UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Utilization, access and outcome of surgical lumbar discectomy in British Columbia Quon, Jeffrey 2007

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

Item Metadata

Download

Media
831-ubc_2007-319116.pdf [ 9.86MB ]
Metadata
JSON: 831-1.0100710.json
JSON-LD: 831-1.0100710-ld.json
RDF/XML (Pretty): 831-1.0100710-rdf.xml
RDF/JSON: 831-1.0100710-rdf.json
Turtle: 831-1.0100710-turtle.txt
N-Triples: 831-1.0100710-rdf-ntriples.txt
Original Record: 831-1.0100710-source.json
Full Text
831-1.0100710-fulltext.txt
Citation
831-1.0100710.ris

Full Text

U T I L I Z A T I O N , A C C E S S A N D O U T C O M E O F S U R G I C A L L U M B A R D I S C E C T O M Y I N B R I T I S H C O L U M B I A by J E F F R E Y Q U O N M H S c , The University o f Bri t ish Columbia , 1999 D C , The Canadian Memor ia l Chiropractic College, 1986 A T H E S I S S U B M I T T E D IN P A R T I A L F U L F I L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R O F P H I L O S O P H Y in T H E F A C U L T Y O F G R A D U A T E S T U D I E S (Health Care and Epidemiology) T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A August 2007 © Jeffrey Quon, 2007 ABSTRACT Lumbar discectomy is the most commonly performed operation on the spine. This thesis includes four studies on the utilization, access and outcome of lumbar discectomy in British Columbia. Studies 1 and 2 estimate temporal trends and regional variations in age- and sex-adjusted lumbar discectomy rates using administrative data from the BC Linked Health Database. Study 1 shows that discectomy rates in public hospitals declined by 61% between 1990/91 and 2003/04. Rates declined faster after 2000/01 when lumbar discectomy became accessible in private clinics, preferentially for workers' compensation beneficiaries. Evidence of the diversion of compensated surgical patients from the public to private sectors was observed, however with no obvious attenuation of the decline in surgical rates for noncompensated patients remaining in the public system. In Study 2, rates of lumbar discectomy varied significantly between health service delivery areas (HSDAs), by almost five-fold during 2000/01-2003/04, up from 3.4 fold in 1990/91-1993/94. Studies 3 and 4 are based on prospective registry data on surgically treated lumbar disc patients at Vancouver General Hospital between November 1999 and December 2003. Study 3 identified significant sociodemographic and clinical determinants of waiting times in this population. Clinical severity (symptom duration, pain intensity) were appropriately associated, while most sociodemographic variables (age, sex, compensation status) were appropriately not associated with access to surgery. However professional occupation predicted waiting time, suggesting that access to surgery may not be based on clinical need alone. Study 4 estimated the effect of waiting 12 weeks or longer on the odds of pain improvement after lumbar discectomy. In propensity score-adjusted ordinal regression models, patients waiting 12 weeks or longer for surgery had a 41% lower odds of pain improvement. In the absence of declining disease prevalence, the implications of declining back surgery rates possibly include declining access to ii care and rising prevalence of disability in the community. This thesis research indicates that delayed access matters, and that "treatment within 12 weeks" may have utility as both a quality-of-care measure in health services research and as a benchmark for defining a maximum appropriate wait for lumbar discectomy. i i i TABLE OF CONTENTS ABSTRACT i i T A B L E OF C O N T E N T S iv LIST OF TABLES xii LIST OF FIGURES xiv A C K N O W L E D G E M E N T S x v DEDICATION x v i CO-AUTHORSHIP S T A T E M E N T x v i i i C H A P T E R 1 1 1.0 I N T R O D U C T I O N 1 1.1 Operational Definitions 1 1.2 Back Pain Epidemiology 1 1.3 Economic Burden of Back Pain 2 1.4 Lumbar Disc Herniation 3 1.5 Lumbar Discectomy 4 1.6 Clinical Uncertainty and Regional Variations in Rates of Utilization 7 1.7 Study Objectives and Hypotheses 8 1.7.1 Study 1: Temporal Trends in Utilization of Lumbar Discectomy 9 1.7.2 Study 2: Regional Variations in Utilization of Lumbar Discectomy 9 1.7.3 Study 3: Determinants of Waiting Time for Lumbar Discectomy 10 1.7.4 Study 4: Effect of Waiting Time on Outcome of Lumbar Discectomy 10 1.8 Thesis Organization 10 iv 1.9 Original Contributions : 12 1.10 References 14 CHAPTER 2 19 2.0 REGIONAL VARIATIONS AND TEMPORAL TRENDS IN RATES OF UTILIZATION OF LUMBAR DISCECTOMY 19 2.1 Lumbar Discectomy: A Discretionary Procedure 19 2.2 Regional Variations in Lumbar Discectomy 20 2.3 Determinants of Regional Variation in Lumbar Discectomy Rates 20 2.4 Implications of Discretionary Surgery 22 2.5 Access and Provision of Lumbar Discectomy in British Columbia 22 2.6 Current Research Needs 24 2.7 References • 26 CHAPTER 3 29 3.0 LITERATURE REVIEW: EFFECT OF T H E TIMING OF LUMBAR DISCECTOMY ON POSTOPERATIVE OUTCOMES 29 3.1 Clinical Uncertainty and the Optimum Timing O f Surgery 29 3.2 Mechanisms of Chronic Pain 29 3.3 Literature Review: Timing and Outcome of Lumbar Discectomy 30 3.3.1 Preoperative Symptom Duration and Outcome 31 3.3.1.1 Symptom Duration as a Primary Independent Variable..... 32 3.3.1.2 Symptom Duration as a Covariate 33 3.4 Summary: Effect of Timing of Surgery in Terms of Symptom Duration 36 3.5 References 51 C H A P T E R 4 55 4.0 L I T E R A T U R E R E V I E W : E F F E C T O F D E L A Y E D T R E A T M E N T O N B A C K P A I N O U T C O M E S 55 4.1 Waiting Time: A n Indicator of Access to Care 55 4.2 Literature Review: Waiting Time and Outcome of Lumbar Discectomy 56 4.3 Summary 59 4.4 References 64 C H A P T E R 5 68 5.0 T E M P O R A L T R E N D S I N R A T E S O F L U M B A R D I S C E C T O M Y I N B R I T I S H C O L U M B I A 68 5.1 Introduction 68 5.1.1 Epidemiology of Back Pain 68 5.1.2 Economic Costs of Back Pain 69 5.1.3 Lumbar Disc Herniation 69 5.1.4 Effectiveness O f Surgical Discectomy 70 5.1.5 Patterns of Utilization of Lumbar Discectomy 72 5.1.6 Questions About Back Surgery Rates in British Columbia 73 5.1.7 Objectives and Hypotheses 75 5.2 Methods 76 5.2.1 Data Source and Patient Selection 76 5.2.2 Compensation Status Definition 78 5.2.3 Data Analysis 79 5.2.3.1 Descriptive 79 5.2.3.2 Age-Specific and Age-Standardized Rates 79 5.2.3.3 Age-Specific Changes In Discectomy Rates Over 14 Years 79 5.2.3.4 Effect-Measure Modification of Trend In Rates 81 5.2.4 Sensitivity Analysis 81 5.3 Results 81 5.3.1 Descriptive Analyses 81 5.3.1.1 Total Number of Discectomies Before Exclusions 81 5.3.1.2 Frequencies of Diagnosis and Procedure Codes 82 5.3.1.3 Descriptive Analyses of the Study Population 82 5.3.1.4 Age-Specific Counts and Rates of Discectomy 83 5.3.1.5 Standardized Rates of Discectomy for the Study Population 83 5.3.1.6 Smoothed Estimates of the Trend in Rates 84 5.3.1.7 Effect-Modification of Trend in Rates by Treatment Period 84 5.4 Discussion 85 5.4.1 Modification of Trend by Compensation Status 85 5.4.2 Modification of Trend by Treatment Period 88 5.4.3 Strengths and Limitations 91 5.5 Summary 93 5.6 References H I vii CHAPTER 6 120 6.0 REGIONAL VARIATIONS IN RATES OF LUMBAR DISCECTOMY IN BRITISH COLUMBIA 120 6.1 Known and Unknown Sources of Variation 120 6.1.1 Study Objectives 122 6.2 Methods 123 6.2.1 Data Sources/Patient Selection 123 6.2.2 Small Area Definition 125 6.2.3 Data Analysis 126 6.2.4 Statistical Methods 126 6.3 Results 128 6.4 Discussion 130 6.4.1 Strengths and Limitations 133 6.5 References 145 CHAPTER 7 149 7.0 DETERMINANTS OF WAITING TIME FOR LUMBAR DISCECTOMY... . 149 7.1 Introduction 149 7.1.1 Compensation Status and Chronic Pain 150 7.1.2 Objectives and Hypotheses 152 7.2 Methods 153 7.2.1 Study Population 154 7.2.2 Data Collection 155 7.2.3 Exposure (Disability Compensation Status) 155 7.2.4 Outcome 156 7.2.5 Other Variables 156 7.2.6 Statistical Methods 157 7.3 Results 159 7.3.1 Descriptive 159 7.3.2 Access to Surgical Discectomy for Patients Overall 161 7.3.3 Access to Surgical Discectomy by Disability Compensation Status 161 7.3.4 Bivariate Correlates of Waiting Time for Surgical Discectomy 162 7.3.5 Multivariable Correlates of Wait Time 162 7.3.6 Adjusted Effect of Compensation Status 162 7.3.7 Adjusted Effect of Clinical Variables 163 7.4 Discussion 164 7.4.1 Strengths and Limitations 164 7.5 Summary 170 7.6 References • 182 C H A P T E R 8 186 8.0 E F F E C T OF WAITING T I M E O N O U T C O M E OF E L E C T I V E SURGICAL L U M B A R DISCECTOMY 186 8.1 Introduction 186 8.1.1 Defining a Maximum Appropriate Waiting Time for Surgical Discectomy 188 8.1.2 Study Objectives and Hypotheses 190 8.2 Methods 190 8.2.1 Study Population 191 8.2.2 Measurements 192 8.2.2.1 Outcome 192 8.2.2.3 Independent Variable 193 8.2.2.4 Confounders 193 8.2.3 Analysis 194 8.2.3.1 Descriptive Analyses 194 8.2.3.2 Unadjusted Effect of Waiting Time 194 8.2.3.3 Adjusted Effect of Waiting Time 195 8.2.3.4 Sensitivity Analyses - Conventional Multivariable Outcome Model 196 8.3 Results 197 8.3.1 Waiting Times 197 8.3.2 Study Population 198 8.3.3 Crude and Adjusted Effects of "Wait Group" on Ordinal Improvement 199 8.3.4 Propensity Score Model 199 8.3.5 Sensitivity Analysis 201 8.4 Discussion 202 8.4.1 Strengths and Limitations 204 8.5 Summary 206 8.6 References 222 CHAPTER 9 229 9.0 DISCUSSION 229 9.1 Review of Findings and Implications 229 9.1.1 Temporal Trends in Lumbar Discectomy Rates 229 9.1.2 Regional Variations in Lumbar Discectomy Rates 233 9.1.3 Determinants of Waiting Time for Lumbar Discectomy 234 9.1.4 Association Between Waiting Time and Outcome of Lumbar Discectomy 236 9.2 Methodological Strengths of Studies 238 9.2.1 Administrative Data 238 9.2.2 Prospective Treatment Registry Data 239 9.3 Methodological Limitations 240 9.3.1 Administrative Data 240 9.3.2 Registry Data 241 9.4 Original Contributions and Future Research 243 9.5 References .' 248 A P P E N D I C E S 252 Appendix A: Search Strategy - Studies on Symptom Duration 252 Appendix B: Search Strategy - Studies on Waiting Time 253 Appendix C: Carriere's T 2 Statistic 254 Appendix D : U B C Research Ethics Board Certificate of Approval 1 257 Appendix E: U B C Research Ethics Board Certificate of Approval 2 258 xi LIST OF TABLES Table 3.1 Reviewed Studies on Symptom Duration 37 Table 3.2 Results of Symptom Duration Studies 41 Table 4.1: Waitlist Control Studies on Back Pain 61 Table 4.2: Studies of Waiting Time and Back Surgery 63 Table 5.1: Study Inclusion Codes 95 Table 5.2: Study Exclusion Codes 96 Table 5.3: Hospitalizations for Lumbar Discectomy 99 Table 5.4: Frequencies of Diagnoses 100 Table 5.5: Surgical Procedure Codes 101 Table 5.6: Demographics of Study Population 102 Table 5.7: Proportions Hospitalized by Age and Sex 103 Table 5.8: Age-Specific Rates of Discectomy 104 Table 5.9: Directly Standardized Rates 105 Table 5.10: Standardized Rates by Sex and Payor 107 Table 5.11: Smoothed Estimates of Rate Changes 108 Table 5.12: Interaction Models 109 Table 5.13: Annualized Percent Changes in Rates 110 Table 6.1: Inclusion ICD-9 and CCP Codes 135 Table 6.2: ICD-9 and CCP Exclusion Codes 136 Table 6.3: Health Service Delivery Areas in BC 139 Table 6.4: Annualized Area Population Counts 140 Table 6.5: Descriptive Measures of Variation 142 Table 6.6: Standardized Discectomy Rates 143 Table 7.1: Characteristics of Excluded Patients 173 Table 7.2: Characteristics of Study Patients 174 Table 7.3: Clinical Characteristics of Study Patients 175 Table 7.4: Health System-Related Variables 176 Table 7.5: Waiting Times by Compensation Status 177 Table 7.6: Bivariate Correlates of Waiting Time 178 Table 7.7: Crude and Adjusted Effects of Predictor Variables 179 Table 8.1: Characteristics of Excluded Patients 210 Table 8.2: Characteristics of Study Population 212 Table 8.3: Clinical Characteristics of Study Patients 213 Table 8.4: Health System Characteristics of Study Patients 214 Table 8.5: Frequencies of Numbers of Patients in Each Recovery Category 215 Table 8.6: Variables in the propensity score model 216 Table 8.7: Balance in Covariates Before and After Adjusting for Propensity Score 217 Table 8.8: Unadjusted and Adjusted Effects of Waiting Time 219 Table 8.9: Effect of Waiting Time - Multivariable Proportional Odds Model 220 Table 8.10: Effect of Waiting Time: Multivariable Continuation Ratio Model 221 xiii LIST OF FIGURES Figure 5.1: Study Patient Selection Algorithm 98 Figure 5.2: Plot of Standardized Rates 106 Figure 6.1: Patient Selection Algorithm 138 Figure 6.2: Bar Chart of Area Discectomy Rates 141 Figure 6.3: Dotplot of Area Rates 144 Figure 7.1: Patient Flow Diagram 171 Figure 7.2: Patient Sampling Scheme 172 Figure 7.3: Kaplan-Meier Time-to-Surgery Functions 177 Figure 7.4: Covariate-Adjusted Time-to-Surgery For Compensated Patients 180 Figure 7.5: Covariate-Adjusted Time-to-Surgery for Professional Workers 181 Figure 8.1: Patient Flow Diagram 207 Figure 8.2: Patient Sampling Scheme 208 Figure 8.3: Kaplan-Meier Time-to-Surgery Curve 209 Figure 8.4: Histograms of Propensity Score Distributions 218 xiv ACKNOWLEDGEMENTS I am greatly indebted to my thesis supervisor, Adrian Levy, for his inspiring guidance, mentorship and unwavering support; for his unprecedented accessibility during all hours of the day and night; and for his incredible practicality when it came to solving—at least what always appeared to be at the time—monumental problems. I would like to thank Martin Schechter, my very first hero in the epidemiology world, for his enduring support and inspiration. Special thanks as well to the other members of my supervisory committee: Boris Sobolev, a marvelous statistician, scientist and philosopher—for showing me the path to clearer thinking; Charles Fisher, a superb surgeon and clinical epidemiologist—for fabulous content expertise, personal time and generous access to infrastructure at Vancouver General Hospital; and last, but certainly not least, Jacek Kopec, mentor and epidemiologist extraordinaire—for his seemingly unlimited knowledge of all things methodological. I am grateful to the many research assistants and coordinators at the Vancouver Spine Research Centre; Kim McGrail and Denise Morettin at the Centre for Health Services and Policy Research; Rita Sobolev, Karissa Johnston, and Kathy Lepik at the Center for Health Evaluation and Outcome Sciences at St. Paul's Hospital; all of the physicians and surgeons in the Combined Neurosurgical and Orthopaedic Spine Program at Vancouver General Hospital; and all study participants for their time and generous support of the Acute Lumbar Disc Study. I owe special appreciation to many faculty and staff members in the Department of Health Care and Epidemiology at the University of British Columbia. I can not list everyone, but I must thank Rick Mathias for facilitating my original foray into graduate school; Shirley Naismith for making sure I did not miss my first class; and both Laurel Slaney and Virginia Anthony for their collective advice and reliable shepherding over the years. I would also like to thank Sam Sheps for his calming influence around the time that I was anxiously preparing for my thesis defense. I am very grateful for the financial support of the Michael Smith Foundation for Health Research, the Canadian Institutes of Health Research, and both the Canadian Chiropractic Research Foundation (CCRF) and the British Columbia College of Chiropractors. A special mention definitely goes to Allan Gotlib for his dedicated work with the CCRF. Many thanks also to my chiropractic colleagues in Canada, especially in BC, for their support. I would like to thank my wonderful patients for tolerating my frequent absences from private practice, and for continuing to be so sincerely interested in my research work. I would also like to thank my incredible friends and extended family members for their love, encouragement, and constant presence in my life, all despite my somewhat distracted demeanor and lack of reciprocation during "PhD life". My deepest and heartfelt appreciation goes to my precious children, Erin and Marcus, and my wife, Pauline, who is my very best friend and the absolute love of my life. xvi DEDICATION To my wife, Pauline; children, Erin and Marcus; father, foe; and mother-in-law, Sheila. Also to the loving memory of my mother, Mai; and a true scholar in life, my father-in-law, Sam. xvii C O - A U T H O R S H I P S T A T E M E N T The candidate was responsible for conceptualizing and designing the four studies that are included in this thesis, for the acquisition and analysis of all data, and for the authorship of this entire document. For studies 1 and 2, these activities included the preparation and submission of an application for access to administrative data from the British Columbia Linked Health Database (BCLDH), and the development of specific coding criteria to create the analytic datasets. For studies 3 and 4 the candidate was responsible for the negotiation of access to medical charts at Vancouver General Hospital, the creation of a relational database, modification of existing standardized patient assessment forms, and the preparation of renewal applications for ethical approval from the University of British Columbia Clinical Research Ethics Board. While conducting all of the analyses in the thesis, the candidate was also responsible for the conceptualization, definition and creation of all variables from these data to characterize independent variables, study outcomes and confounders in each of the four studies. Members of the thesis committee who are listed as co-authors on the four manuscript chapters provided guidance in identifying and designing the program of research for the thesis (Drs. Levy and Sobolev), advice on acquiring access to administrative data (Dr. Levy), clinical expertise about the surgical waitlist and the treatment of lumbar disc disease in BC (Dr. Fisher), and statistical and methodological expertise in the conduct of the research (Drs. Levy, Sobolev, Kopec, and Schechter). For studies 3 and 4, access to primary data from a prospective treatment registry for surgical lumbar discectomy was granted by Dr. Fisher and the Vancouver Spine Program at Vancouver General Hospital (VGH). Dr. Fisher also facilitated an extension of the period of data collection at V G H specifically to accommodate the candidate's thesis research. All xviii co-authors provided expertise in the interpretation of the results, and contributed comments and revisions to the manuscripts in the thesis. xix CHAPTER 1 1.0 INTRODUCTION 1.1 Operational Definitions The terms "back pain" and "low back pain" are often used interchangeably in the literature.1^ In this thesis low back pain refers specifically to the localization of pain, muscle tension, or stiffness below the costal margin and above the inferior gluteal folds, with or without leg pain (i.e. sciatica).2;3 Unless otherwise indicated, the term back pain (unspecified) is used as a general reference to patients suffering mostly from low back pain, only with the acknowledgement that patients with symptoms above the costal margin are not explicitly excluded from this population of interest. 1.2 Back Pain Epidemiology Back pain is an important public health problem in all industrialized nations. The estimated prevalence of back pain varies depending on case definition and period of surveillance, however, investigators from various countries report a point prevalence of between 15% and 30%, an annual prevalence of 22% to 65%, and a lifetime prevalence of between 50% and 80%.4"7 In North America, national statistics indicate an annual prevalence of 14% in the United States (US),8 and between 12% and 14% among working-aged males and females, respectively, in Canada.9 In Saskatchewan, the point prevalence of "low back pain" (otherwise unspecified) was 28% in 1995, and the lifetime cumulative prevalence (persons reporting back pain at least once their lives) was 84% among adults aged 20 to 69 years.6 The annual cumulative incidence of low back pain in this population was 19 %. 6 1 Disability arising from back pain can manifest as an inability to perform various activities ranging from occupational, recreational or sports-related activities to basic activities of daily living (such as bathing, dressing, walking, or even getting into or out of bed). Anecdotaliy, it is reported that 90% of back pain patients regain normal function within three months of suffering an acute episode of back pain, however as many as 75% experience recurring or persisting symptoms that result in more disability within a year of the initial onset of pain.10 In Saskatchewan, studies of the prevalence of back pain according to graded severity indicated that approximately 49% of working aged adults have experienced so-called "low intensity/low disability" back pain, and that 12% and 11% experienced "high-intensity/low-disability" and "high-disability" low back pain, respectively, over a six-month period.6 The prevalence of high disability low back pain was also found to have increased with age (from 8% among patients aged 20-29 years to 14% among those aged 60-69 years).6 1.3 Economic Burden of Back Pain Studies have estimated the cumulative annual direct and lost productivity costs of back pain of over $50 billion in the US. 1 1 This may be lower than per capita costs elsewhere because the average duration of related work absence may be greater in other countries including Canada.12 In 1992, the estimated annual incidence of disabling back pain in the work force was 2% throughout the US, the United Kingdom, and Canada,12 however the estimated rate of job absenteeism due to back pain was only 9 days per patient per year in the US, compared to 20 and 30 days of absenteeism per patient per year in the United Kingdom and Canada, respectively. In British Columbia (BC), Workers' Compensation Board (WCB) statistics showed that both the number and proportion of total lost work days due to back claims remained stable between 1989 and 1998. In 1998, a total of 72,795 claims resulted in 761,800 lost work days.13 2 "Back strains" accounted for 26% (18,710) of these claims and 23% of total lost work days. In that same year, total compensation costs were C D N $837 million. While it is unclear what proportion of this amount was attributable to back claims alone, other studies of the costs of lumbar spine disorders suggest that approximately 10% of patients—typically with chronic conditions—account for more than 80% of total compensation costs, and that the one percent who undergo operative treatment make up the most expensive group.14 In-hospital surgical procedures account for up to one-third of the total costs of all spinal disorders.15 1.4 Lumbar Disc Herniation Of the few specific causes of low back pain that can be objectively identified and diagnosed, lumbar disc herniation is arguably the most important. Lumbar disc herniation is a particularly disabling cause of back pain and the most common indication for a back-related surgical operation in North America. The yearly incidence of herniated disc causing sciatica (pain with altered sensation and/or motor function in the leg) is 1 to 3%.16 The highest prevalence of lumbar disc herniation is seen among persons 30 to 50 years of age with twice as many men as women afflicted.17 In persons 25 to 55 years of age, about 95 percent of herniated discs occur in the lower lumbar spine (L4 and L5 level), whereas disc herniation above these levels is more common in persons older than 55 years.7;8 Less than 5% of all disc herniations take place in the thoracic region.18'19 Surgical operations for disc herniation account for approximately 73% of low back operations in the US. 2 0 Over 280,000 of these procedures are performed annually in the US alone.5 The natural history of disc herniation is not universally progressive, but does allow for spontaneous recovery.17 Sequential magnetic resonance images have shown that the herniated portion of the disc tends to regress over time, with partial to complete resolution after six months 3 in two thirds of afflicted persons.17 However, it is thought that approximately 10% of patients with acute lumbar disc herniation have pain that is still severe enough after six weeks to consider surgery.17 Absolute indications for surgical lumbar discectomy (warranting emergency surgery) include a progressive motor weakness and/or bowel or bladder dysfunction (cauda equina syndrome). However, as these indications are relatively rare (cauda equina syndrome occurs in less than 2% of cases), surgical treatment of lumbar disc herniation is usually performed on an elective basis. The usual indications for elective surgical lumbar discectomy are pain and functional disability in the presence of radicular signs and symptoms, a concordantly located disc herniation on imaging, and a failed response to conservative therapy no less than 6 weeks in duration.21 While widely acknowledged and accepted by clinicians, these indications are prone to subjective interpretation. Therefore, clinically similar patients can and do receive disparate opinions from different surgeons about their suitability for operative treatment. This uncertainty in interpretation is thought to contribute to wide geographic variations in discectomy rates internationally.22 1.5 Lumbar Discectomy The benefits from surgical versus non-surgical treatment of lumbar disc herniation is controversial. Observational studies have suggested that surgically treated patients have better outcomes than non-surgically treated patients,23124 and until recently, a small number of randomized controlled trials had also shown that lumbar discectomy was superior to non-surgical care25 as well as minimally invasive procedures such as chemonucleolysis (chemical discectomy) and percutaneous lumbar discectomy.14;23;26 One of the published randomized trials was conducted by Weber (1983) on 126 patients in Norway who were randomized to surgical discectomy or in-hospital conservative care, which 4 reported that conservatively treated patients experienced more pain and disability on average than surgically treated patients within the first 24 months of follow-up.25 At four years' follow-up there was no difference between groups, however a potentially clinically important trend still existed in favour of surgery.25;27 Recently, the results of a second randomized trial from the US were published, which also compared surgical lumbar discectomy to non-surgical treatment among 501 patients.28 In that study, both groups improved substantially over a two-year period. Although differences in mean changes in primary outcomes (SF-36 bodily pain and physical function scales, and modified Oswestry Disability Index) consistently favoured surgical treatment at all time points (six weeks, three months, six months, one year and two years), these differences were not statistically significant. In contrast, mean changes in secondary outcomes (including the Sciatica Bothersomeness Index) were significantly better for surgical patients. However the interpretation for actual practice is complicated by the fact that a large proportion of patients "crossed over" in both directions with 40% of surgical patients declining surgery and 45% of non-surgical patients eventually receiving surgery. It is noteworthy that patients who declined randomization in this trial were recruited into a concurrent observational study with identical selection criteria and outcome assessments. In the observational study, patients who chose surgical treatment reported greater improvement in all primary outcomes (body pain, general physical function, and back-specific function) and all but one of the secondary outcomes (work status) at six months', and one and two years' follow-up.28 That surgery appeared to be significantly more effective in the observational study than in the randomized trial points to the possibility of residual confounding in the observational study. For example, surgical patients undergoing an invasive intervention may hold higher expectations for a successful outcome than non-surgical patients,28'29 and this potential expectation bias could have had a liberal effect favouring surgery in the observational study.30 On 5 the other hand, the high proportion of crossovers in both treatment groups raises concerns about inevitable contamination—a conservative bias that would have necessarily attenuated the true benefit of surgery in the randomized study. One inference that can be drawn from this randomized trial was that the "intention" to perform surgery had no clear benefit over the intention to treat conservatively. However, a consistent benefit of surgery has been observed over non-surgical treatment in this and other observational studies.23;24 Furthermore, given only the possibility of residual confounding in the latest positive observational study as opposed to the certainty of contamination in the latest negative randomized study, it seems credible that surgical discectomy is more effective than non-surgical treatment, especially within the first year postoperatively. In the week prior to the final editing of this thesis, another randomized controlled trial from the Netherlands of early surgery versus prolonged conservative treatment for sciatica appeared in the literature.31 Surgery within two weeks of randomization was compared to prolonged care by general practitioners. Prolonged conservative care patients were offered surgery if they developed increasing leg pain not responsive to medication, or progressive neurologic deficits. Conservative care patients who had persistent sciatica six months after randomization were also offered surgery. Only 11% of patients who were assigned to surgery declined the operation due to spontaneous recovery within two weeks of randomization, whereas 39% of conservative therapy patients eventually underwent surgery. At one-year after randomization, there were no significant differences in disability scores or leg or back pain intensity between groups; however, rates of pain relief and global perceived recovery were faster for those assigned to early surgery. 6 In summary, surgically treated patients appear to improve more quickly than non-surgically-treated patients though the differences in outcomes between groups is seemingly attenuated over time. 1.6 Clinical Uncertainty and Regional Variations in Rates of Utilization The efficacy of lumbar discectomy depends upon careful selection of patients with appropriate indications.32 However, in the presence of uncertainty about the efficacy of treatment; and the interpretation (and implications) of clinical symptoms, signs and radiological tests for lumbar disc herniation, surgeons must make decisions under conditions of uncertainty when selecting patients for lumbar discectomy.33 Unresolved differences of opinion about how to value different clinical outcomes and complications (i.e. preferences), and their timing, are subject to various potential influences on physicians' diagnostic and therapeutic assessments.33 At the level of geographic jurisdictions, clinical uncertainty such as that observed with lumbar discectomy is probably the most important contributor to regional variations in per capita utilization rates of therapeutic procedures.25 To the extent that utilization rates may also be an indicator of access to medical procedures, uncertainty may also be a major contributor to disparities in access to health care. Consequendy, at the individual level, uncertainty can lead to the unnecessary treatment of inappropriately selected recipients as well as denial of effective treatment from inappropriately excluded patients. In a climate of constrained healthcare resources - as in BC, the rest of Canada and elsewhere - each instance of unnecessary surgery not only exposes a patient to unnecessary surgical risks, but also deprives a more appropriate surgical candidate of the treatment. Therefore, identification, and reductions in the number of unnecessary lumbar discectomies might improve 7 access to appropriate operations by freeing up constrained resources for patients most likely to benefit from treatment. As one author has commented, medical diagnoses are not circumscribed entities and therefore clinical uncertainty is inevitable in health care.34 Diagnostic tests do not perform with perfect sensitivity and specificity; therapeutic interventions are not equally effective for all individuals; and prognoses are not factual, but are rather only probabilistic, predictions of patients' expected outcomes.34 While such uncertainty cannot be eliminated in health care, the goal of clinical and administrative decision-makers is typically to find ways that unexplained variability be reduced and, ultimately, contained. Valid health services research therefore has an important role in informing practice and health care in the area. 1.7 Study Objectives and Hypotheses Studies of time trends in the utilization rates of known discretionary procedures are useful for informing policymakers of about secular variations in resource consumption. Additionally, studies of the regional variations in these procedures can help to identify areas in which utilization or access is potentially excessive or inadequate. That utilization rates are sometimes modifiable by very simple measures such as academic detailing means that this information is potentially an effective path to improved delivery of health services.35 Another avenue of potentially useful health services research is suggested by studies examining the significance of sociodemographic, clinical and health system determinants on timely access to lumbar discectomy. Studies of the "effects" of specific measures of access to healthcare are important as indicators of the functioning of the health care system and ultimately, for determining whether or not disparities in access ultimately matter in terms of patient-centred outcomes. 8 With that in mind, the overarching aims of this thesis were to document the patterns of utilization of lumbar discectomy in BC, and to determine if delayed access to lumbar discectomy was associated with a worse clinical outcome. Four related studies were conducted. 1.7.1 Study 1: Temporal Trends in Utilization The objectives of Study 1 were to: (1) estimate the temporal trend (i.e. time-related change) in population-based rates of lumbar discectomy in BC during 1990-2003; and (2) determine if the temporal trend in provincial lumbar discectomy rates differed according to time period (before and after the time that lumbar discectomy became available in private clinics in BC) and responsible insurer (workers' compensation versus provincial medical plan). The a priori hypotheses were that: (1) the temporal trend in rates of lumbar discectomy declined between 1990 and 2003 (within the context of reductions in acute care hospital resources in BC during the 1990's); and (2) the decline in rates was greater for compensation-insured than provincial-medical-plan-insured hospitalizations (with the reasoning that during a period of declining acute care resources and available bookings for surgical discectomy in BC, patients receiving workers' compensation would be less likely to be offered surgery due to their greater propensity for inferior outcomes). 1.7.2 Study 2: Regional Variations in Utilization In Study 2 the objectives were to: (1) quantify the variation in population-based lumbar discectomy rates across health service delivery areas (HSDAs) in BC in 2000-2003; and (2) compare the variation in rates in 2000-2003 to that of an earlier time period (1990-1993). In Study 2, the a priori hypotheses were that: (1) age- and sex-standardized rates differed significantly between provincial HSDAs; and (2) the variation in rates between HSDAs in BC was less in 2000-2003 than in 1990-1993 (based on the reasoning that greater constraints on acute care resources in 2000-2003 would motivate less discretionary [i.e. more stringent] selection of patients, and therefore less variation in the utilization rates of lumbar discectomy). 9 1.7.3 S t u d y 3: D e t e r m i n a n t s o f W a i t i n g T i m e a n d A c c e s s The objectives of Study 3 were to: (1) identify sociodemographic, clinical and local health care system determinants of waiting time for lumbar discectomy; and (2) determine if "receiving workers' compensation" was associated with longer waiting times and therefore reduced access to lumbar discectomy. The related a priori hypotheses in this study were that: (1) patients receiving compensation will have longer waiting times than non-compensated patients (assuming reduced access to surgical discectomy for compensated patients due to their reputation for inferior outcomes); and (2) patients with higher pain intensity at waitlist enrolment will have shorter waiting times than patients with lower pain intensity (assuming that greater symptom severity would be positively associated with quicker access to lumbar discectomy). 1.7.4 S t u d y 4: W a i t i n g T i m e a n d O u t c o m e The objective of Study 4 was to: determine if longer waiting times were associated with a worse clinical outcome after lumbar discectomy. The primary a priori hypothesis was that: patients who waited 12 weeks or longer had lower odds of experiencing pain reduction than patients who waited less than 12 weeks for lumbar discectomy. 1.8 T h e s i s O r g a n i z a t i o n This thesis is organized into eight chapters. The first chapter sets the stage by introducing public health and the economic importance of low back pain, lumbar disc herniation and lumbar discectomy. Chapter 2 provides a summary of what is known, and not known, about past and current trends, and geographic variations in utilization rates of lumbar discectomy in parts of Canada and the US. Chapter 3 reviews the literature on the association between the timing of lumbar discectomy and the clinical outcome of the intervention. The literature review begins with a discussion on the biological mechanisms that are potentially responsible for the association 10 between longer symptom duration and inferior outcomes, and then goes on to show the extant literature on the effect of preoperative symptom duration on the outcome of surgical lumbar discectomy. Chapter 4 discusses the unique attributes of waiting time for surgery and its distinction from other time intervals of interest along the symptom duration continuum. The chapter concludes with a review of the relatively few studies that provide any insight into the association specifically between waiting time for surgery and the outcome of lumbar discectomy. The four original studies for this thesis are presented in Chapters 5 through 8. Chapter 5 (Study 1) quantifies the magnitude of time-related changes in population-based rates of lumbar discectomy over a 14-year period in BC. Chapter 6 (Study 2) presents the results of the study on geographic variations in lumbar discectomy rates between health service delivery areas in the province. Chapter 7 (Study 3) examines the significance of sociodemographic, clinical and potential health system-related determinants of waiting time for lumbar discectomy. Chapter 8 (Study 4) presents the results of the study on the association between longer waiting times—a measure of delayed access to treatment—and the clinical outcome of lumbar discectomy. The final chapter (Chapter 9) summarizes the findings of the four studies, provides further interpretative information, and discusses the methodological strengths, limitations, and implications of this research. Portions of the introduction and literature reviews in Chapters 2 through 4 are repeated within the introduction sections of the articles in Chapters 5 through 8. Similarly, portions of the discussion sections of Chapters 5 through 8 are repeated in a summary discussion section in Chapter 9. This redundancy is necessary in a manuscript-based thesis because the articles are meant to be published and to stand on their own without reference to other chapters in the thesis. 11 1.9 Original Contributions All four studies in this thesis advance original knowledge in several areas, both substantive and methodological. The first study (Chapter 5) generated previously unknown information about the temporal trend in population-based utilization rates of lumbar discectomy during a period of restructuring and downsizing of acute care resources in BC. A study of the trends in back surgery rates in a Canadian jurisdiction has not been published in nearly a decade. Further, no previous publications have examined the trends in rates in BC. The second study (Chapter 6) provided new information on the extent of variation in lumbar discectomy rates in the BC using regions of analysis (health service delivery areas) that were representative of the level at which comprehensive core medical services (including spinal surgery) are ensured throughout the province. Although a previous review paper on low back pain36 reported the range of rates of "discectomies" across selected local health areas in BC, neither a reference for the source of these figures, nor a description of methods underlying these estimates was included. Furthermore, it was not clear whether the rates referred to lumbar and thoracic "discectomies" exclusively, or if cervical discectomies were also included in the rate estimates. A second report of discectomy rates in BC (and other parts of Canada) used census enumeration areas as the unit of analysis, which, again, do not represent the geographic level at which core medical services are provided throughout the province. The third study in the thesis (Chapter 7) presents new knowledge on the specific determinants of waiting times for surgical lumbar discectomy in BC using prospectively collected clinical data. This study addressed a previously unanswered question about whether or not compensation patients—due to their greater propensity for poorer outcomes—encountered a barrier to access to lumbar discectomy even after being placed on the surgical wait list. In contrast to an earlier study of the determinants of waiting time for lumbar discectomy 3 7 that used a 12 retrospective approach, the use of prospective data collection in the current thesis permitted complete follow-up of patients who received surgery as well as patients who did not receive surgery, thereby avoiding biased ascertainment of patients waiting times for surgery. 3 8 Extensive measures of clinical severity also allowed for case-mix to be controlled in the analysis of waiting times. The fourth and final study in the thesis (Chapter 8) presents new knowledge about the significance and magnitude of the association between longer waiting time and the clinical outcome of lumbar discectomy. An estimate of the size and significance of the effect of waiting time on the clinical outcome of surgical lumbar discectomy has not been previously reported. 13 1.10 References 1. Abenhaim L, Rossignol M , Valat JP et al. The role of activity in the therapeutic management of back pain - Report of the International Paris Task Force on Back Pain. Spine 2000;25:1S-33S. 2. Manek NJ and MacGregor AJ. Epidemiology of back disorders: prevalence, risk factors, and prognosis. Current Opinion in Rheumatology 2005;17:134-40. 3. van Tulder M W and Koes BW. Low back pain. American Family Physician 2002;65:925-8. 4. Frymoyer JW. Back Pain and Sciatica. New England Journal of Medicine 1988;318:291-300. 5. Andersson GBJ. Epidemiological features of chronic low-back pain. Lancet 1999;354:581-5. 6. Cassidy JD, Carroll LJ, and Cote P. The Saskatchewan health and back pain survey - The prevalence of low back pain and related disability in Saskatchewan adults. Spine 1998;23:1860-6. 7. Walker BF. The prevalence of low back pain: A systematic review of the literature from 1966 to 1998. Journal of Spinal Disorders 2000;13:205-17. 8. Deyo RA and Tsuiwu YJ. Descriptive Epidemiology of Low-Back-Pain and Its Related Medical-Care in the United-States. Spine 1987;12:264-8. 14 9. Cole DC, Ibrahim SA, Shannon HS, Scott F, and Eyles j . Work correlates of back problems and activity restriction due to musculoskeletal disorders in the Canadian national population health survey (NPHS) 1994-5 data. Occup Environ Med 2001;58:728-34. 10. Croft PR, Macfarlane GJ, Papageorgiou AC, Thomas E, and Silman AJ. Outcome of low back pain in general practice: a prospective study. British Medical Journal 1998;316:1356-9. 11. Frymoyer JW and Catsbaril W L . An Overview of the Incidences and Costs of Low-Back-Pain. Orthopedic Clinics of North America 1991;22:263-71. 12. Nachemson A L . Newest Knowledge of Low-Back-Pain - A Critical-Look. Clinical Orthopaedics and Related Research 1992;8-20. 13. Workers' Compensation Board of British Columbia: Statistics '98. Workers' Compensation Board of British Columbia . 1999. 14. Gibson J N A and Waddell G. Surgical interventions for lumbar disc prolapse. Cochrane Database of Systematic Reviews 2007. 15. Cats-Baril FJ. The economics of spinal disorders. In: Frymoyer JW, ed. The adult spine: principles and practice. New York: Raven Press, 1991:85-105. 16. Nachemson A. Low-Back-Pain - Are Orthopedic Surgeons Missing the Boat. Acta Orthopaedica Scandinavica 1993;64:1-2. 17. Jordon J, Morgan TS, Weinstein J, and Konstantinou K. Herniated lumbar disk. American Family Physician 2006;73:1240-2. 15 18. Oppenheim JS, Rothman AS, and Sachdev VP. Thoracic Herniated Disks - Review of the Literature and 12 Cases. Mount Sinai Journal of Medicine 1993;60:321-6. 19. Rossvoll I, Benum P, Bredland TR, Solstad K, Arntzen E, and Jorgensen S. Incapacity for Work in Elective Orthopedic-Surgery - A Study of Occurrence and the Probability of Returning to Work After Treatment. Journal of Epidemiology and Community Health 1993;47:388-94. 20. Taylor V M , Deyo RA, Cherkin DC, and Kreuter W. Low-Back-Pain Hospitalization -Recent United-States Trends and Regional Variations. Spine 1994;19:1207-13. 21. LarequiLauber T, Vader JP, Burnand B et al. Appropriateness of indications for surgery of lumbar disc hernia and spinal stenosis. Spine 1997;22:203-9. 22. Cherkin DC, Deyo RA, Loeser JD, Bush T, and Waddell G. An International Comparison of Back Surgery Rates. Spine 1994;19:1201-6. 23. Stevens CD, Dubois RW, Larequi-Lauber T, and Vader JP. Efficacy of lumbar discectomy and percutaneous treatments for lumbar disc herniation. Sozial-und Praventivmedizin 1997;42:367-79. 24. Atlas SJ, Deyo RA, Keller RB et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine 1996;21:1777-86. 25. Weber H . Lumbar-Disk Herniation - A Controlled, Prospective-Study with 10 Years of Observation. Spine 1983;8:131-40. 16 26. Hoffman R M , Wheeler KJ, and Deyo RA. Surgery for Herniated Lumbar Disks - A Literature Synthesis. Journal of General Internal Medicine 1993;8:487-96. 27. Bessette L, Liang M H , Lew RA, and Weinstein JN. Classics in Spine - Surgery literature revisited. Spine 1996;21:259-63. 28. Weinstein J N , Tosteson TD, Lurie JD et al. Surgical vs nonoperative treatment for lumbar disk herniation - The Spine Patient Outcomes Research Trial (SPORT): A randomized trial. Jama-Journal of the American Medical Association 2006;296:2441-50. 29. Flum DR. Interpreting surgical trials with subjective outcomes - Avoiding UnSPORTsmanlike conduct. Jama-Journal of the American Medical Association 2006;296:2483-5. 30. Lutz G K , Butzlaff M E , Atlas SJ, Keller RB, Singer D E , and Deyo RA. The relation between expectations and outcomes in surgery for sciatica. Journal of General Internal Medicine 1999;14:740-4. 31. Boszczyk B, Timothy J, Peul W, and Casey A T H . Neurosurgical training and the spine: Reflections on EANS Winter Meeting Luxembourg, February, 2006. Acta Neurochirurgica 2007;149:339. 32. Fisher C, Noonan V, Bishop P et al. Outcome evaluation of the operative management of lumbar disc herniation causing sciatica. Journal of Neurosurgery 2004;100:317-24. 33. Balsa AI, Seiler N , McGuire T G , and Bloche M G . Clinical uncertainty and healthcare disparities. American Journal of Law & Medicine 2003;29:203-19. 17 34. West A F and West RR. Clinical decision-making: coping with uncertainty. Postgraduate MedicalJoumal 2002;78:319-21. 35. Keller RB, Adas SJ, Singer D E et al. The Maine Lumbar Spine Study .1. Background and concepts. Spine 1996;21:1769-76. 36. Wing PC. Rheumatology: 13. Minimizing disability in patients with low-back pain. Canadian Medical Association Journal 2001;164:1459-68. 37. Shortt SED and Shaw RA. Equity in Canadian health care: Does socioeconomic status affect waiting times for elective surgery? Canadian Medical Association Journal 2003;168:413-6. 38. Sobolev B, Brown P, Zelt D, and Shortt S. Bias inherent in retrospective waiting-time studies: experience from a vascular surgery waiting list. Canadian Medical Association Journal 2000;162:1821-2. 18 CHAPTER 2 2.0 REGIONAL VARIATIONS AND TEMPORAL TRENDS IN SURGICAL LUMBAR DISCECTOMY RATES 2.1 Lumbar Discectomy: A Discretionary Procedure High variations in per capita use rates of lumbar discectomy across geographic regions suggest that despite the existence of generally accepted criteria for lumbar discectomy a wide range of discretion is inherent to the surgical decision making process.1 There is uncertainty in the interpretation and ascertainment of the clinical indications1 and differences in opinion about the long-term efficacy of lumbar discectomy.1 In addition, the choice between surgical and non-surgical treatment is complicated by differences in preferences of patients for alternative treatments. Afflicted individuals weigh differendy the trade-off between the desirable outcomes of rapid pain relief and functional recovery on the one hand, and the risks and costs of poor outcomes—including persistent pain, disability, operative complications and re-operations—on the other.2 Immediate operative complications from lumbar discectomy are rare but potentially serious. In the US, the incidence of mortality during the operation is estimated to be 0.07%, infection necessitating intravenous antibiotics is 0.3%, and major neurologic complication (cauda equina syndrome, nerve root damage, and worsening radiculopathy) also 0.3%. A more frequent complication, however, is recurring herniation or persistent pain requiring repeat-surgery in about 10% of patients.3 In the high quality Maine Lumbar Spine Study, the rate of re-operation after one year was 6.5% and climbed to 19.4% at five years.4;5 19 2.2 Regional Variations in Lumbar Discectomy Rates Given that lumbar discectomy, and back surgery in general, is a discretionary procedure, it is not surprising that high variations in lumbar discectomy rates are observed between geographic regions.1'6 This finding is observed for both large and small geographic jurisdictions. Studies from over a decade ago revealed five to fifteen-fold variations in the rates of back surgery between industrialized countries and between large and small regions of Canada and the US. 5 ; 7 ; 8 The age-standardized rate of operations in the US was at least 40% higher than in any other country and was more than five times that of the rate observed in England and Scotland.9 Back surgery rates in Ontario and Manitoba were only one-half to one-third the US rate although greater than the U K . 9 Notwithstanding lower surgery rates on average in Canada than the U.S., large variations in discectomy rates are observed within provinces. In a review paper on low back pain, Wing reported that rates of back surgery in BC varied from 21 per 100,000 residents to 180 per 100,000 residents between local health areas of 50,000 or greater population.8'8 However the methodology underlying these important estimates was not described. Upon personal communication with the author, it was confirmed that these rates represented direcdy age- and sex-standardized rates of utilization, and that the lowest and highest discectomy rates were observed in the Nanaimo and Kamloops areas, respectively. Also, as LHAs with population counts below 50,000 were associated with even lower surgery rates, the apparent 9-fold difference in rates between larger LHAs was a conservative estimate of the variation across all LHAs in BC. An important methodological issue is that LHAs also historically coincide with school planning districts and currendy do not adequately represent the geographic level at which access to core health care services (including spinal surgery) are ensured in BC, (ie HSDAs). 