendoorn WE, van Poppel M N M , Bongers P M , Koes BW, Boulter L M . Systematic review of psychosocial factors at work and private life as risk factors for low back pain. Spine. 2000;25(16):2114-25. 10. Ovadia D, Steinberg EL, Nissan M N , Dekel S. Whiplash injury~a retrospective study on patients seeking compensation. Injury. 2002;33(7):569-73. 11. Khan S, Bannister G, Gargan M , Asopa V, Edwards A. Prognosis following a second whiplash injury. Injury. 2000;31(4):249-51. 12. Miettinen T, Lindgren K A , Airaksinen O, Leino E. Whiplash injuries in Finland: a prospective 1-year follow-up study. Clin Exp Rheumatol. 2002;20(3):399-402. 98 13. Soderlund A , Lindberg P. Whiplash-associated disorders—predicting disability from a process-oriented perspective of coping. Clin Rehabil. 2003;17(1): 101-7. 14. Hartling L , Brison RJ, Ardern C, Pickett W. Prognostic value of the Quebec Classification of Whiplash-Associated Disorders. Spine. 2001;26(1):36-41. 15. Kyhlback M , Thierfelder T, Soderlund A . Prognostic factors in whiplash-associated disorders. Int JRehabil Res. 2002;25(3): 181-7. 16. Sterner Y , Toolanen G, Gerdle B, Hildingsson C. The incidence of whiplash trauma and the effects of different factors on recovery. J Spinal Disord Tech. 2003;16(2):195-9. 17. Cote P, Hogg-Johnson S, Cassidy JD, Carroll L, Frank JW. The association between neck pain intensity, physical functioning, depressive symptomatology and time-to-claim-closure after whiplash. J Clin Epidemiol. 2001;54(3):275-86. 18. Suissa S, Harder S, Veilleux M . The relation between initial symptoms and signs and the prognosis of whiplash. Eur Spine J. 2001 ;10(l):44-9. 19. Freeman M D , Croft A C , Rossignol A M . "Whiplash associated disorders: redefining whiplash and its management" by the Quebec Task Force. A critical evaluation. Spine. 1998;23(9): 1043-9. 20. Beaton DE, Tarasuk V , Katz IN, Wright JG, Bombardier C. "Are you better?" A qualitative study of the meaning of recovery. Arthritis Rheum. 2001;45(3):270-9. 21. Satoh S, Naito S, Konishi T, et al. An examination of reasons for prolonged treatment in Japanese patients with so called whiplash injuries. J Musculoskeletal Pain. 1997;5:71-84. 22. Borchgrevink GE, Lereim I, Royneland L, Bjorndal A , Haraldseth O. National health insurance consumption and chronic symptoms following mild neck sprain injuries in car collisions. ScandJSoc Med. 1996;24(4):264-71. 23. Sturzenegger M , Radanov BP, Di Stefano G. The effect of accident mechanisms and initial findings on the long-term course of whiplash injury. J Neurol. 1995;242:443-49. 24. Radanov BP, Sturzenegger M , Di Stefano G. Long-term outcome after whiplash injury. A 2-year follow-up considering features of injury mechanism and somatic, radiologic, and psychosocial findings. Medicine (Baltimore). 1995;74(5):281-97. 25. Radanov BP, di Stefano G, Schnidrig A, Ballinari P. Role of psychosocial stress in recovery from common whiplash. Lancet. 1991;338(8769):712-5. 99 26. Norris SH, Watt I. The prognosis of neck injuries resulting from rear-end vehicle collisions. J Bone Joint Surg Br. 1983;65(5):608-11. 27. Radanov BP, Sturzenegger M , Di Stefano G, Schnidrig A , Aljinovic M . Factors influencing recovery from headache after common whiplash. BMJ. 1993;307(6905):652-5. 28. Hall H , Mcintosh G, Wilson L, Melles T. Spontaneous onset of back pain. Clin J Pain. 1998;14(2): 129-33. 29. Malanga G. Cervical Flexion-Extension/Whiplash Injuries. Spine State of the Art Reviews. 1998;12(2). 30. Greenough CG, Fraser RD. Assessment of outcome in patients with low-back pain. Spine. 1992;17(1):36-41. 31. Mcintosh G. Back pain prognostic factors: A cohort study of 1,752 patients of a national rehabilitation clinic system. Master's Thesis - University of Toronto. 1999. 32. Mcintosh G, Frank J, Hogg-Johnson S, Bombardier C, Hall H . Prognostic factors for time receiving workers' compensation benefits in a cohort of patients with low back pain. Spine. 2000;25(2): 147-57. 33. Simon R, Altman DG. Statistical aspects of prognostic factor studies in oncology. BrJ Cancer. 1994;69(6):979-85. 34. Ales K L , Charlson M E . In search of the true inception cohort. J Chronic Dis. 1987;40(9):881-5. 35. Gargan M F , Bannister GC. Long-term prognosis of soft-tissue injuries of the neck. JBone Joint Surg Br. 1990;72(5):901-3. 36. Gargan M F , Bannister GC. The rate of recovery following whiplash injury. Eur Spine J. 1994;3(3): 162-4. 37. Connell FA, Diehr P, Hart L G . The use of large data bases in health care studies. Annu Rev Public Health. 1987;8:51-74. 38. Greenland S, Finkle WD. A critical look at methods for handling missing covariates in epidemiologic regression analyses. Am J Epidemiol. 1995;142(12):1255-64. 39. Miettinen OS. Theoretical epidemiology : principles of occurrence research in medicine. New York: Wiley; 1985. 100 40. Vach W, Blettner M . Biased estimation of the odds ratio in case-control studies due to the use of ad hoc methods of correcting for missing values for confounding variables. Am J Epidemiol. 1991;134(8):895-907. 41. Rothman K J , Greenland S. Modern epidemiology. 2nd ed. Philadelphia: Lippincott-Raven; 1998. 42. Holt A E , Shaw NJ, Shetty A, Greenough CG. The reliability of the Low Back Outcome Score for back pain. Spine. 2002;27(2):206-10. 43. Khatri M , Murray M , Greenough CG. Minimum clinically important difference in low back outcome score in individuals with low back pain. (Abstract) International Society for the Study of the Lumbar Spine: 30th Annual Meeting, Vancouver: 2003. 44. Bombardier C, Hayden J, Beaton DE. Minimal clinically important difference. Low back pain: outcome measures. J Rheumatol. 2001;28(2):431-8. 45. Hagg O, Fritzell P, Nordwall A. The clinical importance of changes in outcome scores after treatment for chronic low back pain. Eur Spine J. 2003; 12(1): 12-20. 46. U.S. Department of Labour Handbook for Analyzing Jobs. Washington: Manpower Administration, U.S. Department of Labour. 1972. 47. Waddell G, McCulloch J, Kummel E, Venner R. Nonorganic physical signs in low-back pain. Spine. 1980;5(2):117-25. 48. Sobel J, Sollenberger P, Robinson R, Polatin P, Gatchel R. Cervical nonorganic signs: a new clinical tool to assess abnormal illness behavior in neck pain patients: a pilot study. Arch Phys Med Rehabil. 2000;81:170-5. 49. Price DD, McGrath PA, Rafii A , Buckingham B. The validation of visual analogue scales as ratio scale measures for chronic and experimental pain. Pain. 1983;17(l):45-56. 50. Huskisson EC. Measurement of pain. J Rheumatol. 1982;9(5):768-9. 51. Hagen K B , Hilde G, Jamtvedt G, Winnem MF. The Cochrane review of advice to stay active as a single treatment for low back pain and sciatica. Spine. 2002;27(16):1736-41. 52. Peeters GG, Verhagen AP, de Bie RA, Oostendorp R A . The efficacy of conservative treatment in patients with whiplash injury: a systematic review of clinical trials. Spine. 2001;26(4):E64-73. 101 53. Kneipp S, Mcintosh M . Handling missing data in nursing research with multiple imputation. Nurs Res. 2001;50(6):384-89. 54. Last J, Association IE. Guidelines on ethics for epidemiologists. Int J Epidemiol. 1990;19(l):226-9. 55. Hosmer DW, Lemeshow S. Applied logistic regression New York: Wiley; 1989. 56. Miller M E , Hui SL, Tierney W M . Validation techniques for logistic regression models. Stat Med. 1991;10(8):1213-26. 57. Bagley S, White H , Golomb B. Logistic regression in the medical literature: Standards for use and reporting, with particular attention to one medical domain. J Clin Epidemiol. 2001 ;54:979-85. 58. Hosmer DW, Taber S, Lemeshow S. The importance of assessing the fit of logistic regression models: a case study. Am J Public Health. 1991 ;81(12): 1630-5. 59. Altman DG, Royston P. What do we mean by validating a prognostic model? Stat Med. 2000; 19(4):453-73. 60. Efron B. How biased is the appartent error rate of a prediction rule? Journal of the American Statistical Association. 1986;81(394):461-70. 61. Steyerberg EW, Harrell FE, Jr., Borsboom GJ, Eijkemans M J , Vergouwe Y , Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. 2001;54(8):774-81. 62. Harrell FE, Jr., Lee K L , Mark DB. Multivariate prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361-87. 63. Feinstein AR. Principles of medical statistics. Boca Raton, FL: CRC Press; 2002. 64. Janssen I, Heymsfield SB, Ross R. Low relative skeletal muscle mass (sarcopenia) in older persons is associated with functional impairment and physical disability. J Am Geriatr Soc. 2002;50(5):889-96. 65. Janssen I, Heymsfield SB, Wang Z M , Ross R. Skeletal muscle mass and distribution in 468 men and women aged 18-88 yr. JAppl Physiol. 2000;89(1):81-8. 66. Keogh E, Herdenfeldt M . Gender, coping and the perception of pain. Pain. 2002;97(3):195-201. 102 67. Rosenfeld M , Gunnarsson R, Borenstein P. Early intervention in whiplash-associated disorders: a comparison of two treatment protocols. Spine. 2000;25(14):1782-7. 68. McKinney L A . Early mobilisation and outcome in acute sprains of the neck. BMJ. 1989;299(6706):1006-8. 69. Borchgrevink GE, Kaasa A , McDonagh D, Stiles TC, Haraldseth O, Lereim I. Acute treatment of whiplash neck sprain injuries. A randomized trial of treatment during the first 14 days after a car accident. Spine. 1998;23(1):25-31. 70. Radanov BP, Begre S, Sturzenegger M , Augustiny K F . Course of psychological variables in whiplash injury—a 2-year follow-up with age, gender and education pair-matched patients. Pain. 1996;64(3):429-34. 71. Sturzenegger M , DiStefano G, Radanov BP, Schnidrig A. Presenting symptoms and signs after whiplash injury: the influence of accident mechanisms. Neurology. 1994;44(4):688-93. 72. Partheni M , Constantoyannis C, Ferrari R, Nikiforidis G, Voulgaris S, Papadakis N . A prospective cohort study of the outcome of acute whiplash injury in Greece. Clin Exp Rheumatol. 2000;18(l):67-70. 73. Obelieniene D, Schrader H, Bovim G, Miseviciene I, Sand T. Pain after whiplash: a prospective controlled inception cohort study. J Neurol Neurosurg Psychiatry. 1999;66(3):279-83. 74. Hadler N . If you have to prove you are i l l , you can't get well: the object lesson of fibromyalgia. Spine. 1996;21(20):2397-2400. 75. Binder L M , Rohling M L . Money matters: a meta-analytic review of the effects of financial incentives on recovery after closed-head injury. Am J Psychiatry. 1996;153(1):7-10. 76. Hazard RG, Haugh L D , Reid S, Preble JB, MacDonald L. Early prediction of chronic disability after occupational low back injury. Spine. 