8 Inefficiencies in the planning and management of health care resources, and in the related accountability for planning decisions, at the level of somewhat autonomous LHAs helped to motivate the 20 establishment of one provincial and five geographic health authorities (HAs) in 2001. While HAs are now responsible for larger-area budgeting and planning responsibilities, their component HSDAs represent the level at which comprehensive medical services are provided, based not only upon provincial geography, but also natural patient and physician referral patterns throughout the province. While variations must exist at the HSDA level, studies at this level of aggregation are needed, first, because no studies yet exist regarding lumbar discectomy in BC at this level of analysis and, secondly, because this provides a preliminary, yet important, assessment of how equitably lumbar discectomy is utilized across different service regions created under the latest reforms. 2.3 Determinants of Regional Variation in Lumbar Discectomy Rates Some investigators have found that "supply-side" variables, such as the number of physicians and hospital beds within regions, and "demand" variables, such as age and sex, have explained some of the variation in surgical rates. However, these factors collectively account for only up to 15% of the variability in rates of back surgery.6 In contrast, physicians' personal beliefs and clinical uncertainty are strongly associated with clinical decision making and, therefore, low rate and high rate areas potentially indicate areas of inappropriate under- and over-servicing, respectively.10111 Only a few studies have examined the effect of physician discretion on regional variations direcdy. However, there exists an extensive body of literature showing that physicians' personal beliefs and clinical uncertainty can result in differences in clinical decision-making.11 Furthermore, geographic variations in the use of operations are consistendy larger when the level of uncertainty about their effectiveness is high and least when it is low. 1 0 , 1 1 Consequently, many investigators 21 currently believe that practice style, or physician discretion and their perceptions of the value of the operation in question is probably an important factor contributing to geographic variations.1'11 2.4 Implications of Discretionary Surgery Low back pain is responsible for an important social and economic burden in all industrialized countries. Lumbar disc herniation is a particularly disabling cause of low back pain and is one of the few specific back pain diagnoses with a readily identifiable pathologic basis. In-hospital procedures for back pain account for up to one-third of total costs for back pain12, and lumbar disc herniation is the most common reason for an in-hospital diagnostic or therapeutic procedure for back pain.13 Although the results of randomized trials suggest that the long-term outcomes of surgical lumbar discectomy are not superior to those of non-surgcal care, functional improvement is typically faster with surgery,14 and this has potentially important implications for policymakers interested in reducing the huge economic losses associated with disability from lumbar disc hernation in the workplace. However, the discretionary component of surgical lumbar discectomy raises understandable concerns about potentially excessive utilization in some areas exhibiting higher utilization rates. 2.5 Access to, and Provision of, Lumbar Discectomy in British Columbia In BC, residents receive universal insurance coverage for "medically necessary" hospital services. Patients have access to services including surgical lumbar discectomy within public hospitals at no direct costs to themselves. Since 2000, for-profit clinics have been licensed to perform single-level surgical discectomies in BC. Only two, both located in Vancouver, perform lumbar discectomies with any regularity. Despite being partially publicly funded (for procedures involving workers' compensation beneficiaries, for example) for-profit clinics do not report utilization statistics to the BC Ministry of Health or the Canadian Institute for Health Information 22 (CIHI). While the proportion of province-wide surgical discectomies performed in those clinics is thought to be minimal based on anecdotal reports, precise estimates of this caseload are not available. As with many surgical procedures in Canada, access to surgical discectomy in BC is managed through waitlists. According to administrative staff at one tertiary care centre, the flow of candidates for surgical lumbar discectomy through the system generally proceeds in a standardized fashion. Patients are usually referred to a surgeon by a primary care physician, and at the time of the surgical consultation, those considered appropriate for operative care and who are willing to undergo surgical discectomy are assigned a priority rating based on the severity of their symptoms and signs, and functional disability. Usually on the same day that surgery is decided upon, patients are placed onto a waitlist and given a tentative procedure date based on priority rating, length of the existing queue, and availability of operating room time. At a subsequent date, patients are contacted and given a confirmed date of surgery that may or may not coincide with their original tentative date, depending upon the inflow and outflow of other (especially higher priority) patients in the queue. According to a surgeon at that same facility, maximum waiting time targets for lumbar discectomy by priority ratings prior to 2003 were as follows: eight hours for emergency, 24 hours for semi-emergency, 48 hours for urgent, and "whenever a bed (was) available" for non-urgent patients. However it was also acknowledged that, actual waiting times regularly exceed these targets due to lack of hospital capacity. Furthermore, this "generally standardized flow" of patients through the system is likely influenced by physician practice style. In 1992, a survey of secondary and tertiary care facilities in BC indicated that waiting lists were compiled by individual surgeons offices in ten, by surgical departments in four and by admissions/operating room departments in three out of 17 participating facilities.15 In 13 facilities, the authority making the decision for the next patient to be admitted was an individual surgeon.15 In five facilities, physician 23 preferences were reported as "a first priority" for admission decisions. Queue jumping was reported by delegated representatives of all facilities surveyed, and nearly 80% of queue jumping instances were reportedly not on the basis of emergency indications, but were instead "on the basis of physician/surgeon preferences, requests from senior ministry of health officials and, sometimes, from members of the legislature." 2.6 Current Research Needs Rates of back surgery including lumbar discectomy have increased in the US over the past decade despite lack of evidence of parallel increases in the prevalence of back pain.1 3 ; 1 6 However, litde is known about the recent trends in back surgery rates in Canada. Over a 10-year period, aggregated rates of neck and back surgical procedures in Ontario increased modestly (from 72.1 per 100,000 in 1982 to 82.3 in 1992),17 however estimates for the trend after 1992, and for other provinces, are not known. In BC, changes in the health care system occurred throughout the 1990's, which conceivably reduced the province's overall capacity to perform lumbar discectomies in public hospitals. Reductions in real spending in the hospital sector resulted in overall reductions in the number of staffed beds in short-term care units. , 8 ;19 As discussed above, for-profit surgery clinics also began to perform uncomplicated, single-level lumbar discectomies in 2000. Notwithstanding evidence of increasing utilization in jurisdictions outside of BC, the declining trend in the supply of acute care beds in BC over the 1990s may have facilitated a decline in lumbar discectomy rates (and therefore access) over that time period. The impact, if any, of these health system changes on lumbar discectomy rates has not been studied. Concerns about declining utilization rates and, potentially, access to lumbar discectomy are warranted because delayed access to treatment may negatively affect patients' outcomes. However, this relationship is poorly understood because the 24 effect of delayed access to treatment on the outcome of lumbar discectomy has never before been assessed. Previous investigators have demonstrated the effect of another measure of the timing of treatment—preoperative symptom duration—on the outcome of lumbar discectomy20 and shown that longer symptom duration is associated with a poorer outcome. However the specific effect of waiting time (i.e. time between waitlist enrolment and surgery) has not been studied. This distinction is important because there are some important differences between the two measures of the timing of treatment (preoperative symptom duration and waiting time for surgery). These differences are the subject of Chapter 3. 2 5 2.7 References 1. Keller RB, Atlas SJ, Singer D E et al. The Maine Lumbar Spine Study: 1. Background and concepts. Spine 1996;21:1769-76. 2. Hoffman R M , Wheeler KJ, and Deyo RA. Surgery for Herniated Lumbar Disks - A Literature Synthesis. Journal of General Internal Medicine 1993;8:487-96. 3. Yorimitsu E, Chiba K, Toyama Y , and Hirabayashi K. Long-term outcomes of standard discectomy for lumbar disc herniation - A follow-up study of more than 10 years. Spine 2001;26:652-7. 4. Adas SJ, Deyo RA, Keller RB et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine 1996;21:1777-86. 5. Adas SJ, Keller RB, Chang Y C , Deyo RA, and Singer D E . Surgical and nonsurgical management of sciatica secondary to a lumbar disc herniation - Five-year outcomes from the Maine Lumbar Spine Study. Spine 2001;26:1179-87. 6. Taylor V M , Deyo RA, Cherkin DC, and Kreuter W. Low-Back-Pain Hospitalization -Recent United-States Trends and Regional Variations. Spine 1994;19:1207-13. 7. Volinn E, Mayer J, Diehr P, Vankoevering D, Connell FA, and Loeser JD. Small Area Analysis of Surgery for Low-Back-Pain. Spine 1992;17:575-81. 8. Wing PC. Rheumatology: 13. Minimizing disability in patients with low-back pain. Canadian Medical Association Journal 2001;164:1459-68. 26 9. Gherkin DC, Deyo RA, Loeser JD, Bush T, and Waddell G. An International Comparison of Back Surgery Rates. Spine 1994;19:1201-6. 10. Stano M . Further Issues in Small Area Variations Analysis. Journal of Health Politics Policy and Law 1991;16:573-88. 11. Leape LL . Unnecessary Surgery. Annual Review of Public Health 1992;13:363-83. 12. Cats-Baril FJ. The economics of spinal disorders. In: Frymoyer JW, ed. The adult spine: principles and practice. New York: Raven Press, 1991:85-105. 13. Weinstein JN , Tosteson T D , Lurie JD et al. Surgical vs nonoperative treatment for lumbar disk herniation - The Spine Patient Outcomes Research Trial (SPORT): A randomized trial. Journal of the American Medical Association 2006;296:2441-50. 14. Boszczyk B, Timothy J, Peul W, and Casey A T H . Neurosurgical training and the spine: Reflections on EANS Winter Meeting Luxembourg, February, 2006. Acta Neurochirurgica 2007;149:339. 15. Amoko D H , Modrow RE, and Tan J K H . Surgical waiting lists II: current practices and future directions using the province of British Columbia as a test study. Healthcare Management F O R U M 1992;5:3439. 16. Weinstein JN , Lurie JD, Tosteson T D et al. Surgical vs nonoperative treatment for lumbar disk herniation - The Spine Patient Outcomes Research Trial (SPORT) observational cohort. Journal of the American Medical Association 2006;296:2451-9. 27 17. Lavis J N , Maker A, Anderson G M et al. Trends in hospital use for mechanical neck and back problems in Ontario and the United States: discretionary care in different health care systems. Canadian Medical Association Journal 1998;158:29-36. 18. Detsky AS and Naylor CD. Canada's health care system - Reform delayed. New England Journal of Medicine 2003;349:804-10. 19. Sheps SB, Reid RJ, Barer M L et al. Hospital downsizing and trends in health care use among elderly people in British Columbia. Canadian Medical Association Journal 2000;163:397-401. 20. Ng L C L and Sell P. Predictive value of the duration of sciatica for lumbar discectomy - A prospective cohort study. Journal of Bone and Joint Surgery-British Volume 2004;86B:546-9. 28 CHAPTER 3 3.0 LITERATURE REVIEW: EFFECT OF THE TIMING OF LUMBAR DISCECTOMY ON POSTOPERATIVE OUTCOMES 3.1 Clinical Uncertainty and the Optimum Timing Of Surgery There is a need for research to identify modifiable risk factors that can be unambiguously defined and ascertained in patients being considered for elective lumbar discectomy. This could reduce a major source of uncertainty in surgical decision making, and help to refine current guidelines for managing lumbar disc herniation. Three separate systematic reviews of studies on the efficacy of lumbar discectomy have concluded that there is a specific need for better evidence on the optimal selection and timing of surgical treatment.1"3 3.2 Mechanisms of Chronic Pain The timing of surgery is a potentially important predictor of outcome after lumbar discectomy. Prolonged conservative therapy averts the potential complications from an operation, yet a longer duration of symptoms may be associated with poorer outcomes in those eventually opting for surgery.2 Pain reduction, and reduction of disability due to pain, are the primary reasons why patients with sciatica from lumbar disc herniation agree to operative treatment. Pain is a complex experience involving both sensory neuroanatomical and neurophysiological mechanisms. Together, these result in nociception (i.e. perception of injurious stimuli by the nervous system) and emotional, mood, and cognitive states, which influence the intensity of suffering experienced by the individual. These pain mechanisms are relevant to the 29 timing of lumbar discectomy because some authors believe that earlier excision of inflammatory disc material reduces the likelihood of chronic, or repetitive inflammation and consequent damage to (i.e. nociception from) the nerve root environment.4 Nystrom showed that early connective tissue reaction around the nerve root is seen as early as one month after the first symptoms of disc herniation.5 Subsequent studies have also supported the hypothesis that herniated disc tissue provokes an aggressive inflammatory response and consequent irritation of the adjacent nerve roots/-8 In contrast, as nerve root inflammation is a ubiquitous finding at surgery, the impact of nerve root inflammation becomes difficult to tease out. Consequendy, other investigators cite an additional hypothesis that prolonged stimulation of primary afferent (i.e. sensory) nerve fibres leads to long-term plastic changes in the nervous system, particularly within neurons that are responsible for processing nociceptive information within the spinal cord. These plastic changes are thought to result in a form of memory within the dorsal horn cells and, consequendy, entrenchment of chronic pain.9 ,10 3.3 L i t e r a t u r e R e v i e w : T i m i n g a n d O u t c o m e o f L u m b a r D i s c e c t o m y Through the mechanisms explained above, it is conceivable that delayed lumbar discectomy results in prolonged nociception at the level of the affected nerve root(s), which in turn begets a chronic pain experience and, ultimately, chronic pain (behaviour) in patients with sciatica due to lumbar disc herniation. If that is the case, it seems reasonable that earlier treatment reduces the duration of nociception and thereby forestalls the entrenchment of chronic pain. A consistent finding in the literature is that patients with chronic, longstanding low back pain of greater than three months in duration exhibit poorer outcomes after conservative therapy than patients with acute symptoms of less than three months in duration.11 Similarly, poorer outcomes have been 30 reported for patients with a longer duration of sciatica symptoms among both non-surgically,12 and surgically-treated4*13 groups. The following section provides a review of the literature on the timing of surgery and its potential effect on outcomes after lumbar discectomy. 3.3.1 Preoperative Symptom Duration and Outcome Preoperative symptom duration necessarily depends upon the date of onset (the "inception point") of symptoms which is typically determined based on recall as well as an operational definition of a discrete episode of back pain or other symptoms. In addition to the choice of the inception point, an estimate of preoperative symptom duration also depends on the choice of an endpoint. Studies of the effect of symptom duration on outcome are not always clear in their descriptions of chosen inception points and endpoints. For example, in some studies it is not clear whether the duration of current symptoms refers to the current duration of leg pain, back pain, or both, or if symptom duration refers to the most recent attack of back only, or a series of attacks over a determinable period of time. A search of EMBASE, M E D L I N E , BIOSIS, and C I N A H L was performed for all years covered by each of these databases. The search strategy is described in Appendix A , and the methodological features and results of these studies are summarized in Tables 3.1 and 3.2, respectively. Only studies that included an internal comparison group were included. Almost all studies were described as prospective cohort designs and only one was retrospective. Most study populations involved a consecutive series of patients undergoing lumbar discectomy. There was wide variability in the specific outcomes and the duration of follow-up that was used. Few studies used outcome measures that had been psychometrically validated.11 Most relied on crude patient-reported measures of improvement. The criteria for ascertaining a successful outcome also varied considerably between studies. Symptom duration was defined 31 differently between studies, and sometimes not at all. Some authors used the estimated duration of symptoms at study intake while others estimated the duration of symptoms at the time of surgery. With the exception of one study,4 authors did not provide information about the process of care or flow of patients between the time of the first encounter with a surgeon and the actual date of surgery. Consequently, it was not possible to determine from any study whether or not patients were actually placed on a wait list. Some authors simply reported "duration of current symptoms" or "duration of preoperative symptoms" without clarification as to when, or how, these measures were ascertained. Finally, symptom duration, however conceived, was usually analysed as a categorical variable without explanation of how or why specific cut-points were chosen. 3.3.1.1 Symptom Duration as a Primary Independent Variable Only three studies examined symptom duration as a variable of primary interest.4;9;13 In Nygaard's first study,4 the proportion of subjects reporting "good" or "fair" outcomes was greatest in those with a duration of sciatica prior to surgery less than six months, compared to those with a duration of six to 12, or greater than 12 months. This was a retrospective cohort study in which bivariate comparisons were statistically significant, but no adjustment for potential confounders was performed. In a second study involving a prospective cohort design, Nygaard showed that after controlling for age, sex, and duration of back pain and sick leave, longer duration of leg pain at the time of surgery was associated with worse outcomes on a composite index of pain, physical findings, functional status, and medication use.9 The authors claimed they had identified an eight-month window for surgery, even though their analysis did not control for prognostic variables between groups. 32 A third study examined the effect of symptom duration as a primary independent variable of interest, however the methods were not described in detail.'3 A mean follow-up time of 9.9 months was reported in the absence of any measure of dispersion. The follow-up period was different for each patient. Outcome was assessed on a Prolo scale,14 a composite measure of pain and work status (2 = incapacitation, 10 = perfect function). This scale appears to correlate with self-reported improvement in a number of surgical case series;13 however its other psychometric properties are not documented. Nonetheless, the mean Prolo score for patients with symptom duration at surgery greater than 60 days (7.0) was worse than that of patients with a duration of either 30 to 60 days (7.4), or less than 30 days (7.9). The results were not adjusted for other variables. As sciatica patients generally improve over time, the length of follow-up between subjects in each of the groups could have confounded the association of interest. Ignoring that issue may have led to biased results. 3.3.1.2 Symptom Duration as a Covariate More often than not, symptom duration was examined as one of many factors in the search for a prognostic model for surgical patients with lumbar disc herniation. Junge and colleagues15 used discriminant analysis to show that a "bad" outcome on a modified Stauffer-Coventry (surgical outcome) scale16 was associated with a longer duration of acute back pain at admission, however, a recommended time frame for surgery was not determined. The Maine Lumbar Spine Study17 originally involved 275 surgically treated, and 232 non-surgically treated, patients. At one year of follow-up from the time of study intake, symptoms of less than six months at intake was associated with good outcomes at a bivariate level, however it was not clear if the effect was maintained after controlling for other confounders.11 Neither a description of waiting times after study enrolment were reported, nor was the effect of symptom duration on outcomes reported separately by treatment group (surgical versus non-surgical). 33 In a related paper, however, Adas et al examined whether the Quebec Task Force criteria11 for classifying patients with activity-related spinal disorders could discriminate between patients with different prognoses. The outcomes for non-surgical patients ended up being associated with baseline duration of symptoms, but not for surgically treated patients. This result was adjusted for baseline quality of life scores, but not for other potential confounders. Subsequent publications based on the Maine Lumbar Spine Study cohort reported that symptoms for more than six months at study entry independendy increased the odds of a patient receiving disability compensation by three-fold at four years,18 and that "shorter duration of symptoms" (not defined in the paper) was independendy associated with improvement in the pain symptom at five years' follow-up.12 Again, this latter finding referred to both surgical and non-surgical patients combined, in which case, the effect of symptom duration among surgical patients alone was not reported. In another study, two different ordinal scales (one measuring function during activities of daily living, the other measuring pain and working capacity) were each dichotomized and used as outcome measures for a predictive model.19 So-called "best outcomes" were independendy predicted by a current duration of sciatica of less than two months, and preoperative sick leave of less than one month. The mean duration of symptoms for subjects overall was three years from the onset of the first episode of back pain, one year from the onset of the first episode of sciatica, and three months for the current episode of back pain. At six months postoperatively, 92% overall reported being either "better" or "much better" on a four-item self-rating scale. Furthermore, in contrast to most of the other studies, patients with severe symptoms of disc herniation including cauda equina syndrome and acute massive paresis, were not excluded. Six percent of subjects received emergency surgery, but since their data were not presented separately it is difficult to know if the early window for surgery in the study is attributable solely to the inclusion of emergency cases. 34 Three other studies discussed the effect of symptom duradon in the context of multivariate prognostic models.20'22 In one study, mean duration of motor deficit "before surgery" was associated with Low Back Outcome Scores23 at one year, but, again, no optimum window for surgery was defined.20 In another study, duration of sciatica "before surgery" was dichotomized at a median value of seven months.21 This cut-off apparendy predicted return-to-work, but not postoperative symptom relief at two years. The effect of symptom duration at other cut-off values was not explored. Furthermore, symptom relief was measured on a crude four-item scale, which was then collapsed into a dichotomous, and possibly less responsive, outcome measure. Two other studies failed to detect an association between symptom duration and outcome after lumbar discectomy.24'25 In the first of these studies, patients with multiple-level herniation and severe neurologic deficits (including cauda equina syndrome) were included. Presumably, some of these patients had surgery on an emergency basis, however no description of the number of such patients was provided. Symptom duration was determined at the time of a medical history, but it was not clear how close the time of assessment was to the time of surgery. In the second study, the primary objective was to determine the ability of psychological measures to predict the outcome of lumbar discectomy, however the effects of demographic, clinical history and physical assessment variables were also examined.25 The study population and methods were well-described, however the sample size was only 46 patients. Duration of symptoms at preoperative assessment predicted neither postoperative pain relief, improvement of disability in daily activities, return to work, nor surgical outcome (as assessed by Stauffer and Conventry's criteria16). However, the coefficients from multiple regression analyses did not include statistically nonsignificant variables. Given the small sample size, clinically significant associations between symptom duration and each outcome might have existed despite the lack of statistically significant 35 associations, however the reader has no opportunity to assess this. Reporting could have been improved using measures of effect and their 95% confidence intervals. Finally, in a systematic review of the efficacy of lumbar discectomy and percutaneous treatments for lumbar disc herniation1 the randomized trial by Weber26 was cited as having shown a worse outcome for subjects with greater than three months of symptoms. However, no such statement appeared in the original article. Rather, on the contrary, the article clearly stated that, "Variables without prognostic importance were... earlier attacks of sciatica (and) duration of radiating symptoms...". 3.4 Summary: Effect of Symptom Duration as a Measure of the Timing of Surgery In summary, the evidence regarding the association between the timing of lumbar discectomy in terms of symptom duration must be interpreted cautiously. Most studies report a significant association between symptom duration and postoperative outcome. In most positive studies, symptom duration longer than six months predicted a worse outcome after lumbar discectomy. Some studies detected an association despite the use of coarse outcome measures. In at least one example, a significant association was detected despite a small sample size. On the other hand, many of the studies were not described well enough for readers to understand how symptom duration was actually calculated, or how long patients had to wait for surgery if symptom duration was not estimated in relation to the date of surgery. Al l of the reviewed studies also had clear methodological limitations, some of which include a lack of acceptable comparison group, adjustment for potential confounders, or other limitations in reporting (i.e. omission of point estimates and 95% confidence intervals for statistically nonsignificant, but potentially clinically significant effects). 36 Table 3.1 Reviewed Studies on Symptom Duration Characteristics of studies reporting effect of symptom duration on outcome of surgical lumbar discectomy. S t u d y S t a t e d d e s i g n S t u d y p o p u l a t i o n O u t c o m e s F o l l o w - u p Hurme 1987 Prospective 215 of 357 consecutive surgical (122 (Turku, conservative) pts; university hospital Finland) setting; admitted for 1s t surgery for "suspected" lumbar disc herniation (LDH); 113 male, 102 female; mean (range) age: 39.3 (16-54) Lewis 1987 (Edmonton, Canada) Nygaard 1994 (Tromso, Norway) Atlas 1996" MLSS-I* (Maine, USA) *Maine Lumbar Spine Study: 1 year followup Prospective 100 consecutive patients, 1st lumbar discectomy for back and/or leg pain, all confirmed on myelography; 75 males, 25 females; 28% receiving workers' compensation Retrospective 93 consecutive pts; 1st operation for "unambiguous" L D H (1988-1990); 57 males, 36 females; Mean age (range; s.d.): 41 (19-66; 10.2) Prospective 507 patients with sciatica recruited from occupational medicine clinics throughout the state of Maine; 275 patients treated surgically, 232 patients treated conservatively; Females 34.6% of surgical group. Mean age (range) in surgical group 42.4 (18-85) years. 6 months postoperatively Patient rating of -postoperative status -change in pain -activities of daily living (ADL) -working capacity -pain & working capacity (PWC) scale Subjectively reported pain relief on 1,5 and 10 4-item scale (complete relief, years partial relief, the same, or worse) Patient rating (good or fair= satisfactory, poor or worse=unsatisfactory) 1 and 3 months postoperatively Improvement in most 1 year predominant symptom (leg or back pain at study entry) on a patient-rated 7-point scale; Back-specific functional status using Roland back disability scale; Generic health status using SF-36 —] S t u d y S t a t e d d e s i g n S t u d y p o p u l a t i o n O u t c o m e s F o l l o w - u p Adas 199627 QTFSf (Maine, USA) -j-Quebec Task Force (Criteria) Study Junge1996 (Hamberg, Germany & Zurich Switzerland) Graver 199822 (Oslo, Norway) Woertgen 199828 (Regensburg, Germany) Prospecdve Prospecdve Prospective Prospective 516 participants from the Maine Lumbar Spine Study, with sciatica and lumbar spinal stenosis treated surgically and nonsurgically. Frequency & "bothersomeness" of symptom scores; Mean change scores from baseline to 1 year for (1) Modified Roland Disability scale and (2) SF-36 589 patients—400 of whom had a "major Good, moderate or bad outcome diagnosis" was disc herniation for which on Stauffer-Coventry Scale they underwent standard discectomy— 381 of whom did not require reoperation & are the basis for this analysis; Males 60.1%; Mean age 44.7(11.2) 122 consecutive pts; 1st operation for "unambiguous" L D H ; 66 males, 56 females; Mean (range; s.d.) age 40.8 (18-66; 10.9) years 121 consecutive patients operated on for 1s t time for lumbar disc herniation; sciatica in 96%; Males 72%; Median age (range): 44.4 (15-76) years 1 year 6 & 12 months postoperatively 1 year Physiotherapist & patient rating of "clinical overall score" (COS [summarizes pain, signs, Oswestry score, use of analgesics]); Return to work Low Back Outcome Score 1 year (LBOS): favourable outcome being > 49 versus < 48 points; Prolo Score: favourable outcome being 8-10 points oo S t u d y S t a t e d d e s i g n S t u d y p o p u l a t i o n O u t c o m e s F o l l o w - u p Schade 199925 (Berne, Switzerland) Prospective 46 consecutive patients, 1s t surgery for confirmed L D H ; Mean age (range): 35.2 (20-50) years Pain relief: 100 mm visual analogue scale; Improved disability: Roland-Morris disability scale; Return to work; Surgical outcome: Stauffer-Coventry criteria. 2 years Vucetic 199921 (Huddinge, Sweden) Woertgen I99920 (Regensburg, Germany) Atlas 200029 LTD:): (Maine, USA) ^Long-Term Disability and Return to Work Study Prospective 160 consecutive pts; 1st surgery for "suspected" L D H , undergoing primary surgery for relief of sciatica; 85 males, 75 females; Mean age: 43 (s.d. 10); median age: 41 Prospective 98 out of 121 consecutive pts operated on for neurologic deficit or radicular pain "resistant to conservative therapy"; 70% of subjects were male; Median (range) age: 43 (15-70) years Prospective 326 patients with sciatica sampled from the Maine Lumbar Spine Study cohort; 199 were receiving compensation Patient rating of relief of sciatica (incapacitating [incapacitating or severe] versus nonincapacitating [moderate or no pain]); Return to work Low Back Outcome Score 2 years All outcomes stratified according to Workers' Compensation status (yes, no) at study entry 2.5 years 4 years S t u d y S t a t e d d e s i g n S t u d y p o p u l a t i o n O u t c o m e s F o l l o w - u p Nygaard 20009 Prospective 132 surgical patients with radicular pain Clinical overall score (COS): 1 year (Trondheim, and radiographically confirmed lumbar composite measure of pain Norway) disc herniation; intensity (100 mm VAS), spinal deviation, straight leg raising, Oswestry Disability Index score, and type and dose of analgesic use Adas 2001 ( 2 Prospective 402 of 507 initially enrolled patients from Frequency & "bothersomeness" of 5 years MLSS-II** the Maine Lumbar Spine Study; 220 symptom scores (Maine, USA) treated surgically, 182 treated nonsurgically Mean change scores from baseline **Maine to 1 year for modified Roland back Lumbar Study: disability scale and SF-36 5 year follow-up Rothoerl 200213 Prospective 219 consecutive pts with main diagnosis Patient rating of work status and Mean follow-up: (Regensburg, of L D H (stenosis patients excluded), symptoms & function (Prolo scale: 298 days (9.9 Germany) 93% (205) of whom were followed-up; 8-10=favourable, 2- months) 74% (162) males, 26% (57) females; mean 7=unfavourable)) age 44.5 Ng 200430 Prospective 113 consecutive patients, "radiculopathy Oswestry Disability Index (ODI); 1 year (Leicester, due to lumbar disc herniation", Low Back Outcome Score England) radiologically confirmed; (LBOS); 45 males, 58 females; Sciatica intensity assessed by visual Mean age (range): 37.8 (12-66) analogue scale (VAS) o Table 3.2 Results of Symptom Duration Studies Results of studies reporting the effect of symptom duration on outcome of surgical lumbar discectomy S t u d y G e n e r a l r e s u l t s R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Hurme 198719 (Turku, Finland) Lewis 1987 (Edmonton, Canada) Overall success rate at 6 months: % of patients better or much better: 92%; % of patients able to work:: 85%; Overall mean pre-op duration: -since first back pain: 3 years; -since first sciatica: 12 months; -current attack of back pain: 3 months; -average pre-op sick leave: 54 days (0-275); Mean postoperative sick leave: 78 days (10% of patients needed an average of 45 days of additional sick leave) Preoperative factors associated with complete back (and leg*) pain relief at 5 years: -no previous hospitalization* -negative straight leg raising test -restricted forward bending* -absence of osteophytes on x-ray -atypical disc protrusion at surgery* -duration of surgery < mean (35 minutes)* Predictors of poor result (low score) on PWC index: -high somatization scale (psychological Symptom Check List [SCL]) score -pre-op duration of sciatica greater than 6 months -divorced or widowed status -perception of own work being physically strenuous "Best result" achieved when: -pre-op duration of sciatica is less than 2 months' duration -pre-operative sick leave is less than 1 month Duration of leg pain < sample mean (16.1 months) predicted complete relief of leg pain, but not back pain, at 1 and 5-10 years' follow-up Inclusion of patients w/CES. 6% of surgeries were on emergency basis. Stepwise logistic regression used. No description of how cut-points were chosen for dichotomizing continuous outcomes (such as PWC index) or independent variables (such as pre-op duration of sciatica). S t u d y G e n e r a l r e s u l t s R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Nygaard 19944 Overall results at 3 years: Mean preoperative duration of sciatica by No stratified analyses (Tromso, -% reporting a satisfactory result = outcome: or multivariate Norway) 89% -satisfactory result = 33 weeks adjustment for -% reporting an unsatisfactory result = -unsatisfactory = 56 weeks confounders or 11% prognostic variables % of patients with good outcome by pre-Overall mean preoperative duration (range; operative duration of current sciatica: No ROC analysis to s.d.): < 6 months = 56% detect best cut-off for -of current LBP = 43 wks (3-146; 6-12 months = 32% distinguishing 45.6) > 12 months =16% satisfactory from (i.e. 10 months) unsatisfactory -of current sciatica = 36 weeks (3-131; Therefore, as duration of preoperative 31.3) symptoms increase, outcomes are less Small number of -of present sick leave prior to surgery favourable subjects in = 26 weeks (0-104; 23.7) unsatisfactory group Adas 1996" % of patients who improved at 1 year: At univariate level, duration of current No report of whether MLSS-I* -Surgically treated = 71% symptoms of less than 6 months at study symptom duration (Maine, USA) -Nonsurgically treated =43% entry was associated with greater was independendy Larger % of surgically treated patients were improvement. associated with *Maine Lumbar better compared to nonsurgical patients, improvement in final Spine Study: 1 both functionally and symptomatically. logistic regression year followup models. Relaive odds of definite improvement with surgery went from 3.2 to 4.3 after adjustment for baseline variables Little difference in employment and WCB status between surgical and nonsurgical groups S t u d y G e n e r a l r e s u l t s R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Adas 1996" QTFSf (Maine, USA) -[Quebec Task Force (Criteria) Study Longer duration of symptoms at study entry were associated with less improvement in symptoms and functional status for nonsurgical patients only. % of nonsurgical patients who were asymptomatic (i.e. predominant pain symptom at study entry [either leg or back pain] either much better or completely gone) at 1 year, by duration of symptom: -Sx duration < 6 wks = 56% asymptomatic at 1 year -Sx duration 6 wks - 3 months= 50% -Sx duration > 3 months = 31% % of surgical patients who were asymptomatic at 1 year, by baseline work status: —working at baseline = 80% asymptomatic at 1 year —idle/not at baseline = 55% asymptomatic at 1 year Therefore, patients who were working at baseline were more likely to be asymptomatic at 1 year. Duration of symptoms at baseline were not associated with symptomatic or H R Q O L outcome for surgical patients at 1 year. Study population included patients with spinal stenosis A N C O V A used to adjust for baseline differences in H R Q O L scores only. Results were not adjusted for other potential confounders or prognostic variables. Maine Study apparendy had low participation rate. Duration of symptoms was not associated w/baseline function or symptom severity, possibly also because of very few patients in the sample with short duration symptoms 4^ S t u d y G e n e r a l r e s u l t s Junge 199631 Overall results at 12 months: (Hamberg, -% classified as good result Germany & 51.5% Zurich -% classified as moderate result Switzerland) 28.4% -% classified as poor result 20.1% 4^ R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Duration of symptoms (s.d.) by outcomes at 12 months: -Acute back pain: —Good outcome = 6.64 weeks (6.64) —Moderate outcome = 9.27 weeks (11.7) —Bad outcome = 20.8 weeks (40.2) (p=0.006) -Acute leg pain: —Good outcome = 12.5 weeks (28.3) —Moderate outcome = 11.4 weeks (12.0) —Bad outcome = 23.4 weeks (45.4) (chi square n.s.) Duration of working disability by good, moderate, or bad outcome on Stauffer-Coventry Scale: -Medically attested working disability (s.d.) —Good outcome = 4.85 wks (8.3) —Moderate outcome= 7.57 wks (12.2) -Bad outcome =7.86 wks (8.23) -Patient reported reduced working ability (s.d.) —Good outcome = 12.6 wks (19.4) —Moderate outcome= 14. wks (16.5) —Bad outcome = 25.7 wks (32.0) No definition of acute versus chronic-persistent and chronic-recurrent symptoms Canonical discriminant analysis performed. Duration of acute low back pain and duration of reduced working ability were important predictors of outcome, but not as strong as physical mobility, intensity of acute pain, and number of other pain locations in the body before surgery. Therefore, longer duration back pain and working disability associated with bad outcomes at 6 & 12 months S t u d y G e n e r a l r e s u l t s R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Graver 199822 (Oslo, Norway) Woertgen 199828 (Regensburg, Germany) Mean preoperative duration: -of present attack of sciatica was 8.8 months -of present sickness absence was 5.5 months Return to work (RTW): -73% rate of RTW overall -RTW was negatively associated with females, longer sickness leave, and physically strenuous work 55% of patients reported feeling absolutely painfree; 27% still had back pain, 18% still had sciatica, and 14% had unequivocal radicular radiation. 4% needed reoperation (at 1 year) Duration of symptoms was not associated with outcome (Clinical Overall Scores) in univariate or multivariate regression analysis. Outcome Measure Good Outcome Bad Outcome Mean duration of pain Prolo LBOS 77 days 84 days 165 days 159 days Mean duration of paresis Prolo LBOS 56 days 60 days 120 days 123 days All differences in means between good and bad groups statistically significant Therefore, duration of symptoms were Study population was atypical given that it included patients w / cauda equina syndrome (CES) and other severe neurologic deficits. No description of what proportion of patients made up this group of severe cases. Results analysed by A N O V A and various tests for multiple post-hoc pairwise comparisons. Results not adjusted for possible confounders. associated with outcomes on the Prolo scale and LBOS S t u d y G e n e r a l r e s u l t s Schade 199925 (Berne, Switzerland) Pain relief, predicted by: -MRI-identified nerve root compromise -social support at home Improved disability, predicted by: -MRI-identified nerve root compression -MRI-identified disc extrusion Return to work, predicted by: -depression -occupational mental stress Surgical outcome, predicted by: -MRI-identified extent of herniation -depression R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Duration of "current spell" of symptoms (not associated with any of the four outcomes (pain relief, disability improvement, return to work, or surgical outcome score) Small sample size (46) may have precluded detection of clinically significant effect of symptom duration. Not clear if symptom duration (in months) was analysed as continuous or categorical variable Study General results Results regarding symptom duration Comments Vucetic 199921 Relief of sciatica predicted by: ruptured (Huddinge, annulus on exploration, male gender, no Sweden) preoperative comorbidity. Median duration of leg pain 7 months. Duration of sciatica < 7 months predicted return to work: (OR 3.8 [1.6, 9.2]) at 2 years Duration of sciatica did not predict relief of leg pain or relief of back pain at 2 years. Categories from a highly qualitative outcome measure may have been inaapropriately collapsed into 2 groups: -incapacitating group=incapacitating & severe pain groups -nonincapacitating group=moderate & no pain groups Not clear if duration of leg pain refers to duration at baseline assessment or duration immediately pre-operatively 4^ S t u d y G e n e r a l r e s u l t s R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Woertgen ! 99920 (Regensburg, Germany) Atlas 200029 L T D t (Maine, USA) ^Long-Term Disability and Return to Work Study Favourable outcome in 66% of pts. Majority of clinical improvement occurs w/in 3 mos of Rx At 4 years, unadjusted relative odds (95% CI) of receiving disability benefits was 5.0 (2.6, 10.3) for padents receiving compensation at study entry. When adjusted for possible confounders, relative odds of receiving disability decreased to 3.5 (1.76, 7.6) In multivariate regression analysis, mean duration of paresis (at time of admission) was independendy associated with outcome (LBOS score) at 2.5 years. 23 patients excluded for not completing all three follow-up visits. (Regression coefficient for duration of paresis *No description of = 0.0231; therefore, OR = 1.02, which describes the relative odds of a 1-point change in LBOS per unit change in duration of paresis)* In multivariate models, "symptoms for more than 6 months" was independendy associated with a relative odds of 3.1 (1.6, 6.1), indicating a 3-fold increase in the odds of a patient receiving disability compensation at 4 years if on compensation at study entry. how time/duration of paresis was coded for regression analysis. Separate results were not presented for surgical patients only. Effect of duration of symptoms was not investigated using different cut-offs Nygaard 20009 (Trondeim, Norway) Age, weakly associated; sex, not associated; Leg symptom duration quartile (pain less than Analysis controlled sick leave, signicandy associated with clinical 4 months, 4 to 8 months, 8 to 12 overall score (COS). months, and greater than 12 months), strongly associated (p = 0.0016); back pain duration, weakly associated (p = 0.10) with COS. for age, sex, sick leave, and time since initial onset of leg and back symptoms, but no other confounders. -1^ 0 0 S t u d y G e n e r a l r e s u l t s R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Adas 2001" MLSS-IP* (Maine, USA) **Maine Lumbar Study: 5 year follow-up Rothoerl 200213 (Regensburg, Germany) 70% of patients initially treated surgically reported improvement in their predominant symptom (back or leg pain) versus 56% of nonsurgical^ treated patients. Above differences persisted after adjustment for other determinants of outcome Overall 60% of patients had a favourable result (40% unfavourable) Mean duration "prior to surgery": -Pain: 86 days -Hypaesthesia: 69 days -Paresis 68 days Baseline factors associated with improvement in predominant pain symptom included: -younger age -not receiving compensation -shorter duration of symptoms -less severe SF-36 bodily pain score Longer duration of symptoms was associated with an unfavourable outcome. Mean duration of symptoms by outcome: -unfavourable Prolo ~ 114 days -favourable Prolo 69 days Monotonic decrease in Prolo scores as categorical symptom duration increases: Time category Prolo score < 30 days 7.9 30-60 days 7.4 > 60 days 7.0 No discussion on what defined/constituted a "shorter duration of symptoms". Pts w/concentric stenosis or receiving laminectomy of hemilaminectomy were excluded Not clear whether duration of symptoms prior to surgery is defined as time from symptom onset to baseline visit or onset to time of operation. No discussion of why 30 & 60 days were chosen as cut-offs for categorizing symptom duration Univariate analyses only; no stratification or multivariate adjustment on prognostic variables 4 ^ S t u d y G e n e r a l r e s u l t s Ng 200430 77% of patients had significant reduction > (Leicester, 20% change on O D I score England) Overall duration (range) of radiculopathy: 11 (2-60) months O R e s u l t s r e g a r d i n g s y m p t o m d u r a t i o n C o m m e n t s Significant association between continuous "duration of radiculopathy at surgery" and Oswestry disability and Low Back Outcome scores after multivariable adjustment; Analysis by categorical duration: Mean change in ODI for duration > 12 months was significantiy less than change in ODI for duration groups "< 4", "4-8" and "8-12" months. Symptom duration explicidy defined; Valid, sensitive outcomes measures; Multivariable adjustment performed; No association between preoperative sciatica duration and postoperative sciatica intensity (on VAS) 3.5 References 1. Stevens C D , Dubois RW, Larequi-Lauber T, and Vader JP. Efficacy of lumbar discectomy and percutaneous treatments for lumbar disc herniadon. Sozial-und Pravendvmedizin 1997;42:367-79. 2. Hoffman R M , Wheeler KJ, and Deyo RA. Surgery for Herniated Lumbar Disks - A Literature Synthesis. Journal of General Internal Medicine 1993;8:487-96. 3. Gibson J N A and Waddell G. Surgical interventions for lumbar disc prolapse. Cochrane Database of Systematic Reviews 2007. 4. Nygaard OP, Romner B, and Trumpy JH. Duration of Symptoms As A Predictor of Outcome After Lumbar Disc Surgery. Acta Neurochirurgica 1994;128:53-6. 5. Nystrom B. Experience of Microsurgical Compared with Conventional Technique in Lumbar-Disk Operations. Acta Neurologica Scandinavica 1987;76:129-41. 6. Olmarker K, Blomquist J, Stromberg J, Nannmark U , Thomsen P, and Rydevik B. Inflammatogenic Properties of Nucleus Pulposus. Spine 1995;20:665-9. 7. Nygaard OP, Mellgren SI, and Osterud B. The inflammatory properties of contained and noncontained lumbar disc herniation. Spine 1997;22:2484-8. 8. Woertgen C, Rothoerl RD, and Brawanski A. Influence of macrophage infiltration of herniated lumbar disc tissue on outcome after lumbar disc surgery. Spine 2000;25:871-5. 51 9. Nygaard OP, Kloster R, and Solberg T. Duration of leg pain as a predictor of outcome after surgery for lumbar disc herniation: a prospective cohort study with 1-year follow up. Journal of Neurosurgery 2000;92:131-4. 10. Coderre TJ, Katz J, Vaccarino A L , and Melzack R. Contribution of Central Neuroplasticity to Pathological Pain - Review of Clinical and Experimental-Evidence. Pain 1993;52:259-85. 11. Atlas SJ, Deyo RA, Keller RB et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine 1996;21:1777-86. 12. Atlas SJ, Keller RB, Chang Y C , Deyo RA, and Singer D E . Surgical and nonsurgical management of sciatica secondary to a lumbar disc herniation - Five-year outcomes from the Maine Lumbar Spine Study. Spine 2001;26:1179-87. 13. Rothoerl RD, Woertgen C, and Brawanski A. When should conservative treatment for lumbar disc herniation be ceased and surgery considered? Neurosurgical Review 2002;25:162-5. 14. Prolo DJ, Oklund SA, and Butcher M . Toward Uniformity in Evaluating Results of Lumbar Spine Operations - A Paradigm Applied to Posterior Lumbar Interbody Fusions. Spine 1986;11:601-6. 15. Junge A , Frohlich M , Ahrens S et al. Predictors of bad and good outcome of lumbar spine surgery - A prospective clinical study with 2 years' follow-up. Spine 1996;21:1056-64. 16. Stauffer R N and Coventry MB. Anterior Interbody Lumbar Spine Fusion - Analysis of Mayo Clinic Series. Journal of Bone and Joint Surgery-American Volume 1972;A 54:756-&. 52 17. Keller RB, Atlas SJ, Singer D E et al. The Maine Lumbar Spine Study .1. Background and concepts. Spine 1996;21:1769-76. 18. Atlas SJ and Keller RB. Effect of disability compensadon on padent outcomes - Reply. Journal of Bone and Joint Surgery-American Volume 2000;82A:1360. 19. Hurme M and Alaranta H . Factors Predicting the Result of Surgery for Lumbar Intervertebral-Disk Herniation. Spine 1987;12:933-8. 20. Woertgen C, Rothoerl RD, Breme K , Altmeppen J, Holzschuh M , and Brawanski A . Variability of outcome after lumbar disc surgery. Spine 1999;24:807-11. 21. Vucetic N , Astrand P, Guntner P, and Svensson O. Diagnosis and prognosis in lumbar disc herniadon. Clinical Orthopaedics and Related Research 1999;116-22. 22. Graver V, Ljunggren A E , Loeb M , Haaland A K , Lie H , and Magnaes B. Background variables (medical history, anthropometric and biological factors) in relation to the outcome of lumbar disc surgery. Scandinavian Journal of Rehabilitation Medicine 1998;30:221-5. 23. Greenough C G and Fraser RD. Assessment of Outcome in Patients with Low-Back-Pain. Spine 1992;17:36-41. 24. Graver V , Ljunggren A E , Malt UF et al. Can Psychological Traits Predict the Outcome of Lumbar Disc Surgery When Anamnestic and Physiological Risk-Factors Are Controlled for - Results of A Prospective Cohort Study. Journal of Psychosomatic Research 1995;39:465-76. 53 25. Schade V , Semmer N , Main CJ, Hora J, and Boos N . The impact of clinical, morphological, psychosocial and work-related factors on the outcome of lumbar discectomy. Pain 1999;80:239-49. 26. Weber H . Lumbar-Disk Herniation - A Controlled, Prospective-Study with 10 Years of Observation. Spine 1983;8:131-40. 27. Atlas SJ, Deyo RA, Patrick D L , Convery K , Keller RB, and Singer D E . The Quebec Task Force classification for spinal disorders and the severity, treatment, and outcomes of sciatica and lumbar spinal stenosis. Spine 1996;21:2885-92. 