1996;21(8):945-51. 77. Stucki G, Daltroy L, Katz IN, Johannesson M , Liang M H . Interpretation of change scores in ordinal clinical scales and health status measures: the whole may not equal the sum of the parts. J Clin Epidemiol. 1996;49(7):711-7. 78. Dionne CE, Koepsell TD, Von Korff M , Deyo R A , Barlow WE, Checkoway H. Predicting long-term functional limitations among back pain patients in primary care settings. J Clin Epidemiol. 1997;50(1):31-43. 103 79. Airman DG, Lyman GH. Methodological challenges in the evaluation of prognostic factors in breast cancer. Breast Cancer Res Treat. 1998;52(l-3):289-303. 80. Sackett D. Bias in analytical research. J Chronic Dis. 1979;32:51-63. 81. Mondloch M , Mckenzie D. Quality Management System for Soft Tissue Injuires (QMS-STI). A pilot project with rehabilitation providers in Ontario and the Institiute for Work & Health. 1998. 82. Raymond M . Missing data in evaluation research. Evaluation & the Health Professions. 1987;9(4):395-420. 83. Rubin DB. Multiple imputation for nonresponse in surveys. New York: Wiley; 1987. 104 Appendices 105 Appendix A: The Quebec Task Force Classification of Whiplash-Associated Disorders (1) Grade Clinical Presentation 0 No complaint about the neck No physical sign(s) I Neck complaint of pain, stiffness or tenderness only No physical sign(s) II Neck complaint and Musculoskeletal sign* III Neck complaint and Neurological sign** IV Neck complaint and Fracture or dislocation * Musculoskeletal signs include decreased range of motion and point tenderness. ** Neurological signs include decreased or absent deep tendon reflexes, weakness and sensory deficits. 106 Appendix B: Criteria for the Appraisal of the Methodological Quality of Cohort Studies 1) Research question and objective are clearly defined 2) Source population is identified 3) Inclusion and exclusion criteria are described and appropriate 4) Participation rate is reported and appropriate (at least 60%) or a comparative analysis of participants and non-participants 5) Follow-up is reported, explained and reasonable 6) Where applicable loss to follow-up is equal in both groups 7) Sample size is pre-planned based on the objective of the study 8) Statistical analysis is appropriate for the research question and objective of the study 9) Adjustment is made for important variables* 10) Zero time is identified 11) Baseline comparability of various group is reported where applicable 12) Same data collection procedures are conducted for all members of the cohort 13) Important baseline variables are measured, valid, and reliable 14) A l l aspects of a prognostic factor are measured (dose, level, duration) and done so adequately (previous, baseline, follow-up) 15) Regular follow-up are accomplished 16) Other prognostic factors are measured 17) Duration of follow-up is adequate for the objective of the study ** 18) Outcome is defined and measurable 19) Outcome is valid 20) Outcome assessment was blind *Based on the previous systematic reviews of the whiplash literature (1,6) the following prognostic factors were considered important for adjustment: age, gender, initial pain intensity (neck and headache), and radicular signs and symptoms. **Based on the objectives of all studies reviewed, a follow-up duration of at least one year was considered appropriate. 107 Appendix C: Canadian Back Institute Spinal Assessment Form - page 1 SPINAL ASSESSMENT CANADIAN BACK INSTITUTE PATIENT, A S S E S S M E N T D A T E MM/DD/YY A R E A A S S E S S E D • B A C K O N L Y • N E C K O N L Y • B A C K & N E C K • R E - A S S E S S M E N T T Y P E O F A S S E S S M E N T • C A • C R A • F C E O IA • R A • W A • O T H E R • . rELEMENTS OH IHIS EglSlElE SITE O F D O M I N A N T PAIN S Y M P T O M L O C A T I O N D U R A T I O N • B A C K • B A C K • B U T T O C K • D A Y S • L E G • TH IGH O C A L F • F O O T O W E E K S • N E C K • A R M • N E C K • A R M • INTER S C A P • F O R E A R M • T R A P R I D G E • H A N D • M O N T H S • Y E A R ( S ) • H / A B A C K / L E G PAIN N E C K / A R M PAIN D A Y - A . M . D A Y - P . M . PAIN D ISTURBING S L E E P • C O N S T A N T • C O N S T A N T • B E T T E R • B E T T E R D Y E S • INTERMITTENT • N O N E A T P R E S E N T • INTERMITTENT • N O N E A T P R E S E N T • W O R S E • S A M E • W O R S E • S A M E • N O B L A D D E R C O N T R O L • U N C H A N G E D • C H A N G E D B O W E L C O N T R O L • U N C H A N G E D • C H A N G E D SIGNIFICANT M E D . H ISTORY . R E C E N T U N E X P L A I N E D W E I G H T L O S S • Y E S • N O D A T E THIS E P I S O D E S T A R T E D M M / D D / Y Y W A S T H E R E A N E V E N T T H A T C A U S E D THIS E P I S O D E ? IF Y E S , H O W L O N G A F T E R E V E N T DID PAIN S T A R T ? M E C H A N I S M • N O • Y E S • M I N U T E S • H O U R S • D A Y S • W E E K S WORK HISTORY O C C U P A T I O N : T Y P E O F W O R K • H O M E M A K E R • S E D E N T A R Y / L I G H T • L IGHT • M E D I U M O S T U D E N T • R E T I R E D • H E A V Y • V E R Y H E A V Y NO PAIN 0 1 2 BACK PAIN 3 4 5 6 7 WORST PAW 8 9 10 0 1 2 L E G PAIN 3 4 5 6 7 e 9 10 0 1 2 NECK PAIN 3 4 5 6 7 S 9 10 0 1 2 ARM PAIN 3 4 5 6 7 8 9 10 0 1 2 HEADACHE 3 4 5 6 7 8 9 10 C U R R E N T L Y W O R K I N G • Y E S • N O IF Y E S , • W O R K I N G WITH S Y M P T O M S • W O R K I N G W I T H O U T S Y M P T O M S D • P R E V I O U S O C C U P A T I O N N E W O C C U P A T I O N • • F U L L H O U R S • F U L L D U T Y PARTIAL H O U R S • MODIFIED D U T Y IF N O . L A S T D A Y W O R K E D M M / D D / Y Y • POSITION A V A I L A B L E • • POSITION T E R M I N A T E D L O O K I N G F O R W O R K • DISABILITY P E N S I O N • N O DIRECTION LUMBAR i :* CERVICAL EFFECT ON PAIN INVESTIGATIONS • N O N E • X - R A Y • M Y E L O / C T / M R I O B L O O D W O R K • B O N E S C A N • E L E C T R I C A L S T U D I E S PREVIOUS EPISODES INVESTIGATIONS • N O N E • X - R A Y O M Y E L O / C T / M R I • B L O O D W O R K D B O N E S C A N • E L E C T R I C A L S T U D I E S O N O N E • B A C K D O M I N A N T • L E G DOMINANT T IME S I N C E 1st E P I S O D E • < 1 YR. D 1 - 5 Y R S . D > 5 Y R S . IN T H E P A S T Y E A R F R E Q . • I N C R E A S E • D E C R E A S E • S A M E • NIL DUR. • I N C R E A S E • D E C R E A S E • S A M E • NIL SIM. T O P R E S E N T E P I S O D E • Y E S • N O N O N E • N E C K D O M I N A N T • A R M DOMINANT T IME S I N C E 1st E P I S O D E • < 1 Y R . D 1 - 5 Y R S . IN T H E P A S T Y E A R F R E Q . • I N C R E A S E • D E C R E A S E D U R . SIM. T O P R E S E N T E P I S O D E • I N C R E A S E • Y E S • D E C R E A S E • N O • >5YRS. • S A M E • NIL O S A M E • NIL P R E V I O U S B A C K • N O N E • D E C O M P R E S S I O N • F U S I O N S U R G E R Y • M U L T I P L E P R O C E D U R E S P R E V I O U S N E C K • N O N E • D E C O M P R E S S I O N • F U S I O N S U R G E R Y • M U L T I P L E P R O C E D U R E S 108 Appendix C: Spinal Assessment Form - page 2 LUMBAR CFRVICAI PHYSICAL EXAMINATION O B S E R V A T I O N . 1 SITTING | P O S T U R E FLEXED/FLAT S T A N D I N G P O S T U R E 1 NEUTRAL 2 FLEXED/FLAT NEUTRAL LORDOTIC S T R U C T U R A L S C O L I O S I S • Y E S • N O R A N G E O F M O V E M E N T US F L E X I O N • N O R M A L • R E D U C E D E X T E N S I O N • N O R M A L • R E D U C E D SHIFT • SHIFT • DEVIATION • DEVIATION • R • R • R • • O B S E R V A T I O N SITTING P O S T U R E S T A N D I N G P O S T U R E O H E A D F O R W A R D SHIFT • L • H E A D F O R W A R D SHIFT • L • E A R O V E R S H O U L D E R • R • E A R O V E R S H O U L D E R • R R A N G E O F M O V E M E N T C / S F L E X I O N • N O R M A L E X T E N S I O N • N O R M A L SIDE B E N D • N O R M A L R O T A T I O N • N O R M A L • R E D U C E D • R E D U C E D O R E S T R I C T I O N L DEVIATION O DEVIATION • • • R • • • • STEST MOVEMENTS DISTAL SYMPTOM • BACK • BUTTOCK • THIGH • CALF • FOOT DISTAL • NECK • SCAPULA • TRAP • ARM • FOREARM • HAND • H/A BEFORE B w s c p NIL PDM ERP NIL AFTER SYMPTOM BEFORE B W S c p NIL PDM ERP NIL AFTER STAND FLEXION /10 /10 SIT FLEXION /10 /10 EXTENSION /10 /10 PROTRACTION /10 /10 SHIFT L /10 /10 RETRACTION /10 /10 SHIFT R /10 /10 EXTENSION no /10 LYING FLEXION /10 /10 LYING RETRACTION no OS mm /10 EXTENSION /10 /10 EXTENSION no no SHIFT L /10 /10 SIDE BEND L no /10 SHIFT R /10 /10 SIDE BEND R no /10 NEW LOCATION OF DISTAL SYMPTOM NEW LOCATION OF DISTAL SYMPTOM lNEURO_OGICAL EXAMINATION IRRITATION T E S T S S L R L4-S2 • • _° • • C O N D U C T I O N T E S T S N O R M A L L+ R+ • W E L L L E G LIFT j A N T . D E L T O I D C 6 • • • O C R O S S O V E R N O R M A L R+ I B I C E P R E F L E X C 6 • • • F S T L2-4 • • • I B I C E P P O W E R C 6 • • C O N D U C T I O N T E S T S N O R M A L L+ R+ T R I C E P R E F L E X C 7 • o • K N E E R E F L E X L3-4 • • • T R I C E P P O W E R C 7 • • • T R E N D E L E N B U R G L5 • • • A N K L E D O R S I F L E X I O N L4-5 • • • E X T . DIG. L O N G . C 7 • • • E X T . HAL. L O N G . L5 • • • P L A N T A R R E S P O N S E C O R D • • • A N K L E R E F L E X S1 • • • P L A N T A R F L E X I O N S1 • • n *-• G L U T E U S M A X I M U S S1 • • • S A D D L E S E N S A T I O N S3-5 • N O R M A L • R E D U C E D P L A N T A R R E S P O N S E C O R D • n • HIP JO INTS • N O T T E S T E D • N O R M A L • L+ O R+ S.I. JOINTS • N O T T E S T E D • N O R M A L • L+ • R+ S H O U L D E R JOINTS • N O T T E S T E D • N O R M A L • L+ • R+ B P T • N O T T E S T E D O N O R M A L • L+ • R+ T H O R A C I C S P I N E • N O T T E S T E D O N O R M A L • + • P A T T E R N 1 • P A T T E R N 2 • P A T T E R N 3 • P A T T E R N 4 • P A T T E R N 5 IF P A T T E R N 3 • L E F T • R IGHT • B I L A T E R A L IF P A T T E R N 4 • L E F T O R IGHT D B I L A T E R A L IF P A T 3 O R 4 C O N D . L O S S • L 3,4 • L5 • S1 P A T 5 C O M B I N E D WITH O l • 2 • 3 • 4 • N O N E • P A T T E R N 1 • P A T T E R N 2 • P A T T E R N 3 • P A T T E R N 4 • P A T T E R N 5 IF P A T T E R N 3 • L E F T D R IGHT • B I L A T E R A L IF P A T T E R N 4 • L E F T • R IGHT • B I L A T E R A L IF P A T 3 O R 4 C O N D . L O S S • C 6 • C 7 • O T H E R P A T 5 C O M B I N E D WITH r j , r j 2 p 3 r j „ r j N 0 N E •ALTERNATIVEiFINblNGS* •>":i"., - •>,:-' '* •: .'•.'