28. Woertgen C, Rothoerl RD, Holzschuh M , Breme K , and Brawanski A . Are prognostic factors still what they are expected to be after long-term follow-up? Journal of Spinal Disorders 1998;11:395-9. 29. Atlas SJ, Chang Y C , Kammann E, Keller RB, Deyo RA, and Singer D E . Long-term disability and return to work among patients who have a herniated lumbar disc: The effect of disability compensation. Journal of Bone and Joint Surgery-American Volume 2000;82A:4-15. 30. Ng L C L and Sell P. Predictive value of the duration of sciatica for lumbar discectomy - A prospective cohort study. Journal of Bone and Joint Surgery-British Volume 2004;86B:546-9. 31. Junge A , Frohlich M , Ahrens S et al. Predictors of bad and good outcome of lumbar spine surgery - A prospective clinical study with 2 years' follow-up. Spine 1996;21:1056-64. 54 CHAPTER 4 4.0 LITERATURE REVIEW: EFFECT OF DELAYED TREATMENT ON BACK PAIN OUTCOMES 4.1 Waiting Time: An Indicator of Access to Care In the Canadian healthcare system, access to many diagnostic and surgical procedures is managed through waitlists.1 A waitlist is can be used as a roster to which patients are assigned as they await a particular health service.2 Waiting time refers to the length of time required for a patient on the list to receive a desired service. Waiting time can reflect delays in service delivery at different stages of a patient's encounter with the healthcare system. Patients with non-emergency symptoms and signs due to lumbar disc herniation typically await the following care processes: to be assessed by a orthopedic/spinal surgeon upon referral from a family physician; for advanced diagnostic imaging with computerized tomography or magnetic resonance imaging; and, for lumbar discectomy after the decision to proceed with surgery is agreed. The duration of waiting during any of these stages has many determinants. Preoperative symptom duration subsumes the duration of waiting at other stages of service delivery that are sensitive to patients' and physician's subjective preferences, which in turn influence how long patients remain on conservative therapy, how quickly they are referred to a specialist, and how quickly they receive, and accept, an offer for surgical intervention. Preoperative symptom duration also subsumes the enormity of psychological, social and environmental influences that determine how long different individuals wait before pursuing medical treatment and subsequendy utilizing the healthcare system for their current predicament.314 55 Furthermore, back pain is typically a recurring condition with new, discrete episodes that may be indistinguishable from flare-ups related to pre-existing episodes.5 Consequently, misclassification of patients according to preoperative symptom duration may be considerable,6 and the lack of explicit, let alone standardized, definitions of a discrete episode of back pain in the literature further hinder attempts to compare different studies on symptom duration.7 Finally, the inception point for symptom duration is typically dependent upon the recall ability of patients, the reliability of which has been shown to be acceptable in some studies,8"10 but not in others.6;11 Although preoperative symptom duration may still have some utility as a predictor of outcome of lumbar discectomy despite these limitations, waiting time subsumes fewer subjective determinants and is potentially more directly modifiable by changes in the capacity of the health care system. The waiting time for the surgery is an objectively measurable outcome and a meaningful performance measure for the system.2 It is also the most readily calculated with data collection systems in place. 4.2 Literature Review: Waiting Time and Outcome of Lumbar Discectomy A search of E M B A S E and M E D L I N E for all years covered by these databases was performed using the terms and primary search strategy described in Appendix B. Abstracts from all identified articles were scanned, and full-text electronic or paper versions were acquired for any potentially relevant articles. References that were cited in each full-text article were also scanned for additional, relevant articles not captured by the primary search strategy. The Web of Science database functions (i.e. "Cited References" and "Times Cited" links) were also used in an iterative fashion to search for additional articles and/or references that were related to the most relevant papers identified by the primary search strategy. 56 Fifty potentially relevant articles were identified. That there were very few studies even potentially devoted to the effect of waiting time on surgically treated back pain was not surprising given the similar experience of other reviewers who have examined the effect of waiting on outcomes for other conditions including chronic pain.12 While reviewing these abstracts, articles that were focused on conditions other than back-related conditions were excluded. Through "cited reference" searches in the Web of Science database, a limited number of studies on the effect of waiting time on the outcome of hip replacement surgery were found.13'13'17 Some consideration was given to studies of the effect of waiting time on the outcome of orthopaedic hip surgery after reasoning that they might be relevant to the issue of waiting and the outcome of orthopaedic spinal procedures. However, the natural histories of surgically treatable hip disease and surgically-treatable back pain are very different. Surgically treatable hip disease tends to worsen steadily over time, so that even when progression is slow, the potential for spontaneous recovery is virtually non-existent. For surgically treatable back pain on the other hand, and sciatica due to lumbar disc herniation in particular, 75% of patients are said to improve within three months of the original onset.18 Full-text articles were reviewed for 12 remaining abstracts that appeared to be studies of delayed treatment for low back pain interventions; closer review showed that none of these articles direcdy addressed the issue of the association between waiting time and outcome of lumbar discectomy. Three of these full-text articles referred to waiting times only conceptually, and were excluded as they did not report relevant data. Of the remaining nine full-text articles, eight were randomized controlled trials of various non-surgical interventions (mostly physiotherapy or psychological therapies), and mosdy for chronic nonspecific low back pain.19"26 These are summarized because the putative mechanisms of the effect of physiotherapy and psychological therapies for nonspecific low back pain are not comparable to that of lumbar discectomy for disc 57 herniation. Also these studies have not been designed to examine the effect of early versus delayed treatment of low back pain specifically, but are concerned primarily with evaluating the efficacy of experimental treatment against a no-treatment (i.e. waitlist) control group. Additionally, chronic nonspecific low back pain (i.e. back pain without a readily identifiable anatomical cause) is relatively stable condition that is less likely to remit spontaneously.27 Therefore comparisons between surgically treatable lumbar disc patients—consisting of both acute and chronic patients— on the one hand, to chronic nonspecific back pain patients must be done cautiously. However, because nonspecific back pain is not universally progressive it is more akin to lumbar disc herniation than degenerative hip disease. Table 4.1 shows that seven of the eight studies involved patients with chronic back pain (typically meaning the presence of continuous pain for more than three months). The experimental treatments in four studies involved one or more forms of behavioural therapy. In other studies, the experimental treatments were early physical therapy, physical therapy and behavioural therapy combined, back school, and acupuncture. The duration of time on the waitlist in these studies was anywhere between four and ten weeks. All reported a benefit in favour of experimental groups over waitlist control groups immediately after the initial experimental treatment period. The studies by Norderman (2006), Newton-John (1995) and Vlayen (1995) are noteworthy however because they followed up patients well past the initial treatment/waitlist control period and therefore report comparisons between experimental patients who received active treatment early, and control patients who received active treatment in a delayed manner. In all three of these studies with longer-term follow-up, control patients who eventually received delayed active treatment faired worse than experimental patients who received active treatment early. 58 Table 4.2 summarizes the results of two remaining studies that were found in the scientific literature that specifically examined the association between waiting time for surgery and the outcome of orthopaedic surgery. Both studies included patients who underwent surgical discectomy. The primary objective of the Norwegian study by Rossvoll28 was to document the extent to which patients were incapacitated for work during the period before surgery as well as for the year afterward. Patients in the study were waiting for a variety of procedures, including total hip arthroplasty, total knee arthroplasty, arthroscopic surgery of the knee, ligament reconstruction of the knee, ankle and foot surgery, shoulder surgery, surgery of the hand, forearm and elbow, and removal of osteosynthesis material. The effect of waiting time on outcome for individual procedures was not reported. Nonetheless, elective orthopaedic surgery patients in general had significandy higher odds of failing to work in the first year after surgery if they had been on the waitlist for more than nine to 12 months (OR: 6.2 [1.9-20.0]), or more than 12 months (OR: 9.2 [95% CI: 3.2-26.5]).28 * In the Ontario study by Quan,29 the effect of waiting time for five different high volume procedures, including lumbar discectomy, showed no association between longer waiting and increased utilization of health care services before and after surgery.29 In that study, clinical outcome was determined using health care utilization as a proxy measure. The use of administrative data also prevented the investigators from adjusting for preoperative disease severity and other prognostic variables at an individual level. 4 . 3 Summary Orthopaedic surgical procedures are undertaken for conditions that, Eke lumbar discectomy, are often very disabling and rarely life-threatening. Delayed access to surgery for orthopaedic conditions prolongs pain and disability at the risk of perpetuating secondary social and 59 economic hardships.15 The current literature on the potential effect of waiting time for some orthopaedic procedures, such as hip surgery, suggests that delayed treatment results in a worse outcome.17 Limited studies on the association between delayed nonsurgical care and the outcome for chronic nonspecific low back pain provide similar results. To our knowledge, no studies have been published within the indexed literature specifically about the impact of waiting time on symptomatic improvement specifically after surgical lumbar discectomy. This represents an important gap in the knowledge base. Given the burden of lumbar disc herniation on individuals and society, studies addressing the question are required. 60 Table 4.1: Waitlist Control Studies on Back Pain Summarized characteristics of waitlist control studies of non-surgically treated back pain Author Study Design Patients Intervention? Control? Outcomes / Results Nordeman 2006 (Berlin, Germany) Randomized controlled (RCT) Acute low back trial pain (3-12 weeks in duration) Early physical therapy (PT) w/in 2 days "Waitlist control" but PT permitted after 4-weeks No difference at post-PT discharge; Early PT significandy better pain reduction than delayed PT at 6 months Smeets 2006 (Eindhoven, Netherlands) RCT Chronic nonspecific back pain Active PT low Cognitive behavioural therapy PT & Cognitive Waitlist control (10 weeks) Significandy greater decrease in "pain catastrophizing" for all 3 active treatment groups compared to waitlist control group at 10 weeks Brinkhaus 2006 (Berlin, Germany) RCT Chronic nonspecific pain back Acupuncture Superficial acupuncture Waitlist control weeks) (8 Significantly better pain reduction for acupuncture and minimal acupuncture than waitlist control at 8 weeks Spinhoven 2004 (Leiden, Netherlands) RCT Chronic low back pain Operant behavioural therapy plus coping skills (OBTCS) Waitlist control (10 Significandy greater reduction in weeks) "pain catastrophizing" in OBTCS group at 10 weeks Author Study Design Patients Intervention? Control? Outcomes/Results Hodselman 2001 (Haren, Netherlands) RCT Chronic Back school nonspecific low back pain (> 3 months in duration) Waitlist control (4 Significandy improve functional weeks) capacity and functional health status at 1 month Kole-Snijders 1999 (Maastricht, Netherlands) RCT (stratified by agegroup) Chronic low back pain Newton-John, 1995 (London, United Kingdom) Vlaeyen 1995 (Hoensboek, Netherlands) RCT (however some patients assigned to control group when initially unable to attend active treatment RCT Chronic low back pain Chronic low back pain Operant behavioural therapy (OBT) plus coping skills (5 weeks inpatient, plus 3 weeks outpatient treatment) Cognitive behaviour therapy (CBT) for 8 weeks Biofeedback Operant conditioning (OC) alone OC &Cognitive OC & biofeedback (1) OBT plus attention control (2) OBT as usual (3) "Waitlist control" but OBT as usual after 8 weeks "Waitlist control" but interventions (1) or (2) permitted after 4 weeks "Waitlist control" Intervention group and control group (1) significandy better reduction in "negative affect" than control groups (2) and (3) immediately posttreatment (at 8 weeks) and at 6 months and 12 months Significant improvement in pain intensity, perceived disability, "pain beliefs" and depression for intervention groups (1) and (2) compared to waitlist control, immediately posttreatment & at 6 months Al l intervention groups significantly better than waitlist control during treatment period. At 6 and 12 months, groups (2) and (3) more efficacious than OC alone in decreasing pain and increasing health behaviours T a b l e 4 . 2 : S t u d i e s o f W a i t i n g T i m e a n d B a c k S u r g e r y R e s u l t s o f s t u d i e s o n e f f e c t o f w a i t i n g t i m e o n o u t c o m e o f o r t h o p a e d i c s u r g e r y n o t e x c l u s i v e t o l u m b a r d i s c e c t o m y . A u t h o r S t u d y D e s i g n P a t i e n t s W a i t i n g t i m e O u t c o m e R e s u l t s Rossvoll 1993 (Trondheim, Norway) Prospective cohort Patients waitlisted for "chronic orthopaedic disorders"; 1,333 employed; 380 incapacitated from working; 37/380 for spinal surgery Categorical waiting time: < 1 month 1-3 months 3-6 months 6-9 months 9-12 months > 1 year (1) occurrences of sickness certification while waitlisted; (2) return to work within 1st year post-surgery for patients initially incapacitated Compared to patients waiting < 1 month, odds ratios (95% CI) of not returning to work within 1 year were: Reference 2.2 (0.9-5.4) 2.0 (0.7-5.3) 4.9 (1.3-18.1) 6.2 (1.9-20.0) 9.2 (3.3-26.5) Quan 2002 Calgary, Alberta Retrospective cohort (administrative data) 4,441 Elective orthopaedic surgical patients First elective surgical procedure for either: (1) cholecystectomy (2) discectomy (n = 352) (3) hysterectomy (4) total knee (5) total hip Continuous waiting time days) Physician claim costs (in during: (1) year before surgery; (2) year after surgery; and (3) actual waiting period No significant association between continuous waiting time (days) and dollar increases in physician claims, 4.4 References 1. Naylor CD. A Different View of Queues in Ontario. Health Affairs 1991;10:110-28. 2. Burstein J, Lee DS, and Alter DA. Do case-generic measures of queue performance for bypass surgery accurately reflect the waiting-list experiences of those most urgent? Journal of Evaluation in Clinical Practice 2006;12:87-93. 3. Lim K L , Jacobs P, and Klarenbach S. A population-based analysis of healthcare utilization of persons with back disorders - Results from the Canadian Community Health Survey 2000-2001. Spine 2006;31:212-8. 4. Cote P, Baldwin ML, and Johnson WG. Early patterns of care for occupational back pain. Spine 2005;30:581-7. 5. Von Korff M . Low back pain recollection versus concurrent accounts - Outcomes analysis -Point of view. Spine 2002;27:994. 6. Wasiak R, Pransky G, Verma S, and Webster B. Recurrence of low back pain: Definition-sensitivity analysis using administrative data. Spine 2003;28:2283-91. 7. De Vet HCW, Heymans MW, Dunn K M et al. Episodes of low back pain - A proposal for uniform definitions to be used in research. Spine 2002;27:2409-16. 8. Raphael K G and Marbach JJ. When did your pain start? Reliability of self-reported age of onset of facial pain. Clinical Journal of Pain 1997;13:352-9. 64 9. Salovey P, Smith A F , Turk DC, Jobe JB, and Willis GB. The Accuracy of Memory for Pain -Not So Bad Most of the Time. APS (American Pain Society) Journal 1993;2:184-91. 10. McGorry RW, Webster BS, Snook SH, and Hsiang SM. Accuracy of pain recall in chronic and recurrent low back pain. Journal of Occupational Rehabilitation 1999;9:169-78. 11. Bolton JE. Accuracy of recall of usual pain intensity in back pain patients. Pain 1999;83:533-9. 12. Lynch, M . E., Campbell, F, Clark, AJ et al. Towards Establishing Evidence Based Benchmarks for Acceptable Waiting Times for Treatment of Pain. <http://www.canadianpainsociet5'.ca/WaitTimes_ForPainTreatment.pdf>. Accessed 2007 08 April. 13. Williams O, Fitzpatrick R, Hajat S et al. Mortality, morbidity, and 1-year outcomes of primary elective total hip arthroplasty. Journal of Arthroplasty 2002;17:165-71. 14. Nilsdotter A K and Lohmander LS. Age and waiting time as predictors of outcome after total hip replacement for osteoarthritis. Rheumatology 2002;41:1261-7. 15. Mahon JL, Bourne RB, Rorabeck C H , Feeny D H , Stitt L, and Webster-Bogaert S. Health-related quality of life and mobility of patients awaiting elective total hip arthroplasty: a prospective study. Canadian Medical Association Journal 2002;167:1115-21. 16. Ostendorf M , Buskens E, van Stel H et al. Waiting for total hip arthroplasty - Avoidable loss in quality time and preventable deterioration. Journal of Arthroplasty 2004;19:302-9. 65 17. Garbuz DS, Xu M , Duncan CP, Masri BA, and Sobolev B. Delays worsen quality of life outcome of pri'malry total hip arthroplasty. Clinical Orthopaedics and Related Research 2006;79-84. 18. Deyo RA. Back Surgery - Who Needs It? New England Journal of Medicine 2007;356:2239-43. 19. Nordeman L, Nilsson B, Moller M , and Gunnarsson R. Early access to physical therapy treatment for subacute low back pain in primary health care: A prospecdve randomized clinical trial. Clinical Journal of Pain 2006;22:505-11. 20. Smeets RJEM, Vlaeyen JWS, Hidding A et al. Acdve rehabilitation for chronic low back pain: Cognitive-behavioral, physical, or both? First direct post-treatment results from a randomized controlled trial [ISRCTN22714229J. BMC Musculoskeletal Disorders.Vol.7, 2006Article Number: 5. Date of Publication: 20 J A N 2006. 21. Brinkhaus B, Witt C M , Jena S et al. Acupuncture in patients with chronic low back pain: A randomized controlled trial. Archives of Internal Medicine 2006:166;450-7. 22. Spinhoven P, Ter Kuile M , Kole-Snijders AMJ, Hutten M M , Den Ouden D-J, and Vlaeyen JWS. Catastrophizing and internal pain control as mediators of outcome in the multidisciplinary treatment of chronic low back pain. European Journal of Pain 2004:8;211-9. 23. Hodselmans AP, Jaegers SM, and Goeken L N . Short-term outcomes of a back school program for chronic low back pain. Archives of Physical Medicine & Rehabilitation 2001:82;1099-105. 66 24. Kole-Snijders AMJ, Vlaeyen JWS, Goossens MEJB et al. Chronic low-back pain: What does cognitive coping skills training add to operant behavioral treatment? Results of a randomized clinical trial. Journal of Consulting and Clinical Psychology 1999:67;931-44. 25. Newton-John TRO, Spence SH, and Schotte D. Cognitive-behavioural therapy versus E M G biofeedback in the treatment of chronic low back pain. Behaviour Research and Therapy 1995:33;691-7. 26. Vlaeyen JWS, Kolesnijders AMJ, Boeren RGB, and Vaneek H . Fear of Movement (Re)Injury in Chronic Low-Back-Pain and Its Relation to Behavioral Performance. Pain 1995;62:363-72. 27. van Tulder MW, Koes BW, Metsemakers JFM, and Bouter L M . Chronic low back pain in primary care: a prospective study on the management and course. Family Practice 1998;15:126-32. 28. Rossvoll I, Benum P, Bredland TR, Solstad K, Arntzen E, and Jorgensen S. Incapacity for Work in Elective Orthopedic-Surgery - A Study of Occurrence and the Probability of Returning to Work After Treatment. Journal of Epidemiology and Community Health 1993;47:388-94. 29. Quan H , Lafreniere R, and Johnson D. Health service costs for patients on the waiting list. Canadian Journal of Surgery 2002;45:34-42. 67 CHAPTER 5 5.0 TEMPORAL TRENDS IN RATES OF LUMBAR DISCECTOMY IN BRITISH COLUMBIA2 5.1 Introduction 5.1.1 Epidemiology of Back Pain Back pain poses an enormous burden to individuals and society in all industrialized countries. In some studies the lifetime prevalence of back pain exceeds 80%.1;2 In the United States (US), back pain is the second most common reason for consulting a physician3 and the second most common reason for being absent from work.4 In Canada back pain is the leading cause of activity limitation in adults younger than 45 years, and the second leading cause among those aged 45 to 64 years.5 The unadjusted cumulative incidence of back pain among adults is estimated to be 42 and 47 per 1,000 person-years for males and females, respectively.6 The highest incidence of back pain is observed among adults aged 45 to 64 years, and the lowest among those aged 18 to 24 years.6 In Saskatchewan, the annual age- and sex-adjusted incidence (proportion) in the general population is 18.6%.7 Most episodes are described as mild in severity, however 28.7% of those with prevalent back pain experience a recurrence within six months, and 40% of prevalent cases have pain that either worsens or does not improve after one year.7 " A version of this chapter will be submitted for publication. Quon J, Levy A, Sobolev B, Kopec J, Fisher C, Schechter M. (2007) Temporal Trends in the Rates of Surgical Lumbar Discectomy in British Columbia, Canada, 1990-2003. 68 5.1.2 Economic Costs of Back Pain Work-related back injuries account for approximately 30% of total workers' compensation costs in Quebec.8 According to Annual Reports at WorkSafeBC (formerly Workers' Compensation Board of British Columbia [BC]) back injuries accounted for approximately 20% of almost $3.5 billion (CAD) for disability and survivor benefits between 2000 and 2004. Additional figures not represented in these data are the costs of treating workers who developed symptoms while employed by industries not covered by compensation insurance.9 Compensation costs also do not include productivity losses of employers or the pain and suffering of claimants and their families. These costs typically exceed direct medical costs.10 In 1994, the total cost of spinal disorders in Canada was estimated to be $8.1 billion (CAD). Indemnification for time loss and other indirect costs account for much of this total, and in-hospital diagnostic and surgical procedures account for up to one-third of direct costs.11 The most common condition associated with in-hospital procedures is herniated lumbar disc.1 2 ; 1 3 5.1.3 Back Pain from Lumbar Disc Herniation Pain from lumbar disc herniation can manifest as acute or chronically intermittent back pain and/or "sciatica", that is radiating pain in a part of the leg supplied by a specific nerve root (a "dermatome"). A diagnosis of sciatica from lumbar disc herniation is commonly based on subjectively reported symptoms and physical examination findings alone. However, an in-hospital procedure such as computerized tomography (CT) or magnetic resonance imaging (MRI) is necessary if an operative treatment—surgical discectomy—is being considered, or if other serious spinal conditions, such as a spinal fracture, infection or tumor, need to be ruled-out. The natural history of lumbar disc herniation is generally favourable, such that patients often improve without surgery regardless of treatment modality. Common non-surgical treatments for disc herniation include activity modification, various forms of physical therapy, 69 exercises, spinal manipulation and pharmacotherapy. ; Surgical treatment (discectomy) is warranted for padents with either a progressive neurologic deficit, intractable pain, or symptoms that do not improve after a trial of medical or other non-surgical therapy no less than six-weeks in duration. Rarely, emergency discectomy is required in the presence of a rapidly progressive neurologic deficit or cauda equina syndrome (characterized by bowel or bladder dysfunction, a bilateral neurologic deficit, and/or saddle anaesthesia), however the majority of procedures are elective in nature. In North America, surgical discectomy accounts for about 80% of all in-hospital spinal procedures.16 5.1.4 Effectiveness Of Surgical Discectomy Until recently, surgical discectomy was widely considered more effective than non-surgical treatment for disc herniation.17 Nonrandomized studies had consistendy shown a benefit in favour of surgical over non-surgical treatment,18"20 and the first published randomized trial direcdy comparing surgical to non-surgical treatment indicated that discectomy was superior to medical treatment at one year.21 The difference between groups was not statistically significant at four and 10 years' follow-up, however independent reviewers noted a clinically important trend in favour of surgical treatment.17122 In a recendy published randomized trial comparing surgical discectomy to non-surgical treatment of lumbar disc herniation, both groups improved substantially over a two-year period, however in an intention-to-treat analysis, no important differences were found in mean changes in primary outcomes (SF-36 bodily pain and physical function scales, and modified Oswestry Disability Index) even though the effect of treatment was slighdy greater for surgery at three, 12, and 24 months' follow-up.23 Mean improvements in secondary outcomes (including the Sciatica Bothersomeness Index) were significandy greater for surgical patients, but because of the large numbers of patients who crossed over in both directions (40% and 45% in the surgical and non-70 surgical groups, respectively) clear conclusions about the genuine effectiveness/ineffectiveness of surgery could not be made. Patients who declined randomization in this latest trial were recruited into a concurrent observational study in which otherwise identical inclusion/exclusion criteria and outcome assessments were used.24 In the observational study, patients who chose surgical treatment reported greater improvement in all primary outcomes (bodily pain, general physical function, and back-specific function) and all but one secondary outcome (work status), again at three, 12 and 24 months' follow-up. That surgery appeared to be effective in the observational study, but not in the randomized trial, raises concerns about possible confounding. As surgical patients undergoing an invasive intervention may hold higher expectations for a successful outcome than non-surgical patients,25 the greater benefit of surgery in the observational study may be spurious due to a liberal, expectation bias. On the other hand, the large number of crossovers raises equal concerns about contamination—a conservative bias—and therefore, an attenuation of the benefit of surgery in the randomized trial. On the balance of the information currently available, surgical discectomy appears to result in quicker pain relief and functional recovery in the short-term, even if its superiority over nonoperative treatment is uncertain in the long-term.16 Despite that conclusion, the fact that lumbar discectomy is performed so commonly suggests that many clinicians and patients believe that surgery is genuinely effective. However, as patients improve with observation alone, and intra- and post-operative complications are possible,23 the decision to proceed with surgical discectomy still depends on a combination of clinical as well as nonclinical factors, including the surgeon's personal enthusiasm or threshold for recommending an operation and the patient's personal preference for other treatments. Some indications for discectomy ("intractable pain" and 71 "progressive neurologic deficit") are subject to personal interpretation and therefore contribute further to the so-called "discretionary" nature of this procedure. 5.1.5 Patterns of Utilization of Lumbar Discectomy Regional variations and overall increases in the rates of discretionary procedures such as discectomy (and spinal fusion) continue to fuel concerns about inappropriate treatment and unnecessary expenditures for hospital care.26"28 In at least one study, higher utilization has actually been associated with worse outcomes after elective back surgery.29 In Maine, population-based rates of surgery ranged from 38% below to 72% above the average rate in the state (a greater than four-fold difference).29 Patients who were treated in the lowest rate spine service area faired better in terms of several outcomes, including: 1) the mean reduction in their frequency of symptoms; 2) the mean reduction in back pain-specific functional status (Roland Disability scores); 3) the proportion of patients satisfied with the degree of improvement in their quality of life; 4) the proportion of patients satisfied with their current state of recovery; and 5) the proportion of patients willing to have the same operation again.29 Differences in baseline symptoms of patients across different spine service areas suggested that the superior outcomes in the lowest rate area may have been related to the more stringent selection of candidates for back surgery in that area.30 These patients were more likely to have prognostically favourable indications for surgery, including symptoms that were of a shorter duration, more likely to be unilateral and generally less severe in the back. These patents also had more severe findings (i.e. clearer pathological indications for back surgery) on independendy evaluated imaging studies. Finally, only 59% of patients in the highest rate area had positive straight leg raising (a specific clinical test for nerve root irritation due to lumbar disc hernation) at baseline, compared to 79% of patients in the other two (lower rate) service areas. 72 Successful outcomes after any treatment are obviously dependent on informed clinical decision making and the appropriate selection of patients according to known predictors of recovery. As others have noted, examining patterns of spine surgery across time (and geographic regions) can lend insight into variations and changes in clinical decision making about a common set of procedures.28 Such studies have important implications for the quality and cost of care because they may suggest over- or under-utilization of certain treatments, areas of limited professional consensus, and opportunities for improving consistency in clinical care.28 For such studies, population-based data sets can be helpful in identifying and quantifying trends in the use of spine surgery, and in identifying important questions for future research. To date, much is known about the trend in back surgery rates in the US where rates of utilization had generally increased throughout the 1990's even though the prevalence of back pain and the proportion of all physician visits related to back pain were stable.28'31 Less is known about the trend in back surgery rates in Canada. Over a 10-year period, aggregated rates of neck and back procedures in Ontario increased mildly (from 72.1 per 100,000 in 1982 to 82.3 in 1992),32 however estimates for the trend after 1992 and for other provinces are lacking. 5.1.6 Questions About Lumbar Discectomy Rates in British Columbia Lumbar discectomy is considered a "high-volume" procedure in BC, 3 3 and despite evidence of an increasing trend in Ontario and the US, the pattern in BC could be different. Canada experienced major reforms in its health care system throughout the 1990's, including important changes in funding. In BC, the proportion of health expenditures covered by federal transfer payments fell from 30.6% in 1980 to an average of 21.5% by 1996.34 Among other significant changes, reductions in real spending in the hospital sector resulted in overall reductions in the number of staffed beds in short-term care units.35 From 1990/91 to 1996/97, the number of staffed short-stay beds declined by 30%, the number of acute days per 1000 population declined 73 by 29%, and the average length of stay declined by 13%.3 6 ; 3 7 The impact—if any—of these changes on lumbar discectomy rates is not known. In a study of the trends in hospital use for mechanical neck and back problems in Ontario and the US, higher admission rates for surgery in the US were thought to reflect a larger supply of surgical specialists and spinal imaging units.32 Similarly, in a comparison of hospitalization rates between Ontario and Washington State, lower rates of surgery in Ontario were attributed to global controls on hospital budgets and access to technology in Ontario.32 In BC, all "necessary medical and surgical procedures" are reimbursed by the government's health insurance plan, with the exception of procedures for work-related injuries, which are covered by workers' compensation insurance. Back injuries account for the majority of workers' compensation claims in BC (and other jurisdictions). Surgical treatment of workers' compensation claimants typically requires approval by a compensation board representative and an attending surgeon. Given this dual-authorization requirement and the extra oversight of patients by claims adjudicators and medical advisors within the compensation system, it is possible that discectomies paid for by the Workers' Compensation Board (WCB) of BC are performed on a less discretionary basis than non-compensation related procedures. On the one hand, the more stringent application of surgery—expected for compensation claimants—could manifest as greater stability of surgical rates over time compared to rates of noncompensation related surgeries. On the other hand, compensated claimants are also generally less responsive to treatment than their non-compensated counterparts.38"41 A negative perception of compensation claimants by surgeons may present a barrier to access to surgical care for compensation beneficiaries in BC. With regard to the association between compensation status and poorer outcomes, back pain is a "biopsychosocial" condition with risk factors and prognostic determinants that involve psychological and social as much as biological variables.42 Historically, so-called "compensation 74 neurosis" was one of several terms used to describe the occurrence of symptoms in circumstances leading to either litigation or entidement to compensation payments where such symptoms could not be attributed to an objectively demonstrable physical injury.43 This concept originally assumed that pain behaviour was motivated primarily by the prospect of financial gain, albeit at an unconscious level. However, it is clear that some compensated patients suffer serious financial losses and may therefore be motivated instead by non-pecuniary incentives, including a genuine fear of re-injury and dissatisfaction with work or an employer.40'43 Mechanisms currently proposed to explain compensation-related disability also acknowledge that the biological components of chronic back pain are modifiable by psychosocial variables, and vice versa.39'44'45 5.1.7 Objectives and Hypotheses The primary objective of this study was to estimate the change in sex- and age-standardized specific lumbar discectomy rates in BC between 1990 and 2003. We hypothesized that rates would decline as a result of reduced hospital sector funding. A secondary objective was to estimate the change in worker's compensation related discectomy rates during the study period. The reasoning is that physicians may offer surgery more willingly to noncompensated patients due to a belief that these patients will have superior outcomes. Therefore, the second a priori hypothesis was that the trend in rates differs by payor group status; specifically, the anticipated decline in rates will be greater for patients receiving workers' compensation. BC offers universal insurance coverage for "medically necessary" hospital services. The entire population has access to services within public hospitals without direct costs borne by patients. Private clinics have been licensed to perform surgical discectomies in BC since 2000, however to date only two facilities throughout the entire province (Cambie Surgery Centre and False Creek Surgical Centre) perform this procedure with any regularity. Notwithstanding the small number of facilities involved, it was also of interest to determine whether the emergence of 75 private clinics from 2000 onward had a discernible impact on the trend in discectomy rates. Consequendy, one other a priori hypothesis was that the trend in rates should differ before and after 2000, the year that private clinics started to become cerdfied to perform lumbar discectomies. 5.2 Methods 5.2.1 Data Source and Patient Selection Hospital discharge abstracts from the BC Ministry of Health were obtained from the BC Linked Health Database (BCLHD) through the University of Bridsh Columbia's Centre for Health Sevices and Policy Research (CHSPR). 4 6 These include a record of all persons discharged from all acute care hospitals in BC. The abstract includes information on patient demographics, diagnoses, and treatment. Cherkin et al previously developed an algorithm for identifying surgical and non-surgical hospitalizations for mechanical, non-traumatic low back disorders in administrative databases using the International Classification of Diseases, 9 t h Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes.47 While ICD-9 C M codes include a second decimal (specifying a more precise anatomic location or surgical procedure), in BC, some hospitals do not record the second decimal. We therefore used a combination of single and double decimal ICD-9 codes. To record treatments, BC hospitals employ the Canadian Classification of Diagnostic and Therapeutic Procedures (CCP) 4 8 surgical codes instead of ICD-9 C M procedure codes. We used a published algorithm from Ontario to match CCP codes to ICD-9 C M surgical codes.49'50 Given our focus on discectomy procedures, we restricted diagnosis codes to those indicating "herniated disc". While the use of single decimal codes allowed for inclusion of herniated disc cases that possibly included thoracic (as well as lumbar) procedures, researchers in Ontario49 reported that 76 the use of four-digit diagnosis codes yielded information that closely matched the results that had been obtained using the algorithm originally described by Cherkin.49 Hospital records of all inpatient and day surgery performed between April 1, 1990 and March 31, 2003 inclusive were searched. Table 5.1 lists the ICD-9 and CCP codes that were used to identify patients hospitalized for an intervertebral disc disorder and corresponding discectomy. Laminectomy ("exploration and decompression of spinal canal") is a procedure that is sometimes performed for spinal stenosis, however it is also used somewhat interchangeably with discectomy within administrative databases.32 Table 5.2 lists the diagnosis codes that were used to exclude patients with nonmechanical back pain, known cervical or thoracic disorders, and/or spinal fusion procedures. The record selection algorithm employing these codes is further described in Figure 5.1. A previous study of lumbar discectomies using hospital discharge data found that the principal diagnosis and procedure were reliably coded in 96% of abstracts.51 Patients included were residents of BC with either a principal or a primary diagnosis of herniated disc, and a procedure code for either disc excision, chemonucleolysis or "other destruction of intervertebral disc".b To avoid potential confounding, we focused on the largest homogenous subgroup of back surgery patients and excluded occurrences of spinal fusion regardless of anatomic reference and order of appearance within the same discharge abstract. Patients older than 79 years were also excluded since they are more likely to suffer from spinal stenosis and degenerative disc disease rather than acute disc herniation. To limit the study population to mechanical back pain patients, those with diagnosis codes for nonmechanical causes of back pain including neoplasms, spinal b In the hospital separations file, a principal diagnosis—indicated by an "M" code in the DIAGNOSIS TYPE field—represents the ICD-9 diagnosis most responsible for a patient's hospital stay. A primary diagnosis—indicated by a " 1 " code in the DIAGNOSIS TYPE field—represents a comorbid condition existing before admission and which usually has a significant influence on the patient's length of stay. For each patient only the first hospitalization was included. 77 infection, pregnancy, inflammatory spondyloarthropathies, congenital anomalies, acute trauma or dislocations and spinal fractures were excluded. 5.2.2 Compensation Status Definition Compensation status/responsible payor group was extracted from the "Responsibility for Payment" field in the hospital discharge data. Hospital episodes that were covered by the Workers Compensation Board (WCB) of BC were coded "1" while those covered for by other sources, including insurance plans from other provinces or territories, veteran affairs, the Medical Services Plan of BC, other federal insurance (RCMP, National Defense, etc.), and private self-paying Canadian, or other, residents were coded "0". c Within the B C L H D , the reliability of the responsible payor field as an indicator for compensation-related lumbar discectomy has not been evaluated. However in California, principal payor codes from hospital discharge data have been used to identify work-related eye injuries.52 In New Jersey, "good" agreement beyond chance (kappa = 0.78) was found between self-reports of serious work-related injuries and the presence of workers' compensation codes on a population-based discharge database.53 In a comparison between the use of hospital separations and compensation claims records in BC, the responsibility for payment field was a comparable surrogate to using ICD-9 C M " E " (external cause of injury and poisoning) codes in identifying serious work related injuries.54 There was good agreement beyond chance between the presence of a "WCB" code on the hospitalization record and a matching compensation claims record (kappa = 0.84). The concordance rate between of ICD-9 E coding for determining general cause-of-injury was 81%.5 5 c The coding scheme for the payment field changed in 2001, however the data values for provincial hospital programs and W C B — the two largest payor groups within the data set—remained the same. 78 5 . 2 . 3 D a t a A n a l y s i s 5 . 2 . 3 . 1 D e s c r i p t i v e A n a l y s i s For descriptive purposes, we tabulated the total number of patients and discectomies for the adult population of BC by including all repeat hospitalizations and patients older than 19 years. 5 . 2 . 3 . 2 A g e - S p e c i f i c a n d A g e - S t a n d a r d i z e d R a t e s For the study population—that is, patients aged 20 to 79 years—age- and year-specific rates of disecectomy were calculated separately for women and men. The numerators for these rates were the number of discharges, and the denominators were the annual age-, sex- and year-specific population counts in BC based on 1991, 1996 and 2001 censuses, and intercensal midyear estimates. Age- and sex-specific rates were standardized to the 1996 census-estimated population of BC 1 6 using the direct method. For each sex, age-standardized rate ratios (SRR) were obtained by dividing the standardized discectomy rate for each year by the standard population discectomy rate. Since the standardized rates in each year are based on the same population weights, this is equivalent to a comparative mortality figure obtained by dividing the expected number of events in each group (i.e. year in the current study) by the observed number of events in the standard population.56 5 . 2 . 3 . 3 A g e - S p e c i f i c C h a n g e s I n D i s c e c t o m y R a t e s O v e r 1 4 Y e a r s To estimate the annual change in the rate of discectomy for each age-, sex- and payor-specific group over the 14-year study period, negative-binomial regression was used to model the natural logarithm of the age-, sex-, and payor group-specific rates.d A smoothed estimate of the D A Poisson regression showed significant overdispersion indicating that the mean and variance of the rates were not equal. Negative binomial regression is similar to Poisson regression, but corrects for overdispersion by allowing the variance of the discectomy rates to exceed the mean.8 79 total percent change in stratum-specific rates over the 14-year study period was obtained by using the formula: lOOx {[exp(/?/x 14)]-1}, where B is the regression coefficient for the change in the discectomy rate over time, and i indicates the ith age-, sex- and compensation group-specific stratum. An approximate 95% confidence interval was estimated by replacing B with its corresponding upper and lower 95% confidence limits. The variable T R E A T M E N T PERIOD was created by assigning patients to one of two "intervention" groups based on the fiscal year of procedure. The period 1 group consisted of patients who were operated on between fiscal years 1990 and 1999 when discectomies in private clinics were not available in BC. Period 2 patients were those treated during fiscal year 2000 or later, when private discectomies in BC were available. Age-specific annual changes in rates were estimated by regressing the logarithm of rates for each age group at a time on continuous year. The overall crude annual changes in rates were estimated by including the rates for all age groups combined. Age-adjusted annual changes in rates were estimated by including dummy variables for each age-group in the model, while both age-and period-adjusted changes in rates were obtained by further including treatment period. Patients aged 20 to 64 years accounted for 91 and 99% of discectomies in the non-compensation and compensation groups, respectively. To minimize the number of empty strata, the study population was restricted to patients aged 20-64 years for any analyses that were stratified by compensation status/responsible payor group. In analyses that preceded stratification by compensation status, the standardized rates for each sex did not appreciably change when patients of retirement age (65-79 years) were excluded. 80 5.2.3.4 Effect-Measure Modification of Trend In Rates To test for effect-measure modification of the trend in rates over time by treatment period, sex, and WCB status, the variable Y E A R was combined with each relevant covariable as a series of two-way interaction terms in a multivariable regression model. Since the effect of compensation status was also expected to vary by sex, an additional two-way interaction term between WCB status and sex was included. Inclusion of interaction terms indicated that the trend in rates over time (effect of YEAR) was dependent on both WCB status and treatment period, and that the effect of WCB status was in turn dependent upon sex. To facilitate the interpretation of these multiple interactions, WCB status was used as a stratification variable so that only one remaining interaction term, Y E A R x PERIOD, needed to be retained in separate compensation status-specific models. 5.2.4 Sensitivity Analysis The case selection algorithm was repeated to include back-related diagnoses under the index heading "Injury and Poisoning" (ICD-9 codes 846.0-847.9), however this resulted in the addition of only three extra patients. Given the small number, and negligible impact on the results, we therefore adhered to our original case selection algorithm, which was consistent with that of existing studies of surgically treated mechanical back pain. 5.3 Results 5.3.1 Descriptive Analysis 5.3.1.1 Total Number of Discectomies Before Exclusions Table 5.3 shows the total number of discectomies for mechanical back pain identified during the entire 14 year study period. Columns 2 and 3 show the numbers of patients and hospitalizations, according to frequency of surgery. Between 1990 and 2003 inclusive 18,394 81 patients aged 20 years or more, with a probable mechanical intervertebral disc disorder in the back, accounted for 20,240 discectomies. About 91% of patients were hospitalized once for this procedure, while 8% had two hospitalizations during the study period. In 2003 (the most recent year of data), 24 surgeons, representing 10 public hospitals in BC, billed "more than occasionally" (at least once a month on average) for a discectomy procedure. This group accounted for a total of 821 admissions for lumbar (or thoracic) disc surgery that year. 5.3.1.2 Frequencies of Diagnosis and Procedure Codes Tables 5.4 and 5.5 show the frequency counts for the most common diagnosis and procedure codes. The ICD-9 code for "displaced thoracic or lumbar disc without myelopathy" accounted for 90% of all principal diagnoses, and the CCP code for "excision of intervertebral disc" accounted for more than 98% of all procedure codes. The CCP code for "intervertebral chemonucleolysis" (chemical discectomy) accounted for less than 1% of procedure codes. 5.3.1.3 Descriptive Analysis of the Study Population The demographic characteristics of the study population (primary discectomies for age groups between 20 and 79 years) are shown in Table 5.6 by compensation status/responsible payor group. From 1990 to 2003, provincial insurance and WCB together were responsible payors for more than 99% of total hospitalizations for discectomy. Compensation beneficiaries accounted for 15% of the study population. This distribution of procedures by responsible payor group was the same before and after the emergence of private clinics in 2000 (data not shown). On average, compensation beneficiaries were slightly younger and much more likely male than non-compensation patients (p < 0.0001). The distribution of patients across socioeconomic status (SES) quintiles were similar overall (p = 0.17). Compensated patients were represented only slightly more in the second SES quintile (22% versus 18%), and slightly less in the fourth quintile (15% versus 20%). 82 5.3.1.4 Age-Specific Counts and Rates of Discectomy Table 5.7 shows the proportions hospitalized in each age group by fiscal year, for women and men separately. The distribution of discectomies across age groups was relatively consistent in each year with few exceptions: a decrease in the proportion of discectomies for women aged 20-29 years, and a corresponding increase among women aged 50-64 years was observed in 2003. For men, an increase in the proportion of discectomies among age groups 20-29 and 50-64 years was seen in 2003, with corresponding decreases among age groups 30-39 and 50-64 years. Table 5.8 shows that age-specific and crude overall discectomy rates generally declined over time for all age groups and both sexes. 5.3.1.5 Standardized Rates of Discectomy for the Study Population Direcdy age-standardized discectomy rates for the study population are shown for each year, by sex, in Table 5.9. For women and men in each year, the direct standardized rates showed a declining trend over 14 years (Figure 5.2). The SRR estimates indicated that compared to the BC population in 1996, rates were 56% higher in 1990 for both women and men, and by 2003, they were approximately 34% and 44% lower for women and men, respectively. In Table 5.10, direct standardized rates and SRRs are displayed for women and men by compensation status/responsible payor group. Compared to the 1996 reference population, standardized discectomy rates for women and men were higher in 1990 and lower in 2003 for both payor groups. The SRRs at the top of each panel of Table 5.10 show that the difference between the 1990 and 1996 discectomy rates was greatest for compensated men. The rate in 1990 had been 105% higher than the 1996 rate for compensated men, while only 53% higher for noncompensated men. For women, the rate in 1990 had been 38% higher than the 1996 rate for compensated women, but 67% higher than the 1996 rate among noncompensated women. 