•"S,,r.r;.:^ .r, • N O N E • A C U T E C E S • S U S P E C T S Y S T E M I C D I S E A S E O H I P J O I N T ( S ) rJsiJOINT(S) O N O N M E C H A N I C A L H E A D A C H E • T H O R A C I C S P I N E • S H O U L D E R JOINT(S) • B P T NON ORGANIC / BEHAVIOURAL FINDINGS >•„ -n ^ ™ , IF POSIT IVE, • S U P E R F I C I A L T E N D . • AXIAL L O A D I N G LI N O T T E S T E D O POSITIVE N 0 N A N A T O M I C T E N D . • A C E T A B . R O T A T I O N • S L R D I S C R E P A N C Y • S E N S O R Y DISTURB. • O V E R - R E A C T I O N • D O U B L E S L R • C O G W H E E L R E C O M M E N D A T I O N S T H E R A P I S T S I G N A T U R E 109 Appendix D: Canadian Back Institute Questionnaire (CBIQ) | INFORMATION ON YOUR INJURY ' - | THE FOLLOWING INFORMATION IS NECESSARY TO PROVIDE THE THERAPIST WITH A BETTER UNDERSTANDING OF YOUR INJURY. 1. A R E A ( S ) A F F E C T E D • N E C K • A R M • B A C K • L E G • O T H E R 2. C U R R E N T E P I S O D E C A U S E D B Y • W O R K A C C I D E N T • M O T O R V E H I C L E A C C I D E N T • U N K N O W N • O T H E R 3. O N S E T O F P A I N • S U D D E N • G R A D U A L 4. L E N G T H O F T I M E IN P A I N • 2 W E E K S O R L E S S • 3-10 W E E K S • 11 W K S - 6 M O • + 6 M O N T H S 5. A B I L I T Y T O D O D O M E S T I C C H O R E S • N O R M A L • M O S T B U T S L O W E R O F E W • N O N E 6. S P O R T S A N D A C T I V I T I E S • N O R M A L • L E S S • N O N E 7. IS A L A W Y E R I N V O L V E D W I T H Y O U R I N J U R Y ? • N O • Y E S 8. S M O K E R O N O • Y E S 9. N E E D F O R R E S T D U R I N G T H E D A Y O U N C H A N G E D T • R E S T L E S S T H A N 1/2 D A Y • R E S T M O R E T H A N 1/2 D A Y 10. I V IS IT M Y D O C T O R • N E V E R • R A R E L Y • O N C E A M O N T H • M O R E T H A N O N C E A M O N T H 11. N E E D F O R P A I N M E D I C A T I O N • N E V E R • O C C A S I O N A L L Y • O N C E A D A Y • S E V E R A L T I M E S / D A Y P L E A S E I N D I C A T E H O W Y O U R P A I N H A S A F F E C T E D T H E F O L L O W I N G A C T I V I T I E S : 12. W A L K 13. S I T 14. S T A N D 15. L I F T 16. D R E S S 17. W O R K 18. T R A V E L 19. S L E E P N O E F F E C T • • • • • • • • M I L D O O D • • • • • D I F F I C U L T • • • • • • • • I M P O S S I B L E • • • • • • • • Note that question 1, 2, 3, 4, 7 and 8 are not included in the calculation of questionnaire function score. The result is the 14-item questionnaire similar to the Low Back Outcome Score (30) 110 Appendix E: Alternate Approaches of Imputation for Missing Data Values 1) Central Tendency. One alternate approach uses imputed data to replace the missing values. An overall mean or group-specific (i.e. age or gender specific) mean may be appropriate. The median value could be used for highly skewed data. 2) Regression. In this method results are obtained using regression estimates for the missing values based on the other information that is provided within the record (39). 3) Indicator Method. Miettnen suggests that a modeling approach using indicator variables for missing values is a suitable alternative (39). In this method for each variable (X) with missing values, a new variable indicating "missing" (M) is created. The M variable would take the value of 1 when the value is missing and 0 other wise. Next the missing values of X are replaced with a constant Finally the regression is conducted with both the original variable with imputed values and the M variable (82). Generally, the constant imputed into the X variable is a product of the X and (1-M). Thus, i f the X variable is missing the product of X * ( l - M ) would be zero. When dealing with categorical variables this method is equivalent to creating a new missing category for the covariate (i.e. commonly coded as 999 or 99) (38). 4) Multiple Imputation. More complex statistical procedures have also been developed. Multiple imputation techniques create multiple data sets using a number of plausible imputations for each missing value (83). Each of these datasets area analyzed as i f complete, and then the results are combined in a manner that takes into account the variability that was generated (38). I l l Appendix F: Physical Examination Conduction Tests for Each Anatomical Neurological Area The dataset contains the physiotherapist's clinical interpretation of whether the neurological examination (conduction tests) was positive or negative. The conduction tests were categorized into summary variables representing their specific neurological areas: C6 (bicep reflex, anterior deltoid, bicep), C8 (triceps reflex, extensor digitorum longus, triceps power), L4 (knee reflex, ankle dorsiflexion), and SI (ankle reflex, plantar flexion, gluteus maximus). After the tests were categorized, they were further classified as either positive or negative based on the results of the individual conduction tests. For each anatomical area there typically is a deep tendon reflex test and 1 or 2 motor power tests. If both the deep tendon reflex and the motor power tests were positive, the anatomical region was presumed positive. If both the deep tendon reflex and the motor power tests were negative then the anatomical region was considered negative. If either the tendon reflex test or both of the motor power tests were positive then the anatomical region was presumed positive. If the deep tendon reflex was not tested, and discordant results remain among the motor power tests than the anatomical region was presumed negative (i.e. normal). Note: -ve = normal, +ve = abnormal, nt = not tested Patients were categorized as negative for C6 based on the fo lowing test results: Bicep reflex Anterior deltoid Bicep -ve -ve -ve -ve -ve nt -ve nt -ve nt -ve -ve -ve nt nt nt -ve nt nt nt -ve No other combinations were observed, and therefore are not listed. Patients were categorized as positive for C6 based on the fo lowing test results: Bicep reflex Anterior deltoid Bicep +ve +ve +ve +ve nt +ve +ve nt nt +ve -ve -ve +ve nt -ve nt +ve +ve -ve +ve +ve -ve +ve nt -ve nt +ve No other combinations were observed, and therefore are not listed. 112 Patients were categorized as negative for C8 based on the fo lowing test results: Tricep reflex Extensor digitorum longus Tricep -ve -ve -ve -ve -ve nt -ve nt -ve nt -ve -ve -ve nt nt nt -ve nt nt nt -ve nt +ve -ve nt -ve +ve No other combinations were observed, and therefore are not listed. Patients were categorized as positive for C8 based on the following test results: Tricep reflex Extensor digitorum longus Tricep +ve +ve +ve +ve nt +ve +ve- +ve nt +ve nt nt +ve -ve -ve +ve -ve nt +ve nt -ve -ve +ve nt -ve nt +ve No other combinations were observed, and therefore are not listed. Patients were categorized as negative for L4 based on the following test results: Knee reflex Ankle dorsiflexion -ve -ve -ve nt nt -ve No other combinations were observed, and therefore are not listed. Patients were categorized as positive for L4 based on the following test results: Knee reflex Ankle dorsiflexion +ve +ve +ve -ve +ve nt nt +ve -ve +ve No other combinations were observed, and therefore are not listed. 113 Patients were categorized as negative for SI based on the fo lowing test results: Ankle reflex Plantar flexion Gluteus maximus -ve -ve -ve -ve -ve nt -ve nt -ve nt -ve -ve -ve +ve -ve -ve -ve +ve No other combinations were observed, and therefore are not listed. Patients were categorized as positive for SI based on the fol owing test results: Ankle reflex Plantar flexion Gluteus maximus +ve +ve +ve +ve -ve -ve +ve -ve nt +ve nt -ve nt nt +ve nt +ve nt No other combinations were observed, and therefore are not listed. 114 Appendix G : Waddell's Non-organic Signs • Tenderness: •^Superficial - the patient's skin is tender to light pinch over a wide area of lumbar skin •^Non-anatomic - deep tenderness felt over a wide area, not localized to one structure • Simulation Tests: Axial Loading - light vertical loading over patient's skull in the standing position causes typical lumbar pain -^Acetabular Rotation - back pain is reported when the pelvis and shoulders are passively rotated in the same plane as the patient stands. This is considered to be a positive test i f pain is reported in the first 30 degrees. • Distraction Tests: •^Straight Leg Raise Discrepancy - marked improvement of straight leg raising on distraction as compared to formal testing •^Double Leg Raise - when both legs are raised after straight leg raising, the organic response would be a greater degree of double leg raising. Patients with a non-organic component demonstrate significantly less double leg raise as compared to the single leg raise Regional Disturbances: T> Weakness - cogwheeling or giving way of many muscle groups that cannot be explained on a neurological basis •^Sensory disturbance - diminished sensation fitting a "stocking" rather than a dermatomal pattern • Overreaction: •^disproportionate verbalization, facial expression, muscle tension and tremor, collapsing or sweating 115 Appendix H: Common Techniques of Internal Validation for Logistic Regression 1) Single sample statistics. Statistics such as Akaike's Information Constant (AIC) or Schwarz Bayesian Criterion (SC) provide simple estimators of the generalization error. These methods use the entire sample to generate an estimator. Model validation based on these indices may over estimate performance (60, 61). For instance, i f another sample of patients (even when drawn from the same population as the original) is tested with the model, typically the discriminative ability is reduced (61). This is because the same data that was used to build the model and generate parameter estimates is also being used to test the model (60). Several methods have been described to attempt to address this limitation (i.e. split-sample, leave-one-out, or external validation). 2) Split-Sample Cross Validation. A split-sample approach can be employed, in which a percentage (typically Vi or 3A) of the data is randomly selected to build the model (training sample), while the remainder of the data is used for test purposes or validation (55). The test, sample is not used in any way during the training. The disadvantage of this approach is that the smaller sample size will result in loss of some precision in the coefficient estimates. In addition this method of validation has shown to underestimate performance in logistic regression models (61). 