83 By comparison, the SRRs at the bottom of each panel of Table 5.10 show that discectomy rates in 2003 were lower than rates in 1996 and that the difference between the 2003 and 1996 rates was greatest for compensated women. The rate in 2003 was 61% lower than the 1996 rate for compensated women, while only 32% lower than the 1996 rate for noncompensated women. For compensated men, the rate in 2003 was 53% lower than the 1996 rate, and for noncompensated men it was 43% lower than the 1996 rate. 5.3.1.6 Smoothed Estimates of the Trend in Rates Table 5.11 shows the smoothed estimates of percent change in discectomy rates over 14 years. Between 1990 and 2003, crude and age-adjusted rates declined by 55% (95% CI: -68, -35) in women and 61% (95% CI: -71, -48) in men. Compensation beneficiaries exhibited larger declines than noncompensated patients, and compensated men showed a larger decline (77%) than compensated women (65%) over the 14 years. 5.3.1.7 Effect-Modification of Trend in Rates by Treatment Period In multivariable regression analyses, we found a statistically significant trend in discectomy rates over time (p < 0.0001) even after adjusting for age group, sex and WCB status (data not shown). In an interactions model, the trend in rates differed significantly by WCB status (p = 0.01) and treatment period (p = 0.02), and the effect of WCB status in turn differed significandy by sex (p < 0.0001). To simplify the presentation of multiple interactions, parsimonious models were created by stratifying on WCB status. The coefficients and corresponding rate ratios from these models are presented in Table 5.12. The significant Year x Period interaction in each of the stratified models indicated that the time-related trend in discectomy rates was different before and after 2000 (when private discectomies became available in BC) in both payor groups. Rate ratios for the main effects of 84 these variables are omitted since they can not be interpreted in isolation from their interaction. However, Table 5.13 shows the rate ratios for the time-related trend in rates along with the annualized % change in rates (with 95% confidence intervals) according to treatment period and compensation status. The differences in the rate ratios between treatment periods and payor groups illustrate the modifying effect of both variables. Al l rate ratios for continuous year were less than one, indicating declines in the rates in both payor groups, and during both treatment periods. Vertical comparisons within Table 5.13 show again that the decline in rates was greater for compensated than noncompensated patients. The difference between compensated and non-compensated patients was greater during period 2 than period 1. The confidence intervals around each point estimate were wider during period 2 due to the smaller number of observations during that period. Horizontal comparisons within Table 5.13 show again that the decline in rates was greater during treatment period 2 than treatment period 1 for both payor groups. However the accentuated decline in rates during period 2 was greater among compensated than noncompensated patients (annualized % changes for compensated patients before and after 2000: -9 [95% CI: -10, -7] and -18 [95% CI: -25, -10], respectively). 5.4 Discussion From 1990 to 2003, the rate of lumbar discectomies in BC declined in similar fashion for both men and women. Similarities between the crude rates and direct age-standardized rates within years for both sexes indicated that these declines were not merely due to changes in the age distribution of the BC population over time. As the crude and direct standardized rates within years are calculated using the same age- and sex-specific rate, similarities between the two types of 85 rates confirmed that the population distribution in each study year was similar to that of the reference year (1996). Also, differences between the crude rate for the reference year and the direct age-standardized rates in other years were necessarily due to genuine differences in the age-specific rates between 1996 and these other years. Compared to 1990/91, the decline in lumbar discectomy rates was approximately 55% for women and 61% for men over 14-years. Therefore, the annualized rates of decline were approximately 3.9% ( 55%/14-years) for women and 4.4% (61%/14-year) for men. The decline in the number of staffed short-stay beds in BC was approximately 30% between 1990/91 and 1996/97, which also translates into an approximate annualized rate of decline of 4% (30%/7 years), which possibly reflects a causal association between the two. A temporary increase in age-adjusted rates for both women and men was observed in 1999 when new government funding was allocated for 58,000 additional surgical and medical procedures (which was announced in the 1999 BC Budget Speech).57 We also found that after controlling for age and sex, the decline in lumbar discectomy rates was greater from 2000 onward when spinal surgery became accessible through private clinics in BC. Although the decline in age-specific rates was particularly low in the final year of the study period, sensitivity analyses with and without this final year of data generated similar results (data not shown), indicating that the greater decline after 2000 was not an artifact of one particularly influential year. The overall decline in discectomy rates was hypothesized mainly in consideration of the downsizing in the number of staffed short-stay beds and other hospital resources in BC during the study period.37 However, other factors may have contributed to this decline as well. It is possible that the prevalence of back pain in general—and lumbar disc herniation in particular—declined 86 during the study period even though existing studies (mosdy from the Nordic countries) do not show a consistent trend in back pain prevalence during other time periods. An increasing trend in back pain prevalence was reported in Finland from 1964 to 1987,58 however a stationary trend was reported subsequendy in Finland from 1978 to 1992.59 Two additional longitudinal studies from Finland reported a slight decline in the overall prevalence of back pain between 1979 and 199260 and an overall decline only among men between 1972 and 1992.61 A potentially valuable future investigation would be to examine and quantify the trend in back pain prevalence in Canada, perhaps using serial panels of the National Population Health Survey. We were unable to identify studies of temporal changes in back pain prevalence specifically in the Canadian general population. We tried to obtain a crude indicator of disease prevalence by querying the BC Medical Services Plan (MSP) database for counts of unique study identification numbers according to ICD-9 diagnosis and fiscal year of service (data not shown). In contrast to the hospital discharge data, most records in the MSP database listed ICD-9 codes for back pain only to the three digit level. Consequendy, we were not able to separate an MSP record with a cervical and/or thoracic disc diagnosis from one with only a lumbar disc disorder. Notwithstanding this limitation, the aggregated counts of patients with a three-digit ICD-9 code for "intervertebral disc disorder" (in any location of the spine) decreased by about 35% from 4,137 in 1991 to 2,686 in 2002. Summary statistics for the number of new time loss claims for "back strain" were also obtained from WCB as a proxy indicator of the annual incidence of work-related back pain. From 1990 to 2003, the number of new time loss claims for back strain decreased by approximately 30% from 20,601 in 1990 down to 14,060 in 2003. This decline may have been the result of public education and injury prevention initiatives implemented by WCB through the media, particularly in 1996 (from WCB Annual Reports 1996 and 2004) However, because the incidence (and 87 prevalence) of identified disease does not necessarily reflect the true burden of underlying illness, inferences about rates of acute lumbar disc herniation based solely on annual numbers of new time loss claims must be interpreted cautiously. In Ontario, 29 to 32% reductions in the number of new work-related injuries during 1993 to 1998 were documented in three independent data sources: 1) claims statistics from the Workplace Safety and Insurance Board; 2) self-reported data from the Ontario sample of the National Population Health Survey (NPHS); and 3) self-reported data from the Survey of Labour and Income Dynamics (SLID).6 2 Regarding time-related changes in the underlying burden of illness, as well as changes in the incidence of claims actually filed for compensation, Mustard and colleagues (2003) found that the decline in rates of WSIB-documented time loss claims was steeper than the decline in rates of long-duration work absences documented by SLID. 6 2 The authors suggested that long duration work absences were a more accurate indicator of the underlying burden of illness and that the decline in claims for time loss had therefore been greater than the decline in the actual incidence of morbidity attributable to occupational exposures.62 In future studies, it would be useful to estimate temporal trends in the incidence and prevalence of back disorders according to subgroups defined by type of injury (e.g. work-related versus other), diagnosis (e.g. lumbar disc disorder versus nonspecific back strain), and compensation status. 5.4.1 Modification of Trend by Compensation Status We also found that for both sexes, the decline in rates was greater for compensated patients than non-compensated patients. Our finding of a greater decline in discectomy rates for compensated patients could have been attributable to parallel declines in any of the following: 1) access to surgical care; 2) patient or physician demand for surgical care; 3) prevalence of occupational injuries; and/or 4) incidence of claims filed by compensated patients. We hypothesized that access to care may have been more limited for compensated than 88 noncompensated padents had surgeons either consciously or unconsciously allocated operating room time preferentially to those deemed most likely to benefit from treatment. From an anecdotal perspective, surgeons and non-surgical spinal care specialists at a major tertiary care centre in Vancouver indicated at the inception of this study that although they would not consciously deny surgical treatment to any patients with appropriate indications, they might unconsciously persist with non-operative treatment longer for a compensated patient. Further studies on the effect of compensation status on waiting time and other measures of access to surgery using a clinical data set would be useful in clarifying this issue further. Over the past decade, small area variations in the rates of back surgery have been viewed as an indictment of highly variable practice styles, the discretionary nature of surgical decision-making, and a lack of definitive evidence of the superiority of surgical discectomy over nonoperative treatment.16 A greater decline among WCB-patients might have been possible given the potential for differential adherence to evidence based care (which specifically discourages inappropriate surgery) within versus outside of the compensation system. Certainly, education of primary care physicians about the appropriate management of back pain has been associated with large reductions in the rates of diagnostic imaging and referrals to specialists,63 and therefore, the extra oversight of patients as well as physicians could have resulted in the more stringent selection of patients (and therefore higher threshold) for surgery for compensation beneficiaries. In future, a study of temporal variations in the severity of symptoms of patients at the time of placement onto the waitlist for discectomy would help to clarify whether the threshold for surgical treatment has increased over time, particularly among compensated surgical patients. In a climate of service rationing, a higher threshold for elective surgical discectomy over time is 89 conceivable since surgeons would likely devote an increasing proportion of their limited operating room time to the treatment of patients with the most urgent conditions. The effect of compensation status on the temporal trend in discectomy rates must be interpreted with caution for one other reason. Compared to data from hospitalization records, recent workers' compensation statistics in BC were found to have underestimated the number of probable work-related injuries by at least 10%.(Hasnat 2006) Other studies report that as few as 11% of injured workers receive compensation coverage.64 Some injured workers choose not to report their injuries at all, while others file for compensation and then subsequently have their claims rejected.65 That compensated patients might have been misclassified as noncompensated patients is less concerning if misclassification had occurred at a constant rate over the entire study period. On the other hand, if the rate of misclassification had been increasing in the same direction over time, then a bias would have existed in favour of a greater decline in discectomy rates for the compensation group. It has been hypothesized by at least one author that the influence of the adversarial policies and practices of (compensation) insurers, regulators and employers has raised the threshold of disability for a compensation claim, and that this increasing disability threshold may be responsible for a decline in compensable workplace injury and illness over the past decade.62 In BC, however, we have little reason to believe that the compensation system became increasingly adversarial towards workers during the study period. Since 1995, the WCB of BC has made conscious efforts to improve its service satisfaction levels among claimants. Between 1996 and 1998, the percentage of injured workers who said they were satisfied with WCB's services increased from 75% to 84% (from WCB Annual Report 1998), and between 1998 and 2003 injured workers consistently rated their level of satisfaction with WCB's services highly (average ratings on a scale out of 10 were 8.1, 7.9, 7.8, 8.0, 7.6, 7.5 from 1998 through 2003 respectively according to in-house 1998-2003 Annual Reports} 90 5.4.2 Modification of Trend by Treatment Period While the diversion of surgical padents to private clinics may pardy explain the greater decline in surgical rates among compensation beneficiaries after 2000, it is less clear why a greater decline was also observed for noncompensated patients. Possibly, private clinics attract surgeons from the public system, who end up reducing the time that they normally spend operating in public hospitals. In both Manitoba and Alberta, where cataract surgery is available through private clinics, median waiting times for surgery in the public sector are longer for surgeons who operate in the both the private and public sector than for surgeons who operate only in the public sector.66 This raises questions about whether or not surgeons who practice in both sectors have less operating room time in the public sector. Some researchers suggest that, based on the patterns of waiting times across surgeon-types in Manitoba, Alberta and the United Kingdom, surgeons who are allowed to operate in both sectors have an underlying incentive to encourage long waits in the public sector to encourage more patients to seek care in the private sector.66 However, future studies of the effects of private clinics on surgeons' practice patterns and caseloads are needed to determine with greater certainty whether private clinics actually erode, rather than enhance, access to already limited services within public hospitals. 5.4.3 Strengths and Limitations This study was based on administrative data with person-specific information which permitted us to use individual patients rather than procedures as the unit of analysis. Also, in Canada, health care insurance coverage is universal and therefore the use of province-wide hospital discharge data allowed us to capture virtually all patients who received a surgical discectomy for back pain in a public hospital in BC. We studied a homogenous population by focusing on one specific diagnosis group (intervertebral disc disorders in the back) and one specific procedure group (discectomy and/or 91 laminectomy) while simultaneously excluding patients with diagnoses of nonmechanical back pain. We intentionally focused on discectomy procedures, since they account for the majority of low back surgeries in Canada.13 We also looked specifically at the trends in surgical discectomy rates within the compensation population, a group for which discectomy utilization rates have not been documented in the existing literature. Due to the use of single-decimal instead of double-decimal ICD-9 codes, we were not always able to tell if patients with a diagnosis of "intervertebral disc disorder - region unspecified" were treated for cervical, thoracic or lumbar pain. On the other hand, the frequency distribution of ICD-9 codes in this study was very similar to that of another study which specifically validated Cherkin's algorithm for identifying cases of mechanical low back pain using administrative data.49 Again, Ontario researchers obtained "closely matching" data sets from Washington State when using single and then double-decimal diagnosis codes.49 We did not validate the responsible payment field in the B C L H D , nor did we rely on records from WCB to validate compensation status codes in the hospital discharge records. However, other investigators showed that the responsible payment schedule in the hospitalization discharge abstracts (i.e. B C H L D data) captured a greater number of "serious work-related injuries" among sawmill workers than the claims records form WCB. Of all patients whose hospitalization record indicated WCB as the responsible payor, 91% had a matching compensation claims record, suggesting that the provincial compensation board underreported at least 9% of serious work related injuries between 1989 and 1998.67 Future studies on the temporal trend in the incidence and prevalence of lumbar disc herniation in BC would add context to the findings of this study and would help to elucidate whether the current trend in surgical discectomy is, in fact, related to reductions in the underlying 92 burden of disease. Also, studies on the trends in clinical severity of padents at the time of placement onto the waitlist would help to determine whether limited access to surgical discectomy in BC has resulted in a higher prevalence of patients with severe pain and/or functional disability from their condition. This study also raises questions about whether compensated patients access surgical care within public hospitals in an equitable manner. While compensated patients appear to have superior access to care within private clinics it would be useful to determine in a future study how compensation status affects waiting time and other measures of access to surgical discectomy in public hospitals. An assessment of the variation in discectomy rates between sub-provincial regions of BC would also help to provide an initial overview of how equitable access to this procedure is for different communities throughout the province. 5.5 Summary Prior to this study, it was already known that population-based rates of back surgery had been increasing in the US despite stability in the prevalence of back pain over the past decade. Similar, smaller increases in rates of back surgery had also been documented in Ontario, but only prior to 1994. What was not known was how population-based rates of back surgery had changed, if at all, since the early 1990's in any part of Canada. This study showed that in BC the rates of lumbar discectomy—the most commonly performed surgical procedure in the low back—actually declined steadily from 1990 over a 14-year period. For women and men, lumbar discectomy rates declined more rapidly after 2000 when access to lumbar discectomy became available through private clinics. This accentuated decline was greater among patients receiving workers' compensation, who are already known to be the primary patrons of private surgery clinics. However as noncompensated patients also showed a steeper decline in back surgery rates after 93 2000, the diversion of workers' compensadon padents to private clinics did not have a discernible impact on improving the supply of lumbar discectomies to noncompensated patients remaining within the public system. 94 Table 5.1: Study Inclusion Codes Cases were included if they had one ICD-9 code indicating herniated disc as either a "principal" (diagnosis type = "M") or a "primary" (diagnosis type = "1") diagnosis, and one CCP code in any procedure field indicating discectomy, chemical discectomy, or laminectomy. ICD-9 codes for clinical category "herniated disc (principal or primary diagnosis): 722.1 Displacement of thoracic or lumbar disc without myelopathy 722.10 Displacement of lumbar disc without myelopathy 722.2 Displacement of unspecified disc without myelopathy 722.7 Disc disorder with myelopathy 722.70 Disc disorder with myelopathy, site unspecified 722.73 Lumbar disc disorder with myelopathy CCP codes for discectomy and laminectomy (any field): 16.0 Exploration and decompression of spinal canal 89.08 Sequestrectomy other site 89.09 Sequestrectomy unspecified site 92.31 Excision of intervertebral disc 92.33 Intervertebral chemonucleolysis 92.34 Other destruction of intervertebral disc 95 Table 5.2: Study E x c l u s i o n Codes Exclusion codes for diagnoses or procedures associated with non-mechanical back pain in any field, spinal fusions, repeat back surgery, known cervical or thoracic principal diagnoses, and principal or primary diagnoses of back pain due to traumatic injury. ICD-9 codes for "herniated disc" in cervical or thoracic region (as principal diagnosis): 722.71 Cervical disc disorder with myelopathy 722.72 Thoracic disc disorder with myelopathy CCP code for repeat laminectomy: 16.02 Re-opening of laminectomy site ICD-9 codes for diagnoses of non-mechanical back pain (in any diagnosis field): 140-239.9 Neoplasms 324.1 Intraspinal infection 630-677 Complications related to pregnancy 720* Inflammatory disease 730* Osteomyelitis 805.1, 805.3, 805.5, Open vertebral fractures without spinal cord injury 805.7, 805.9 806* Fractures of vertebral column with spinal cord injury 839-839.5 Vertebral dislocations E800-E849.9 Motor vehicular accidents (i.e. trauma-related back pain) CCP codes for procedure involving non-mechanical back pain (in any procedure field): 16.2 Chordotomy ICD-9 codes for non-mechanical back pain (as principal or primary diagnosis): 733.1 Pathological fracture 805.0, 805.2, 805.4, Closed vertebral fracture without spinal cord injury 805.6, 805.8 96 Table 5.2: Exclusion codes (continued) ICD-9 codes for mechanical pain in cervical or thoracic region (as principal diagnosis): 353.2 Cervical root lesions, not elsewhere classified 353.3 Thoracic root lesions, not elsewhere classified 721.0 Cervical spondylosis, without myelopathy 721.1 Cervical spondylosis, with myelopathy 721.2 Thoracic spondylosis, without myelopathy 721.41 Thoracic or lumbar spondylosis with myelopathy (thoracic region) 722.0 Displacement of cervical intervertebral disc without myelopathy 722.11 Displacement of thoracic intervertebral disc without myelopathy 722.4 Degeneradon of cervical intervertebral disc 722.71 Intervertebral disc disorder with myelopathy-cervical region 722.72 Intervertebral disc disorder with myelopathy-thoracic region 722.81 Posdaminectomy syndrome-cervical region 722.82 Posdaminectomy syndrome-thoracic region 723.0 Spinal stenosis, other than cervical-thoracic region 723.4 Brachial neuritis or radiculitis NOS 724.01 Spinal stenosis, other than cervical-thoracic region*/ CCP codes for procedures involving spinal fusion (in any procedure code field): 93.0* Spinal fusion listed before or in absence of listing for discectomy 97 Figure 5.1: Study Patient Selection Algorithm Examine all hospital separations during fiscal years 1990/91-2003/04 Include only residents of BC at time of admission Exclude patients with incomplete age, sex or residency information Include cases meeting ICD-9 and CCP coding criteria described in Table 5.1 Exclude cases using ICD-9 and CCP coding criteria described in Table 5.2 Include only patients between 20-79 years of age If more than one hospitalization per patient, include only the first hospitalization 98 Table 5.3: Hospitalizations for Lumbar Discectomy Number of patients and frequency of hospitalizations for discectomy (or laminectomy) in BC, 1990/91-2003/04, including all age groups > 20 years. Back procdures excluding fusions, known "repeat laminectomy" and known cervical spine, thoracic spine, or non-mechanical principal/primary diagnoses** Frequency of surgery Number of unique study IDs Number of procedures 0) (2) (3) = (l)x(2) 1 16,717 (91) 16,717 (83) 2 1,525 (8) 3,050 (15) 3 136 408 (2) 4 15 60 5 1 5 6 0 0 Total 18,394 (100) 20,240 (100) *Frequency counts with column percentages shown in parentheses. Column percentages less than 0.5 are omitted. | C / S = cervical spine, T/S = thoracic spine. 99 Table 5.4: Frequencies of Diagnoses Frequencies of diagnosis codes for first-hospitalizadons, by treatment period, BC, Canada, 1990/01 -2003/04. Code Total* Principal diagnoses 722.1 Displaced thoracic or lumbar disc without myelopathy 16,544(90) 722.7 Intervertebral disc disorder with myelopathy 647 (4) 722.10 Lumbar intervertebral disc without myelopathy 583 (3) 724.0 Spinal stenosis, other than cervical 304 (2) 722.2 Displaced intervertebral disc, site unspecified, withoug 55 (0.3) 722.5 Degeneration of thoracic or lumbar intervertebral disc 42 (0.2) Other 219 (1) Total** 18,394 (100) Primary diagnoses (n = 500)f 722.1 Displaced thoracic or lumbar disc without myelopathy 417(83) 722.7 Intervertebral disc disorder with myelopathy 25 (5) 722.10 Lumbar intervertebral disc without myelopathy 16(3) 724.0 Spinal stenosis, other than cervical 7 (1) 722.2 Displaced intervertebral disc, site unspecified, without 5 (1) 344.6 Cauda equina syndrome 2 (0.4) 724.3 Sciatica 2 (0.4) 738.4 Acquired spondylolisthesis 2 (0.4) Other 24 (5) Total** 500 (100) *Counts with column percentages in parentheses. **Row percentages in parentheses. fWhere principal diagnosis was not coded 722* for an intervertebral disc disorder. 100 Table 5.5: Surgical Procedure Codes Frequencies of CCP codes for primary hospitalizations, including all age groups >20 years, by treatment period, BC, 1990/01 - 2003/04. CCP Code Description of Procedure Total* N = 18,394 9231 Excision of intervertebral disc 18,085 (98) 9233 Intervertebral chemonucleolysis 39 (< 1) 9234 Other destruction of intervertebral disc 272 (1) 160 Exploration and decompression 1,184 (6) 8908 Sequestrectomy, other specified site 23 (< 1) 8909 Sequestrectomy, unspecified site 0 Total 19,603 (107) *Frequency counts with column percentages shown in parentheses (not including percentages less than 0.5). +Total count exceeds 18,394, and total percentage exceeds 100 because some hospitalizations involved more than one procedure. 101 Table 5.6: Demographics of Study Population Demographic characteristics of study population, aged 20 to 79 years at time of first-hospitalization for discectomy, BC, Canada, 1990/01 - 2003/04 Characteristicf WCB Provincial Program Total p value* N (row %) 2,709 (15) 15,336 (85) 18,045 (100) Age* Mean (SD) 39.3 (9.3) 44.6 (13.3) 43.8 (12.9) < 0.0001 Median (IQR) 38.0 (25.0) 42.0 (24.0) 42.0 (18.0) Age (categorical) < 0.0001 20-29 381 (14) 1,673 (11) 2,054 (11) 30-39 1,116 (41) 4,557 (30) 5,673 (31) 40-49 809 (30) 4,280 (28) 5,089 (28) 50-64 387 (14) 3,246 (21) 3,633 (20) 65-79 16(1) 1,480 (10) 1,596 (9) Sex 0.0004 Women 444 (16) 6,528 (43) 6,972 (39) Men 3,365 (84) 8,808 (57) 11,073 (61) Neighbourhood SES 0.17 Quintile 1 508 (19) 2,860 (19) 3,368 (19) Quintile 2 593 (22) 2,799 (18) 3,392 (19) Quintile 3 531 (20) 2,865 (19) 3,396 (19) Quintile 4 514 (19) 3000 (20) 3,514 (19) Quintile 5 412 (15) 3080 (20) 3,492 (19) Missing 151 (6) 732 (5) 883 (5) •(•Frequency counts with column percentages shown in parentheses. +.SD = standard deviation. IQR = interquartile range. *p values based on t test for continuous variables, Fisher's Exact test for dichotomous variables, and chi-square tests for polychotomous variables 102 T a b l e 5.7: P r o p o r t i o n s H o s p i t a l i z e d f o r L u m b a r D i s c e c t o m y b y A g e a n d S e x Annual percentages of hospitalizations for primary discectomies, by sex and age, BC, Canada, 1990/01 — 2003/04 Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Total Women N 627 598 559 574 552 499 475 472 446 495 464 419 449 343 6,972 Age 20-29 11 12 11 10 10 7 9 6 8 9 10 10 9 5 9 30-39 29 28 29 28 29 25 30 28 29 33 26 29 27 26 29 40-49 28 27 31 31 30 30 31 33 30 29 27 28 30 33 30 50-64 21 21 20 19 20 22 18 20 22 21 23 22 22 26 21 65-79 11 12 9 12 11 16 12 13 11 9 13 11 12 10 11 Total* 100 100 100 100 100 100 100 100 100 101 99 100 100 100 100 Men N 1,019 971 944 934 893 878 758 767 696 804 692 645 616 456 11,073 Age 20-29 16 13 14 14 13 13 10 11 13 10 12 10 10 14 13 30-39 32 36 34 37 35 33 35 34 33 32 31 31 31 25 33 40-49 26 27 27 28 27 26 28 28 27 28 27 28 30 27 27 50-64 19 18 19 16 19 19 20 19 19 22 20 23 22 26 20 65-79 7 5 6 5 6 9 8 8 8 8 9 9 8 9 7 Total* 100 99 100 100 100 100 101 100 100 100 99 101 101 101 100 • S u m o f c o l u m n p e r c e n t a g e s n o t a l w a y s e q u a l t o 1 0 0 d u e t o r o u n d i n g . o Table 5.8: Age-Specific Rates of Discectomy Age-specific rates (per 100,000 person-years) of discectomy for women and men, BC, Canada, 1990/91 — 2003/04. Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Women Age 20-29 26.5 27.7 23.9 21.8 22.1 14.0 15.0 11.3 13.2 15.9 17.3 15.4 15.6 6.6 30-39 60.9 55.8 52.8 50.8 49.1 38.1 42.4 39.9 39.4 49.3 36.6 37.8 37.7 28.2 40-49 75.9 67.5 68.1 66.4 58.4 51.5 47.4 49.3 40.7 42.9 37.9 33.6 38.9 32.6 50-64 54.4 54.5 45.3 43.3 43.9 41.1 31.6 31.7 32.0 32.6 33.4 27.7 27.6 23.8 65-79 35.5 36.3 27.2 36.1 29.9 39.0 27.1 29.1 23.5 21.4 28.9 21.2 24.9 16.3 Overall 51.0 48.8 44.4 44.4 41.7 36.9 33.8 33.2 30.9 34.0 31.5 28.1 29.8 22.6 Men Age 20-29 59.1 46.7 50.4 47.4- 42.5 40.9 27.4 31.1 30.9 29.4 31.6 24.2 22.5 22.8 30-39 112.0 117.1 104.1 109.8 94.2 85.8 76.8 75.2 68.9 77.8 65.8 61.4 60.3 37.4 40-49 112.3 105.0 99.0 96.1 86.2 77.2 67.9 68.8 58.4 69.0 57.3 52.9 54.3 35.3 50-64 78.3 71.6 69.7 58.3 62.8 62.4 55.1 49.6 44.5 55.5 42.0 43.3 37.6 31.8 65-79 44.6 28.8 36.5 28.8 32.5 47.3 34.4 35.6 30.9 35.6 32.4 28.7 24.7 19.8 Overall 84.7 78.1 75.9 73.1 67.6 65.2 55.0 54.6 49.0 56.1 47.6 44.0 41.8 30.5 o -1^  T a b l e 5.9: D i r e c t l y S t a n d a r d i z e d R a t e s Annual directly age-standardized rates of discectomy, direct standardized rate ratios (SRR) and indirect standardized morbidity ratios (SMR) for women and men, aged 20 to 79 years, BC, 1990/91 -2003/04 Direct Standardization Year Rate* 95% CI SRRf Women 1990 52.6 48.5 - 57.0 155.5 1991 49.8 45.9 - 54.0 147.4 1992 45.4 41.7-49.4 134.3 1993 45.0 41.4-48.8 133.0 1994 42.1 38.7 - 45.8 124.5 1995 36.9 33.8 - 40.3 109.2 1996 33.8 30.9 - 37.0 100.0 1997 33.1 30.3 - 36.3 98.0 1998 30.8 28.1 - 33.8 91.1 1999 34.1 31.2-37.2 100.8 2000 31.3 28.6 - 34.3 92.6 2001 28.1 25.5-31.0 83.2 2002 29.8 27.1 - 32.7 88.2 2003 22.3 20.1 - 24.9 66.1 Men 1990 85.6 80.4-91.1 156.1 1991 79.3 74.4 - 84.5 144.6 1992 76.2 71.4-81.2 138.9 1993 73.0 68.5 - 77.9 133.2 1994 67.4 63.1 - 72.0 123.0 1995 65.0 60.8-69.5 118.6 1996 54.8 51.1-58.9 100.0 1997 54.5 50.8-58.5 99.4 1998 48.9 45.4-52.7 89.3 1999 56.0 52.2 - 60.0 102.1 2000 47.8 44.3-51.5 87.2 2001 44.0 40.7 - 47.6 80.3 2002 42.1 38.9-45.5 76.7 2003 30.5 27.8-33.5 55.6 *Rate per 100,000 population, age groups 20 to 79. Standard population is 1996 Census-estimated population of BC. ySRR = direct standardized rate ratio multiplied by 100. 105 Figure 5.2: Plot of Standardized Rates Plots o f directly standardized rates o f discectomy (with 9 5 % confidence bands), by sex. CD Direct S tandard i zed Rates of Lumbar D iscectomy in British Co l umb i a 1990 - 2003 100 90 80 70 Men Upper 9 5 % confidence limit Point estimate Lower 9 5 % confidence limit 204===^ ~ i — | — i — . — i — i — i — i — i — i — i — | — i — . — i — . — i — i — i — i — i — | — i — , — . — i — i — i — . — i — i — r 1990 1992 1994 1996 1998 Yea r 2000 2002 2004 p Table 5.10: Standardized Rates by Sex and Payor Direcdy age standardized rates of discectomy for working aged men and women (20 to 65 years), by responsible payor, BC, 1990/91 - 2003/04. Year WCB Provincial/Other All f Rate 95% CI SRR Rate 95% CI SRR Rate 95% CI SRR Women 1990 4.6 3.4 6.1 138 53.6 49.1 58.4 167 43.9 40.1 48.0 160 1991 3.5 2.5 4.9 106 49.7 45.5 54.3 154 42.0 38.4 46.0 153 1992 4.1 3.0 5.5 123 45.2 41.3 49.5 141 38.3 34.9 42.0 140 1993 2.5 1.7 3.6 75 44.8 41.0 49.0 139 37.9 34.6 41.5 138 1994 5.0 3.8 6.5 150 39.5 36.0 43.4 123 33.7 30.7 37.1 123 1995 2.4 1.6 3.4 71 34.3 31.1 37.9 107 29.6 26.8 32.7 108 1996 3.3 2.4 4.5 100 32.2 29.1 35.5 100 27.4 24.8 30.3 100 1997 1.9 1.3 2.8 57 32.0 29.0 35.4 100 27.5 24.9 30.4 100 1998 2.4 1.7 3.4 72 29.9 27.0 33.1 93 25.5 23.0 28.3 93 1999 2.3 1.6 3.3 69 34.0 30.9 37.5 106 29.2 26.5 32.2 107 2000 2.8 2.0 3.8 83 29.2 26.4 32.4 91 25.1 22.6 27.8 91 2001 1.6 1.0 2.4 47 27.9 25.1 31.0 87 23.9 21.5 26.5 87 2002 1.5 1.0 2.4 46 29.2 26.4 32.3 91 25.1 22.7 27.8 92 2003 1.3 0.8 2.1 39 22.0 19.6 24.7 68 19.0 16.9 21.4 69 Men 1990 25.1 22.2 28.4 205 70.5 65.5 76.0 153 58.7 54.4 63.3 148 1991 22.3 19.6 25.4 182 68.1 63.2 73.3 148 57.0 52.9 61.5 144 1992 20.1 17.6 23.0 164 63.7 59.1 68.7 138 54.0 50.0 58.3 136 1993 17.9 15.6 20.6 146 63.3 58.7 68.2 137 54.1 50.2 58.3 137 1994 20.4 17.9 23.2 166 53.9 49.7 58.3 117 45.8 42.2 49.6 116 1995 17.8 15.5 20.4 145 50.5 46.6 54.8 110 43.4 40.0 47.1 110 1996 12.3 10.4 14.4 100 46.0 42.4 50.0 100 39.6 36.4 43.0 100 1997 10.2 8.5 12.1 83 47.4 43.7 51.4 103 40.8 37.6 44.3 103 1998 10.5 8.8 12.5 85 41.5 38.1 45.3 90 35.7 32.7 38.9 90 1999 13.3 11.4 15.5 108 46.1 42.5 50.1 100 39.7 36.6 43.1 100 2000 9.3 7.7 11.2 75 41.3 37.9 45.1 90 35.5 32.5 38.7 90 2001 7.9 6.5 9.7 64 38.9 35.5 42.5 84 33.4 30.5 36.5 84 2002 5.9 4.7 7.4 48 38.9 35.6 42.5 84 33.6 30.8 36.7 85 2003 5.9 4.7 7.4 47 26.3 23.6 29.3 57 22.7 20.4 25.3 57 *Rates per 100,000 population. Standard population is 1996 Census-estimated BC population. SRR = direct standardized discectomy rate ratio multiplied by 100 107 Table 5.11: Smoothed Estimates of Rate Changes Smoothed estimates of percent change in discectomy rates, with 95% confidence intervals, for WCB and non-WCB related hospitalizations, projected over 14 years, in BC, Canada, relative to 1990. Age group WCB Provincial/Other Al l % decline 95% CI* % change 95% CI* % decline 95% CI* Women 1) Age-specific 20-29 29 (-52) - 67 63 47-74 61 44-73 30-39 80 64-89 40 28-50 45 34-54 40-49 75 50-88 58 52-64 60 56-64 50-64 16 58-56 58 53-63 56 51 -62 2) Crude 66 45-79 54 33-68 55 35-68 3) Age-adusted 65 49-76 54 49-58 55 51-59 Men 1) Age-specific 20-29 84 77-89 54 43-63 64 57-70 30-39 76 70-81 53 42-62 61 54-67 40-49 73 66-78 61 54-66 64 58-68 50-64 76 56-87 52 46-57 56 50-62 2) Crude 78 65-86 55 41 -66 61 48-71 3) Age-adusted 77 73-80 55 51-59 61 58-64 *95% CI = 95% confidence interval. 108 T a b l e 5 . 1 2 : I n t e r a c t i o n M o d e l s Interaction models for the effect of period on the trend (annualized change) in discectomy rates, stratified by WCB status group* Variable Parameter estimate 95% CL11 Rate ratio 95% CL" p value Responsible payor = WCB Year (annual change) -0.090 -0.107, -0.073 < 0.0001 Period (> 2000 vs < 2000) 1.033 -0.014, 2.080 0.05 Sex (Women vs Men) -1.612 -1.723,-1.502 0.20 0.18, 0.22 < 0.0001 Age 30-39 (vs 20-29) 0.910 0.778,1.041 2.48 2.18, 2.83 Age 40-49 0.732 0.593, 0.870 2.08 1.81,2.39 Age 50-64 0.031 -0.124, 0.187 1.03 0.88,1.21 < O.OOOlf Year x Period -0.103 -0.196, -0.009 0.03 Responsible payor = Provincial/N on-WCB Year (annual change) -0.062 -0.071, -0.054 < 0.0001 Period (> 2000 vs < 2000) 0.516 0.073, 0.958 0.02 Sex (Women vs Men) -0.331 -0.374, -0.288 0.72 0.69, 0.75 < 0.0001 Age 30-39 (vs 20-29) 0.844 0.777, 0.911 2.33 2.17,2.49 Age 40-49 0.883 0.816, 0.951 2.42 2.26, 2.59 Age 50-64 0.642 0.573, 0.711 1.90 1.77, 2.04 < O.OOOlf Year x Period -0.041 -0.080, -0.002 0.04 *Stratification of the analysis by responsible payor group eliminates the need for two interaction terms (WCB x year, and WCB x sex). fFrom type 3 test of effect of all age groups overall. "95% CL = 95% confidence limits o Table 5.13: Annualized Percent Changes in Rates Rate ratios representing the change in the rate of surgical discectomy per one-year increment, and corresponding annualized percent changes in rates during each period of treatment, by compensation status. Period 1 Period 2 1990/91-1999/00 2000/01-2003/04 Insurance Rate ratio Annualized Rate ratio Annualized Group (95% CI) % decline* (95% CI) % decline* (95% 0 ) t (95% CI)f WCB 0.91 (0.90 - 0.93) 9 (7 -10) 0.82 (0.75 - 0.90) 18 (10-25) Non-WCB 0.94 (0.93 - 0.95) 6(5-7) 0.90 (0.87 - 0.94) 10(6-13) *Percent decline in the rate of discectomy for every one-year increment. f95% CI = 95% confidence interval 110 5.6 References 1. Andersson GBJ. Epidemiology of low back pain. Acta Orthopaedica Scandinavica 1998;69:28-31. 2. Walker BF. The prevalence of low back pain: A systematic review of the literature from 1966 to 1998. Journal of Spinal Disorders 2000;13:205-17. 3. Loisel P, Durand MJ, Berthelette D et al. Disability prevention - New paradigm for the management of occupational back pain. Disease Management and Health Outcomes - 2001;9:351-60. 4. Sternbach RA. Pain and Hassles in the United-States - Findings of the Nuprin Pain Report. Pain 1986;27:69-80. 5. Badley E M , Rasooly I, and Webster G K . Relative Importance of Musculoskeletal Disorders As A Cause of Chronic Health-Problems, Disability, and Health-Care Utilization - Findings from the 1990 Ontario Health Survey. Journal of Rheumatology 1994;21:505-14. 6. Kopec JA, Sayre EC, and Esdaile JM. Predictors of back pain in a general population cohort. Spine 2004;29:70-7. 7. Cassidy JD, Cote P, Carroll LJ, and Kristman V . Incidence and course of low back pain episodes in the general population. Spine 2005;30:2817-23. 8. Scientific approach to the assessment and management of activity-related spinal disorders. A monograph for clinicians. Report of the Quebec Task Force on Spinal Disorders. Spine 1987;12:Sl-59. I l l 9. Frank JW, Kerr MS, Brooker AS et al. Disability resulting from occupational low back pain -What do we know about primary prevention? A review of the scientific evidence on prevention before disability begins. Spine 1996;21:2908-17. 10. Maetzel A and Li L. The economic burden of low back pain: a review of studies published between 1996 and 2001. Best Practice and Research Clinical Rheumatology 2002;16:23-30. 11. Cats-Baril FJ. The economics of spinal disorders. In: Frymoyer JW, ed. The adult spine: principles and practice. New York: Raven Press, 1991:85-105. 12. Cherkin DC, Deyo RA, Loeser JD, Bush T, and Waddell G. An International Comparison of Back Surgery Rates. Spine 1994;19:1201-6. 13. Gentleman JF, Vayda E, Parsons GF, and Walsh M N . Surgical rates in subprovincial areas across Canada: Rankings of 39 procedures in order of variation. Canadian Journal of Surgery 1996;39:361-7. 14. Vroomen PCAJ, De Krom MCTF, Slofstra PD, and Knottnerus JA. Conservative treatment of sciatica: A systematic review. Journal of Spinal Disorders 2000;13:463-9. 15. Lisi AJ, Holmes EJ, and Ammendolia C. High-velocity low-amplitude spinal manipulation for symptomatic lumbar disk disease: A systematic review of the literature. Journal of Manipulative and Physiological Therapeutics 2005;28:429-42. 16. Volinn E, Turczyn K M , and Loeser JD. Patterns in Low-Back-Pain Hospitalizations -Implications for the Treatment of Low-Back-Pain in An Era of Health-Care Reform. Clinical Journal of Pain 1994;10:64-70. 112 17. Gibson J N A and Waddell G. Surgical interventions for lumbar disc prolapse. Cochrane Database of Systematic Reviews 2007. 18. Adas SJ, Singer D E , Keller R, Convery K, Mooney N , and Deyo R. The Efficacy of Lumbar Disc Surgery in Patients with Sciatica - Preliminary-Results from the Maine Lumbar Spine Study. Clinical Research 1993;41:A200. 19. Adas SJ, Deyo RA, Keller RB et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine 1996;21:1777-86. 20. Atlas SJ, Keller RB, Chang Y C , Deyo RA, and Singer D E . Surgical and nonsurgical management of sciatica secondary to a lumbar disc herniation - Five-year outcomes from the Maine Lumbar Spine Study. Spine 2001;26:1179-87. 21. Weber H . Lumbar-Disk Herniation - A Controlled, Prospective-Study with 10 Years of Observation. Spine 1983;8:131-40. 22. Bessette L, Liang M H , Lew RA, and Weinstein JN. Classics in Spine - Surgery literature revisited. Spine 1996;21:259-63. 23. Weinstein JN , Tosteson TD, Lurie JD et al. Surgical vs nonoperative treatment for lumbar disk herniation - The Spine Patient Outcomes Research Trial (SPORT): A randomized trial. Journal of the American Medical Association 2006;296:2441-50. 24. Weinstein JN , Lurie JD, Tosteson T D et al. Surgical vs nonoperative treatment for lumbar disk herniation - The Spine Patient Outcomes Research Trial (SPORT) observational cohort. Journal of the American Medical Association 2006;296:2451-9. 113 25. Flum DR. Interpreting surgical trials, with subjective outcomes - Avoiding UnSPORTsmanlike conduct. Journal of the American Medical Association 2006;296:2483-5. 26. Deyo RA, Cherkin D, Conrad D, and Volinn E. Cost, Controversy, Crisis - Low-Back-Pain and the Health of the Public. Annual Review of Public Health 1991;12:141-56. 27. Taylor V M , Deyo RA, Cherkin DC, and Kreuter W. Low-Back-Pain Hospitalization -Recent United-States Trends and Regional Variations. Spine 1994;19:1207-13. 28. Deyo RA, Mirza SK, and Martin BI. Back pain prevalence and visit rates - Estimates from US national surveys, 2002. Spine 2006;31:2724-7. 29. Lutz G K , Butzlaff M E , Atlas SJ, Keller RB, Singer D E , and Deyo RA. The relation between expectations and outcomes in surgery for sciatica. Journal of General Internal Medicine 1999;14:740-4. 30. Keller RB, Adas SJ, Singer D E et al. The Maine Lumbar Spine Study: 1. Background and concepts. Spine 1996;21:1769-76. 31. Jordon J, Morgan TS, Weinstein J, and Konstantinou K. Herniated lumbar disk. American Family Physician 2006;73:1240-2. 32. Lavis JN , Maker A, Anderson G M et al. Trends in hospital use for mechanical neck and back problems in Ontario and the United States: discretionary care in different health care systems. Canadian Medical Association Journal 1998;158:29-36. 33. Wright CJ, Chambers G K , and Robens-Paradise Y. Evaluation of indications for and outcomes of elective surgery. Canadian Medical Association Journal 2002;167:461-6. 114 34. Naylor CD. Health care in Canada: Incrementalism under fiscal duress. Health Affairs 1999;18:9-26. 35. Detsky AS and Naylor CD. Canada's health care system - Reform delayed. New England Journal of Medicine 2003;349:804-10. 36. Cloutier-Fisher D, Penning MJ, Zheng C, and Druyts EBF. The devil is in the details: trends in avoidable hospitalization rates by geography in British Columbia, 1990-2000. B M C Health Services Research 2006;6. 37. Sheps SB, Reid RJ, Barer M L et al. Hospital downsizing and trends in health care use among elderly people in British Columbia. Canadian Medical Association Journal 2000;163:397-401. 38. Adas SJ, Chang YC, Kammann E, Keller RB, Deyo RA, and Singer D E . Long-term disability and return to work among patients who have a herniated lumbar disc: The effect of disability compensation. Journal of Bone and Joint Surgery-American Volume 2000;82A:4-15. 39. Harris I, Mulford J, Solomon M , van Gelder JM, and Young J. Association between compensation status and outcome after surgery - A meta-analysis. Journal of the American Medical Association 2005;293:1644-52. 40. Teasell RW. Compensation and chronic pain. Clinical Journal of Pain 2001;17:S46-S51. 41. Young JN , Shaffrey CI, Laws ER, and Lovell LR. Lumbar disc surgery in a fixed compensation population: A model for influence of secondary gain on surgical outcome. Surgical Neurology 1997;48:552-8. 115 42. Waddell G. Biopsychosocial Analysis of Low-Back-Pain. Baillieres Clinical Rheumatology 1992;6:523-57. 43. Mendelson G. Chronic Pain, Compensation and Clinical Knowledge. Theoretical Medicine 1991;12:227-46. 44. Weighill V E . Compensation Neurosis - A Review of the Literature. Journal of Psychosomatic Research 1983;27:97-104. 45. Winkelstein BA. Mechanisms of central sensitization, neuroimmunology & injury biomechanics in persistent pain: implications for musculoskeletal disorders. Journal of Electromyography and Kinesiology 2004;14:87-93. 46. Chamberlayne R, Green B, Barer ML, Hertzman C , Lawrence WJ, and Sheps SB. Creating a population-based linked health database: A new resource for health services research. Canadian Journal of Public Health-Revue Canadienne de Sante Publique 1998;89:270-3. 47. Cherkin DC, Deyo RA, Volinn E, and Loeser JD. Use of the International Classification of Diseases (Icd-9-Cm) to Identify Hospitalizations for Mechanical Low-Back Problems in Administrative Databases. Spine 1992;17:817-25. 48. Blais R. Variations in Surgical Rates in Quebec - Does Access to Teaching Hospitals Make A Difference. Canadian Medical Association Journal 1993;148:1729-36. 49. Hu RW, Jaglal S, Axcell T, and Anderson G. A population-based study of reoperations after back surgery. Spine 1997;22:2265-70. 116 50. Taylor V M , Anderson G M , McNeney B et al. Hospitalizations for back and neck problems: A comparison between the Province of Ontario and Washington State. Health Services Research 1998;33:929-45. 51. Ramirez LF and Thisted R. Using A National-Health Care Data-Base to Determine Surgical Complications in Community Hospitals - Lumbar Discectomy As An Example. Neurosurgery 1989;25:218-25. 52. Baker RS, Wilson MR, Flowers CW, Lee D A , and Wheeler N C . Demographic factors in a population-based survey of hospitalized, work-related, ocular injury. American Journal of Ophthalmology 1996;122:213-9. 53. Sorock GS, Smith E, and Hall N . An Evaluation of New-Jersey Hospital Discharge Database for Surveillance of Severe Occupational Injuries. American Journal of Industrial Medicine 1993;23:427-37. 54. Alamgir H , Koehoom M , Ostry A, Tompa E, and Demers P. An evaluation of hospital discharge records as a tool for serious work related injury surveillance. Occupational and Environmental Medicine 2006;63:290-6. 55. Smith SM, Colwell LS, and Sniezek JE. An Evaluation of External Cause-Of-Injury Codes Using Hospital Records from the Indian-Health-Service, 1985. American Journal of Public Health 1990;80:279-81. 56. Julious SA, Nicholl J, and George S. Why do we continue to use standardized mortality ratios for small area comparisons? Journal of Public Health Medicine 2001;23:40-6. 117 57. McPhail, J. Budget '99: Improving health care, helping small business create jobs. Speech to British Columbia Legislative Assembly 1999 30 April. <http://www.fin.gov.bc.ca/archive/budget99/pdf/b99_sp.pdf>. Accessed 2007 08 April. 58. Sievers K and Klaukka T. Back Pain and Arthrosis in Finland - How Many Patients by the Year 2000. Acta Orthopaedica Scandinavica 1991;62:3-5. 59. Leino PI, Berg M A , and Puska P. Is Back Pain Increasing - Results from National Surveys in Finland During 1978/9-1992. Scandinavian Journal of Rheumatology 1994;23:269-76. 60. Manninen P, Riihimaki H , and Heliovaara M. Has musculoskeletal pain become less prevalent? Scandinavian Journal of Rheumatology 1996;25:37-41. 61. Heistaro S, Vartiainen E, Heliovaara M , and Puska P. Trends of back pain in eastern Finland, 1972-1992, in relation to socioeconomic status and behavioral risk factors. American Journal of Epidemiology 1998;148:671-82. 62. Mustard C, Cole D, Shannon H , Pole J, Sullivan T, and Allingham R. Declining trends in work-related morbidity and disability, 1993-1998: A comparison of survey estimates and compensation insurance claims. American Journal of Public Health 2003;93:1283-6. 63. Klein BJ, Radecki RT, Foris MP, Fell EI, and Hickey M E . Bridging the gap between science and practice in managing low back pain - A comprehensive spine care system in a health maintenance organization setting. Spine 2000;25:738-40. 64. Morse T, Dillon C, Warren N , Hall C, and Hovey D. Capture-recapture estimation of unreported work-related musculoskeletal disorders in Connecticut. American Journal of Industrial Medicine 2001;39:636-42. 118 65. Rosenman K D , Gardiner JC, Wang J et al. Why most workers with occupational repetitive trauma do not file for workers' compensation. Journal of Occupational and Environmental Medicine 2000;42:25-34. 66. DeCoster C, Carriere K C , Peterson S, Walld R, and MacWilliam L. Waiting times for surgical procedures. Medical Care 1999;37:JS187-JS205. 67. Alamgir H , Koehoorn M , Ostry A, Tompa E, and Demers PA. How many work-related injuries requiring hospitalization in British Columbia are claimed for workers' compensation? American Journal of Industrial Medicine 2006;49:443-51. 119 C H A P T E R 6 6.0 R E G I O N A L V A R I A T I O N S I N R A T E S O F L U M B A R D I S C E C T O M Y I N B R I T I S H C O L U M B I A 2 6.1 Known and Unknown Sources of Variation Surgical lumbar discectomy is considered a "high volume" hospital procedure in British Columbia (BC).1 Elsewhere in North America, this procedure accounts for approximately 80% of all spinal operations, and approximately 250,000 are performed annually in the US alone.2 In all regional variation studies, per capita rates of back surgery, including disc surgery, vary widely compared to other procedures such as surgery for hip fracture. For surgical procedures in general, regional variations in utilization rates can result from either random variation over time and place or systematic differences in the determinants of surgical treatment between areas. Factors that have been shown to be systematically associated with geographic variations can be categorized into one of three major groups: (1) patient characteristics—such as disease incidence (a function of demography and determinants of disease), patient preferences and decisions to contact the health care system (a function of sociocultural characteristics, income and insurance status); (2) health care supply characteristics—such as the number and accessibility of physicians and surgeons, and the available supply of hospital beds; and(3) physician discretion—or uncertainty in clinical decision making and the corresponding difficulty that physicians have in distinguishing their own preferences and values from those of patients. Studies show that standard supply and demand variables represented in groups (1) and (2) are associated with surgical rates, a A version of this chapter will be submitted for publication. Quon J, Levy A, Sobolev B, Fisher C, Kopec J, Schechter M. (2007) Regional Variations in the Rates of Lumbar Discectomy in British Columbia, Canada, 1990-2003. 