3) Leave- One-Out Cross Validation. Other methods of cross-validation leave other fractions of the data out for validation purposes. The dataset may be divided into k subsets, of approximately equal size, each of which will be used to both train and test the data. The model is trained k times with one of the subsets left out to compute the error criterion of interest. If k is equal to the entire sample then the validation is termed "leave-one out" cross validation, because each time only one subject is left out for test purposes (62). During this procedure one record is omitted from the data. The regression model is fit with the remaining n-1 records. The parameters of the model are estimated using this analysis dataset and then these parameters are used to estimate the outcome on the single holdout observation. Note that the single holdout observation was not used to generate the parameter estimates. The process is repeated by removing a different observation each time. The accuracy of the model is estimated by comparing the predicted to observed outcome observation in the holdout samples. 116 Appendix I: Table Comparing the Study Population (in the Comparison Study of Acute vs. Chronic Injury) and Those That Were Excluded Due to Not Responding to Either the Entry or Exit Patient Questionnaire Table 1 - Characteristics associated with inclusion in comparison study based on completion of both initial and discharge patient questionnaire. Variable Respondents Excluded due to p value (n=3472) missing questionnaire (n=5473) Age - mean yrs (SD) 35.1 (11.1) 35.9(11.4) 0.0088 Gender - % female 55.8 53.8 NS Work status - % not working 56.0 58.3 0.0326 Lawyer retention - % with lawyer 36.1 35.6 NS Lagtime - mean no. of days (SD) 30.4 (23.5) 31.6(24.7) NS Smoking status - % current or former 34.4 39.6 <0.0001 Province - % British Columbia 31.9 25.8 <0.0001 Alberta 18.5 20.8 0.0079 Saskatchewan 3.0 4.2 0.0048 Ontario 40.8 42.3 NS Quebec 1.2 2.1 0.0025 Nova Scotia 4.5 4.5 NS Previous treatment - % Yes 19.6 21.1 NS Concurrent treatment - % Yes 13.2 11.2 O.0001 Comorbid medical conditions - % yes 8.3 10.9 0.0241 Non-organic signs >3 positive - % 5.0 6.7 0.0008 CBIQ raw score on presentation No. of respondents 7698 Mean score (SD) 35.6(11.6) 33.7(11.7) O.0001 Initial pain intensity No. of respondents 2500 3907 Mean score (SD) 5.9 (2.0) 6.0(2.1) NS Median 6.0 6.0 Duration of treatment program Mean days (SD) 72.1 (66.0) 61.7(71.0) <0.0001 117 Appendix J: Comparison of the Outcome, Personal, Clinical and Treatment Related Characteristics in Acute, Early Chronic, and Chronic Patient Populations After a Motor Vehicle Collision Table 1 - Outcome information for 5581 patients presenting with acute, early chronic or chronic injuries following a motor vehicle collision Outcome Acute Early Chronic Chronic £91 days >91 days £6mos > 6 mos (N = 3075) (N = 958) (N = 1548) Clinically Important Change in CBIQ Positive Outcome (>10%) - no./total no. (%) 2224 / 3075 (72.3) 588/958 (61.4) 806 / 1548 (52.1) Negative Outcome (<10%) 681/ 3075 (27.7) 370 / 958 (39.6) 742/ 1548(47.9) Mean Change in CBIQ Mean (SD) 13.8 (11.1) 9.1 (9.4) 6.8 (8.4) Median 13.0 8.0 6.0 CBIQ Raw Score at Presentation Mean (SDO 29.3 (10.5) 30.1 (9.6) 32.2 (9.0) Median 29.0 30.0 33.0 CBIQ Raw Score at Discharge Mean (SD) 43.1 (12.2) 39.2 (11.0) 39.0(10.2) Median 43.0 39.0 39.0 Bold indicates p<0.05 with Bonferroni correction for multiple comparisons applied. Comparisons made between adjacent columns only. If all three columns bold, then both the Acute vs. Early Chronic comparison and Early Chronic vs. Chronic columns were significant. 118 Appendix J: Comparison of the Outcome, Personal, Clinical and Treatment Related Characteristics in Acute, Early Chronic, and Chronic Patient Populations After a Motor Vehicle Collision Continued Table 2 - Personal characteristics of 5581 patients presenting with acute, early chronic or chronic injury after a motor vehicle collision Personal Characteristics Acute Early Chronic Chronic £91 days >91 days £ 6 mos > 6mos (N = 3075) (N = 958) (N = 1548) Age no. of respondents 3075 958 1548 Mean age - yr(SD) 35.7(10.6) 36.1 (10.8) 36.2(10.9) Female Sex - no./total no. (%) 1719/3075 (55.9) 536/ 958 (54.9) 954/1548(61.6) Lagtime no. of respondents 3053 867 1548 Mean no. of days (SD) 31.4 (23.5) 133.6(26.1) 434.0(321.6) Median 25 132 348 Off work because of collision no. / total no. (%) 1749/3075 (56.9) 506 / 958 (52.8) 506 /1548 (32.7) Physical Demands of Work - no./total no (%) Sedentary 447/2484(18.0) 133 /733 (18.1) 239/ 1203 (19.9) Light 480/2484(19.3) 129/733 (17.6) 282/ 1203 (24.4) Medium 682 / 2484 (27.5) 192/733 (26.2) 305/ 1203 (25.4) Heavy 875 / 2484 (35.2) 279/733 (38.1) 377/ 1203 (31.3) If remained at work has modified work duties because of collision. no. / total no (%) 466/ 1310(35.6) 187 / 490 (38.2) 279 / 1024 (27.2) Retained a lawyer on or before first visit to clinic - no./total no. (%) 913 / 3075 (29.7) 453 / 958 (47.3) 943/1548(60.9) Retained a lawyer at some point prior to discharge - no./total no. (%) 1127/3075(36.7) 474 / 958 (49.5) 935 /1548 (60.4) Smoking Status current or former - no./total no. (%) 1054/3072 (34.3) 336/957(35.2) 561 / 1547 (36.3) non-smoker - no./total no/ (%) 2018/3072 (65.7) 621 /957(64.8) 986/ 1547 (63.7) Province - no./total no. (%) British Columbia 1020/3075(33.2) 428 / 958 (44.7) 573 /1548 (37.0) Alberta 575/3075 (18.7) 261 / 958 (27.2) 563 /1548 (36.4) Saskatchewan 81 /3075(2.6) 84/958 (8.8) 96/1548 (6.2) Ontario 1216/3075(39.5) 148/958 (15.5) 259/ 1548 (16.7) Quebec 43/3075 (1.1) 15 /958 (1.6) 36/ 1548 (2.3) Nova Scotia 149/3075 (4.9) 22 / 958 (2.3) 21 / 1548 (1.4) Bold indicates p<0.05 with Bonferroni correction for multiple comparisons applied. Comparisons made between adjacent columns only. If all three columns bold, then both the Acute vs. Early Chronic comparison and Early Chronic vs. Chronic columns were significant. Lagtime statistical significance is not indicated as this was the manner for establishing the three groups. 