120 however these variables usually account for only a small proportion of the variance in the rates within multivariable regression analyses.3 In Washington State, for example, a comprehensive list of explanatory variables accounted for less than 15% of variability in rates of surgery for back pain. While few studies have examined the effect of physician discretion, or group 3) variables, on regional variations directly, an extensive amount of literature has shown that physicians' personal beliefs and clinical uncertainty do result in differences in clinical decision-making.3 Furthermore, geographic variations in the use of operations are consistendy greater when the level of uncertainty about their value is high and least when it is low. Many investigators currendy believe that physician discretion and their perceptions of the value of the operation in question is probably the most important factor contributing to geographic variations.3 Lumbar discectomy, and back surgery in general, is a highly discretionary procedure, which is associated with correspondingly high variations in utilization rates between large and small areas.4"6 In BC, population-based rates of discectomy reportedly vary by eight-fold between local health areas,7 and in neighbouring Washington State—where access to hospital procedures is generally less restricted than in BC—approximately 15-fold variations in discectomy rates are observed between counties. According to Wennberg, conditions that result in low variation rates of utilization share certain characteristics: 1) the symptoms or dysfunction are of sufficient severity so that individuals will almost always contact the health care system; 2) they pose relatively litde diagnostic difficulty; and 3) physicians' treatment decisions are constrained by a firm consensus on correct treatment.8 Utilization rates for back operations are highly variable as the underlying condition (back pain) satisfies none of the above criteria. Flexibility in the interpretation of otherwise widely accepted indications for discectomy likely contributes to the observed variations in rates of back surgery, including discectomy, across 121 geographic regions.9'10 While physicians may agree on the general indications for discectomy, some indications (e.g. a progressive neurological deficit) may be defined differentiy by surgeons. Clinicians also disagree about the relative efficacy of other treatment options and therefore about how long to persist with nonoperative care. Furthermore, although complications from surgical discectomy are rare, they can be potentially serious, and up to 15% of patients require a second operation within three years.11 Individual surgeons may weigh the risk of these complications against the desired potential for symptom resolution differentiy. Similarly, patients may have different levels of aversion to risk (of complications) and preferences for surgery and other treatment options. To the extent that regional variations reflect the degree of clinical uncertainty, practice style, or physician discretion underlying surgical decision-making, it stands to reason that systematic changes in practice style or surgical decision-making could potentially manifest as changes in rate variations. Wennberg has shown that in the absence of targeted efforts to change physician behaviour, patterns of utilization (i.e. geographic variation in the rates) of a procedure are remarkably stable over time.8'12 However, other studies have shown that with minimal intervention—such as academic detailing—variations in the rates of back surgery can be reduced.13 6.1.1 Study Objectives In BC, no initiatives have been undertaken to target surgeon's practice patterns specifically, however rates of surgical discectomy have declined steadily since 1990, likely, although not necessarily as a result of health care regionalization14 and reductions in the numbers of short-stay beds within hospitals.15 The purpose of this study was to obtain descriptive information for a future in-depth study of the determinants of regional variations in back surgery rates in BC. The main objectives in the current study were to quantify the variation in, specifically, surgical lumbar discectomy rates in BC during 2000-2003, and to compare the variation in surgical lumbar 122 discectomy rates during this time period to that of a period when regionalization of health care was still just beginning, and the supply of short-stay beds in BC was less constrained (1990-1993). From an exploratory perspective, we hypothesized that the variation in surgical discectomy rates between health service delivery areas (HSDAs) in BC in 2000-2003 should be significandy lower than the variation in 1990-1993. We reasoned that in a climate of increasingly constrained resources, reductions in surgical discectomy rates during 2000-2003 should occur at the greater expense of patients with uncertain indications for surgery, which should, in turn, reduce the degree of physician (and patient) discretion underlying the surgical decision-making process and any corresponding variation in surgical rates during the more recent study period. 6.2 Methods 6.2.1 Data Sources/Patient Selection Medically necessary hospital services in BC are covered by universal health insurance. Al l BC residents have access to services within public hospitals without direct costs borne by patients. Private clinics have been licensed to perform surgical discectomies in BC since 2000, however to date only two facilities throughout the entire province (Cambie Surgery Centre and False Creek Surgical Centre) perform this procedure with any regularity. Information on all inpatient and day surgery separations in BC is available through the British Columbia Linked Health Database. This database is maintained by the University of British Columbia's (UBC) Centre for Health Services and Policy Research (CHSPR). 1 6 Originally, hospital records of all inpatient and day surgery procedures were searched between 1990 and 2003, however for the current study, only data between 1990 and 1993, and between 2000 and 2003 inclusive were used. 123 Cherkin et al previously developed an algorithm to identify surgical and non-surgical hospitalizations for low back problems in administrative databases using diagnosis and procedure codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). 1 7 Compared to ICD-9 codes without the clinical modification, ICD-9 C M diagnosis codes provide additional morbidity detail by including a second decimal specifying a more precise anatomic location or procedure. In BC, some hospitals do not record the second decimal. We therefore used a combination of single- and double-decimal ICD-9 codes. To record treatments, BC uses the Canadian Classification of Diagnostic and Therapeutic Procedures (CCP) surgical codes instead of ICD-9 C M procedure codes. We used a published algorithm from Ontario to match CCP codes to ICD-9 C M surgical codes.4;18 Given our focus on discectomy procedures, only diagnosis codes under the clinical category "herniated disc" were used. While the use of single-decimal codes allowed for inclusion of herniated disc cases that possibly included thoracic procedures, researchers in Ontario showed that the use of four-digit diagnosis codes yielded information very similar to the information extracted through Cherkin's original algorithm employing five-digit codes.18 Table 5.1 lists the ICD-9 and CCP codes that were used to identify patients hospitalized for an intervertebral disc disorder and corresponding discectomy. Laminectomy ("exploration and decompression of spinal canal") is a procedure that is sometimes administered for spinal stenosis, however it is also used interchangeably with discectomy in administrative databases.19 Appendix 2 lists the diagnosis codes that were used to exclude patients with nonmechanical back pain, known cervical or thoracic disorders, and/or spinal fusion procedures. The record selection algorithm employing these codes is further described in Figure 5.1. A previous study of lumbar discectomies using hospital discharge data found that the principal diagnosis and procedure were reliably coded in 96% of abstracts.20 124 Patients had to be residents of BC and have both a principal or primary diagnosis of herniated disc, and a procedure code for either disc excision, chemonucleolysis or "other destruction of intervertebral disc". In the hospital separations file, a principal diagnosis—indicated by an " M " code in the CHSPR DIAGNOSIS T Y P E field—represents the ICD-9 diagnosis most responsible for a patient's hospital stay. A primary diagnosis—indicated by a "1" code in the DIAGNOSIS T Y P E field—represents a comorbid condition existing before admission and which usually has a significant influence on the patient's length of stay. For each patient only the first hospitalization was included. We focused on the largest homogenous group of back surgery (discectomy) patients and therefore excluded occurrences of spinal fusion regardless of anatomic reference and order of appearance within the same discharge abstract. Furthermore, to limit the study population to mechanical back pain patients, those with diagnosis codes for nonmechanical causes of back pain including neoplasm, spinal infection, pregnancy, inflammatory spondyloarthropathy, congenital anomaly, acute trauma or dislocation and spinal fracture were excluded. Only patients aged 20 to 64 years at the time of surgery were included. In a previous study using hospital discharge data in BC, this age group accounted for 91% of the total number of discectomies over a 14 year period. Furthermore, patients older than 65 years are more likely to suffer from sciatica due to degenerative disc disease and/or concomitant spinal stenosis, the surgical indications for which are more discretionary than for acute disc herniation without age-related degenerative changes. 6.2.2 Small Area Definition Health service delivery area (HSDA) was the area unit of analysis in this study. In BC, health services are managed and administered by six major health authorities, five of which are 125 geographically distinct. Each of the five geographically distinct authorities is divided into three or four HSDAs that are also geographically distinct and capture the natural patterns of referrals for health services within each larger authority. Patients were assigned to HSDAs according to forward sortation area (FSA) of residence. 6.2.3 Data Analysis Age was categorized primarily by decade into the following groups: 20-29, 30-39, 40-49 and 50-64 years. Four-year age-specific rates of discectomy were calculated for women and men combined. The numerators for these rates were the number of age- and sex-specific discharges in each region, and the denominators were the corresponding age- and sex-specific population counts in BC during the periods 1990-1993 and 2000 to 2003. The denominator counts were based on 1986, 1991, 1996 and 2001 censuses, inter-censal midyear estimates and post-censal midyear projections. Direct standardized rates of discectomy for each region were calculated by applying the age-specific rates to the 1996 Census-estimated population of BC. 1 1 Age- and sex-standardized four-year discectomy rates were calculated for each HSDA and BC overall, for both time periods. 6.2.4 Statistical Methods For descriptive purposes, the extremal quotient (EQ), coefficient of variation (CV, weighted and unweighted), and systematic component of variation were calculated for each period. The E Q is the ratio of the highest to lowest standardized discectomy rate. The CV is the ratio of the standard deviation of 16 HSDA rates to the mean rate, weighted by the population in each area (to account for unequal population sizes), and multiplied by 100. The SCV estimates the variance of the rates among HSDAs that cannot be accounted for by the variability within each HSDA. As a matter of convention, the SCV is multiplied by 1000.21 126 For the primary analysis, we tested for an overall difference in the four-year regional rates during 2000-2003, for women and men combined, using the T 2 statistic.22 Carriere and Roos originally presented a method for comparing standardized rates that does not rely on parametric assumptions such as the sample variance to estimate the variance of rates.22 If h — 1,2, H - l , and H , where H is the total the number of regions being compared, then the T 2 statistic is calculated by taking the difference between the standardized rate in region h (Rj) and the standard population rate (R0), squaring the difference, dividing by the variance of the standardized rate in region h, then summing up this resulting quotient over H regions. This statistic is described in further detail in Appendix 3. Given the low procedure rates in some regions (less than 50 per 100,000), a logarithmic transformation of the rates was used to correct for skewness in the data.23 The T 2 statistic has a chi-square distribution with degrees of freedom equal to one minus the number of groups being compared. Provided that the T 2 test was significant at p < 0.05, "overall" 95% confidence intervals for the rate of each region were examined to identify individual service delivery areas that differed significantly from the provincial mean rate. These overall confidence intervals were adjusted for multiple comparisons using the Bonferroni method. For the primary analysis alpha was divided by 16 (the total number of HSDAs), and the confidence intervals for the four-year rates were adjusted conservatively to control the overall significance level at 5%. In the secondary analysis, the variation in discectomy rates between HSDAs during 2000-2003 was compared to the variation in rates between areas during 1990-1993. To test for a significant difference in variation between the two time periods, the T 2 statistic measuring the variation in rates during 2000-2003 was divided by the T 2 statistic representing the variation in 127 1990-1993. This ratio was then compared to an F-distribution with 15-numerator and 15-denominatorb degrees of freedom.22;23 6.3 Results The names and identification codes of the 16 geographically distinct HSDAs in BC and the smaller local health areas they encompass are described in Table 6.3. Annualized population counts of each HSDA, for the period 2000 to 2003, appear in Table 6.4. From 2000 to 2003 there were 10,206,243 person-years of observation and 3,669 primary discectomy procedures that had been performed in public hospitals. The bar chart in Figure 6.2 depicts the standardized four-year rates in each service delivery area during 2000-2003. For context, the four-year rates for 1990-1993 are also depicted. Table 6.5 shows descriptive measures of the regional variation in discectomy rates in BC for both time periods. The overall standardized surgical lumbar discectomy rate during 2000-2003 was 36.0 per 100,000, down from 67.9 per 100,000 during 1990-1993. In both time periods, the area with the highest surgical lumbar discectomy rate was Thompson Cariboo Shuswap (HSDA 14) whereas the lowest rate area was Richmond (HSDA 31). All conventional descriptive measures of variation (except for the standard deviation of the rates) indicated a slight increase from 1990-1993 to 2000-2003. The E Q indicated a 4.7-fold variation in rates during 2000-2003, up by 38% from a 3.4 fold variation in 1990-1993. Similarly, the CV and SCV statistics were 37% and 79% larger, respectively, in 2000-2003 compared to 1990-1993. Table 6.6 lists the standardized discectomy rate and adjusted confidence interval for each HSDA for women and men combined. These confidence limits have a coverage of 99.7% due to the inclusion of a Bonferroni correction for multiple comparisons, which controlled the h 15 equals the total number of regions minus 1. 128 overall significance level in comparisons between each of the 16 regions and the provincial mean at p < 0.05. The T 2 statistic summarizing the regional variation in rates during each time period was highly significant (p < 0.0001) indicating simply that the rate of surgical lumbar discectomy was not the same in all regions in 2000-2003 or 1990-1993. Focusing on the four-year rates during 2000-2003, the standardized surgery rates were highest in Thompson Cariboo Shuswap (70.2 [99.7% CI: 59.9-82.4] per 100,000) followed closely by Northern Interior (66.5 [99.7% CI: 55.0-80.3] per 100,000). The area with the lowest rate in this period was Richmond (15.0 [99.7% CI: 10.4-21.7] per 100,000) with North Vancouver Island showing the second-lowest rate (16.8 [99.7% CI: 10.7-26.4] per 100,000). Figure 6.3 shows in graphical form the standardized discectomy rate and overall confidence interval for each region during 2000-2003. In this figure, HSDA 66 represents the province of BC (and the provincial mean rate). Inspection of the confidence limits for the rate of each HSDA indicated that rates were significantly higher than the provincial mean rate in four of 16 service delivery areas: HSDAs 14-Thompson Cariboo Shuswap, 52-Northern Interior, 41-South Vancouver Island, and 33-North Shore/Coast Garibaldi. Three regions had discectomy rates that were significandy lower than the provincial mean: HSDAs 31-Richmond, 43-North Vancouver Island, and 32-Vancouver. To compare the variation in rates between HDSAs in 2000-2003 to the variation in 1990-1993, Carriere's T 2 statistic was also calculated for the rates from 1990-1993 (bottom of Table 6.6). As with the previous descriptive measures of variation, a qualitative comparison of the T 2 statistic from both time periods suggested that the regional variation in discectomy rates was greater during 2000-2003 than 1990-1993. Carriere showed that the ratio of two T 2 statistics will follow an F-distribution with numerator and denominator degrees of freedom equal to the degrees 129 of freedom for the respective T 2 statistics, and that this test is appropriate for comparing the relative variation in rates for two independent groups, no matter how constituted.22 In studies of nonrecurrent procedures (such as the current study), patients in separate time periods are different individuals and, therefore, represent sufficiendy independent events.23 In the current study, the F-test for a difference between the T 2 statistics from each time period was clearly not significant (p = 0.4). This indicated, at the very least, that the regional variation in rates had not diminished between 1990-1993 and 2000-2003. 6.4 Discussion In this study, we found statistically significant variations in the rate of discectomy for back pain among health service delivery areas in the province of BC, Canada. The Thompson-Cariboo-Shuswap area accounted for the highest four-year utilization rate in 2000-2003. North Shore/Coast Garibaldi and South Vancouver Island areas also had rates that were significandy higher than the provincial rate in this period. At the other end of the spectrum, Richmond had the lowest utilization rate, however North Vancouver Island and Vancouver were also significandy below the provincial mean. We found that the rate of surgical lumbar discectomy across 16 different HSDAs, in BC during 2000-2003 varied by 4.7-fold. Recendy, variations in surgical discectomy rates of almost 8-fold have been reported between hospital referral regions (HRRs) in the US, however that study was of Medicare enrollees, who are aged 65 years or older.24 On the other hand, in that same study, the reported CV for lumbar discectomy was 34.6, which is very similar to our estimated C V of 38.0 in BC. To our knowledge, the variation in rates of lumbar discectomy in BC at the health service delivery area level has not been previously described. HSDAs represent geographic areas in BC that capture the natural patterns of referrals for health services throughout the province, and 130 within which all core medical services—including orthopaedic surgical procedures—are available. Wing previously reported that population-based rates of discectomy operations (not limited to the lumbar spine) performed in 1994 in BC varied between the extremes of 21 per 100,000 to 180 per 100,000.7 Those figures apparently reflected the variation in rates between local health areas (LHAs), which are smaller than HSDAs and generally follow local political and natural boundaries. The rates from smaller areas are also likely unstable compared to rates based on larger HSDAs. We could find no other published reports of the geographic variation in per capita use rates of surgical lumbar discectomy or other surgical procedures in BC to compare our figures to. However compared to published EQ's for various procedures from Ontario25 our estimate of the rate variation for lumbar discectomy at the HSDA level in BC (EQ = 4.7 during 2000-2003) was greater than that of hip replacement (EQ = 1.6), cholecystectomy (EQ = 1.9), knee replacement (EQ = 2.4), hysterectomy (EQ = 2.5) and hospitalization for asthma (EQ = 3.2) at the county level in Ontario. Procedures with higher EQ's in Ontario were carotid endarterectomy (EQ = 9.1), tympanostomy tube procedures (EQ = 9.6) and routine circumcision (EQ = 11.9). Rankings of these procedures based on their weighted CVs suggested that the variation in lumbar discectomy rates in BC (CV = 38.0) was greater than that of all procedures in Ontario except for routine circumcision (CV = 44.9), whereas rankings based on SCV statistics were identical to those based on the E Q statistics. The results of this descriptive comparison are consistent with those of other reports that disc operations (discectomy and/or laminectomy) represent a high variation procedure compared to low variation procedures such as hip replacement or operations for hip fracture.25 Conventional descriptive measures of variation (the E Q , C V and SCV) must be interpreted cautiously. Diehr and co-workers described the E Q as the most intuitively satisfying measure of small-area variation}'' however, they did not recommend its use because of its poor 131 statistical properties. First, because no tables of distributions are available, doing hypothesis tests or constructing confidence intervals is difficult without simulation. In simulations using actual county sizes from Washington State and a fixed hypothetical procedure rate of 50 per 100,000 population in each county, the E Q could be as high as 11 without exceeding the 95 th percentile of the null distribution.26 Second, the E Q is unstable when denominators are low and infinite if the smallest rate is zero (which is possible if the rate of an event is rare and the areas for analysis have small populations). Third, the E Q combines both within-area and across-area variation. That is, extreme rates tend to occur for uncommon events and in areas with small populations, and this random variation leads to spuriously high values. Similar limitations apply to the C V . 2 7 The distribution of this statistic is unknown, and therefore hypothesis testing is not possible without simulation. Even with simulation, the 95 th percentile of the distribution of the C V decreases as the average rate of surgery increases between regions increases. In this study of nonrecurrent events, a 2 x k chi-square statistic utilizing the observed counts (and crude discectomy rates) may have been an appropriate test for differences across regions as well, 2 2 ' 2 6 however, with the T 2 method, the test statistic does not depend on they underlying unknown distribution of the events and does not require restrictive assumptions such as equal variances among the competing rates.22 Also, the variance estimates and confidence intervals generated can be calculated using the adjusted rates rather than relying on the unadjusted rates, and the T 2 statistic allows for correction for highly skewed data and can therefore be applied to very low incidence rates (of approximately 50 per 100,000 or less) whereas other suggested measures, such as the systematic component of variation (SCV) and the coefficient of variation from A N C O V A (CVA), may not be appropriate for low incidence rates. In the current study, our estimation of higher E Q , CV and SCV statistics during 2000-2003 could have resulted simply from the lower "incidence" of surgery during the latter period. 132 We included the SCV as a descripdve measure in the current study. This statistic was originally proposed by McPherson and Wennberg,21 however it too takes on very large values even when there is no underlying variation among the small areas.28 Like the CV, it becomes larger as the mean rate becomes smaller, and is therefore inappropriate for comparing the variation between groups with different mean rates. In contrast, Carriere's T 2 statistic is robust to differences in the underlying rates of different procedures. We did not find evidence of a reduction in the regional variation in lumbar discectomy rates in 2000-2003 compared to 1990-1993. We therefore found no evidence in support of our exploratory hypothesis that a reduction in the numbers of short-stay beds (an important supply variable), particularly after 1996, was associated with less discretionary use of lumbar discectomy in BC. This finding supports Wennberg's "surgical signature" hypothesis, which specifies that utilization rates for individual procedures may vary considerably between regions, yet differences between regions will be relatively stable over time.8 However, to test this hypothesis more direcdy, a future in-depth study is needed to examine the relationship between a broad range of health service supply variables on the one hand and regional variation in the rates of lumbar discectomy on the other hand. Such a study would be possible using administrative database resources already in existence. Another valuable study would involve measuring surgeons' preferences and other determinants of practice style directly in response to changes in health services supply variables, even if only hypothetically in the context of self-administered questionnaires. 6.4.1 Strengths and Limitations In the current study we had access to unique study identification numbers which allowed us identify individual patients and, therefore, to focus on rates of "first" (or primary) lumbar discectomy. Also, as health insurance is universal in Canada, the province-wide hospital separations data that we used included admissions for surgical lumbar discectomy in public 133 hospitals for the entire population of BC. Our data did not include patients seen in private clinics. Lumbar discectomy was accessible in only two private clinics in all of BC beginning in the year 2000. At the time of study inception, no reliable data were available on the numbers patients seen in private facilities. While our study still provides a useful picture of the regional variation in rates of lumbar discectomies performed in public hospitals, a valuable future study will involve the development of a system of surveillance to track the numbers of lumbar discectomies in BC that are performed in private clinics at WorkSafeBC (formerly the Workers' Compensation Board of BC), whose beneficiaries account for the majority of discectomies performed in those clinics. Given that WCB-funded surgeries are ultimately paid for by the public system, our study of patients treated in public hospitals inevitably underestimates regional discectomy rates, but not necessarily the variation in these rates. Finally, in a climate of constrained health care resources, there is a need for allocating existing resources efficiendy. Jin and colleagues have emphasized that, "If efficiency is important,, efforts (at improving efficiency in the health care system) must be directed toward common diseases that have a large economic burden and where practice variation rather than case mix or severity drives the utilization of resources."29 Back pain is a common and expensive problem in industrialized nations. In BC, back pain accounts for more than 680,000 lost work days and approximately $122 million in Workers' Compensation costs per year, and therefore accounts for 25% of both the total number of claims and the total number of lost work days attributable to all work-related injuries in the province per year. In BC, the variation in rates of surgical discectomy for painful disc herniation in the low back is significant between core service delivery areas. The results of this study offer a first look at the current state of rate variations for a commonly performed procedure in BC, and an important baseline for future comparisons. 134 Table 6.1: Inclusion ICD-9 and CCP Codes Cases were included if they had one ICD-9 code indicating herniated disc as either a "principal" (diagnosis type = "M") or a "primary" (diagnosis type = "1") diagnosis, and one CCP code in any procedure field indicating discectomy, chemical discectomy, or laminectomy. ICD-9 and CCP Inclusion Codes ICD-9 codes for clinical category "herniated disc (principal or primary diagnosis) 722.1 Displacement of thoracic or lumbar disc without myelopathy 722.10 Displacement of lumbar disc without myelopathy 722.2 Displacement of unspecified disc without myelopathy 722.7 Disc disorder with myelopathy 722.70 Disc disorder with myelopathy, site unspecified 722.73 Lumbar disc disorder with myelopathy CCP codes for discectomy and laminectomy (any field) 16.0 Exploration and decompression of spinal canal 89.08 Sequestrectomy other site 89.09 Sequestrectomy unspecified site 92.31 Excision of intervertebral disc 92.33 Intervertebral chemonucleolysis 92.34 Other destruction of intervertebral disc 135 Table 6.2: ICD-9 and CCP Exclusion Codes Exclusion codes for diagnoses or procedures for non-mechanical back pain in any field, spinal fusions, repeat procedures, known cervical or thoracic principal diagnoses, and principal or primary diagnoses of back pain due to traumatic injury. ICD-9 and CCP Exclusion Codes ICD-9 codes for "herniated disc" in cervical or thoracic region (as principal diagnosis) 722.71 Cervical disc disorder with myelopathy 722.72 Thoracic disc disorder with myelopathy CCP code for repeat laminectomy 16.02 Re-opening of laminectomy site ICD-9 codes for diagnoses of non-mechanical back pain (in any diagnosis field) 140-239.9 Neoplasms 324.1 Intraspinal infection 630-677 Complications related to pregnancy 720* Inflammatory disease 730* Osteomyelitis 805.1, 805.3, 805.5, 805.7, 805.9 Open vertebral fractures without spinal cord injury 806* Fractures of vertebral column with spinal cord injury 839-839.5 Vertebral dislocations E800-E849.9 Motor vehicular accidents (i.e. trauma-related back pain) CCP codes for procedure involving non-mechanical back pain (in any procedure field) 16.2 Chordotomy ICD-9 codes for non-mechanical back pain (as principal or primary diagnosis) 733.1 Pathological fracture 805.0, 805.2, 805.4, 805.6, 805.8 Closed vertebral fracture without spinal cord injury 136 ICD-9 and CCP Exclusion Codes ICD-9 codes for mechanical pain in cervical or thoracic region (as principal diagnosis) 353.2 Cervical root lesions, not elsewhere classified 353.3 Thoracic root lesions, not elsewhere classified 721.0 Cervical spondylosis, without myelopathy 721.1 Cervical spondylosis, with myelopathy 721.2 Thoracic spondylosis, without myelopathy 721.41 Thoracic or lumbar spondylosis with myelopathy (thoracic region) 722.0 Displacement of cervical intervertebral disc without myelopathy 722.11 Displacement of thoracic intervertebral disc without myelopathy 722.4 Degeneration of cervical intervertebral disc 722.71 Intervertebral disc disorder with myelopathy-cervical region 722.72 Intervertebral disc disorder with myelopathy-thoracic region 722.81 Postlaminectomy syndrome-cervical region 722.82 Postlaminectomy syndrome-thoracic region 723.0 Spinal stenosis, other than cervical-thoracic region 723.4 Brachial neuritis or radiculitis NOS 724.01 Spinal stenosis, other than cervical-thoracic region*/ CCP codes for procedures involving spinal fusion (in any procedure code field) 93.0* Spinal fusion listed before or in absence of listing for discectomy 137 Figure 6.1: Patient Selection Algorithm Examine all hospital separations during fiscal years 1990/91-2003/04 Include only residents of BC at dme of admission Exclude patients with incomplete age, sex or residency information Include cases meeting ICD-9 and CCP coding criteria described in Table 1 Exclude cases using ICD-9 and CCP coding criteria described in Table 2 Include only patients between 20-79 years of age If more than one hospitalization per patient, include only the first hospitalization 138 Table 6.3: Health Service Delivery Areas in B C Health service delivery areas and related local health areas and health authorities (HA), H A Name (Code) Health Service Delivery Area Local Health Area Names (Code) Code Name Interior 11 East Kootenay Fernie (001), Cranbrook (002), Kimberley (003), Windermere (004), Creston (005), Golden (018) Interior 12 Kootenay Boundary Kootenay Lake (006), Nelson (007), Casdegar (009), Arrow Lakes (010), Trail (011), Grand Forks (012), Ketde Valley (013) Interior 13 Okanagan Southern Okanagan (014), Penricton (015), Keremeos (016), Princeton (017), Armstrong-Spallumcheen (021), Vernon (022), Central Okanagan (023), Summerland (077), Enderby (078) Interior 14 Thompson Cariboo Shuswap Revelstoke (019), Salmon Arm (020), Kamloops (024), 100 Mile House (025), North Thompson (026), Cariboo-Chilcotin (027), Lillooet (029), South Cariboo (030), Merritt (031) Fraser 21 Fraser East Hope (032), Chilliwack (033), Abbotsford (034), Mission (075), Agassiz-Harrison (076) Fraser 22 Fraser North New Westminster (040), Burnaby (041), Maple Ridge (042), Coquidam (043) Fraser 23 Fraser South Langley (035), Surrey (201), South Surrey/White Rock (202), Delta (037) Vancouver Coastal 31 Richmond Richmond (038) Vancouver Coastal 32 Vancouver Vancouver-City Centre (161), Vancouver-Downtown Eastside (162), Vancouver-North East (163), Vancouver-Westside (164), Vancouver-Midtown (165), Vancouver-South (166) Vancouver Coastal 33 North Shore/Coast Garibaldi North Vancouver (044), West Vancouver-Bowen Island (045), Sunshine Coast (046), Powell River (047), Howe Sound (048), Bella Coola Valley (049), Central Coast (083) Vancouver Island 41 South Vancouver Island Greater Victoria (061), Sooke (062), Saanich (063), Gulf Islands (064) Vancouver Island 42 Central Vancouver Island Cowichan (065), Lake Cowichan (066), Ladysmith (067), Nanaimo (068), Qualicum (069), Alberni (070) Vancouver Island 43 North Vancouver Island Courtenay (071), Campbell River (072), Vancouver Island West (084), Vancouver Island North (085) Northern 51 Northwest Queen Charlotte (050), Snow Country (051), Prince Rupert (052), Upper Skeena (053), Smithers (054), Kitimat (080), Stikine (087), Terrace (088), Nisga'a (092), Telegraph Creek (094) Northern 52 Northern Interior Quesnel (028), Burns Lake (055), Nechako (056), Prince George (057), Northern 53 Northeast Peace River South (059), Peace River North (060), Fort Nelson (081) 139 Table 6.4: Annualized Area Population Counts Annualized population size of Health Service Delivery Areas in BC, Canada, during the period 2000/01-2003/04 Health Service Delivery Area Women Population Count Men Total East Kootenay 23,678 24,213 47,891 Kootenay Boundary 23,620 24,281 47,901 Okanagan 91,723 86,938 178,661 Thompson Cariboo Shuswap 65,292 65,140 130,432 Fraser East 71,024 72,024 143,048 Fraser North 175,790 175,814 351,603 Fraser South 182,554 187,403 369,957 Richmond 58,366 54,989 113,354 Vancouver 199,615 200,151 399,766 North Shore/Coast Garibaldi 85,149 83,023 168,172 South Vancouver Island 107,278 101,910 209,187 Central Vancouver Island 71,095 68,867 139,962 North Vancouver Island 34,262 34,896 69,157 Northwest 24,083 26,663 50,746 Northern Interior 45,334 48,067 93,401 Northeast 18,253 20,070 38,323 British Columbia (Total) 1,277,115 1,274,446 2,551,561 140 is a LO I—A ro to to to LO LO DO LO LO 4 i ro J* L h ro LH LO W Q LO 4i ro ro ro ro LO ro LO LO ro Ji LO LH ro m LO w Q W CO FJ "2 ^ * r-i-<D O O o o o cr g o rt ° g « H* SI 8 2 £ o East Kootenay (Interior) Kootenay Boundary (Interior) Okanagan (Interior) Thompson Cariboo Shuswap (Interior) Fraser East (Fraser) Fraser North (Fraser) Fraser South (Fraser) Richmond (Vancouver Coastal) Vancouver (Vancouver Coastal) North Shore/Coast Garabaldi (Vancouver Coastal) South Vancouver Island (Vancouver Island) Central Vancouver Island (Vancouver Island) North Vancouver Island (Vancouver Island) Northwest (Northern) Northern Interior (Northern) Northeast (Northern) British Columbia fD O > CO fD I V) a x •> Q . c" </> <•+ fD a Ti a fD c CT 0) O fD O O rt-rt O a. B < bd rt P H rt P ce Er-rs p p C B-o tl. o H 3 rt 13 p a. o 3 o o o I 5' rt a. cr rt o Table 6.5: Descriptive Measures of Variation Descriptive statistics summarizing the regional variation in four-year discectomy rates in BC, 2000-2003. Description 1990-1993 2000-2003 No. of discectomies BC Total 5,595 3,669 Lowest count 72 45 Highest count 772 496 Crude rate BC Overall 67.1 35.9 Lowest rate 38.5 14.6 Highest rate 128.4 69.2 Standardized rate BC Overall 67.9 36.2 Lowest rate 37.4 15.0 Highest rate 128.1 70.2 Standard deviation (of standardized rates) Unweighted 21.9 15.3 Weighted* 18.9 13.8 Extremal Quotient 3.4 4.7 Coefficient of Variation Unweighted 32.8 40.0 Weighted* 27.8 38.0 Systematic Component of Variation 96.9 173.7 Calculations for these statistics were weighted by the population in each health service delivery area. 142 Table 6.6: Standardized Discectomy Rates Direct age-standardized rates of discectomy, per 100,000 person-years, for each HSDA, by time period, all women and men combined. Health Service Delivery Area 1990-1993 2000-2003 Rate L C L U C L Rate L C L U C L Women and Men Combined East Kootenay 46.0 32.7 63.7 37.5 25.9 54.4 Kootenay Boundary 68.5 51.8 90.1 31.9 21.3 47.6 Okanagan 68.7 59.4 80.4 40.7 34.0 48.7 Thompson Cariboo Shuswap 128.2 113.2 145.1 70.2 59.9 82.4 Fraser East 68.7 57.6 81.5 33.6 27.2 41.7 Fraser North 73.5 66.0 81.8 34.8 30.4 39.8 Fraser South 66.3 59.4 74.1 31.9 27.8 36.6 Richmond 45.2 35.8 58.1 15.0 10.4 21.7 Vancouver 52.7 46.7 58.9 20.1 17.0 23.7 North Shore/Coast Garibaldi 88.2 77.7 101.1 46.9 39.5 55.5 South Vancouver Island 61.1 53.3 70.8 48.0 41.2 55.8 Central Vancouver Island 58.4 48.8 70.1 37.1 30.0 45.9 North Vancouver Island 37.8 27.5 50.8 16.8 10.7 26.4 Northwest 63.0 47.2 80.3 34.7 24.3 49.5 Northern Interior 91.5 76.5 107.7 66.5 55.0 80.3 Northeast 52.7 36.3 73.9 45.9 32.3 65.3 British Columbia 68.0 65.3 70.7 36.0 34.3 37.8 T 2 statistic (p value) 426.30 ( < 0.0001) 500.46 (< 0.0001) T 2 ratio (p value)f 1.174 (0.38) *LCL and U C L = lower and upper 99.7% confidence limits controlling overall significance level (for 16 comparisons) at 5% for each time period. fP value based on F-test with 15 numerator and denominator degrees of freedom 143 Figure 6.3: Dotplot of Area Rates Dotplot of discectomy rates and adjusted confidence intervals for 16 health service delivery areas, B C , 2000/01 - 2003/04.* H S D A Number Thompson Car iboo Shuswap Northern Interior •South Vancouver Island •North Shore Coast Garabaldi Northeast Okanagan East Kootenay , •Central Vancouver Island British Columbia Fraser North Northwest Fraser East Fraser South Kootenay Boundary Vancouver North Vancouver Island Richmond 0.1 0.2 0.3 0.4 0.5 0.6 Lumbar Discectomy Rate per 1000 Population 0.7 0.8 *Rates are per 1000 population. Confidence intervals are adjusted to 99.7% to control overall significance of 16 comparisons at 5%. Vertical line represents provincial mean rate. 5 6.5 References 1. Wright CJ, Chambers G K , and Robens-Paradise Y. Evaluation of indications for and outcomes of elective surgery. Canadian Medical Association Journal 2002;167:461-6. 2. Andersson GBJ and Weinstein JN. Disc herniation. Spine 1996;21:S1. 3. Leape LL . Unnecessary Surgery. Annual Review of Public Health 1992;13:363-83. 4. Taylor V M , Anderson G M , McNeney B et al. Hospitalizations for back and neck problems: A comparison between the Province of Ontario and Washington State. Health Services Research 1998;33:929-45. 5. Deyo RA. Back Surgery - Who Needs It? New England Journal of Medicine 2007;356:2239-43. 6. Keller RB, Adas SJ, Singer D E et al. The Maine Lumbar Spine Study: 1. Background and concepts. Spine 1996;21:1769-76. 7. Wing PC. Rheumatology: 13. Minimizing disability in patients with low-back pain. CMAJ 2001;164:1459-68. 8. Wennberg JE, Barnes BA, and Zubkoff M . Professional Uncertainty and the Problem of Supplier-Induced Demand. Social Science and Medicine 1982;16:811-24. 9. Cherkin DC, Deyo RA, Loeser JD, Bush T, and Waddell G. An International Comparison of Back Surgery Rates. Spine 1994;19:1201-6. 145 10. Volinn E, Mayer J, Diehr P, Vankoevering D, Connell FA, and Loeser JD. Small Area Analysis of Surgery for Low-Back-Pain. Spine 1992;17:575-81. 11. Volinn E, Turczyn K M , and Loeser JD. Patterns in Low-Back-Pain Hospitalizations -Implications for the Treatment of Low-Back-Pain in An Era of Health-Care Reform. Clinical Journal of Pain 1994;10:64-70. 12. Wennberg JE, Bunker JP, and Barnes B. The Need for Assessing the Outcome of Common Medical Practices. Annual Review of Public Health 1980;1:277-95. 13. Keller RB, Atlas SJ, Soule D N , Singer D E , and Deyo RA. Relationship between rates and outcomes of operative treatment for lumbar disc herniation and spinal stenosis. Journal of Bone and Joint Surgery-American Volume 1999;81A:752-62. 14. Frankish CJ, Kwan B, Ratner PA, Higgins JW, and Larsen C. Social and political factors influencing the functioning of regional health boards in British Columbia (Canada). Health Policy 2002;61:125-51. 15. Sheps SB, Reid RJ, Barer M L et al. Hospital downsizing and trends in health care use among elderly people in British Columbia. Canadian Medical Association Journal 2000;163:397-401. 16. Chamberlayne R, Green B, Barer ML, Hertzman C, Lawrence WJ, and Sheps SB. Creating a population-based linked health database: A new resource for health services research. Canadian Journal of Public Health-Revue Canadienne de Sante Publique 1998;89:270-3. 17. Cherkin DC, Deyo RA, Volinn E, and Loeser JD. Use of the International Classification of Diseases (Icd-9-Cm) to Identify Hospitalizations for Mechanical Low-Back Problems in Administrative Databases. Spine 1992;17:817-25. 146 18. Hu RW, Jaglal S, Axcell T, and Anderson G. A population-based study of reoperations after back surgery. Spine 1997;22:2265-70. 19. Lavis JN , Maker A , Anderson G M et al. Trends in hospital use for mechanical neck and back problems in Ontario and the United States: discretionary care in different health care systems. Canadian Medical Association Journal 1998;158:29-36. 20. Ramirez LF and Thisted R. Using A National-Health Care Data-Base to Determine Surgical Complications in Community Hospitals - Lumbar Discectomy As An Example. Neurosurgery 1989;25:218-25. 21. Mcpherson K , Strong P M , Epstein A, and Jones L. Regional Variations in the Use of Common Surgical-Procedures - Within and Between England and Wales, Canada and the United-States-Of-America. Social Science & Medicine Part A-Medical Sociology 1981;15:273-88. 22. Carriere K C and Roos LL. Comparing Standardized Rates of Events. American Journal of Epidemiology 1994;140:472-82. 23. Carriere K C and Roos LL . A method of comparison for standardized rates of low-incidence events. Medical Care 1997;35:57-69. 24. Weinstein JN, Lurie JD, Olson PR, Bronner K K , and Fisher ES. United States' trends and regional variations in lumbar spine surgery: 1992-2003. Spine 2006;31:2707-14. 25. Coyte PC, Croxford R, Asche CV, To T, Feldman W, and Friedberg J. Physician and population determinants of rates of middle-ear surgery in Ontario. Journal of the American Medical Association 2001;286:2128-35. 147 26. Diehr P, Cain K, Connell F, and Volinn E. What Is Too Much Variation - the Null Hypothesis in Small-Area Analysis. Health Services Research 1990;24:741-71. 27. Coory M and Gibberd R. New measures for reporting the magnitude of small-area variation in rates. Statistics in Medicine 1998;17:2625-34. 28. Diehr P and Grembowski D. A Small Area Simulation Approach to Determining Excess Variation in Dental Procedure Rates. American Journal of Public Health 1990;80:1343-8. 29. Jin Y, Marrie TJ, Carriere K C et al. Variation in management of community-acquired pneumonia requiring admission to Alberta, Canada hospitals. Epidemiology and Infection 2003;130:41-51. 148 CHAPTER 7 7.0 DETERMINANTS OF WAITING T I M E FOR LUMBAR DISCECTOMY 3 7.1 Introduction Back pain is one of the most common health problems in the industrialized world. Between 65% and 80% of individuals experience at least one episode during adulthood.1"3 It is the most common cause of disability in persons younger than 45 years and the second most common cause of disability in persons 45 to 65 years of age.4 Lumbar disc herniation (LDH) can result in a particularly disabling form of back pain that interferes with a person's ability to function at home and at work.5 Surgical discectomy—the operative procedure for LDH—is effective and likely results in quicker recovery compared to non-surgical treatment.6 As a result, it is one of the most common surgical operations in North America.7;8 However, L D H often improves with conservative therapy alone so the decision to operate is a discretionary one depending on the "practice style" of individual surgeons and preferences by individual patients. In British Columbia (BC), surgical discectomies are performed throughout a network of specialty hospitals. In our study of the temporal trend in surgical discectomy rates in BC, we estimated that in 2003 alone, 24 surgeons billed "more than occasionally" (i.e. at least once a month on average) for a discectomy procedure in the back. This particular group of surgeons a A version of this chapter will be submitted for publication. Quon J, Levy A, Sobolev B, Fisher C, Kopec J, Schechter M. (2007) Determinants of Waiting Time for Lumbar Discectomy in BC, Canada. 149 performed a total of 821 such procedures in ten different public hospitals in 2003. That same year, the crude discectomy rate among adults aged 20 to 79 years was 22.6 per 100,000 and 30.5 per 100.000 person-years for women and men, respecdvely. Access to surgical discectomy is managed through wait lists with patients prioritized according to clinical severity. While sociodemographic and institutional determinants have been associated with the length of waiting times for other elective procedures, the impact of these variables on waiting times for surgical discectomy specifically are not known. Only one Canadian study has reported data on access to surgical lumbar discectomy.9 However, in that retrospective study, only patients who received the operation were sampled whereas wait-listed patients who failed to get surgery were not. Thus the mean and median waiting times may have been underestimated.10 The interpretation was also hindered by the fact that administrative data precluded an assessment of the impact of health status at presentation and other important clinical variables on access. In a Swedish study, preoperative quality of life was not independently associated with waiting times for three groups of orthopaedic procedures, including back surgery.11 However, that study used a generic instrument (EuroQol questionnaire [EQ-5D]) to measure health status, in which case, the relationship between a disease-specific measure of back pain severity and waiting times might have been different.12 Employment status (employed versus unemployed) and hospital type (county versus university) were strongly associated with shorter waits, suggesting that access to back surgery is determined by factors other than clinical severity. 7.1.1 Compensation Status and Chronic Pain L D H is one of the most frequent reasons for disability compensation claims in working individuals.13 While most compensation claimants do not require surgical treatment, it is of 150 interest to know whether those who do undergo surgery have equitable access to services compared to patients not receiving compensation. The notion that receiving compensation may be associated with any barrier to health care access is based on the longstanding depiction of compensated patients in this manner within the literature. Over the years, several mechanisms have been posited to explain the association between compensation and poor outcomes.14 Some of these mechanisms allude to the psychosocial determinants of chronic pain and disability, while others emphasize the role of secondary gain or the effect of the often adversarial compensation process itself on encouraging illness behaviour.15 These theoretical mechanisms are not mutually exclusive, and most respect a number of consistendy observed (and reported) associations between illness behaviour and psychosocial variables. Previous studies on the association between compensation and patient outcomes after lumbar discectomy have shown contradictory results. However, recent systematic reviews16'17 including one review of compensation status and patient outcomes specifically after surgery14 support such an association. In a meta-analysis, Harris and colleagues reported that compensation status was associated with poorer outcomes and that this effect was consistent across different types of interventions, types of compensation, and methodological characteristics (such as the length and completeness of follow-up, prospective versus prospective data collection, and study design).14 However, even if outcomes "on average" are indeed inferior among compensated patients, this is not the same as saying that outcomes are universally negative for claimants. In fact, compensated claimants experience symptomatic relief and improved function in response to the 151 same interventions that noncompensated patients do, only the degree of these improvements may not be as large.18 Compensation status appears to effect clinicians' judgments about the expected prognosis of patients with back pain.19 The nature of physical examination findings detected among back pain patients by physical therapists, for example, is not influenced by compensation status and yet the expectation of a poor prognosis is clearly greater when patients are identified as receiving compensation.19 In a study on the effect of compensation status on patient outcomes for disc-related back pain, a lower proportion of compensated claimants were offered surgery even when the proportion of those exhibiting objective evidence of sciatica was similar between compensated and noncompensated patients.20 Presendy, the impact of compensation status on the actual probability of accessing surgery after enrolment onto a waitlist has not been studied. Our hypothesis is that the negative depiction of compensated patients in the medical literature leads surgeons to prioritize operating room resources less preferentially to compensated patients, who are generally thought to be less likely to benefit from treatment. This study, with a prospective component, considers the wait list experience of all patients queued for surgery and not just those receiving surgery. To our knowledge a wait list study controlling for a range of sociodemographic, clinical and institutional variables has not been conducted in this particular surgical population, nor has the potentially negative impact of compensation status on access to lumbar discectomy specifically after queuing been evaluated. 7.1.2 Objectives and Hypotheses The primary objective was to determine whether compensation status affected waiting time for surgical lumbar discectomy. The secondary objectives were: 1) to ascertain whether 152 waiting time for surgical lumbar discectomy was associated with measures of clinical severity and, therefore, equitably determined by clinical need; and 2) to ascertain whether waiting time for surgical lumbar discectomy was associated with sociodemographic variables. The primary hypothesis was that waiting times for patients receiving disability compensation are longer compared to patients not receiving compensation. The secondary hypotheses were: 1) that shorter waiting times were associated with higher preoperative pain score; and 2) that shorter waiting times were associated with one or more sociodemographic variables (including age, sex, occupation, and ecologic measures of income and education). 7.2 Methods We employed a prospective cohort design using an existing registry of surgical lumbar discectomy patients treated for low back pain between November 1999 and December 2003 in a tertiary care centre in Vancouver, Canada. Details on the functional status and quality of life outcomes have been published on a subset of patients in this cohort.21 The target population was working age adults with non-emergency indications for open or microscopic lumbar discectomy, managed within a publicly funded health care system such as that in the province of BC. This includes the majority of patients with back and/or leg pain caused by a confirmed disc herniation while excluding those requiring operative treatment on an emergency basis. The available population included surgical patients from ten participating orthopaedic and neurologic surgeons at Vancouver General Hospital (VGH), one of two main centres for lumbar disc operations within the Vancouver Coastal Health Authority of BC. About 75% of surgical disc patients at V G H are residents of the city of Vancouver or the surrounding area. Another 23% of patients come from more remote locations within BC, while only 2% are from another territory, 153 province or country. Access to individual surgeons is largely through referral from family physicians in the community or emergency physicians at V G H . 7.2.1 Study Population Between November 1999 and December 2003, patients with back and/or leg symptoms due to suspected lumbar disc herniation were screened for eligibility either exclusively by an attending surgeon or jointly by a surgeon and medical physician specializing in non-surgical back care. Patients were eligible for this study if they were older than 16 years of age, had disabling low back and/or leg pain, and had evidence on computerized tomographic or magnetic resonance imaging of nerve root compression at a level concordant with their physical symptoms and signs. As the registry was originally designed to study the effect of surgical discectomy on quality of life outcomes, patients were excluded if they had other conditions that could independentiy influence their health status. Such conditions included: (1) previous surgery at the current level of herniation; (2) significant co-morbidity (e.g. spinal fracture, inflammatory arthritis, cancer); (3) a structural spinal deformity (e.g. scoliosis, spondylolisthesis); and (4) pregnancy. Patients with emergency indications for lumbar discectomy (i.e. cauda equina syndrome, intractable pain, saddle anaesthesia, or rapidly progressive neurological deficit) were also excluded. All patients met evidence-based indications for surgical lumbar discectomy according to the InterQual Indications for Surgical Procedures (ISP™) criteria. These criteria included the presence of either: (1) A severe but non-progressive motor deficit (< grade 3 muscle strength) in the discrete distribution of a nerve root on physical examination; or (2) A mild motor deficit (again, in the discrete distribution of a nerve root) with continued presence of this deficit after six weeks of medication (NSAIDS, analgesics, corticosteroids, and/or 154 muscle relaxants) and limitation of provocative activities (heavy lifting, repetitive bending, or prolonged standing); or 3) Unilateral lower extremity pain that persisted after six weeks of conservative treatment (medication and limitation of provocative activities). As a requirement for completing pencil-and-paper questionnaires at study enrolment and at post-surgical follow-up, patients had to be proficient in reading and writing English. The protocol for this study was approved by the University of British Columbia's Clinical Research Ethics Board. Al l patients provided written informed consent. 7.2.2 Data Collection Information on sociodemographic variables, symptoms, prior treatment, co-morbidity and physical findings were recorded by a surgeon or physician on standardized assessment forms. Upon booking for surgery, study patients completed a baseline questionnaire package consisting of a pain drawing, 11-point numerical rating scales for back and leg symptoms, items about how pain affected daily activities, an employment information questionnaire, the Beck Depression Inventory-II, and the North American Spine Society (NASS) Lumbar Spine Baseline Questionnaire, which includes the SF-3622 and a modified version of the Oswestry Disability Questionnaire.23'24 By linking personal postal codes to 2001 census enumeration areas, measures of neighbourhood socioeconomic status were also ascertained for each patient. 7.2.3 Exposure (Disability Compensation Status) Disability compensation was defined as any type of financial compensation for current back and/or leg pain including coverage by workers' compensation, motor vehicle collision insurance, or private personal disability insurance. This was ascertained from items on the employment information questionnaire at study enrolment. Patients were asked specifically 155 whether they were receiving either workers' compensation or other disability benefits for their current back/leg condition. Patients acknowledging any one of these sources of benefits were classified as receiving disability compensation for their current surgical condition. Source of payment records from hospital charts were used to validate workers' compensation status for a consecutive sample of 100, and a random sample of another 50, patients. 7.2.4 Outcome Surgical booking and admission dates, and dates of removal from the wait-list for other reasons were recorded for each patient. Waiting time for surgical discectomy was calculated by subtracting the date of booking from the date of surgery. Waiting time was calculated in weeks to reflect the length of the operating room scheduling cycle at V G H . 7.2.5 Other Variables Potentially important determinants of waiting time for elective orthopaedic surgical procedures were identified from the literature. Initially, there were no publications on the predictors of waiting time for lumbar disectomy specifically. However, Canadian studies on waiting times for major joint arthroplasty25 and elective surgical procedures in general9 indicated marital status, socioeconomic status and size of the treating surgeon's practice as potential correlates of waiting time. Later, employment status (employed versus unemployed) was reported as a predictor of shorter waiting times for back surgery in Sweden.11 As a proxy measure of practice size in the current study, surgeons were categorized as having a low, medium or high caseload of lumbar discectomy patients according to the average duration in weeks between patient enrolments. This was estimated for each surgeon by calculating the total number of weeks over which his first through last patients were enrolled, and dividing by his total number of recruited patients. 156 The effects of the following clinically plausible predictors were examined: (1) sociodemographic variables including age, sex, marital status, educadon, type of occupation (professional/managerial versus other), and employment status; (2) clinical variables including severity of pain at enrolment (0 to 10 points), pain duration (< 6 weeks, 6 weeks to < 3 months, 3 months to < 6 months, or > 6 months), motor weakness (present or absent), reflex asymmetry (present or absent), sensory loss (present or absent), and straight leg raising (positive or negative); (3) institutional variables such as hospital service (orthopaedics versus neurosurgery), and urgency status at enrolment (non-urgent versus urgent), and finally (4) neighbourhood socioeconomic conditions including area median income, area average income, urban versus rural neighbourhood, local unemployment rate, and percentage of residents in the area with at a bachelor's level of education. 7 . 2 . 6 S t a t i s t i c a l M e t h o d s For categorical variables frequencies and percentages were tabulated for patients overall, and by compensation status. Variables with sparsely populated categories were recoded and collapsed to eliminate small strata. Continuous variables with skewed distributions were also categorized. Differences in baseline characteristics between compensation status groups were compared using Fisher's exact test for categorical variables and Student's t test for continuous variables. Waiting times were analyzed as prospective observations beginning at the time of booking for surgery. The probabilities of remaining on the surgical waitlist as a function of waiting time were estimated using the Kaplan-Meier method. Patients removed from the list for reasons other than surgery were treated as censored observations. Additionally, patients who were admitted on an emergency basis after study enrolment were censored on their date of surgery while patients who were re-scheduled to a later date were censored on the date of re-scheduling. 157 The log-rank test was used to compare the difference in the distribution of waiting times between compensation status groups. Potential correlates of waiting time were identified initially in bivariate analyses. Kaplan-Meier estimators and corresponding log-rank tests were applied to potentially associated categorical variables, and univariate Cox proportional hazards regression was used to identify potentially associated continuous variables. The adjusted effect of compensation status on time to surgery was estimated using a multivariable Cox proportional hazards model. Variable selection was based on purposeful selection. First, a preliminary multivariable model was constructed by including age, sex, and compensation status with any covariables that had been significant at the 25% level in bivariate analyses. A reduced model was then obtained by deleting covariables that were no longer significant (at the 25% level) within this preliminary multivariable model. Next, a restricted model was created by deleting variables one-by-one in descending order of significance until all remaining variables were significant at the 5% level. Deleted variables that produced a change of more than 20% in the coefficients of other variables were considered important confounders and were retained in the restricted model regardless of their statistical significance. Cross-product terms between compensation status and each main-effect in the restricted model were individually screened for significance at the 5% level. The study protocol included a plan to retain all individually significant interaction terms within our final model. For each categorical variable in the restricted model, adherence to the proportionality assumption was evaluated by confirming the presence of parallel lines in plots of both the Kaplan-Meier survivorship functions, and the log(-log(survival)) versus survival time. In this regard, the survivorship functions for compensated and noncompensated patients converged, but only near the end of the follow-up period where the number of observations was sparse. We also 158 sequentially stratified on compensation status and each categorical covariate in the restricted model, one at a time, and examined for the stability of coefficients between the stratified and non-stratified models for each covariate. Interactions between the natural logarithm of survival time and each covariate in the restricted model were also examined as a screen for significant time dependent covariates. 7.3 Results 7.3.1 Descriptive Results Figure 7.1 and Figure 7.2 describe the flow of patients and the sampling scheme for our study. A total of 576 patients were screened for the study, of which 132 either violated initial screening criteria or refused outright to participate. Of 444 willing participants, a history of prior back surgery at the same level of current disc injury was identified in 42 patients. For nine of the 402 appropriately enrolled patients, either the date of waitlist enrolment or the date of surgery had not been recorded and this information was not recoverable due to misplacement of the booking form and/or operating room report. Demographic data were available, albeit incompletely, for only five of those patients (Table 7.1). Al l nine patients were otherwise excluded from further analysis. In the end, compensation status and valid wait time data were ascertained for 393 patients. For these patients, our analysis focused on the first 52 weeks after placement onto the waitlist for surgery. Fifteen patients who remained on the waitlist at one-year were censored, as were an additional 16 patients who deteriorated and received emergency surgery after initially being booked for elective surgery. Another six patients who were censored were actually admitted but then had to be re-scheduled when the operating room ran late on their originally scheduled date of surgery. 159 Table 7.2 summarizes the sociodemographic characterisdcs of the cohort. The largest subgroups of patients were male (61%), urban residents (76%), married or living with a partner (60%), currendy employed (64%), and working in a professional or managerial position (44%). Overall, 17% of patients were receiving disability compensation for their current condition. Patients receiving compensation were more likely to be male (77%), skilled labourers (59%), and current or recent smokers (55%). Compared to non-compensation patients, they were also more likely to be unemployed specifically due to current back or leg symptoms (35%). The clinical and health system-related characteristics of patients are described in Tables 3 and 4. The largest subgroups among these characteristics were patients who had current symptoms for more than six-months (46%), complained of numbness in the leg (55%), or exhibited a positive straight leg raise test on preoperative evaluation (84%). Most patients had been treated by a neurosurgeon (68%) and were considered non-urgent (66%) at the time of booking for surgery. Few patients had evidence of concurrent central canal stenosis on C T / M R imaging (14%). Regarding symptom severity, mean pain scores were almost identical between compensated and noncompensated patients. The mean Beck Depression Inventory score was slightly higher among compensated patients. Compensated patients were more likely than noncompensated patients to exhibit an abnormal reflex. The groups did not differ significandy with respect to the frequency of motor weakness, sensory numbness, or positive straight leg raising at the time of preoperative assessment. Approximately 69% of noncompensated and 61% of compensated patients were seen by neurologic as opposed to orthopaedic surgeons. Equal proportions of compensated and noncompensated patients were treated by surgeons with low caseloads, however a slighdy greater 160 proportion of compensated patients were treated by medium caseload surgeons whereas a slighdy greater proportion of noncompensated patients were seen by high caseload surgeons. The proportions of urgent versus non-urgent patients at the time of booking, and the proportions of urgent or semi-urgent, versus non-urgent patients at the time of surgery, were similar between compensation groups. 7 . 3 . 2 A c c e s s t o S u r g i c a l D i s c e c t o m y f o r P a t i e n t s O v e r a l l The mean waiting time for the entire cohort was 11.5 weeks (standard deviation: 15.6) while the median waiting time was 6.0 weeks (interquartile range: 12.0). A total of 37 observations were censored: 22 patients were upgraded to emergency status after initially being classified as urgent or nonurgent at the time of queuing, and 15 patients were censored for not having had surgery within 52 weeks. Figure 7.3 shows the Kaplan-Meier estimate of the survivorship functions for compensated and noncompensated patients. At most time points, the survivorship function for compensated patients was above that of noncompensated patients, indicating that the probability of remaining on the waitlist was slightly higher for compensated patients during most weeks. At 36 weeks the curves crossed over minimally and then converged at 52 weeks. 7 . 3 . 3 A c c e s s t o S u r g i c a l D i s c e c t o m y b y D i s a b i l i t y C o m p e n s a t i o n S t a t u s Descriptive statistics for the survivorship experience of both groups are presented in Table 7.5. The median wait times were 6 and 9 weeks for noncompensated and compensated patients respectively. While a difference of this magnitude would be clinically significant, the log rank test for the difference between the two functions was not significant, and therefore the null hypothesis of no difference in waiting times was not rejected. 161 7.3.4 Bivariate Correlates of Waiting Time for Surgical Discectomy Table 7.6 lists each of the sociodemographic, clinical and health system variables that were associated with waiting time in bivariate analyses. Women and patients in professional or managerial occupations had significandy shorter waiting times while other sociodemographic variables were not significandy associated with waiting time at a bivariate level. Ecological measures of socioeconomic status were not associated with waiting time in bivariate analyses and not considered further. In contrast, several clinical variables, including pain duration, pain score (on a continuous scale), and motor weakness were strongly associated with shorter waits. 7.3.5 Multivariable Correlates of Wait Time The adjusted risk ratios for variables in our reduced and restricted models are presented in Table 7.7, alongside their crude effects from the bivariate analyses. To improve model parsimony, age and sex were excluded from the final restricted model because they were clearly not associated with waiting time in either the full or reduced models. Furthermore, their exclusion did not effect the coefficients of remaining variables. The effect of symptom duration of "three-to-six months" did not significantly differ from the effect of symptom duration of "less than six-weeks", therefore these two categories were combined to reduce the overall number of strata in the analysis. 7.3.6 Adjusted Effect of Compensation Status After adjusting for other variables, compensation status was not associated with waiting time for surgical lumbar discectomy. In the reduced model, the effect of compensation status clearly approached the null value (RR, 95% confidence interval: 0.99, 0.72-1.36) despite previously showing a mild association at a bivariate level. Figure 7.4 shows the estimated covariate adjusted survivorship functions for patients with covariate values representing the largest subgroups in the 162 study population, by compensation status. For patients with a symptom duration longer than 6 months, a preoperative pain score of 8, motor weakness in the leg, and non-urgent at the time of booking for surgery, the curves for compensated and noncompensated patients overlap almost completely, indicating no difference in the waitlist experience between payor groups. Of all sociodemographic variables that were examined, only dichotomized occupation type showed an independent association with waiting time at a multivariable level. In the final restricted model (Table 7.7, column 6), patients classified as either professionals or managers had a conditional weekly probability of acquiring surgery that was 29% higher than that of other workers (RR: 1.29 [95% CI: 1.03, 1.62]). Figure 7.5 shows the estimated covariate adjusted survivorship function, again for patients with covariate values representing the largest subgroups in the study population. The separation between the curves for patients doing professional and nonprofessional work reflects the significant effect of occupation on waiting times that we found in this study. 7.3.7 Adjusted Effect of Clinical Variables Most clinical measures of disease severity were associated with waiting time in the expected direction. Patients with higher pain intensity at study enrolment were more likely to access surgery than patients with lower pain intensity during any given week. For each two-point increase in pain intensity at the time of enrolment, the conditional weekly probability of accessing surgery was higher by 19% (RR: 1.19 [95% CI: 1.08-1.33]). Similarly, conditional weekly probabilities of undergoing surgery were 36% higher for patients with muscle weakness (RR: 1.36 [95% CI: 1.08-1.72]), and 44% lower for patients who were non-urgent (versus urgent) at the time of enrolment (RR: 0.54 [95% CI: 0.43-0.69]). 163 Symptom duration at the time of enrolment was also clearly associated with the likelihood of acquiring surgery. Compared to patients with symptom durations of "less than six weeks" and "three to six months" combined, conditional weekly probabilities of undergoing surgery were 75% higher for patients with a symptom duration of "six weeks to three months" (RR: 1.74 [95% CI: 1.25-2.43]) and 32% lower for patients with symptoms "greater than six months" in duration (RR: 0.68 [95% CI: 0.52-0.87]). No significant interactions were found between compensation status and other variables in the final model. 7.4 Discussion This study showed that compensation status was not independently associated with waiting time for surgical lumbar discectomy. We therefore found no evidence that compensated patients encounter a barrier to access to care once they are accepted for surgery and placed on the waitlist. On the other hand, we observed that professionals and managerial workers had, on average (or at every point in time) 29% faster access to surgery than workers in other occupations (including sales or service, skilled labour, or semi-/unskilled labour), suggesting that some sociodemographic variables may play a role in a patient's access to surgical care, even in a publicly funded health care system. As expected, several measures of disease severity were significantly associated with waiting time, indicating that access to surgical discectomy in BC is in fact largely based on clinical need. Statistically significant associations between clinical variables and waiting time were in the expected direction. Patients with more severe pain and nerve root irritation (i.e. positive straight leg raising) accessed surgery in greater proportions each week than those with less severe pain and/or no objective orthopaedic signs. 164 Symptom duration at enrolment was significandy associated with access to surgery, and this relationship followed an inverted "J" pattern. Compared to patients in the shortest duration group (less than 6 weeks), those who had symptoms for 6 weeks to 3 months, 3 to 6 months, and greater than 6 months accessed surgery at rates that were greater, comparable, and lower, respectively. The greater risk of accessing surgery for those with symptom durations between 1.5 and 3 months was expected since clinical guidelines generally discourage the use of surgery prior to six weeks of conservative therapy or observation, and generally approve surgery for symptoms that persist beyond six weeks in duration. Beyond a symptom duration of 3 months, access to surgery declined monotonically with each incremental increase in symptom duration. A symptom duration of 3 to 6 months was associated with a risk of surgery that was comparable, while a symptom duration greater than 6 months was associated with a risk of surgery that was significantly less compared to the shortest symptom duration group. Diminished access to surgery for patients with more chronic symptoms possibly reflects a pattern of surgical decision making that is driven by previous studies showing significant associations between longer symptom duration and poorer outcomes.26^7 In Canada, service rationing within a publicly funded health care system is associated with wait lists for many procedures, yet the distribution and determinants of access have been examined for only a limited number of these. Studies on orthopaedic surgery in Canada, for example, have focused on wait times for non-spinal operations such as total hip or knee arthroplasty. In some of these studies, health status—and therefore, clinical need—was not found to be a predictor of waiting times for orthopaedic surgery in general,12;25;28 but possibly a weak predictor of waiting times for major joint replacement.29 Only one study has been published recendy about the predictors of access to back surgery specifically. In Sweden" sex, age and living arrangement were not associated with wait time for back surgery (or knee or hip replacement surgery), however 165 working (versus non-working) patients, and those admitted into county/district (versus university/regional) hospitals had shorter waiting times. Also, longer time to first contact with health services, but not time from referral to first hospital outpatient visit, was associated with longer waiting time for back surgery. We found that the relative impact of clinical and sociodemographic variables on waiting times may be different for back surgery compared to major joint arthroplasty. Coyte showed differing results between Ontario and the US, however in Ontario, institutional variables including hospital type (teaching versus non-teaching hospitals) and number of beds were important predictors of waiting times for knee-replacement surgery whereas sociodemographic variables including income, education and sex were not.28 Measures of clinical status such as knee pain at rest, and an inability to climb stairs were not associated with waiting times for knee surgery in Ontario but were in the US (where an inability to climb stairs was associated with shorter waiting times). Kelly found that waiting times for hip and knee surgery were associated with both sociodemographic and clinical variables including marital status, primary language (English, French, or other), body mass index, social function (measured on the SF-36 subscale), pain medication use and size of the treating surgeon's practice, however these variables explained only 10% of the total variance in the model. More importantiy, neither severity of pain nor severity of disability predicted waiting times for joint replacement surgery suggesting that in Canada waitlists for these procedures are managed equitably from a social perspective, however inequitably from a clinical perspective.25 Our study of surgical discectomy patients at one hospital precluded an assessment of the effect of hospital type and bed capacity. However, in contrast to existing studies on the predictors of waiting times for major joint arthroplasty, we found that access to surgical discectomy appears to be equitable from a clinical perspective but possibly not from a social perspective (given that 166 professionals accessed surgery more readily than non-professionals). Future research is warranted to confirm whether the balance of clinical and sociodemographic determinants of access genuinely vary across different elective procedures, or if apparent differences in these determinants are a mere artifact of the way in which these variables are ascertained across different studies. 7 . 4 . 1 S t r e n g t h s a n d L i m i t a t i o n s One of the strengths of the current study is that we considered the waitlist experience of all patients queued for surgical discectomy and not just those receiving an operation. We were able to identify those who were rescheduled or removed from the waitlist entirely for reasons other than accessing elective surgery. Patients who experience an intermediate event may have a different admission rate after the intermediate event and therefore it is necessary to account and/or control for these events.30 In the current study, these patients were censored instead of being completely excluded from the analysis. They therefore contributed at least partially to the risk set of patients waiting for surgery up until the time of an intermediate event, but were then censored after the event to eliminate potential biases had they been retained in the data set.30 To our knowledge, no previous study has examined the potentially negative impact of compensation status on access to surgical lumbar discectomy specifically after the time of placement onto the waitlist. Also, no previous study to our knowledge has examined the effect of compensation status on access to surgical lumbar discectomy using primary clinical data. This latter feature of our study enabled us to screen a broad range of sociodemographic, clinical, and institutional/health system variables for potential confounding and to subsequently adjust for identified confounders in our analysis. Another strength is that waiting times were calculated exacdy using actual dates of surgery and dates of placement onto the waitlist. We did not have to rely on a proxy measure of the date 167 of waitlist enrolment (such as date of last outpatient visit to a surgeon) as is commonly necessary in studies relying only on administrative data. We were unable to document the date of referral for each patient and were therefore unable to estimate the time that patients spent waiting to see a specialist. A future study on the determinants of time to first-appointment with a specialist and total time to surgery from the date of referral to a specialist would be a valuable supplement to the literature. The exposure variable, compensation status, was self-reported but also verified through an independent assessment of hospital payment records. As for potential misclassification of the outcome (i.e. waiting time), for six patients both the date of surgical booking and the date of the corresponding preoperative visit to a surgeon were missing, and in each of these instances, the date of surgical booking had to be imputed using the date of completion of the patient's standardized assessment questionnaire as a proxy for the date of surgical booking. This questionnaire was typically administered on the date of surgical booking, however on occasion, it was administered late and therefore closer to the time of surgery. Therefore, at worst, waiting time values for six observations may have been underestimated in this manner, however we had no reason to believe that the missing data values were systematically related to either true waiting time, compensation status, or other potential confounders. Due to relatively small numbers of patients receiving disability compensation, we combined workers' compensation claimants with claimants receiving disability from other sources (private disability or motor vehicle/personal injury insurance). While some studies suggest that the specific type of compensation may be a determinant of outcome during a claim, there is no evidence in the literature to suggest that specific type of compensation influences waiting times for surgery. Furthermore, exploratory analyses of our data showed that the coefficients for the effects 168 of workers' disability compensation and disability compensation from alternate sources were similar in magnitude and statistical significance. Five additional observations were associated with negative wait time values, reflecting errors in the documentation of one or more source variables (date of surgical booking or date of surgery). While this was possibly indicative of other recording errors within the data, we were able to validate our source variables for all other subjects in the database by cross-referencing both the date of surgical booking and the date of surgery against the date of preoperative visit to the surgeon and date of discharge after admission, respectively. Our study population was representative of mosdy urban residents. Study patients were recruited at V G H , a public tertiary care hospital situated within the BC's largest metropolitan centre. V G H provides care to patients within the Vancouver Coastal Health Authority, most of whom are urban residents with only about 8% living in smaller towns or non-urban areas. Vancouver Coastal Health Authority contains 25% of the province's total population. V G H is one of only two public hospitals providing surgical lumbar discectomies regularly within this health authority. Our sample did not capture patients who underwent surgery in private clinics. However, private, expedited lumbar disc surgery was available only after the year 2000, and although there are approximately 20 private medical care clinics in British Columbia currendy, only two private clinics treat back pain patients with any regularity. In any event, the exclusion of privately treated patients from this study did not invalidate our specific objective of identifying the significant determinants of waiting time for surgical lumbar discectomy within the public system. 169 7.5 Summary Surgeons feel ethically obliged to offer surgery to all patients in the presence of appropriate indications regardless of compensation status. However the belief that surgical outcomes are possibly "not as good" in compensated patients may encourage surgeons (as well as general practitioners) to persist with a longer trial of conservative management. While this attitude might conceivably delay a surgeon's decision to offer surgery and to therefore place a compensated patient onto the surgical waitilist, we found no evidence that compensation status affected time to surgery after placement onto the waitlist. Further research is encouraged to determine whether compensation status affects other measures of access to care, such as time to referral to a specialist after first-contact with general practitioner (GPs), time to consultation with a surgeon after referral by a GP, and ultimately, time to placement onto the surgical waitlist after first visit to a surgeon. 170 Figure 7.1: Patient Flow Diagram Enrolment into "Acute" Lumbar Disc Studv Enrolment onto Withdrawn from Surgical Wait List w Wait Est r Baseline Questionnaires Surgery Performed Follow-up at 6 months 171 Figure 7.2: Patient Sampling Scheme 576-Acute disc patients invited for study screening 132-Refused or not eligible to participate 444-Willing to undergo screening Standardized assessment 42-Patients with previous surgery at the same level 402-Appropriately enrolled primary lumbar discectomy patients Data prepation/extraction 9-Patients with missing values for wait time source variables 393-Total number of patients with valid wait time information Table 7.1: Characteristics of Excluded Patients Characteristics of patients excluded due to missing source variables for waiting time calculation.* Characteristic Not receiving compensationf Receiving compensation-)- Overallf Number of patients 7 2 9 Mean age (SD) 42 (9.0) 40 (7.8) 41 (8.1) Sex Males 4 1 5 Females 3 1 4 Marital status Married/with partner 1 0 1 Not married/not with partner 0 0 0 Missing 6 2 8 Currently employed Yes 1 1 2 No 2 1 3 Missing 4 0 4 Usual type of work Semi-/unskilled labour 0 0 0 Skilled labour 1 0 1 Sales or service 0 0 0 Professional/managerial 2 2 4 Student 0 0 0 Missing 4 0 4 Unemployment specifically due to back/leg pain Yes 2 1 3 No 1 1 2 Missing 4 0 4 Smoking status Never/Not in 6 months 2 1 3 Currendy/recent smoker 1 1 2 Missing 4 0 4 *For 4 of these 9 patients, data was almost completely missing. 173 Table 7.2: Characteristics of Study Patients Sociodemographic characteristics of study patients at preoperative evaluation, overall and by compensation status. Characteristic Not receiving compensation)' Receiving compensation-)- Overallf p value* Number of patients 327 66 393 Mean age (SD) 42 (12.2) 43 (11.7) 42 (12.1) 0.9 Sex Males 188 (57) 51 (77) 239 (61) 0.002 Females 139 (43) 15 (23) 154 (39) Marital status Married/with partner 210 (63) 32 (47) 242 (60) 0.02 Not married/not with partner 124 (37) 36 (53) 160 (40) Average median income for area of 25,049 23,159 24,716 0.02 residence (SD)ff (5,839) (5,678) (5,837) Urban area of residence-)-)" Yes 247 (76) 51 (77) 298 (76) 0.9 No 80 (24) 15 (23) 95 (24) Mean % with bachelor's degree (SD) in areaf-)- 25.0 (14.2) 21.1 (11.7) 24.3 (13.9) 0.02 Currendy employed Yes 223 (68) 29 (44) 252 (64) 0.0001 No 104 (32) 37 (56) 141 (36) Usual type of work Semi-/unskilled labour 16(5) 9(14) 25(7) < 0.0001 Skilled labour 69 (21) 36 (55) 105 (27) Sales or service 43 (13) 7(11) 50 (13) Professional/managerial 164 (50) 8 (12) 172 (44) Student 7(2) 1(2) 8(2) Missing 28 (9) 5(8) 33 (8) Main employment activity Sitting 121 (37) 8 (12) 129 (35) < 0.0001 Walking/ standing 49 (15) 4(6) 53 (13) Seldom lifting > 35 kg 78 (24) 20 (30) 98 (25) Awkward lifting > 35 kg 41 (13) 15 (23) 56 (14) Maximal/awkward lifts 20(6) 15 (23) 35(9) Missing 18(6) 4(6) 22(6) Unemployment specifically due to back/leg pain Yes 40 (12) 23 (35) 63 (16) < 0.0001 No 287 (88) 43 (65) 330 (84) Smoking status Never/Not in 6 months 230 (70) 30 (45) 260 (66) < 0.0001 Currendy/recent smoker 97 (30) 36 (55) 133 (34) *Using the Fisher exact test for categorical variables and t test for continuous variables. •(•Number with column percentages in parentheses unless otherwise noted. If Based on census data, linked by postal code, to the enumeration area patient resided in at time of preoperative assessment. 174 Table 7.3: Clinical Characteristics of Study Patients Clinical characteristics of patients at preoperative evaluation. Not receiving Receiving Characteristic compensationf compensationf Overallf p value* Number of patients 327 66 393 Pain duration < 6 weeks 38 (12) 7(11) 45 (11) 0.25 6 weeks — 3 months 59 (18) 7(11) 66 (17) 3 — 6 months 88 (27) 15 (23) 103 (26) > 6 months 142 (43) 37 (56) 179 (46) Mean pain severity (SD) 7.4 (2.2) 7.5 (2.0) 7.4 (2.1) 0.64 Muscle weakness Yes 184 (56) 30 (46) 214 (54 0.13 No 143 (44) 36 (55) 179 (46) Reflex loss Yes 147 (45) 38 (58) 185 (47) 0.08 No 180 (55) 28 (42) 208 (53) Numbness Yes 180 (55) 38 (58) 218 (55) 0.7 No 147 (45) 28 (42) 175 (45) Positive straight leg raising Yes 278 (85) 54 (82) 332 (84) 0.50 No 49 (15) 12 (18) 61 (15) Stenosis on C T / M R imaging Yes 45 (14) 11 (17) 56 (14) 0.56 No 282 (86) 55 (83) 337 (86) Beck Depression Inventory Mean score (SD) 12.7 (80.1) 14.7 (11.0) 13.1 (8.6) 0.17 *Using the Fisher exact test for categorical variables and t test for continuous variables. fNumber with column percentages in parentheses unless otherwise noted. 175 Table 7.4: Health System-Related Variables Health system characteristics of patients at preoperative evaluation. Not receiving Receiving Characterisdc compensation! compensation-!- Overallf p value* Number of patients 327 66 393 Hospital service Orthopaedics 100 (31) 25 (39) 125 (32) 0.24 Neurosurgery 221 (69) 39 (61) 260 (68) Caseload of surgeon (no. of weeks between enrolments§ T l : (11.0-14.0weeks) 15(5) 3(5) 18(5) 0.30 T2: (6.9- 10.9 weeks) 80 (24) 22 (33) 102 (26) T3: (1.5-6.8 weeks) 232 (71) 41 (62) 273 (69) Urgency status at booking Urgent 140 (43) 25 (38) 165 (42) 0.50 Non-urgent 187 (57) 41 (62) 228 (58) Urgency status at surgery Emergency (censored) 11(3) 3(5) 14(4) 0.71 Urgent/Semi-urgent 101 (31) 18 (28) 119 (31) Non-urgent 215 (66) 45 (68) 260 (66) *Using the Fisher exact test unless otherwise noted. f Number with column percentages in parentheses unless otherwise noted. §Tertiles for the average number of weeks between patient enrolments (between dates of first, and last, patients recruited by each surgeon). 176 Figure 7 . 3 : Kaplan-Meier Time-to-Surgery Functions Estimated probabilities of remaining on the waitlist for surgical lumbar discectomy, by compensation status. Unadjusted probabilites of remaining on the waitlist, by compensation status Compensation Noncompensation Noncompensation Compensation STRATA: 10 20 30 40 50 wait_cens Noncompensation ooo Censored Noncompensation Compensation • • • Censored Compensation 60 Table 7 . 5 : Waiting Times by Compensation Status Median waiting time Log-rank Compensation Status n (weeks) 95% CI p value Not receiving compensation 327 6.0 5.0, 8.0 0.19 Receiving compensation 66 9.0 6.0,15.0 Overall 393 6.0 6.0, 8.0 177 Table 7.6: Bivariate Correlates of Waiting Time Point Variable estimate * 95% CI P value§ Age: < 33 8.0 4.0,11.0 0.12 33-37 7.0 5.0,10.0 38-44 6.0 5.0, 9.0 45-55 6.0 3.0, 9.0 > 55 7.0 4.0,12.0 Age (continuous scale, per 5-year increase) 1.03f 0.98,1.07 0.22 Sex Male 8.0 6.0,10.0 0.003 Female 5.0 3.0, 6.0 Occupation Semi-/unskilled labour 13.0 3.0,15.0 0.33 Skilled labour 7.0 5.0, 9.0 Sales or service 8.0 4.0,14.0 Professional/managerial 5.0 4.0, 7.0 Student 9.0 2.0, 25.0 Work (dichotomized) Professional/managerial 5.0 4.0, 7.0 0.045 Other 8.0 6.0,11.0 Duration of symptoms < 6 weeks 3.0 2.0, 6.0 < 0.0001 6 weeks-3 months 3.0 2.0, 4.0 3-6 months 5.0 3.0, 9.0 > 6 months 10.0 7.0,12.0 Pain score < 5 11.0 5.0,15.0 0.031 5 - 6 12.0 6.0,14.0 7 - 8 5.0 4.0, 7.0 9-10 6.0 5.0, 8.0 Pain score (dichotomized) < 7 11.0 7.0,13.0 0.006 > 7 6.0 5.0, 7.0 Pain score (continuous scale, per 2-point increase) 1.16f 1.05,1.28 0.003 Weakness Yes 5.0 4.0, 6.0 0.018 N o 9.0 7.0,12.0 Positive straight-leg raise Yes 6.0 5.0, 8.0 0.05 N o 12.0 7.0,17.0 Urgency status 5.0 4.0, 6.0 < 0.0001 Non-urgent 9.0 6.0,11.0 Specialty Orthopaedics 9.0 6.0,12.0 0.009 Neurosurgery 6.0 5.0, 7.0 Surgeon's frequency of recruitment (average no. weeks between enroments ||) Low frequency (11.0 - 14.0) 1.5 1.0, 4.0 0.003 Medium frequency (6.9 - 10.9) 12.0 8.0,14.0 High frequency (1.5 - 6.8) 6.0 5.0, 7.0 Compensation Not receiving 6.0 5.0, 8.0 0.19 Receiving 9.0 6.0,15.0 *Stratum-specific median wait time (in weeks) for categorical variables, unless otherwise fRelative risk of having surgery per specified change in continuous variable. §Log-rank equality across strata for categorical variables, or Wald Chi-squared test from univariable proportional hazards regression for continuous variables. || Surgeons were divided into three groups based on the average number of weeks between patient enrolments. For each surgeon, the total number of weeks between the dates of first and last patients enrolled was divided by the total number of patients recruited by the surgeon. 178 Table 7.7: Crude and Adjusted Effects of Predictor Variables Crude and adjusted effects of age, sex, compensadon status, and significant multivariable predictors of wait time for surgical lumbar discectomyf Univariable Model Reduced Model Restricted Model Crude RR Adjusted RR Adjusted RR Variable (95% CI) p* (95% CI) p* (95% CI) p* Age Per 5-yr increase 1.03 (0.99,1.07) 0.22 0.98(0.94,1.03) 0.37 Sex Male 1.0 1.0 Female 1.36 (1.10,1.68) 0.0049 1.10(0.86,1.41) 0.46 Compensation status No compensation 1.0 1.0 Compensation 0.83(0.63,1.11) 0.20 0.99 (0.72,1.36) 0.95 : Type of occupation Non-professional 1.0 1.0 1.0 Professional/manager 1.24 (1.00, 1.54) 0.056 1.29 (1.01,1.65) 0.041 1.29 (1.03,1.62) 0.029 Duration of symptoms < O.OOOlf < 6 wks 1.0 < O.OOOlf 1.0 (reference) < O.OOOlf 1.0 (reference) 6 wks-3 months 1.72 (1.10, 2.70) 1.73 (1.24, 2.42) 1.74 (1.25, 2.43) 0.0012 3-6 months 1.03(0.68,1.57) 0.0008 1.0 (reference) 0.0013 1.0 (reference) > 6 mo. 0.66 (0.45, 0.98) 0.057 0.67 (0.52, 0.87) 0.0024 0.68 (0.52, 0.87) 0.0029 Pain score 1.0 Per 2-point increase 1.16(1.05,1.28) 0.0032 1.19 (1.06,1.33) 0.0026 1.19 (1.08,1.33) 0.0009 Muscle weakness Absent 1.0 1.0 1.0 Present 1.28 (1.03,1.59) 0.024 1.36 (1.07,1.73) 0.011 1.36 (1.08,1.72) 0.0093 Urgency at booking Urgent 1.0 1.0 1.0 Non-urgent 0.65 (0.53, 0.81) < 0.0001 0.54 (0.43, 0.69) < 0.0001 0.55 (0.43, 0.69) < 0.0001 *Based on Wald test for stratum-specific parameter estimate, and partial likelihood ratio test for predictors with more than one design variable. fWald Chi-squared (type 3) test for overall effect of categorical variable. Figure 7.4: Covariate-Adjusted Time-to-Surgery For Compensated Patients Covariate adjusted survivorship functions for men and women combined, with symptom duration > 6 months, median pain score (8), muscle weakness, and non-urgent status at booking, according to disability compensation status. Covariate adjusted survivorship curves by disability compensation status Men & women, symptom duration > G months, median pain score (8) , muscle weakness, non-urgent at booking 180 Figure 7.5: Covariate-Adjusted Time-to-Surgery for Professional Workers Covariate adjusted survivorship functions for non-professional men and women combined, aged 42, with symptom duration > 6 months, median pain score (8), and muscle weakness, according to type of occupation. Covariate adjusted survivorship curves by type of occupation Men & women, symptom d u r a t i o n > 6 m o n t h s , m e d i a n p a i n s c o r e ( 8 ) , m u s c l e w e a k n e s s , n o n - u r g e n t a t b o o k i n g 10 20 30 40 50 Ua i t t i me T y p e o f o c c u p a t i o n N o n - p r o f e s s i o n a l Q-«"-G P r o f e s s i o n a l 181 7.6 References 1. Taylor V M , Anderson G M , McNeney B et al. Hospitalizations for back and neck problems: A comparison between the Province of Ontario and Washington State. Health Services Research 1998;33:929-45. 2. Deyo RA, Cherkin D, Conrad D, and Volinn E. Cost, Controversy, Crisis - Low-Back-Pain and the Health of the Public. Annual Review of Public Health 1991;12:141-56. 3. Cassidy JD, Carroll LJ, and Cote P. The Saskatchewan health and back pain survey - The prevalence of low back pain and related disability in Saskatchewan adults. Spine 1998;23:1860-6. 4. Frank JW, Kerr MS, Brooker AS et al. Disability resulting from occupational low back pain -What do we know about primary prevention? A review of the scientific evidence on prevention before disability begins. Spine 1996;21:2908-17. 5. Adas SJ, Deyo RA, Keller RB et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine 1996;21:1777-86. 6. Gibson J N A and Waddell G. Surgical interventions for lumbar disc prolapse. Cochrane Database of Systematic Reviews 2007. 7. Taylor V M , Deyo RA, Cherkin DC, and Kreuter W. Low-Back-Pain Hospitalization -Recent United-States Trends and Regional Variations. Spine 1994;19:1207-13. 8. Andersson GBJ. Epidemiological features of chronic low-back pain. Lancet 1999;354:581-5. 182 9. Shortt SED and Shaw RA. Equity in Canadian health care: Does socioeconomic status affect waiting times for elective surgery? Canadian Medical Association Journal 2003;168:413-6. 10. Sobolev B, Brown P, Zelt D, and Shortt S. Bias inherent in retrospective waiting-time studies: experience from a vascular surgery waiting list. CMAJ 2000;162:1821-2. 11. Lofvendahl S, Eckerlund I, Hansagi H , Malmqvist B, Resch S, and Hanning M . Waiting for orthopaedic surgery: factors associated with waiting times and patients' opinion. International Journal for Quality in Health Care 2005;17:133-40. 12. Kelly K D , Voaklander D, Kramer G, Johnston DW, Redfern L, and Suarez-Almazor M E . The impact of health status on waiting time for major joint arthroplasty. Journal of Arthroplasty 2000;15:877-83. 13. Klein BJ, Radecki RT, Foris MP, Fell EI, and Hickey M E . Bridging the gap between science and practice in managing low back pain - A comprehensive spine care system in a health maintenance organization setting. Spine 2000;25:738-40. 14. Harris I, Mulford J, Solomon M , van Gelder JM, and Young J. Association between compensation status and outcome after surgery - A meta-analysis. Journal of the American Medical Association 2005;293:1644-52. 15. Teasell RW. Compensation and chronic pain. Clinical Journal of Pain 2001;17:S46-S51. 16. Steenstra IA, Verbeek JH , Heymans MW, and Bongers PM. Prognostic factors for duration of sick leave in patients sick listed with acute low back pain: a systematic review of the literature. Occupational and Environmental Medicine 2005;62:851-60. 183 17. van der Hulst M , Vollenbroek-Hutten MMR, and IJzerman MJ. A systematic review of sociodemographic, physical, and psychological predictors of multidisciplinary rehabilitation -or, back school treatment outcome in patients with chronic low back pain. Spine 2005;30:813-25. 18. Atlas SJ, Chang Y C , Kammann E, Keller RB, Deyo RA, and Singer D E . Long-term disability and return to work among patients who have a herniated lumbar disc: The effect of disability compensation. Journal of Bone and Joint Surgery-American Volume 2000;82A:4-15. 19. Simmonds M and Kumar S. Does knowledge of a patient's workers' compensation status influence clinical judgments? Journal of Occupational Rehabilitation 1996;6:93-107. 20. Adas SJ, Singer D E , Keller R, Convery K, Mooney N , and Deyo R. The Efficacy of Lumbar Disc Surgery in Patients with Sciatica - Preliminary-Results from the Maine Lumbar Spine Study. Clinical Research 1993;41:A200. 21. Fisher C, Noonan V, Bishop P et al. Outcome evaluation of the operative management of lumbar disc herniation causing sciatica. Journal of Neurosurgery 2004;100:317-24. 22. Ware JE, Jr. SF-36 health survey update. Spine 2000;25:3130-9. 23. Fairbank JC and Pynsent PB. The Oswestry Disability Index. Spine 2000;25:2940-52. 24. Daltroy L H , Cats-Baril WL, Katz J N , Fossel A H , and Liang M H . The North American spine society lumbar spine outcome assessment Instrument: reliability and validity tests. Spine 1996;21:741-9. 184 25. Kelly K D , Voaklander DC, Johnston WC, and Suarez-Almazor M E . Equity in waiting times for major joint arthroplasty. Canadian Journal of Surgery. 2002;45:269-76. 26. Nygaard OP, Kloster R, and Solberg T. Duration of leg pain as a predictor of outcome after surgery for lumbar disc herniation: a prospective cohort study with 1-year follow up. Journal of Neurosurgery 2000;92:131-4. 27. Ng L C L and Sell P. Predictive value of the duration of sciatica for lumbar discectomy - A prospective cohort study. Journal of Bone and Joint Surgery-British Volume 2004;86B:546-9. 28. Coyte PC, Wright JG, Hawker G A et al. Waiting-Times for Knee-Replacement Surgery in the United-States and Ontario. New England Journal of Medicine 1994;331:1068-71. 29. Williams JI, Llewellyn T H , Arshinoff R, Young N , and Naylor CD. The burden of waiting for hip and knee replacements in Ontario. Ontario Hip and Knee Replacement Project Team. Journal of Evaluation in Clinical Practice. 1997;3:59-68. 30. Sobolev B, Brown P, and Zelt D. Potential for bias in waiting time studies: events between enrolment and admission. Journal of Epidemiology and Community Health 2001;55:891-4. 185 C H A P T E R 8 8.0 E F F E C T OF WAITING T I M E O N O U T C O M E OF E L E C T I V E SURGICAL LUMBAR DISCECTOMY 3 8.1 Introduction Back pain is a highly prevalent condition and a major public health burden in all industrialized nations.1 In Canada, about 80% of people experience back pain at least once in their lives.2 Studies consistendy show that less than 10% of back pain patients account for more than 80% of the total health care and social costs for back pain care, and that the one percent who undergo surgery make up the most expensive group.3 It has been said that inpatient surgical (diagnostic and therapeutic) procedures account for up to one-third of total costs for all spinal disorders.4 In North America, the most common indication for spinal surgery is sciatica, or pain in the leg, resulting from lumbar intervertebral disc herniation.516 Surgical discectomy is effective compared to non-surgical treatment, particularly if patients are willing to accept quicker pain relief and functional recovery over the risk of rare, but not insignificant, surgical risks.7"9 Lumbar discectomy is also probably cost-effective as well. Maker analysed outcome and cost data for 126 patients who had been randomly assigned to medical or surgical treatment for radicular pain that had been unresponsive to initial conservative therapy.10 He found that for carefully selected patients with herniated lumbar disc, surgical discectomy provides substantial benefit with cost-* A version of this chapter will be submitted for publication. Quon J, Sobolev B, Levy A, Fisher C, Kopec J, Schechter M. (2007) Determinants of Waiting Time for Lumbar Discectomy in BC, Canada. 186 effectiveness that is comparable to other widely accepted therapies including coronary artery bypass grafting for single vessel coronary artery disease and medical therapy for moderate hypertension.10 However in Canada limited access to surgeons and operating rooms often results in delayed treatment." Patients who are clinically severe enough to warrant surgery must cope with pain, disability and impaired quality of life while awaiting effective care.12 The impact of delayed access to surgical discectomy on outcomes is unknown. For some orthopaedic conditions there are concerns that once a patient is deemed suitable for operative care, inordinate waiting may reduce the likelihood of a successful outcome after treatment. In studies of nonspinal orthopaedic surgery, patients who waited longer for hip replacement, for example, experienced smaller gains in health-related quality of life after surgery than patients who underwent surgery earlier.1*14 However, for patients awaiting surgical lumbar discectomy, the impact of delayed treatment is not so clear-cut. For example, in contrast to degenerative hip disease, the natural course of acute lumbar disc herniation is not universally progressive and therefore waiting can be beneficial by allowing some patients to recover without an operation. On the other hand, some studies have reported that longer preoperative symptoms is associated with worse outcomes after surgical lumbar discectomy.15 The mechanisms by which delayed surgery might adversely influence postoperative recovery are speculative. Nystrom reported that early connective tissue reaction around the nerve root is seen as early as one month after the first symptoms of disc herniation,16 and other studies have shown that nucleus pulposus tissue—from the centre of the intervertebral disc—provokes an inflammatory response and consequent irritation of the adjacent nerve roots once extruded peripherally into the spinal canal.17" 1 9 It remains a popular hypothesis, therefore, that earlier excision of inflammatory disc material reduces the likelihood of chronic, or repetitive inflammation and irreversible damage of the nerve 187 root environment. Others investigators also speculate that prolonged stimulation of primary afferent nerve fibres may lead to long-term (neuronal) plastic changes in nociceptive transmission in the spinal cord, which results in a form of nociceptive memory within the dorsal horn cells and, ultimately, chronic pain.2 1 ; 2 2 In Canada, waiting lists for elective surgical procedures are common yet the independent effect of waiting time on outcomes has been examined for only a limited number of procedures. To our knowledge, no studies have been published on the impact of waiting time or delayed service delivery on symptomatic improvement specifically after surgical lumbar discectomy. In one study in Ontario, the effect of waiting time for five different high volume procedures, including lumbar discectomy, showed no association between longer waiting and increased utilization of health care services before and after surgery.23 In that same study, clinical outcome was determined using health care utilization as a proxy measure. The use of administrative data also prevented the investigators from adjusting for preoperative disease severity and other prognostic variables at an individual level. In a prospective cohort study from Norway, elective orthopaedic surgery patients had significandy higher odds of failing to return to work within the first year after surgery if they had been on the waitlist for more than nine to 12 months (OR: 6.2) , or more than 12 months (OR: 9.2).24 The effect of waiting time on return-to-work specifically for discectomy patients was not reported. 