119 Appendix J: Comparison of the Outcome, Personal, Clinical and Treatment Related Characteristics in Acute, Early Chronic, and Chronic Patient Populations After a Motor Vehicle Collision Continued Table 3 - Clinical characteristics of 5581 patients presenting with acute, early chronic or chronic injury after a motor vehicle collision Clinical Variables Acute Presentation (N = 3075) Early Chronic (N = 958) Chronic (N = 1548) Location of Pain - no./total no. (%) Neck 916/3075 (29.8) 228 / 958 (23.8) 399/ 1548 (25.8) Neck & Back 1764/3075 (57.4) 520/958 (54.3) 819/ 1548 (52.9) Back 395/3075 (12.8) 210/958 (21.9) 330/ 1548 (21.3) Other Areas Affected - no/total no. (%) Headache 772/3075 (25.1) 219/958 (22.9) 382/ 1548 (24.7) Mid-Back 1227/3075(39.9) 323 / 958 (33.7) 544/ 1548 (35.1) Extremity Pain - no./total no. (%) Upper Limb 422/3075 (13.7) 118/958 (12.3) 196/1548 (12.7) Lower Limb 331 /3075 (10.8) 135/958 (14.1) 280/1548 (18.1) Number of Symptoms Mean no. (SD) 2.47 (1.24) 2.37 (1.26) 2.43.(1.31) Median 2.0 2.0 2.0 Neurological Testing - no./total no. (%) Positive 39/3075 (1.3) 13/958(1.4) 25 / 1548 (1.6) Non-Organic Signs no./total no. (%) 344/3075 (11.2) 139/958 (14.5) 234/1548 (15.1) Any positive 160/3075 (5.2) 62/958 (6.5) 95 / 1548 (6.1) >3 positive Previous Episodes no./total no. (%) Neck Pain 683 /910(75.1) 213 /289 (73.8) 398/527(75.5) Back Pain 471 / 771 (61.2) 159/277 (57.4) 333 / 504 (66.1) Previous Spinal Surgery no. /total no. (%) Cervical 3 /3075 (0.1) 1 / 958 (0.1) 5 /1548 (0.4) Lumbar 12/3075 (0.4) 12/958 (1.3) 15 / 1548 (1.0) Comorbid Conditions no./total no. (%) Yes 309/3075 (10.1) 99/958 (10.3) 196 / 1548(12.7) Presence of Constant Pain (Yes) no./total no. (%) 918/ 1172 (78.3) 282 / 379 (74.4) 490 / 642 (76.3) Intensity of Pain Intensity Mean VAS (SD) 6.0 (2.0) 6.0 (2.0) 5.8(2.1) Median 6 6 6 Bold indicates p<0.05 with Bonferroni correction for multiple comparisons applied. Comparisons made between adjacent columns only. If all three columns bold, then both the Acute vs. Early Chronic comparison and Early Chronic vs. Chronic columns were significant. 120 Appendix J: Comparison of the Outcome, Personal, Clinical and Treatment Related Characteristics in Acute, Early Chronic, and Chronic Patient Populations After a Motor Vehicle Collision Continued Table 4 - Treatment Related Characteristics of 5581 patients presenting with acute, early chronic or chronic injury after a motor vehicle collision Treatment Related Variables Acute Early Chronic Chronic (N = 3075) (N = 958) (N = 1548) Bed rest previously prescribed -no./total no. % 6 / 3075 (0.2) 2 / 958 (0.2) 13 /1548 (0.8) Duration or treatment program Mean days (SD) 73.8 (66.3) 70.6 (60.0) 73.6 (60.0) Median 56 56 59 Number of treatment sessions Mean # (SD) 23.7 (17.1) 22.0 (14.9) 21.0(14.5) Median 20 20 19 Treatment Costs ($) Mean (SD) 2097 (1779) 1888 (1843) 1724(1601) Median 1587 1268 1170 Previous Investigations - no. ./total no. (%) X-rays 737/3075 (24.0) 315/958 (32.9) 572/1548 (37.0) Other 58/3075 (1.9) 50/958 (5.2) 159/1548 (10.3) Previous Treatment - no./ total no. (%) Yes 579/3075 (18.8) 407 / 958 (42.5) 774/1548 (50.0) Concurrent Treatment - no./total no. (%) Yes 251 / 3075 (8.2) 147 / 958 (15.3) 248/ 1548 (15.6) Health care seeking - no./total no. (%) Mean (SD) 1.5 (0.8) 2.0 (1.1) 2.1 (1.2) Median 1 1 2 Bold indicates p<0.05 with Bonferroni correction for multiple comparisons applied. Comparisons made between adjacent columns only. If all three columns bold, then both the Acute vs. Early Chronic comparison and Early Chronic vs. Chronic columns were significant. 121 Appendix K: Provincial Variation in the Management of Whiplash Associated Disorders. Table 1 - Comparison of Insurance System, Lagtime, Duration of Treatment, Treatment Costs, and Treatment Outcome by Province. Province Insurance Lagtime-mean Duration of Treatment % With a clinically System days (SD) treatment- Costs - mean $ important change between injury mean days (SD) (SD) in CBIQ Score date and (positive outcome) presentation date British Columbia Tort 42.0 (23.3) 57.7 (46.3) 1807(1404) 69.0 Alberta Tort 32.7 (24.2) 66.8 (53.2) 1304 (852) 65.7 Saskatchewan No-fault 40.7 (24.9) 76.4 (46.9) 1132(1294) 69.4 Ontario Tort 23.0 (20.4) 84.7 (78.5) 2529 (1927) 75.9 Quebec. No-fault 33.7 (22.3) 89.0(60.1) 2113 (2391) 80.4 Nova Scotia Tort 26.5 (19.8) 89.0 (76.7) 2402 (2538) 88.7 122 

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}]}"
                            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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0091251/manifest

Comment

Related Items