8.1.1 Defining a Maximum Appropriate Waiting Time for Surgical Discectomy In this study, we sought to evaluate the effect of waiting time on the outcome of surgery using a patient-centred definition of an unacceptable delay in service delivery. At the time of study inception, there were no published reports of patients' perceptions of their waiting times for back surgery. During discussions with study surgeons, however, it was suggested that a waiting time of 188 six weeks for elective surgical lumbar discectomy was probably reasonable, and that a wait beyond 12 weeks was probably unreasonable based on local consensus and clinical intuition alone. In the meantime, two studies of patients' perceptions of waiting times for non-spinal orthopaedic surgery had been published. One study reported that 85% of patients in Ontario considered their waiting time for knee replacement as acceptable when the median time to surgery for the group overall was eight weeks.25 In another Ontario study the mean waiting time for knee surgery was 13.2 weeks, however only among patients who specifically described their waiting times as acceptable.26 The median wait in this latter group was not reported, but was conceivably less (than the mean of 13.2) given the typical, positively skewed, distribution of waiting times.27"30 A maximum appropriate waiting time of 3 months has also been proposed in other jurisdictions. In one study, cataract surgery patients in Manitoba, Denmark and Barcelona defined a maximum reasonable wait for "non-emergency surgery of an unspecified nature" (i.e. surgery not limited to cataract procedures) as three months or less.31 Also, in 1992, a maximum waiting time policy of three months was temporarily implemented in Sweden based on an estimation of what the maximum wait would be if "production" (i.e. the number of operations performed) and all waitlisted patients in Sweden were evenly distributed across procedures, hospital departments and 32 time. Subsequent to the time of study inception, other publications appeared in the literature, which further supported the use of 12 weeks (or three months) as a cut-off for an inappropriately long wait. In Ontario, rural and urban patients who were asked to define "an acceptable time" to wait for total hip or knee arthroplasty provided mean responses of 3.55 months (SD: 2.00) and 3.59 months (SD: 2.10), respectively.33 That same year, in a study of patients' perceptions of 189 waiting times for orthopaedic—including spinal—surgery in Sweden, patients who judged their wait for back surgery as acceptable and unacceptable reported median waiting times of 1.6 months and 4.4 months, respectively.34 In that study, 3 months was the midpoint between the median waiting times for the two groups. 8.1.2 Study Objectives and Hypotheses The objective of the current study was to determine if waiting longer than 12 weeks for surgical lumbar discectomy was associated with lower probability of pain reduction, after adjusting for clinical severity at the time of enrolment. 8.2 Methods We employed a prospective cohort design using an existing registry of surgical lumbar discectomy patients treated for low back pain between November 1999 and December 2003 in a tertiary care centre in Vancouver, Canada. Details on the functional status and quality of life outcomes have been published on a subset of patients in this cohort.35 For the current study, we extended recruitment for an additional two-years and focused on postoperative pain reduction as the outcome of interest. The target population for this study was working age adults with non-emergency indications for open or microscopic lumbar discectomy, managed within a publicly funded health care system such as that in the province of British Columbia (BC). This includes the majority of patients with back and/or leg pain caused by a confirmed disc herniation while excluding those requiring operative treatment on an emergency basis. The available population included surgical patients from ten participating orthopaedic and neurologic surgeons at Vancouver General Hospital (VGH), one of two main centres for lumbar disc operations within the Vancouver Coastal Health Authority of BC. About 75% of surgical disc patients at V G H are residents of the 190 city of Vancouver or the surrounding area.5 Another 23% of padents come from more remote locations within BC, while only 2% are from another jurisdiction. Access to individual surgeons is largely through referral from family physicians in the community or emergency physicians at V G H . 8.2.1 Study Population Al l consecutive patients who were scheduled for elective surgical lumbar discectomy at Vancouver General Hospital (VGH) between November 1999 and December 2003 were identified at the time of enrolment onto the surgical waitlist. These patients had been referred to V G H for a consultation with either an orthopaedic or neurological surgeon, or a nonsurgeon-physician specializing in conservative back care. Patients were referred mainly from general medical practitioners in the community or, less frequendy, emergency room physicians at V G H and other nearby hospitals. To be included in the registry, patients had to qualify for elective surgical lumbar discectomy according to the InterQual Indications for Surgical Procedures (ISP™) criteria,36 which included the presence of either: 1) a severe but non-progressive motor deficit (less than grade three strength) in the discrete distribution of a nerve root on examination; 2) a mild motor deficit in the discrete distribution of a nerve root with continued presence of mild deficit after six weeks of medication (NSAIDS, analgesics, corticosteroids, and/or muscle relaxants) limitation of provocative activities (heavy lifting, repetitive bending, etc.); or 3) unilateral leg pain that persisted after six weeks of medication and activity limitation. Patients also had to have evidence of a lumbar disc herniation on computed tomography (CT) or magnetic resonance imaging (MRI) at a location (i.e. disc level and side of the body) that was concordant with their clinical symptoms and signs. 191 Patients who were booked for emergency operations, due to the presence of a sudden onset of severe or progressive muscle weakness, or symptoms or other signs of cauda equina syndrome at the time of preoperative assessment, were excluded from the study. Patients were also excluded if they were younger than 16 years of age, pregnant, had prior surgery at the same level, or had a previous history of spinal malignancy, infection, osteoporosis, spondyloarthropathy, spinal deformity or other co-morbidity that could conceivably influence their outcome independendy of surgery. As a prerequisite for completing our questionnaires at six-months' follow-up, participants had to be proficient in reading and writing English. Al l patients consented to undergo surgical treatment after being carefully informed of the possible risks associated with surgical discectomy, including infection, anesthetic risks, and possible need for a repeat-operation. The protocol for this study was approved by the University of British Columbia's Clinical Research Ethics Board. 8.2.2 Measurements 8.2.2.1 Outcome The outcome of interest was pain improvement at six-months postoperatively. While the biological rationale for surgical discectomy is to relieve pressure on a nerve root from herniated disc material, the primary clinical indication for this procedure is persistent or intractable pain in the back and/or leg (i.e. sciatica). Therefore, back pain and leg pain intensities were measured separately using the 11-point numerical rating scale (NRS-11). Scores on the NRS-11 instrument range from zero, meaning "no pain", to 10 meaning "pain as bad as it could be". The more intensely rated symptom (back or leg pain) at preoperative assessment was designated the predominant symptom and pain improvement was defined as the change in NRS-11 pain ratings for the predominant symptom between enrolment and six-months postoperatively. 192 8.2.2.3 Independent Variable Waiting time, in weeks, was calculated by subtracting the date of enrolment onto the surgical waitlist from the date of surgery. This variable was dichotomized using 12 weeks as the cut-off for defining a long versus short waiting time. 8.2.2.4 Confounders We identified important prognostic variables for surgically and non-surgically treated lumbar disc herniation from the literature.15;37;38 Information on these and other clinical variables were recorded by attending physicians and surgeons during an outpatient visit, typically at the time of study enrolment and placement onto the waitlist. Location of disc herniation and nerve root compression, and presence or absence of bony stenosis were confirmed through computerized tomographic and/or magnetic resonance imaging. Sociodemographic variables, prior and current medical history, and current examination findings were recorded on standardized forms. Neighbourhood socioeconomic status was ascertained prior to data analysis by linking postal codes and 2001 census enumeration areas for each patient. A research assistant administered baseline questionnaires on the date of enrolment onto the waitlist (usually coinciding with the date of assessment by the surgeon). We used the North American Spine Society (NASS) lumbar outcome assessment instrument,39'40 which includes the NASS Pain and Disability scale, the NASS Neurogenic Symptoms scale, and the SF-36.41 Patients also completed a brief employment information questionnaire, the Beck Depression Inventory,42 and a pain drawing. On a section of the employment information questionnaire, patients were asked if they were receiving workers' or other disability compensation for their current condition. Another questionnaire was used to document the duration of the current episode of pain symptoms ("< 6-weeks, "6-weeks to 3-months", 3-months to 6-months" or "> 6 months"), and 193 the number of days within the past four-weeks patients had to "stay in bed", "cut down on regular activities", or "miss work or school" due to back pain or sciatica. Follow-up questionnaires were mailed at six-months after surgery. A research assistant also contacted patients by telephone if they did not return their follow-up questionnaires within a month. 8.2.3 Analysis 8.2.3.1 Descriptive Analyses Univariate distributions of sociodemographic, clinical and health system variables were examined in frequency tables and histograms. Sparsely populated levels within categorical variables were eliminated by regrouping sparse categories. The distributions of baseline sociodemographic, clinical, and health system variables were compared using Fisher's exact or Pearson's chi-squared tests for categorical variables, and Student's t-test for continuous variables. 8.2.3.2 Unadjusted Effect of Waiting Time In the primary analysis, ordinal logistic regression was used to estimate the effect of waiting 12 weeks or longer over multiple, ranked definitions of improvement. A six-point ordinal outcome variable was created by collapsing pain improvement scores into relatively equally populated groups. The use of an ordinal rather than continuous scale of pain improvement was preferable since we could not assume that a one-unit change was uniform over all points on the NRS-11. 4 3" 4 5 Improvement scores less than 0 (i.e. worsening), 0-1, 2-3, 4-5, 6-7 and 8 or greater were recoded to ordinal categories of 0,1, 2, 3, 4 and 5, respectively. First, a proportional odds model was fitted. At each cut-point, patients either achieving or exceeding a given level of improvement (i.e. patients in all categories > J) were compared to reference patients not achieving that level of improvement (i.e. patients in all categories < j). A proportional odds ratio (POR) was then obtained, summarizing the association between a long 194 wait and the likelihood of improvement over five increasingly stringent definitions of improvement on the ordinal scale. The Score test of heterogeneity was used to verify that the proportional odds ratio obtained in this manner was representative over all possible dichotomizations of the ordinal improvement scale.46 Second, a continuation ratio model was fitted. At each cut-point, patients exceeding a given level of improvement (i.e. patients in all categories > j) were compared to reference patients who had achieved only that level of improvement (i.e. patients in category = j). This method operationally involves the creation of a series of 2 x 2 contingency tables such that patients in the lowest level of improvement in each table are used as the reference level, and are subsequently excluded from the next table.47 A continuation ratio (CR) statistic is then obtained, which summarizes the association between a longer wait and the likelihood of pain improvement over all contingency tables and, again, over five increasingly stringent definitions of conditional pain improvement. To test for heterogeneity in the continuation ratio over multiple cut-points, interaction terms between wait group and dummy variables for each contingency table representing each possible cut-point on our ordinal outcome were included in the model. The log-likelihoods were then compared between models including and excluding these interaction terms. 8.2.3.3 Adjusted Effect of Waiting Time We identified from the literature potentially important prognostic variables and determinants of waiting time for surgical lumbar discectomy.15 Some of these potential confounders were known to be correlated with waiting time, the primary exposure of interest. To eliminate collinearity between independent variables we used propensity scores to control for prognostic imbalances between the long and short wait groups.48 The use of propensity scores addressed concerns about the correlation between waiting time and other prognostic variables for 195 surgical discectomy (e.g. pain severity, and urgency status at the time of waitlist enrolment) and removed the limit on the number of covariates that could be included in our analysis. The propensity score, or the probability of membership in the longer wait group conditional upon the patient's individual covariate pattern, was formulated using multivariable logistic regression to model a dichotomous outcome of waiting 12 or more weeks versus less than 12 weeks for each patient. Al l potential confounders were used in a backward stepwise regression. Rosenbaum cautions against selecting only variables that were associated with "treatment" at a conventional significance level.49 We therefore retained variables using p < 0.3 and also included clinically plausible confounders (age and sex) regardless of their statistical significance. Balance in the distribution of retained covariates between wait groups was assessed using univariate logistic regression for categorical variables and univariate linear regression for continuous variables. The degree of overlap, and therefore comparability, in propensity scores between the long and short wait groups was inspected in histograms. Two-way interactions between the covariates were also examined and were retained in the formulation model if they improved the degree of overlap between wait groups. 8.2.3.4 Sensitivity Analyses - Conventional Multivariable Outcome Model We obtained an adjusted estimate of the effect of waiting 12 weeks or longer on pain reduction after surgical discectomy using a conventional multivariable outcome model. Using a process of purposeful selection, bivariate associations between pain improvement and potential confounders were examined within univariable ordinal logistic regression models, and significant bivariate correlates of improvement were examined joindy within a preliminary multivariable model. Any covariates that were no longer associated with improvement at a multivariable level were dropped one at a time, initially by using p < 0.3 and then subsequendy using p < 0.05. As a 196 final assessment of potential confounders, variables that were initially eliminated were re-introduced one at a time, and the coefficients of covariates already in the model were examined for changes exceeding 20%. Age, sex and compensation status were retained in the final outcome model regardless of their statistical significance. We also assessed the significance of two-way interactions between wait group and each covariate within the final model. 8.3 Results Figures 8.1 and 8.2, respectively, show the flow of patients and the sampling scheme for the study. Between November 1999 and December 2003, a total of 576 surgical lumbar discectomy patients were flagged for the study, of which 129 were either clearly ineligible or refused to participate from the outset. Another 42 were subsequendy excluded when a history of prior surgery at the current level of herniation became manifest. Of 402 appropriately enrolled patients, 111 (28%) were excluded from the final analysis because of missing outcome information at six-months follow-up (102 patients). Table 8.1 shows that patients with missing follow-up information did not differ significandy from the rest of the study population in terms of baseline characteristics. Excluded patients were slighdy less likely to be married or to be receiving disability compensation for their current condition. 8.3.1 Waiting Times Figure 8.3 shows the cumulative probability of accessing surgical lumbar discectomy for patients in the study. The distribution of waiting was positively skewed; the median waiting time for the entire study population was 6 weeks (interquartile range P Q R ] : 11.0) while the mean was 11 weeks (SD: 14.7). Seventy-five percent of patients had surgery within 14 weeks of waiting while 95% had their surgeries within 36 weeks. 197 Within individual wait groups, the median and mean waiting times for patients in the short wait group were 3 weeks (IQR: 5) and 3.9 weeks (SD: 3.2), respectively, while the median and mean waiting times for those in the long wait group were 19 (IQR: 16.0) and 25.9 (SD: 17.8) weeks, respectively. 8.3.2 Study Population Patients in this cohort were mostly male (61%), nonsmokers (67%), currendy employed (62%), and married or living with a partner (68%) (Table 8.2). Most described their usual occupation as a professional or managerial position (53%). Despite generally high baseline pain intensity scores in this population, only 17% were currently unemployed specifically because of their current back or leg symptoms. Most patients had had their current back/leg pain for longer than six-months (47%)(Table 8.3). Positive straight leg raising was documented in 89%, and concurrent radiographic evidence of stenosis in only 9% of study patients. The majority of patients were operated on by neurosurgeons (69%). Twenty percent of patients were receiving workers' compensation or other disability benefits for their current condition. Compared to patients who waited less than 12 weeks for surgical discectomy, patients who waited longer were more likely to be male, unemployed, suffering from current symptoms for more than six-months, and receiving disability compensation at the time of queuing. Patients who waited longer were also less likely to have exhibited a positive straight leg raising test or a sensory neurological deficit at baseline, and less likely to be classified as "urgent" or "semi-urgent" (versus non-urgent) at the time of waitlist enrolment. Table 8.4 shows that patients who waited 12 weeks or longer were also less likely to have been treated by a surgeon whose rate of recruitment was in the highest tertile. 198 8.3.3 Crude and Adjusted Effects of "Wait Group" on Ordinal Improvement Table 8.5 shows the frequencies and proportions of patients achieving different degrees of pain improvement on a six-level ordinal scale, according to short and long wait group. Patients who either deteriorated or improved by less than two-points on the NRS-11 (i.e. two lowermost categories in Table 8.5) were more common in the long-wait group. In contrast, patients who experienced either complete pain relief or improvement of at least six-points on the NRS-11 pain scale (i.e. two uppermost categories in Table 8.5) were more common in the short-wait group. The two intermediate categories of improvement were represented almost equally by patients in each group. 8.3.4 Propensity Score Model Table 8.6 lists the variables and interactions that were retained in the final model used to formulate propensity scores. The variables included age, sex, type of occupation (professional/manager versus other), compensation status, symptom duration (< 3 months, 3-6 months, > 6 months), pain score at enrolment, Beck Depression Inventory score at enrolment, motor status, straight leg raising (positive versus negative), and urgency status at enrolment (non-urgent versus urgent). The interactions between occupation type and compensation status, symptom duration and urgency status, and pain score and depression score were included to improve overlap in the distribution of propensity scores between wait groups. Table 8.7 shows the distributions of each covariate and the results of univariate tests of association between variables and dichotomized waiting time. Column 4 shows that prior to adjustment for propensity score the wait groups differed significantly with respect to sex, symptom duration, baseline pain score, and proportions with motor weakness, positive straight leg raising and non-urgent status at enrolment. A trend toward a significant difference between wait groups was also observed for compensation status and Beck Depression score. Although the distribution 199 of age and occupation type did not differ significandy between groups, age was included because of its clinical significance, and occupation type was included because it was a significant predictor of time to surgery in our previous study. Column 5 of Table 8.7 shows that after adjustment for propensity score, no significant difference between groups existed for the main effects of the covariates. An assessment of interactions between wait group and propensity score quintile suggested a possible imbalance only for age. We therefore adjusted for age and propensity score, however the coefficient for waiting time remained the same regardless of whether or not age was included and therefore age was excluded from the final model. There was adequate overlap in the distribution of the logit of the propensity scores between wait groups (Figure 8.4), and estimates of the effect of waiting 12 weeks or longer on the likelihood of ordinal pain improvement at six months after surgical lumbar discectomy are presented in Table 8.8. Both the proportional odds ratio (from the proportional odds model) and continuation ratio (from the continuation ratio model) represent summary estimates of the effect of waiting 12 or more weeks over multiple dichotomizations of the ordinal improvement scale. The upper panel of Table 8.8 shows the unadjusted proportional odds ratios (Column 2) and continuation ratios (Column 4). The middle and lower panels show the adjusted ratios from the propensity score and conventional multivariable adjusted models, respectively. None of the tests for heterogeneity was statistically significant, supporting the assumption of proportionality in the odds ratios over multiple strata. The estimated proportional odds and continuation ratios were therefore representative of the effect of waiting time over multiple dichotomizations of the ordinal scale of improvement. The p values for these tests of heterogeneity were universally higher for the 200 continuation ratio models, suggesting that conformity to the assumption of proportionality was perhaps greater under the continuation ratio model than under the proportional odds model. Both continuation ratios and proportional odds ratios indicated a clinically important effect of waiting longer for surgery. The unadjusted continuation ratio was 0.50 (95% CI: 0.35-0.70), indicating that patients who waited 12 weeks or longer were 50% less likely to improve than patients who waited shorter than 12 weeks for surgical lumbar discectomy. After adjusting for the propensity score the continuation ratio was 0.59 (95% CI: 0.40-0.87), indicating still a 41% reduction in the likelihood of recovery for patients who waited longer for surgery. In comparison, the unadjusted proportional odds ratio was 0.41 (95% CI: 0.26-0.64) while the propensity score adjusted proportional odds ratios was 0.53 (95% CI: 0.32-0.88). 8 .3 .5 Sensitivity Analysis We re-estimated the effect of waiting time while further adjusting for age by including it as a separate covariate in the model. The effect of dichotomized waiting time—adjusted further for age—was 0.58 (95% CI: 40-80), which was almost identical to the effect of wait group without age in the model. Table 8.8 also shows that in a conventional multivariable model (which included age, sex, compensation status, Beck Depression Inventory score, NRS-11 pain score at enrolment, and categorical symptom duration as covariates) the estimate of the effect of a longer wait was also more conservative than its corresponding unadjusted estimate. The multivariable adjusted continuation ratio of 0.65 (95% CI: 0.44-0.96) indicated a 35% lower likelihood of improvement for patients waiting 12 or more weeks for surgery. The corresponding multivariable adjusted proportional odds ratio was almost identical (0.64 [95% CI: 0.40-1.04]), however the precision of the estimate was lower than that which was obtained from the continuation ratio model. 201 In sensitivity analyses in which waiting time was dichotomized at six weeks, conventional multivariable proportional odds and continuation ratio models indicated 35% (POR: 0.65; 95% CI: 0.40-1.06) and 29% (CR: 0.71; 95% CI: 0.48-1.05) reductions, respectively, in the likelihood of ordinal pain improvement for patients waiting longer than six weeks (data not shown). These estimates were not statistically significant at the conventional 5% level. However, when considered in combination with the estimates at 12 weeks, they suggested the presence of a dose-response relationship in the expected direction between waiting time and the probability of improvement. In Tables 8.9 and 8.10, the point estimates for symptom duration indicated an inverted J-shaped dose-response relationship between symptom duration and the likelihood of improvement. However, as is shown in the bottom of both Tables 8.9 and 8.10, patients with a symptom duration of 6 weeks to 3 months were significandy more likely to improve (POR: 3.61 [95% CI: 1.61-8.10]; COR: 2.72 [95% CI: 1.39-5.32]) than reference patients in the shortest symptom duration category (< 6-weeks), while improvement among patients in the two longest symptom duration groups (3 to 5.9 months, and > 6 months) was not significandy different from the reference group. 8.4 Discussion We showed that patients who waited 12 weeks or longer for surgical lumbar discectomy were approximately 40% less likely to experience pain improvement compared to patients waiting less than 12 weeks. This finding was consistent across six increasingly stringent definitions of improvement on an ordinal scale (> 0, > 2, >_4, > 6, >. 8 points), two ordinal regression models (proportional odds and continuation ratio models), and two methods of adjusting for confounders (propensity score adjustment and conventional multivariable adjustment). Our findings suggest 202 that for patients with back pain and/or sciatica due to lumbar disc herniation with both appropriate indications and a willingness to undergo surgery, a delay in treatment of 12 weeks or longer is associated with a clinically important deterioration in the outcome of surgery. To our knowledge, an association between waiting time and the clinical outcome of surgical lumbar discectomy has not been previously reported. Until now, symptom duration at either the time preoperative consultation or at the time of surgery of has been associated with postoperative outcomes in a number of observational studies.35 However, total preoperative symptom duration is comprised of many intervals of potential interest, including, but not limited to: 1) the time between back/leg pain onset and first encounter with the health care system; 2) the time between the first encounter with a primary care physician and the decision to refer to a surgeon; 3) the time between referral and actual consultation with a surgeon; and 4) the time from when surgery is decided upon and the date that surgery is actually performed. This last interval typically corresponds to the time a patient waits for surgery from the time of enrolment onto the waitlist. Time from waitlist enrolment to the date of surgery also represents the delay in access to treatment for those who have reached a critical threshold of disease severity warranting operative treatment. Specifically, the time of waitlist enrolment represents the confluence of two important events: 1) the presence of evidence-based indications meeting a surgeon's threshold for recommending operative treatment, and 2) an occurrence of symptoms that are severe and/or prolonged enough to shift a patient towards accepting (if not, preferring) invasive treatment over less risky non-surgical care. This threshold clearly distinguishes time on the waitlist (i.e. waiting time for surgery) from preoperative symptom duration in general. Also, in contrast to symptom duration in general—which is dependent upon retrospective recall and, often, varying definitions of a discrete episode of back pain at the time of preoperative screening—waiting time for surgery is easily and objectively measurable. In prospective or 203 registry-based studies, this variable is not susceptible to the vagaries of retrospective recall. It is also the period of greatest interest from a service delivery perspective since it represents the delay in access to treatment that is least dependent on patient-related circumstances and most dependent on the capacity of the health care system.50 Waiting times are, therefore, important performance measure for health services research. The results of our current study suggest that the proportion of patients treated within 12 weeks of waitlist enrolment may have utility as an additional performance measure, particularly for health services research involving spinal surgery. 8.4.1 Strengths and Limitations In this study, we focused on the outcome of patients at six months, however other studies of the outcome of surgical discectomy have shown that postoperative recovery is nearly complete at six months, and that outcome measurements at 12 months postoperatively rarely differ significandy from those at six months.15 In the current study, it was possible that short-wait patients differed prognostically from long-wait patients in terms of unknown and/or unmeasured factors. We adjusted for potential confounders and clinical severity using propensity scores, however in the absence of randomization, confounding by indication is always a potential concern. In a study of the effect of waiting time on the outcome lumbar disc herniation, confounding by indication could result in a liberal bias. Patients with greater pain intensity are generally classified as more urgent and are likely to access surgery faster than patients with low pain intensity. Higher preoperative pain intensity is associated with a greater probability of pain reduction after surgery, which could create a bias in favour of a better outcome in the short-wait group. Short-wait patients are also more likely to have shorter symptom duration. As shorter symptom duration is also associated with better outcomes following surgical lumbar discectomy, this could also have generated a bias in favour of our observed result. Ultimately, evidence from a randomized controlled trial of early versus 204 delayed surgical lumbar discectomy is desirable, but ethically challenging as it would require careful justification of an intentional delay in treatment for patients in one of the study arms. Source variables for waiting time measurements were improperly recorded for a small proportion (3%) of study patients. We have no reason to believe that these data entry errors were systematically related to patients' outcomes. In addition, given their small proportion, exclusion of these patients from our analysis should have resulted in litde bias if any. On the other hand, approximately 27% of patients did not respond to the six-month' follow-up questionnaire. Non-respondents did not differ significandy from respondents on most baseline characteristics. However, patients with missing outcome data were less likely to be married or to be receiving disability compensation. Marital status was not found to be associated with either waiting time or pain improvement, however in conventional multivariable models compensation status (i.e. receiving compensation) was associated with a worse outcome. We were not able to control for socioeconomic status (SES) at an individual level, however we were able to link individual postal codes to census enumeration areas and were therefore able to ascertain SES at an ecologic level. We found no association between our ecologic measures of SES and postoperative improvement. While this does not rule-out the possibility of residual confounding by SES at an individual level, the literature suggests that the effect(s) of SES, if any, are modest in comparison to those associated with clinical severity or medical need.51 This study had other notable methodological strengths. This was a prospective study involving primary data collection and the use of both a clinically relevant outcome and a well-defined inception point (date of enrolment onto the surgical waitlist). Waiting times were calculated direcdy from the date of waitlist enrolment and the date of surgery, both of which were documented prospectively within an existing patient registry. Although this was not a randomized 205 trial, we measured a rich assortment of sociodemographic, occupational, clinical, psychological, and health system/institutional variables and were able to adjust for adjust for all statistically significant covariates and clinically plausible predictors, particularly those related to disease severity. We also found a dose-response relationship in the expected direction of effect, further supporting a genuine association between longer waiting time and a lower probability of pain improvement. 8.5 Summary Prior to the completion of this study, it was known that longer symptom duration prior to elective surgical lumbar discectomy was associated with poorer outcomes afterward. What was not known was whether one particular component of symptom duration—waiting time for surgery— was also independently associated with the outcome of surgery. Our study showed that waiting 12 weeks or longer for surgery was associated with an important deterioration in the probability of pain reduction at six months postoperatively. The results of this study support the use of 12 weeks as a clinically relevant benchmark for defining an inappropriate delay in elective surgical lumbar discectomy. Further research, possibly including a randomized trial of early versus delayed surgery, is encouraged to corroborate our findings. Given the existing evidence to date, "the proportion of patients treated within 12 weeks of waitlist enrolment" has potential utility as both a relevant outcome measure in health services research, and as a practical quality performance measure for the public health care system. 206 Figure 8.1: Patient Flow Diagram Patients with back/leg pain and concordant C T / M R I evidence of disc herniation Yes Consent to surgery Yes Booked/waitlisted for surgery Yes Cons follov cnt to v-up? Yes Baseline questionnaires Sur. perfo ?ery rmed Six-month follow-up questionnaires H N o * \ N o +\ N o Ineligible patients 207 Figure 8.2: Patient Sampling Scheme 576-Patients booked for surgical discectomy Solicited & screened for study eligibility 444-Initially enrolled in the study * Completion of data collection forms 402-Appropriately enrolled primary lumbar discectomy patients w 393-Total number of patients with valid waiting time information w 291-Total number of patients with complete waiting time and outcome information 9-Patients with missing or invalid values for waiting time source 102-Patients with missing outcome information at 6 months 208 132-Refused or not eligible to participate 42-Patients with previous surgery at the same level Figure 8.3: Kaplan-Meier Time-to-Surgery Curve Probability of remaining on the waitlist for surgical lumbar discectomy over time for all patients in the study. 209 Table 8.1: Characteristics of Excluded Patients Sociodemographic, clinical, and health system-related factors for included and excluded patients. Characteristic Study population^ Excluded patientsf^ : P value* Number of patients 291 102 Age (SD) 43 (12.8) 41 (9.4) 0.26 Sex Men 179 ,(62) 60 (59) 0.64 Women 112(38) 42 (41) Marital status Not married 103 (35) 49 (48) 0.025 Married 188 (65) 53 (52) Employment status Not employed 109 (38) 32 (33) 0.39 Employed 178 (62) 65 (67) Usual occupation Non-professional 138 (51) 50 (57) 0.27 Professional 135 (49) 37 (43) Smoker Never/Not in last 6 192 (67) 59 (61) 0.27 months 95 (33) 38 (39) Current/recent smoker Duration of current symptoms 33 (11) 7(8) 0.18** < 6 weeks 53 (18) 12 (13) 6 weeks-2.9 months 68 (24) 31 (34) 3-5.9 months 135 (47) 41 (45) > 6 months Pain score at enrolment Mean (standard 7.4 (2.1) 7.4 (2.1) 0.99 deviation) Weakness No 126 (45) 32 (36) 0.18 Yes 157 (55) 57 (64) Abnormal reflexes Absent 144 (51) 41 (46) 0.46 Present 137 (49) 48 (54) Sensory loss No 114(41) 32 (36) 0.45 Yes 162 (59) 56 (64) Straight leg raising Negative 30 (11) 9 (10) 1.0 Positive 251 (89) 81 (90) Beck Depression score 12.6 (8.0) 14.6 (10.0) 0.08 210 C h a r a c t e r i s t i c S t u d y p o p u l a t i o n f E x c l u d e d p a t i e n t s f $ P v a l u e * N u m b e r o f p a t i e n t s 2 9 1 1 0 2 Hospital service Orthopaedics 90 (31) 35 (34) 0.54 Neurosurgery 201 (69) 67 (66) Surgeon's "rate" of recruitment 205 (70) 68 (67) 0.013** High 78 (27) 24 (24) Medium 8(3) 10 (10) Low Urgency at booking Urgent 124 (43) 42 (41) 0.82 Non-urgent 167 (57) 60 (59) Waiting time Long 94 (33) 36 (36) 0.72 Medium 58 (20) 17 (17) Short 134(46) 46 (46) Disability compensation Not receiving 232 (80) 95 (93) 0.001 Receiving 59 (20) 7(7) *Fisher's exact test for categorical variables, or Student's t-test for continuous variables, unless otherwise specified. **Pearson Chi-squared test for categorical variables. fFrequency count with column percentage in parentheses, unless otherwise specified, f Due to missing outcome/follow-up information 211 Table 8.2: Characteristics of Study Population Sociodemographic characteristics of patients at preoperative evaluation by waiting time groups Short Long Characteristic (< 12 weeks) (> 12 weeks) Overallf p value* Number of patients 197 94 291 Age Mean (SD) 42 (11.6) 44 (15.1) 42 (12.8) 0.30 Males 113(57) 66 (70) 179 (62) 0.04 Females 84 (43) 28 (30) 112 (38) Marital status Not married/with partner 75 (38) 28 (30) 103 (35) 0.32 Married/with partner 122 (62) 66 (70) 188 (65) Average neighbourhood income:): Mean (SD) 32,489 (8,744) 33,315 (12,228) 32,775 (9,986) 0.57 Median neighbourhood income^ Mean (SD) 25,002 (5,873) 24,054 (5,762) 24,696 (5,844) 0.20 Living in urban neighbourhood^: N o 38 (20) 24 (27) 62 (22) 0.22 Yes 151 (80) 66 (73) 217 (78) Percent with bachelor's degree in neighbourhood^ Mean (SD) 24 (13.6) 25 (14.7) 24 (14.0) 0.62 Currendy employed N o 67 (34) 42 (46) 109 (38) 0.07 Yes 128 (66) 50 (54) 178 (62) Usual type of work Semi-/unskilled labour 10(5) 8(9) 17(7) 0.45** Skilled labour 55 (30) 26 (30) 81 (30) Sales or service 23 (12) 9(10) 32 (12) Professional/manager/student 97 (53) 45 (51) 142 (52) Main employment activity Sitting 63 (33) 35 (39) 98 (35) 0.52** Walking/standing 21 (11) 13 (14) 34 (12) Seldom lifting > 35 kg 58 (30) 20 (22) 77 (28) Awkward lifting > 35 kg 31 (16) 12 (13) 43 (15) Maximal/awkward lifts 18(9) 10(11) 28 (10) Unemployed specifically due to back/leg pain N o 164(85) 74 (80) 238 (83) 0.40 Yes 30 (15) 18(20) 48 (17) Smoking status Never/Not in last 6 months 133 (68) 63 (67) 196 (67) 1.0 Currendy/recent smoker 64 (32) 31 (33) 95 (33) On disability compensation N o 162 (82) 70 (74) 232 (80) 0.16 Yes 35 (18) 24 (26) 59 (20) *Using the Fisher exact test for categorical variables or Student's t test for continuous variables, unless otherwise noted. **Pearson Chi-squared test for categorical variables. fNumber with column percentages in parentheses unless otherwise noted. ^Based on 2001 census data, linked by postal code, to enumeration area patient resided in at time of preoperative assessment 212 Table 8.3: Clinical Characteristics of Study Patients^ Clinical characteristics of patients at preoperative evaluation by waiting time group. Short Long Characteristic (< 12 weeks) (> 12 weeks) Overall p value* Number of patients 197 94 291 Baseline NRS-11 pain score Mean (SD) 7.6 (2.0) 7.1 (2.3) 7.4 (2.2) 0.04 Pain duration < 6 weeks 27 (14) 6(6) 33 (11) < 0.0001 6 weeks - 3 months 48 (24) 5(5) 53 (18) 3 — 6 months 48 (24) 20 (22) 68 (24) > 6 months 73 (37) 62 (67) 135 (47) Muscle weakness No 72 (37) 54 (61) 126 (45) 0.0003 Yes 122 (63) 35 (39) 157 (55) Reflex loss No 96 (50) 48 (55) 144 (51) 0.52 Yes 97 (50) 40 (45) 137 (49) Sensory deficit No 69 (37) 45 (51) 114 (41) 0.04 Yes 118 (63) 44 (49) 162 (59) Positive straight leg raising No 12(6) 18 (20) 30 (11) 0.0007 Yes 180 (94) 71 (80) 251 (89) Stenosis on C T / M R imaging No 179 (93) 75 (87) 254 (91) 0.17 Yes 14(7) 11 (13) 25(9) Baseline BDI score Mean (SD) 13.1 (8.0) 11.5 (8.2) 12.6 (8.0) 0.13 Urgency status at booking Urgent or semi-urgent 97 (49) 27 (29) 124 (43) 0.0009 Non-urgent 100 (51) 67 (71) 162 (57) **Using the Fisher exact test for categorical variables or one-way analysis of variance for continuous variables, unless otherwise noted. **Pearson Chi-squared test for categorical variables. fNumber with column percentages in parentheses unless otherwise noted. 213 Table 8.4: Health System Characteristics of Study Patients Health system characteristics of patients at preoperative evaluation by waiting time quartile. Characteristicf Short (< 12 weeks) Long (> 12 weeks) Overall p value* Number of patients 197 94 291 Hospital service Orthopaedics 52 (26) 38 (40) 90 (31) 0.02 Neurosurgery 145 (74) 56 (60) 201 (69) Surgeon's "rate" of recruitment^ T3: (1.5-6.8 weeks) 150 (76) 55 (59) 205 (70) 0.008** T2: (6.9-10.9 weeks) 42 (21) 36 (38) 78 (27) T l : (11.0-14.0 weeks) 5(3) 3(3) 8(3) **Using the Fisher exact test for categorical variables or one-way analysis of variance for continuous variables, unless otherwise noted. **Pearson Chi-squared test for categorical variables. fNumber with column percentages in parentheses unless otherwise noted. §Tertiles based on estimated average number of weeks between enrolment of patients into the study (between dates of first, and last, patients recruited by each surgeon) 214 Table 8.5: Frequencies of Numbers of Patients in Each Recovery Category Description of ordinal scale of pain improvement and the frequency (column %) of patients achieving each level of improvement according to wait group. Category of pain Short Long Overall improvement (change in (wait < 12 weeks) (wait > 12 weeks) NRS-11 pain score) 6 ( > 8 points) 32 (16) 6(6) 38 (13) 5 (6-7 points) 56 (28) 14(15) 70 (24) 4 (4-5 points) 39 (20) 23 (25) 62 (21) 3 (2-3 points) 36 (18) 19 (20) 55 (19) 2 (0-1 point) 24 (12) 25 (27) 49 (17) 1 (< 0 points) 10(5) 7(7) 17(6) Total 197 (100) 94 (100) 291 215 Table 8.6: Variables in the propensity score model Multivariable associations between long waiting time (> 12 weeks) and covariates in the model for formuladng propensity scores. Variable Odds 95% CL* p-value Ratio Lower Upper Age (per 10-year increment) 1.12 0.93 1.49 0.18 Female 0.61 0.31 1.18 0.14 Professional/managerial worker 0.88 0.44 1.75 0.71 Receiving disability compensation 1.04 0.45 2.41 0.93 Symptom duration < 3 months 1.0 3 to 5.9 months 2.21 0.35 14.08 0.40 >. 6 months 6.98 1.42 34.20 0.02 Pain score (per 2-point increment) 0.61 0.38 0.98 0.04 Beck depression score (per 15-point increment) 0.31 0.03 3.30 0.33 Motor weakness 0.48 0.26 0.88 0.02 Positive straight-leg raise test 0.30 0.12 0.76 0.01 Non-urgent at enrolment 3.99 0.64 17.80 0.15 Interactions Professional worker x Compensation 4.52 0.69 29.66 0.12 Duration 3 to 5.9 months x Non-urgent 1.04 0.12 8.64 0.97 Duration > 6 months x Non-urgent 0.72 0.12 4.51 0.73 Pain score x Beck depression score 1.32 0.74 2.34 0.35 95% C L = 95% confidence limits 216 T a b l e 8.7: B a l a n c e i n C o v a r i a t e s B e f o r e a n d A f t e r A d j u s t i n g f o r P r o p e n s i t y S c o r e Results of univariate tests of association between long wait (> 12 weeks) and study factors before and after adjustment for propensity score. V a r i a b l e s W a i t G r o u p * p - v a l u e * S h o r t L o n g U n a d j u s t e d A d j u s t e d < 1 2 w k s > 1 2 w k s f o r l o g i t - P S f Age (mean[SD]) 42.0 (11.6) 43.8 (15.1) 0.26 0.93| Female 84 (43%) 28 (30%) 0.04 0.79 Professional/manager 97 (50%) 44 (47%) 0.64 0.90 Receiving compensation 35 (18%) 24 (25%) 0.12 0.85 Symptom duration < 3months 75 (38%) 12(13%) < 0.0001 0.77 3 to 5.9 months 49 (25%) 20 (21%) 0.50 0.68 >. 6 months 73 (37%) 62 (66%) < 0.0001 0.78t Pain score (mean[SD]) 7.6 (2.0) 7.1 (2.3) 0.04 0.95 Depression score (mean [SD]) 13.1 (8.0) 11.5 (8.2) 0.13 0.82 Motor weakness 122 (62%) 35 (37%) < 0.0001 0.93 Positive straight-leg raise 180(91%) 71 (76%) 0.0004 0.98 Non-urgent at enrolment 102 (52%) 67 (71%) 0.002 0.90 *From logistic regression for binary (or indicator) variables, and linear regression for continuous variables. *Logit-PS = logit of propensity score ^Significant interaction between long wait and logit of propensity score 217 Figure 8.4: Histograms of Propensity Score Distributions Comparison of the distribution of propensity scores on a logit scale, between wait groups. clatps Short Wait 0.35" c 0.30-° 0.25-|_ 0.20" •2 0.15-Q. 0.10" 0.05" Short Wait Long Wait 0 0.35-| o;25-o 0.20 " ,S 0.15-CL 0.10" 0.05" Long Wait U 1 1 1 1 1 1; .1 -4 - 3 - 2 - 1 0 1 2 Log ft of Propensity Score 218 Table 8.8: Unadjusted and Adjusted Effects of Waiting Time Adjusted effect of waiting 12 weeks or longer on the probability of pain reduction on an ordinal scale, six months after undergoing surgical lumbar discectomy. Proportional Odds Model Continuation Ratio Model PORf (95% P value CR$ (95% CI) p value CI) Unadjusted Short wait (< 12 weeks) 1.0 1.0 Long wait (> 12 weeks) 0.41 (0.26-0.64) < 0.0001 0.50 (0.35-0.70) < 0.0001 Test of heterogeneity 0.52 0.57 Propensity Score Adjusted* Short wait (< 12 weeks) 1.0 1.0 Long wait (> 12 weeks) 0.53 (0.32-0.88) 0.014* 0.59 (0.40-0.87) 0.008 Test of heterogeneity 0.71 0.79 Conventional Multivariable Adjusted Model Short wait (< 12 weeks) 1.0 1.0 Long wait (> 12 weeks) 0.64(0.40-1.04) 0.07 0.65 (0.44-0.96) 0.03 Test of heterogeneity 0.16 0.67 *Adjusted for logit of propensity score. **Adjusted for age, sex, compensation status, Beck Depression Inventory score, NRS-11 pain score at enrolment, and categorical symptom duration. fPOR = proportional odds ratio. ^CR = continuation ratio 219 T a b l e 8 . 9 : E f f e c t o f W a i t i n g T i m e - M u l t i v a r i a b l e P r o p o r t i o n a l O d d s M o d e l Results of proportional odds ordinal logistic regression "outcome" model showing unadjusted and adjusted effects age, sex, waiting time other significant correlates of ordinal improvement over multiple definitions of improvement, at six-months after surgery. E f f e c t O R U n a d j u s t e d e f f e c t s 9 5 % C I p v a l u e * O R A d j u s t e d e f f e c t s 9 5 % C I p v a l u e * Waiting time Short 1.0 1.0 Long 0.42 0.27-0.65 0.0001 0.64 0.40-1.04. 0.07 Age/10-years 0.91 0.78-1.07 0.25 0.74 0.63-0.88 0.0006 Sex Male 1.0 1.0 Female 1.85 1.21-2.82 0.004 1.04 0.66-1.62 0.88 Compensation status No compensation 1.0 1.0 Compensation 0.26 0.15-0.44 < 0.0001 0.21 0.12-0.37 < 0.0001 Beck score (15-point increase) 1.00 0.69-1.47 0.98 0.48 0.32-0.73 0.0006 Baseline pain score -(2-point increase) 2.64 2.12-3.29 < 0.0001 1.85 1.63-2.10 < 0.0001 Pain duration 0.0006** 0.007** < 6 wks 1.0 1.0 6 wks to < 3 mo.s 3.24 1.48-7.07 0.003 3.61 1.61-8.10 0.002 3 mo.s to < 6 mo.s 1.15 0.55-2.39 0.71 1.86 0.87-3.96 0.11 > 6 mo.s 0.98 0.50-1.93 0.96 1.39 0.69-2.80 0.36 Test of heterogeneity* 0.16 •Likelihood ratio test of the slope coefficient being zero, unless otherwise indicated. **Wald chi-square test for type III analysis of effect. ***Score Test of homogeneity of effect over increasingly stringent definitions of "improvement". t o t o o T a b l e 8 . 1 0 : E f f e c t o f W a i t i n g T i m e : M u l t i v a r i a b l e C o n t i n u a t i o n R a t i o M o d e l Results of continuation ratio ordinal logistic regression "outcome" model showing unadjusted and adjusted effects age, sex, waiting time other significant correlates of ordinal improvement over multiple definitions of improvement, at six-months after surgery. U n a d j u s t e d e f f e c t s A d j u s t e d e f f e c t s E f f e c t O R 9 5 % C I p v a l u e * O R 9 5 % C I p v a l u e * Waiting time Short 1.0 1.0 Long 0.46 0.32-0.66 < 0.0001 0.65 0.44-0.96 0.031 Age/10-years 0.99 0.87-1.13 0.92 0.83 0.72-0.96 0.012 Sex Male 1.0 1.0 Female 1.42 1.02-1.97 0.039 0.98 0.68-1.43 0.96 Compensation status No compensation 1.0 1.0 Compensation 0.39 0.26-0.59 < 0.0001 0.27 0.17-0.43 < 0.0001 Beck score (15-point increase) 1.02 0.74-1.39 0.91 0.56 0.40-0.79 0.001 Baseline pain score (2-point increase) 2.50 2.06-3.03 < 0.0001 3.06 2.46-3.81 < 0.0001 Pain duration 0.0014** 0.014** < 6 wks 1.0 1.0 6 wks to < 3 mo.s 2.47 1.35-4.52 0.0034 2.72 1.39-5.32 0.003 3 mo.s to < 6 mo.s 1.21 0.69-2.12 0.5147 1.54 0.83-2.86 0.16 > 6 mo.s 1.01 0.60-1.69 0.9721 1.29 0.72-2.30 0.36 Heterogeneity Test *** 0.67 *Likelihood ratio test of the slope coefficient being zero, unless otherwise indicated. **Wald chi-square test for type III analysis of effect. ***Test of homogeneity of effect over increasingly stringent definitions of "improvement". to to 8.6 References 1. Hanning M . Maximum waiting-rime guarantee - An attempt to reduce waiting lists in Sweden. Health Policy 1996;36:17-35. 2. Cassidy JD, Cote P, Carroll LJ, and Kristman V. Incidence and course of low back pain episodes in the general population. Spine 2005;30:2817-23. 3. Gibson JNA, Grant IC, and Waddell G. The Cochrane review of surgery for lumbar disc prolapse and degenerative lumbar spondylosis. Spine 1999;24:1820-32. 4. Daltroy L H , Cats-Baril WL, Katz JN, Fossel A H , and Liang M H . The North American spine society lumbar spine outcome assessment Instrument: reliability and validity tests. Spine 1996;21:741-9. 5. Hu RW, Jaglal S, Axcell T, and Anderson G. A population-based study of reoperations after back surgery. Spine 1997;22:2265-70. 6. Atlas SJ, Keller RB, Wu Y A , Deyo RA, and Singer D E . Long-term outcomes of surgical and nonsurgical management of sciatica secondary to a lumbar disc herniation: 10 year results from the Maine Lumbar Spine Study. Spine 2005;30:927-35. 7. Gibson JNA, Grant IC, and Waddell G. The Cochrane review of surgery for lumbar disc prolapse and degenerative lumbar spondylosis. Spine 1999;24:1820-32. 222 8. Weinstein JN , Tosteson T D , Lurie JD et al. Surgical vs nonoperative treatment for lumbar disk herniation - The Spine Patient Outcomes Research Trial (SPORT): A randomized trial. Journal of the American Medical Association 2006;296:2441-50. 9. Boszczyk B, Timothy J, Peul W, and Casey A T H . Neurosurgical training and the spine: Reflections on EANS Winter Meeting Luxembourg, February, 2006. Acta Neurochirurgica 2007;149:339. 10. Maker A D , Larson EB, Urban N , and Deyo RA. Cost-effectiveness of lumbar discectomy for the treatment of herniated intervertebral disc. Spine 1996;21:1048-54. 11. Planning M . Maximum waiting-time guarantee - An attempt to reduce waiting lists in Sweden. Health Policy 1996;36:17-35. 12. Hanning M . Maximum waiting-time guarantee - An attempt to reduce waiting lists in Sweden. Health Pokey 1996;36:17-35. 13. Mahon JL, Bourne RB, Rorabeck C H , Feeny D H , Stitt L, and Webster-Bogaert S. Health-related quakty of Ufe and mobikty of patients awaiting elective total hip arthroplasty: a prospective study. Canadian Medical Association Journal 2002;167:1115-21. 14. Garbuz DS, Xu M , Duncan CP, Masri BA, and Sobolev B. Delays worsen quakty of Hfe outcome of primary total hip arthroplasty. Clinical Orthopaedics and Related Research 2006;447:79-84. 15. Adas SJ, Deyo RA, Keller RB et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine 1996;21:1777-86. 223 16. Nystrom B. Experience of Microsurgical Compared with Conventional Technique in Lumbar-Disk Operations. Acta Neurologica Scandinavica 1987;76:129-41. 17. Olmarker K, Blomquist J, Stromberg J, Nannmark U , Thomsen P, and Rydevik B. Inflammatogenic Properties of Nucleus Pulposus. Spine 1995;20:665-9. 18. Nygaard OP, Mellgren SI, and Osterud B. The inflammatory properties of contained and noncontained lumbar disc herniation. Spine 1997;22:2484-8. 19. Woertgen C, Rothoerl RD, and Brawanski A. Influence of macrophage infiltration of herniated lumbar disc tissue on outcome after lumbar disc surgery. Spine 2000;25:871-5. 20. Nygaard OP, Romner B, and Trumpy JH. Duration of Symptoms As A Predictor of Outcome After Lumbar Disc Surgery. Acta Neurochirurgica 1994;128:53-6. 21. Nygaard OP, Kloster R, and Solberg T. Duration of leg pain as a predictor of outcome after surgery for lumbar disc herniation: a prospective cohort study with 1- year follow up. Journal of Neurosurgery 2000;92:131-4. 22. Coderre TJ, Katz J, Vaccarino A L , and Melzack R. Contribution of central neuroplasticity to pathological pain: review of clinical and experimental evidence. Pain 1993;52:259-85. 23. Quan H , Lafreniere R, and Johnson D. Health service costs for patients on the waiting list. Canadian Journal of Surgery 2002;45:34-42. 224 24. Rossvoll I, Benum P, Bredland TR, Solstad K , Arntzen E, and Jorgensen S. Incapacity for Work in Elective Orthopedic-Surgery - A Study of Occurrence and the Probability of Returning to Work After Treatment. Journal of Epidemiology and Community Health 1993;47:388-94. 25. Coyte PC, Wright JG, Hawker G A et al. Waiting-Times for Knee-Replacement Surgery in the United-States and Ontario. New England Journal of Medicine 1994;331:1068-71. 26. Ho E , Coyte PC, Bombardier C, Hawker G, and Wright JG. Ontario Patients Acceptance of Waiting-Times for Knee Replacements. Journal of Rheumatology 1994;21:2101-5. 27. Sanmartin C, Shortt SED, Barer ML, Sheps S, Lewis S, and McDonald PW. Waiting for medical services in Canada: lots of heat, but little light. Canadian Medical Association Journal 2000;162:1305-10. 28. Martin R M , Sterne JAC, Gunnell D, Ebrahim S, Smith G D , and Frankel S. NHS waiting lists and evidence of national or local failure: analysis of health service data. British Medical Journal 2003;326:188-192A. 29. Sanmartin C, Shortt SED, Barer ML, Sheps S, Lewis S, and McDonald PW. Waiting for medical services in Canada: lots of heat, but little light. Canadian Medical Association Journal 2000;162:1305-10. 30. Sanmartin C, Shortt SED, Barer ML, Sheps S, Lewis S, and McDonald PW. Waiting for medical services in Canada: lots of heat, but little light. Canadian Medical Association Journal 2000;162:1305-10. 225 31. Dunn E, Black C, Alonso J, Norregaard JC, and Anderson GF. Patients' acceptance of waiting for cataract surgery: What makes a wait too long? Social Science & Medicine 1997;44:1603-10. 32. Planning M . Maximum waiting-time guarantee - An attempt to reduce waiting lists in Sweden. Health Policy 1996;36:17-35. 33. Snider M G , MacDonald SJ, and Pototschnik R. Waiting times and patient perspectives for total hip and knee arthroplasty in rural and urban Ontario.[see comment]. Canadian Journal of Surgery 2005;48:355-60. 34. Lofvendahl S, Eckerlund I, Hansagi H , Malmqvist B, Resch S, and Hanning M . Waiting for orthopaedic surgery: factors associated with waiting times and patients' opinion. Int.J.Qual.Health Care 2005;17:133-40. 35. Fisher C, Noonan V, Bishop P et al. Outcome evaluation of the operative management of lumbar disc herniation causing sciatica. Journal of Neurosurgery 2004;100:317-24. 36. McKessock L, Smith B H , Scott A et al. A randomized controlled trial of direct access for laparoscopic sterilization. Fam.Pract. 2001;18:1-8. 37. Kopec JA, Sayre EC, and Esdaile JM. Predictors of back pain in a general population cohort. Spine 2004;29:70-7. 38. Hurwitz E L and Morgenstern H. Correlates of back problems and back-related disability in the United States. Journal of Clinical Epidemiology 1997;50:669-81. 39. Fairbank JC and Pynsent PB. The Oswestry Disability Index. Spine 2000;25:2940-52. 226 40. Daltroy L H , Cats-Baril WL, Katz JN , Fosse] A H , and Liang M H . The North American spine society lumbar spine outcome assessment Instrument: reliability and validity tests. Spine 1996;21:741-9. 41. Ware JE, Jr. SF-36 health survey update. Spine 2000;25:3130-9. 42. Hanning M . Maximum waiting-time guarantee - An attempt to reduce waiting lists in Sweden. Health Policy 1996;36:17-35. 43. Hastie TJ, Botha JL, and Schnitzler C M . Regression with An Ordered Categorical Response. Statistics in Medicine 1989;8:785-94. 44. Scott SC, Goldberg MS, and Mayo N E . Statistical assessment of ordinal outcomes in comparative studies. Journal of Clinical Epidemiology 1997;50:45-55. 45. Salaffi F, Stancati A , Silvestri CA, Ciapetti A, and Grassi W. Minimal clinically important changes in chronic musculoskeletal pain intensity measured on a numerical rating scale. European Journal of Pain 2004;8:283-91. 46. Scott SC, Goldberg MS, and Mayo N E . Statistical assessment of ordinal outcomes in comparative studies. Journal of Clinical Epidemiology 1997;50:45-55. 47. McKessock L, Smith B H , Scott A et al. A randomized controlled trial of direct access for laparoscopic sterilization. Fam.Pract. 2001;18:1-8. 48. Rosenbaum PR and Rubin DB. The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika 1983;70:41-55. 227 49. Rosenbaum PR. Covariance adjustment in randomized experiments and observational studies. Statistical Science 2002;17:286-304. 50. Hanning M . Maximum waiting-time guarantee - An attempt to reduce waiting lists in Sweden. Health Policy 1996;36:17-35. 51. Hanning M . Maximum waiting-time guarantee - An attempt to reduce waiting lists in Sweden. Health Policy 1996;36:17-35. 228 C H A P T E R 9 9.0 DISCUSSION 9.1 R e v i e w o f F i n d i n g s a n d I m p l i c a t i o n s 9.1.1 T e m p o r a l T r e n d s i n L u m b a r D i s c e c t o m y R a t e s The main objective of the first study of the thesis was to estimate changes in the rate of utilization of lumbar discectomy during a period of hospital downsizing in BC. Prior to the completing this study, rates of "non-fusion" neck and back surgery were stable in Washington State and slightly increasing in Ontario between 1982 and 1992.15 In BC, hospital downsizing had resulted in a 30% decline in the number of staffed short-stay beds in BC between 1990/91 and 1996/97.7;18 Private clinics were also authorized to perform lumbar discectomies after 1999. What was not known was the net effect of downsizing and the later emergence of private clinics from 2000 onward on rates of lumbar discectomy/laminectomy procedures within public hospitals in BC. Based on provincial hospital discharge data, calculated annual age- and sex- adjusted rates of lumbar discectomy over a 14-year period showed a decline in lumbar discectomy rates by approximately 62% between 1990 and 2003 inclusive. In multivariable analyses, the rate of decline was greater during 2000-2003 when the discectomies became available in the private sector, and was also greater for Workers' Compensation beneficiaries than for regular publicly-insured patients (under the Medical Services Plan of BC). These results support the hypothesis that utilization rates decreased in accordance with the shrinking supply of staffed short-stay beds in BC throughout the 1990's. While this association 229 would seem too obvious on the surface, examples in the literature exist in which resource downsizing did not automatically lead to reduced rates of in-hospital procedures. In Manitoba, for example restraints in health care spending after 1991 were also achieved by downsizing in the hospital sector. Manitoba had 209 (7.7%) bed closures in 1993 and an additional 188 (4.7%) closures in 1994 and 1995.9 However, over the same four-year period (1992-1995 inclusive), the number of patients treated in-hospital remained stable. Rates of outpatient surgery, which were already on the increase prior to hospital sector downsizing, continued to increase during this period (by 25% between 1992 and 1995), while rates of short-stay inpatient hospitalizations decreased (by 13.7%).9 In the end, shorter stays for surgical inpatients, combined with a shift from inpatient to outpatient surgery, explained a 24% reduction in surgical days per 1000 residents without compromising the number of patients treated in Manitoba between 1992 and 1995.9 Moreover, quality-of-care proxy measures such as mortality rates, and emergency room admissions did not increase. Hospital readmission rates increased for only one type of hospitalization (cesarean sections), however on the whole, no relationship between length of stay and readmission rates was detected.9 Interestingly, regarding Study 1 of the thesis, lumbar discectomy rates in BC declined so steadily throughout the 1990's that the data offers no obvious evidence that similar adjustments were attempted in order to stabilize the rates of lumbar in BC throughout the 1990's. During a presentation of this research at a tertiary care centre in the province, local spine surgeons suggested that declining rates of lumbar discectomy might have reflected surgeons' declining interest in lumbar disc patients. It was suggested that with limited operating room availability, lumbar disc patients are considered low priority in comparison to other spinal or neurosurgical patients with frankly unstable, progressive conditions or life-threatening conditions (e.g. traumatic injuries, infections, and tumors). 230 A serious implication of any genuine decline in surgeons' interest in lumbar disc patients in BC would be a likely increase in the prevalence of disabling sciatica in the community. While not life-threatening, sciatica (and back pain) due to disc herniation substantially diminishes functional capacity and quality of life.10 Untreated, or delayed treatment of pain and disability predisposes to chronic pain, which tends to be refractory to otherwise effective treatment once present.5 Future surveillance studies are warranted to determine if reduced utilization of lumbar discectomy indeed reflects decreased interest (even neglect) of patients and consequent increases in rates of back-related disability in the community. The results of Study 1 also reveal a larger rate of decline in lumbar discectomy rates within public hospitals, which were conceivably facilitated by the availability of lumbar discectomy in the private sector from 2000 onward. Rates of lumbar discectomy in the public sector declined significandy in this later period, especially among workers' compensation beneficiaries. Admittedly, this was not anticipated as the candidate originally hypothesized that compensation-related surgeries should be less discretionary in nature than non-compensation-related surgeries (due to the overseeing of patient care by claims managers and medical advisors at WorkSafeBC, even if only passively) and that, therefore, rates of surgery among non-compensated patients— again, only assumed to be more discretionary in nature—would decline faster during a period of downsizing and service rationing. However, Study 1 showed the opposite trend; lumbar discectomy rates for non-compensated patients actually declined less quickly, thus failing to support one of the candidate's original hypotheses. One possible conclusion might be that surgical decisions for compensated patients are contrary to what was hypothesized and are, in fact, more dependent upon physician discretion or "practice style" than surgical decisions regarding non-compensated patients. However, on closer examination of the trend estimates according to treatment period and responsible payor group, it 231 was apparent that the steeper decline in compensation-related (compared to non-compensation-related) rates of discectomy was only apparent during the later treatment period, which was when lumbar discectomy was available in the private sector. Given this finding, it seemed reasonable that the greater decline in surgery rates for compensated patients in public hospitals likely represented the diverting of compensated-related operation in the public system to private for-profit clinics after 1999/2000. Another important finding from Study 1 is that there was no evidence of an attenuation of the decline in rates among non-compensated patients even during the period that compensated patients were diverted to the private sector. This additional finding supports already existing assertions that in a system of "roughly fixed per capita human resources... where physicians and nurses are already in short supply", the reallocation of human capacity to the private system does not (necessarily) improve access to public sector services.17 The documentation of an overall 61% reduction in per capita use rates of lumbar discectomy since 1990/91 raises the possibility of potentially serious implications. First, reduced utilization may reflect reduced access to lumbar discectomy due to increasing displacement of acute disc patients, pardy because of reduced capacity for operations in general in BC, but also partly due to further exclusion of lumbar disc patients from surgeons' waiting lists in efforts to maintain access to surgery for patients with higher priority conditions. Second, and as a result of the first condition, preferential access to spine surgery for higher priority patients (e.g. tumors, and trauma-related conditions) at the necessary expense and displacement of lower priority—but still profoundly disabled—acute disc patients possibly increases the prevalence of disabling lumbar disc disease in the community. Further studies of the prevalence of disabling back pain over time would provide insight into this important issue. 232 9.1.2 Regional Variations in Lumbar Discectomy Rates The main objectives of the second study were to quantify the variation in population-based lumbar discectomy rates across geographic areas according to contemporary units of regional health services planning and service provision (i.e. HSDAs). The other objective was to compare recent area variation to that of an earlier time period (1990-1993) when acute care resources were less constrained. Prior to this study, an eight-fold difference in rates of discectomy (not restricted to lumbar procedures) had been reported between LHAs. 6 However, neither the methodology underlying these important estimates, nor a discussion of their significance in relation to random variation was provided. Since the inception of HSDAs in 2001, an assessment of the variation in lumbar discectomy rates at this contemporary level of service planning and coordination had not been reported. This study revealed that age- and sex-adjusted rates of discectomy across 16 geographically distinct HSDAs varied by almost five-fold (4.7) in 2000-2003. This was qualitatively much less than the amount of variation previously reported among LHAs at an earlier time.28 However, previously reported rates at the L H A level were based on discectomies for all spinal regions whereas in this thesis, discectomies primarily involving the lumbar region were analysed. Formal hypothesis tests confirmed that the variation in rates in 2000-2003 exceeded the amount expected by random variation alone. Also, compared to the variation in rates for 1990-1993 inclusive, the variation in 2000-2003 was slightly greater, although not (statistically) significantly so. This finding refutes the hypothesis that regional variation—a proxy measure of physician discretion—is reduced in a climate of constrained resources, thus fortifying Wennberg's concept of the "surgical signature" hypothesis: that in the absence of initiatives specifically targeting clinical practice behaviour, diagnostic and therapeutic procedure rates may vary substantially across regions, yet this variation across regions will be ironically stable over time.25126 233 That geographical differences in discectomy rates exist in BC is not a novel finding. However an estimate of rate variations specifically in BC for a homogenous procedure group— recipients of lumbar discectomy as opposed to discectomy or other procedures anywhere in the spine—at a geographic level supposedly representative of health services planning and provision in the community does fill an important information gap. The documentation of relatively stable variations near the onset of, and later on into, the process of regionalization of health care in BC is also important in suggesting that the surgical signature hypothesis holds true, even with considerable changes within the underlying health care system. As with other rate variation studies, a potential practical implication of this research is that areas of high and/or low rates may reflect either areas of higher and/or lower need on the one hand, or alternatively over- and/or under-utilization of lumbar discectomy, respectively, on the other hand. Suggestions for future research aimed at identifying the "right" rate of surgery in BC are discussed in a later section of this chapter. 9.1.3 Determinants of Waiting Time for Lumbar Discectomy In Study 3, the primary objective was to determine whether disparities in access to lumbar discectomy exist in BC, particularly in association with compensation status. Prior to this study, determinants of waiting time for other orthopaedic procedures had been documented. One of these studies included "back surgery" patients,16 but none examined a more homogenous group such as lumbar discectomy recipients exclusively. Furthermore, studies on orthopaedic surgery from Canada have focused on waiting times for non-spinal operations such as total hip or knee arthroplasty. These studies suggest that health status is not a consistent predictor of waiting times for orthopaedic procedures in general,8'13'14 although for knee and hip replacements health status is modesdy associated27. 234 In Sweden16 some demographic variables (such as sex, age and living arrangement) were not associated with wait time for back surgery, or knee or hip replacement surgery while others (such as work status) were. Also, only one system-related variable, hospital type (county/district versus university/regional) also predicted waiting times. Time from referral to first hospital outpatient visit with a surgeon (waiting time to see a specialist) was not associated with longer waiting time for back surgery. Study 3 of this thesis showed that measures of clinical severity were clearly associated with waiting time at a large tertiary care centre in BC. This confirmed that access to lumbar discectomy was appropriately dependent on clinical need. Generally, sociodemographic variables in this study were not associated with waiting times for lumbar discectomy, suggesting that access to surgery generally equitable from a social perspective. This included compensation status, which is a health system variable from a payor perspective but was examined in this study as a demographic/psychosocial variable. Despite having a reputation for being less responsive to otherwise effective treatment,3 compensated patients did not exhibit a statistically significant delay in accessing surgery. On the other hand, the median waiting time for compensation patients was three weeks longer than the median time for non-compensated patients, which is potentially clinically important. However, this clinically significant difference at the univariate level was not at all apparent when access to surgery was modeled and adjusted for other prognostic variables. Previous studies have indicated that clinicians' judgments about patients may be influenced by compensation status and that compensated patients with lumbar disc herniation are less likely to be offered surgery than noncompensated patients.3 However, during a meeting with study clinicians, surgeons alleged that knowledge of a worker being on compensation might conceivably delay their decision to offer surgery, however this did (should) not result in prejudicial treatment of compensated patients or interfere with prioritization of these patients on the waitlist after a decision 235 to proceed with surgery. The candidate's study findings supports this nodon of equitable treatment of compensation patients after the time of enrolment onto the waitlist. The one sociodemographic variable that was associated with waiting time in this thesis was a coarse measure of occupation type. Patients who classified themselves as either professionals or managerial workers had a conditional weekly probability of acquiring surgery that was 29% higher than that of workers in other types of occupations. Given the discretionary nature of lumbar discectomy and previous suggestions of queue jumping within the public system,2 this finding deserves attention in future studies using finer measures of social and/or occupational status and greater representation of patients in a variety of other occupational groups. It has been suggested that waitlists for joint replacement surgery in Canada have been managed equitably from a social perspective (by not being dependent upon social status variables) but perhaps unfairly from a clinical perspective (by not being dependent upon clinical severity).14 This is not consistent with the findings from Study 3 of this thesis, which show that access to lumbar discectomy at a major tertiary care centre in BC appears to be strongly, and appropriately, dependent on clinical variables, but still potentially inappropriately dependent upon social status (i.e. professional or managerial occupation). 9.1.4 Association Between Waiting Time and Outcome of Lumbar Discectomy The literature reviewed in Chapter 3 provided some evidence that longer symptom duration prior to elective surgical lumbar discectomy has been associated with poorer outcomes. However, Chapter 4 also reviewed the methodological challenges associated with measurements involving symptom duration and the methodological and conceptual advantages of focusing on waiting time for surgery as a potential measure of the optimal timing of treatment. 236 The primary objective of Study 4 was to estimate the significance and magnitude of the effect of waiting time on pain improvement after lumbar discectomy. Study 4 of this thesis shows that waiting 12 weeks or longer for surgery is associated with an important deterioration in the odds of pain reduction at six months postoperatively, even after controlling for detailed measures of clinical severity. This result suggests that waiting time for lumbar discectomy is a meaningful predictor of outcome for patients actually scheduled for lumbar discectomy. This finding also motivates future research on the potential utility of using a 12-week cutoff as a clinically relevant benchmark for defining a maximum appropriate waiting time for elective surgical lumbar discectomy. The implications of such a benchmark however are controversial in a climate of rationed health care in BC. If advocated as a benchmark for the timely delivery of surgical services, the provision of more discectomies within this benchmark may not be realistic without first examining the current leeway for accommodating more discectomies on a day-surgical as opposed to an inpatient basis (in order to shorten lengths-of-stay). On the other hand, benchmarks of this nature may also serve as a useful start to identifying other benchmarks for lumbar discectomy, such as the maximum waiting time, if any, beyond which a lack of benefit—if not outright harm—is likely, if not certain. Operationally-defined, and well-rationalized time points for predicting intermediate outcomes between certainty of harm on the one hand and certainty of benefit on the other hand would also be useful in this regard. Waiting time for surgery represents the last pre-operative stage of waiting on the total symptom-duration continuum. It also begins after two important clinical conditions are met: (1) indications crossing a surgeon's threshold for recommending lumbar discectomy, and (2) symptoms crossing the patient's threshold for becoming willing to undergo invasive treatment. At the inception point of this interval, the treatment preferences of both physicians and patients are in agreement, and any further delay in care is typically attributable to the limited capacity of the 237 healthcare system (patients can take themselves off the wait list for many reasons including fear, symptom resolution, travel plans, family emergency, no caregiver after they get done, etc). For this reason, waiting time for surgery can be viewed as a reasonable proxy for the delay in service that is imposed by the healthcare system. To the extent that capacity for individual procedures in healthcare can be manipulated, it is also modifiable. This responsibility of such modification lies jointly in the hands of policymakers who fund and allocate the resources necessary for health care delivery, and providers (physicians and surgeons) who are in charge of maximizing health outputs on the ground.24 9.2 Methodological Strengths of Studies 9.2.1 Administrative Data The first two articles of the thesis were completed through the use of administrative data. Use of administrative data is associated with certain limitations which will be discussed in a later section, however several strengths are noteworthy here. Health insurance is universal in Canada, and therefore the province-wide hospital separations data that were used included admissions for lumbar discectomy in public hospitals for the entire population of BC, except for private sector patients after 2000/01. Use of administrative data for the first two studies therefore represented all procedures of interest in hospitals throughout the entire province, which increases the generalizability of the research findings back to the majority of the population, except for the population treated in private sector clinics. Similarly, data for the entire provincial population allowed for an assessment of service utilization within communities larger than any one restricted region of the province. Unique, scrambled, study identification numbers linked multiple records to individual patients over time. The longitudinal, person-level, nature of this data permitted the identification 238 of hospital admissions for repeat-procedures. While only primary lumbar discectomies were used in the studies, the person-specific nature of the data permitted a direct assessment of the extent of multiple admissions and potentially correlated observations during statistical analyses. A homogenous study population was chosen intentionally. The focus was on one specific diagnosis group (intervertebral disc disorders in the back) and one specific procedure group (discectomy and/or laminectomy) while simultaneously excluding patients with diagnoses of nonmechanical back pain. Patients with cervical and known thoracic procedures were excluded. We also intentionally focused on discectomy procedures, which account for the majority of low back surgeries in Canada." Also, in the assessment of temporal and regional variations in Studies 1 and 2, the homogenous nature of surgical patients, together with direct age- and sex-standardization techniques enhanced the comparability in procedure rates over time, and between areas, that were compared. 9.2.2 Prospective Treatment Registry Data In Study 3 (Chapter 7), due to the use of prospective treatment registry data, the waitlist experience of all patients registered for lumbar discectomy could be documented regardless of whether they actually received surgery or were removed from the waitlist for other reasons (e.g. spontaneous recovery, going on vacation, or change of mind for unspecified reason). Patients who experienced these intermediate events were censored instead of being completely excluded from the analysis. This permitted subjects to contribute partially to the risk set of patients up until the time of censoring, which reduced potential bias in the wait time estimates (toward underestimation) had they been simply excluded.19 The use of a prospective registry allowed for data collection on a broad range of sociodemographic, clinical, and institutional/health system variables, all of which were screened 239 for potential confounding and subsequendy adjusted for if associated (clinically or statistically) with the outcome. Another strength is that waiting times were calculated exacdy using actual dates of surgery and dates of placement onto the waitlist. For only a small number of patients was it was not necessary to rely on a proxy measure of the date of waidist enrolment (e.g. date of preoperative assessment), as is commonly necessary in studies relying only on administrative data. The exposure variable, compensation status, while self-reported was verified through an audit of hospital payment records. Study 4 study had some notable methodological strengths. This was a prospective study involving primary data collection and the use of both a clinically relevant outcome and a well-defined inception point (date of enrolment onto the surgical waitlist). Wait times were calculated directly from the date of waitlist enrolment and the date of surgery, both of which were documented prospectively within an existing patient registry. Although this was not a randomized trial, detailed sociodemographic, occupational, clinical, psychological, and health system/institutional variables were all measured, and screened as potential confounders. Both statistically significant and clinically plausible predictors were adjusted for. A dose-response relationship in the expected direction of effect, further supported a genuine association between longer waiting time and a lower odds of pain improvement. 9.3 Methodological Limitations 9.3.1 Administrative Data In Studies 3 and 4 (Chapters 5 and 6) the predominance of single-decimal diagnosis codes in the data prevented the identification and exclusion of thoracic disc patients who might have been coded as simply "intervertebral disc disorder — region unspecified". However, in the 240 literature, thoracic disc herniation reportedly accounts for only 0.15% to 4% of all disc herniations.21 Furthermore, the "closely matching" nature of datasets defined by single-decimal and double-decimal ICD-9 coding has been documented.12 For Study 1, the candidate did not specifically validate the responsible payment field—the source of the compensation status field—in the B C L H D . However, other investigators have reported the B C L H D information is actually more complete than that of WCB's own documentation of, at least, "severe" work-related injuries. (Other researchers suggest that the BC provincial compensation board underreported at least 9% of serious work related injuries between 1989 and 1998.1) The data from B C L H D does not capture patients treated in the private sector. Lumbar discectomy was accessible in only two private clinics in all of BC beginning in the year 2000. At the time of study inception (and completion), no reliable data even from WorkSafeBC (a principal patron of these private clinics) were available on the numbers patients seen in these private facilities. The administrative data used for this thesis therefore, underestimates total discectomy rates throughout the province, but still accurately presents the rates of lumbar discectomies performed in all public hospitals in BC. 9.3.2 Registry Data For Study 3, both the date of surgical booking and the date of the corresponding preoperative visit to a surgeon were missing for six patients. In each of these instances, the date of surgical booking had to be imputed using the date of completion of the standardized assessment questionnaire as a proxy for the date of surgical booking. This questionnaire was typically administered on the date of surgical booking, however on occasion, it was administered late and therefore closer to the time of surgery. At worst, however, waiting time values for six observations 241 may have been underestimated in this manner. The candidate had no reason to suspect that these missing data values were systematically related to either true waiting time, compensation status, or other potential confounders. Due to relatively small numbers of patients receiving disability compensation, workers' compensation claimants were combined with claimants receiving disability from other sources (private disability or motor vehicle/personal injury insurance). Some studies suggest that the specific type of compensation may be a determinant of outcome during a claim, however there is no evidence in the literature to suggest that the different types of compensation affect waiting times for surgery differendy. Also, in exploratory analyses the coefficients for the effects of workers' disability compensation and disability compensation from alternate sources were similar in magnitude and statistical significance. The study population for Study's 3 and 4 were underrepresented by rural patients. Study patients were recruited at V G H , a public tertiary care facility located in a large metropolitan centre. V G H serves the Vancouver Coastal Health Authority, inhabited mostly urban residents with only approximately 8% living in smaller towns or non-urban areas. Vancouver Coastal Health Authority contains 25% of the province's total population. V G H is one of two public hospitals performing surgical lumbar discectomies regularly within this health authority. The prospective registry data did not capture patients who had surgery in a private clinic. However, private, expedited lumbar disc surgery was available only after the year 2000, and although' there are approximately 20 private medical care clinics in British Columbia currendy, only two private clinics treat back pain patients with any regularity. In the meantime, exclusion of privately treated patients from this study did not undermine the specific objective of identifying 242 significant determinants of waiting time for lumbar discectomy patients treated within the public system. Potential confounders and clinical severity were adjusted for using propensity scores, however in the absence of randomization, confounding by indication is still a potential concern. In this observational study, it was possible for short-wait patients to differ prognostically from long-wait patients in terms of unknown and/or unmeasured confounders. In a study of the effect of waiting time on the outcome lumbar disc herniation, confounding by indication could result in a liberal bias. Patients with greater pain intensity are generally classified as more urgent and are likely to access surgery faster than patients with low pain intensity. Higher preoperative pain intensity is associated with a greater probability of pain reduction after surgery, which could create a bias in favour of a better outcome in the short-wait group. Short-wait patients are also more likely to have shorter symptom duration. As shorter symptom duration is also associated with better outcomes following surgical lumbar discectomy, this could also have generated a bias in favour of our observed result. 9.4 Original Contributions and Future Research The research conducted for this thesis advances original knowledge in several areas. The first study on the temporal trends in lumbar discectomy rates in BC (Chapter 5) generated new information about the temporal trend in population-based utilization rates of lumbar discectomy during a period of restructuring and downsizing of acute care resources in BC. A study of the trends in back surgery rates in a Canadian jurisdiction has not been published in nearly a decade. Furthermore, no previous publications have examined the trends in rates in BC at the level of HDSAs, which currently represent the level at which comprehensive services are delivered to the community. 243 Given that a sustained decline in lumbar discectomy rates was observed over the entire study period, a useful future study would be to estimate waiting times for surgery and per capita surgical days for lumbar discectomy patients in BC over recent and remote time periods. This will help to confirm if substantial changes in either waiting times or in-hospital lengths-of-stay have occurred over the past decade of regionalization and hospital downsizing. Furthermore, similar to the Manitoba example,9 an assessment of the rates of emergency room admissions and hospital readmissions for lumbar disc herniation over time will also help to provide insight into potential changes in the quality of hospital care. The second study (Chapter 6) provided previously unknown information on the extent of variation in lumbar discectomy rates in the BC using regions of analysis (health service delivery areas) that are representative of the level at which comprehensive medical services (including spinal surgery) are coordinated and ensured throughout the province. A study of the regional variation in lumbar discectomy rates specifically has never before been conducted at this geographic level. A useful extension of this current study would be to try to estimate the prevalence of back pain conditions throughout sub-provincial regions using a combination of WCB, provincial Medical Services Plan (MSP) and hospital separations data. The effects of alternative definitions and follow-up parameters on various estimates of low back pain incidence, prevalence, recurrences and episodes of care could be performed using administrative data and a health services research framework.22;23 Estimates, particularly of the prevalence of back conditions in various communities could then be used for regional variations studies based on age- sex-, as well as prevalence-adjusted rates of back-related procedures. Such information would remove a major source of uncertainty (about the potential effect of regional differences in disease prevalence) from future related studies in BC. 244 The third study in the thesis (Chapter 7) presented new knowledge on the specific determinants of waiting times for surgical lumbar discectomy in BC using prospective treatment registry data. Prospective data collection enabled follow-up of patients who received surgery as well as patients who were removed from the waitlist for other reasons, thereby avoiding biased ascertainment of patients waiting times for surgery.20 Extensive measures of clinical severity also allowed for case-mix to be controlled for in the analysis of waiting times. This study addressed a previously unanswered question about whether or not compensation patients—due to their greater propensity for poorer outcomes—encountered a barrier to access to lumbar discectomy even after being placed on the surgical wait list. Access (waiting time) to lumbar discectomy appeared to be equitable from a clinical perspective but possibly not from a social perspective. A significant association between type of occupation (professional and/or managerial work) and waiting time was observed in this study, which deserves further corroboration (or refutation) in a study utilizing more detailed measures of occupation type and other individual-level measures of social and economic status. New knowledge about the significance and magnitude of the association between longer waiting time and the clinical outcome of lumbar discectomy was generated by the fourth study in the thesis (Chapter 8). Reports of the effect of waiting time for surgery on outcomes are relatively rare in the literature. An estimate of both the significance and magnitude of the effect of waiting time on the clinical outcome of lumbar discectomy specifically has not been previously reported. The results of this study fill an important knowledge gap by suggesting that a disparity in access to lumbar discectomy really does matter. The finding that patients experience an approximate 41% reduction in the odds of pain improvement after 12 weeks on the waitlist has potential utility as benchmark for defining a maximum appropriate waiting time for elective surgical lumbar discectomy. However in a climate of constrained resources, a 12-week window of opportunity for 245 a better outcome is more likely to be viewed as a 12-week threshold beyond which the decision to undertake surgery has to be weighed more stringendy—by patients as well as providers—against a lower odds improvement. In the meantime, a benchmark such as "the proportion of patients receiving lumbar discectomy within 12 weeks on the waitlist" may have potential utility as a quality performance measure for the public health care system. Future research could extend this work by helping to define other benchmarks of interest including the maximum waiting time (if any) beyond which a lack of benefit-—if not harm—from surgery is certain. Ultimately, however, the evidence of any effect of waiting time on outcome will require corroboration from a randomized controlled trial of early versus delayed surgery. Very recently, a randomized trial of early surgery (within two weeks of randomization) versus "prolonged conservative treatment with eventual surgery if needed" was published.4 At one-year, the outcomes were similar between groups, however the rates of pain relief and of perceived recovery were higher for those assigned to early surgery. Quicker recovery through surgery with convergence of outcomes over the longer-term appears to be a recurring finding within randomized trials of surgery versus no surgery, or, now, early surgery versus surgery either not at all or possibly delayed "if necessary". Given current waiting times for lumbar discectomy at one hospital in BC (i.e. a median time of 11.5 weeks when patients are censored at one year), a randomized controlled trial in BC of early surgery versus delayed surgery would be ethically less challenging if in the delayed surgery group, treatment was contingent only upon the "usual" waiting time in the public system (11.5 weeks). However, a trial of early surgery may not be useful in BC unless the potential for making early surgery publicly available (and only in the event of positive trial) is feasible. In a climate of constrained resources, early surgery, defined as surgery within 2 weeks, may not be realistic. Early surgery defined alternatively as falling within six weeks of waitlist enrolment would appear to be a 246 more reasonable on the surface, but would still require an assessment of available resources and downstream implications of such a policy on other surgical programs/departments within administrative regions. In light these unknowns, the logical "next study" in lieu of a randomized trial of early or late surgery may have to be a study on the feasibility of implementing expedited surgery within the current system. 247 9.5 References 1. Alamgir H , Koehoom M , Ostry A, Tompa E, and Demers P. An evaluation of hospital discharge records as a tool for serious work related injury surveillance. Occupational and Environmental Medicine 2006;63:290-6. 2. Amoko D H , Modrow RE, and Tan J K H . Surgical waiting lists II: current practices and future directions using the province of British Columbia as a test study. Healthcare Management F O R U M 1992;5:3439. 3. Adas SJ, Chang Y C , Kammann E, Keller RB, Deyo RA, and Singer D E . Long-term disability and return to work among patients who have a herniated lumbar disc: The effect of disability compensation. Journal of Bone and Joint Surgery-American Volume 2000;82A:4-15. 4. Boszczyk B, Timothy J, Peul W, and Casey A T H . Neurosurgical training and the spine: Reflections on EANS Winter Meeting Luxembourg, February, 2006. Acta Neurochirurgica 2007;149:339. 5. Carey TS, Garrett JM, and Jackman A M . Beyond the good prognosis - Examination of an inception cohort of patients with chronic low back pain. Spine 2000;25:115-20. 6. Carrel T and Pfammatter JP. Complete transposition of the great arteries: Surgical concepts for patients with systemic right ventricular failure following intraatrial repair. Thoracic and Cardiovascular Surgeon 2000;48:224-7. 248 7. Cloutier-Fisher D, Penning MJ, Zheng C, and Druyts EBF. The devil is in the details: trends in avoidable hospitalization rates by geography in British Columbia, 1990-2000. B M C Health Services Research 2006;6. 8. Coyte PC, Wright JG, Hawker G A et al. Waiting-Times for Knee-Replacement Surgery in the United-States and Ontario. New England Journal of Medicine 1994;331:1068-71. 9. DeCoster C, Carriere K C , Peterson S, Walld R, and MacWilliam L. Waiting times for surgical procedures. Medical Care 1999;37:JS187-JS205. 10. Fisher C, Noonan V, Bishop P et al. Outcome evaluation of the operative management of lumbar disc herniation causing sciatica. Journal of Neurosurgery 2004;100:317-24. 11. Gentleman JF, Parsons GF, Walsh M N , and Vayda E. High and low surgical procedure rates in census divisions across Canada. Health Reports 1994;6:403-40. 12. Hu RW, Jaglal S, Axcell T, and Anderson G. A population-based study of reoperations after back surgery. Spine 1997;22:2265-70. 13. Kelly K D , Voaklander D, Kramer G, Johnston DW, Redfern L, and Suarez-Almazor M E . The impact of health status on waiting time for major joint arthroplasty. Journal of Arthroplasty 2000;15:877-83. 14. Kelly K D , Voaldander DC, Johnston WC, and Suarez-Almazor M E . Equity in waiting times for major joint arthroplasty. Canadian Journal of Surgery. 2002;45:269-76. 249 15. Lavis JN , Maker A , Anderson G M et al. Trends in hospital use for mechanical neck and back problems in Ontario and the United States: discretionary care in different health care systems. Canadian Medical Association Journal 1998;158:29-36. 16. Lofvendahl S, Eckerlund I, Hansagi H , Malmqvist B, Resch S, and Hanning M . Waiting for orthopaedic surgery: factors associated with waiting times and patients' opinion. International Journal for Quality in Health Care 2005;17:133-40. 17. Sanmartin C, Shortt SED, Barer ML, Sheps S, Lewis S, and McDonald PW. Waiting for medical services in Canada: lots of heat, but little light. Canadian Medical Association Journal 2000;162:1305-10. 18. Sheps SB, Reid RJ, Barer M L et al. Hospital downsizing and trends in health care use among elderly people in British Columbia. Canadian Medical Association Journal 2000;163:397-401. 19. Sobolev B, Brown P, and Zelt D. Potential for bias in waiting time studies: events between enrolment and admission. Journal of Epidemiology and Community Health 2001;55:891-4. 20. Sobolev B, Brown P, Zelt D, and Shortt S. Bias inherent in retrospective waiting-time studies: experience from a vascular surgery waiting list. Canadian Medical Association Journal 2000;162:1821-2. 21. Stillerman CB, Chen TC, Couldwell WT, Zhang W, and Weiss M H . Experience in the surgical management of 82 symptomatic herniated thoracic discs and review of the literature. Journal of Neurosurgery 1998;88:623-33. 22. Wasiak R, Pransky G, Verma S, and Webster B. Recurrence of low back pain: Definition-sensitivity analysis using administrative data. Spine 2003;28:2283-91. 250 23. Wasiak R, Pransky GS, and Webster BS. Methodological challenges in studying recurrence of low back pain. Journal of Occupational Rehabilitation 2003;13:21-31. 24. Welch H G . Should the Health-Care Forest be Selectively Thinned by Physicians Or Clear Cut by Payers. Annals of Internal Medicine 1991;115:223-6. 25. Wennberg JE, Barnes BA, and Zubkoff M . Professional Uncertainty and the Problem of Supplier-Induced Demand. Social Science & Medicine 1982;16:811-24. 26. Wennberg JE, Bunker JP, and Barnes B. The Need for Assessing the Outcome of Common Medical Practices. Annual Review of Public Health 1980;1:277-95. 27. Williams JI, Llewellyn T H , Arshinoff R, Young N , and Naylor CD. The burden of waiting for hip and knee replacements in Ontario. Ontario Hip and Knee Replacement Project Team. Journal of Evaluation in Clinical Practice 1997;3:59-68. 28. Wing PC. Rheumatology: 13. Minimizing disability in patients with low-back pain. Canadian Medical Association Journal 2001;164:1459-68. 251 APPENDICES Appendix A: Search Strategy - Studies on Symptom Duration Primary search strategy for E M B A S E and M E D L I N E databases on the association between symptom duration and outcomes of lumbar discectomy An initial search of the EMBASE, M E D L I N E databases (using medical subject headings: lumbar vertebrae, lumbosacral region and intervertebral disc; then cross-referenced by the keywords: surgery, laminectomy, discectomy, outcome, and timing OR duration of symptoms) initially identified only one study that explored symptom duration as a possible predictor of the outcome of first-surgery for LDH(30). A forward search for articles citing this reference was then carried out on the Science Citation Index Expanded (SCI-EXPANDED®) database. Fifty-four articles were identified, the titles and abstracts of which were scanned for discussions on the effect of symptom duration and/or timing of surgery on the outcomes of patients with L D H . In this manner, 2 more articles of direct relevance were identified(22;31). Additional articles were then obtained iteratively by hand searching the references within these original papers and those of other articles obtained subsequently. 252 Appendix B: Search Strategy - Studies on Waiting Time Primary search strategy for E M B A S E and M E D L I N E databases on the effect of waiting/waitlist time on outcomes of back pain treatments 1. ("low back pain" or "back pain").mp. [mp=title, original title, abstract, name of substance word, subject heading word] 2. ("lumbar dis$" or "orthopl surgery" or "spin$ surgery").mp. [mp=tide, original tide, abstract, name of substance word, subject heading word] 3. ("wait list$" or ("waiting listf" or "wait rime?" or "waiting time" or waitlist$ or awaitf or queu$)).mp. [mp=title, original tide, abstract, name of substance word, subject heading word] 4. (outcome or Q O L or H R Q O L or HRQL).mp. [mp=tide, original tide, abstract, name of substance word, subject heading word] 5. 1 or 2 6. 3 and 4 and 5 7. from 6 keep 1-45 8. from 7 keep 14,39 9. Find similar to Incapacity for work in elective orthopaedic surgery: a study of occurrence and the probability of returning to work after treatment. 10. Find similar to Waiting for orthopaedic surgery: factors associated with waiting times and patients' opinion. 253 Appendix C: Carriere's T Statistic As summarized by To {To 2003}, Carriere and Roos originally presented a method that does not rely on parametric assumptions such as the sample variance to estimate the variance of rates.(16) The authors use a T 2 statistic to test the following null hypothesis: Rh Rp *-• MRJ -> % w i t h H - l d f H 0 : R , - R 2 - R 3 - . . . — R H ~ RQ where RQ is the event rate to be compared against and can be the known event rate for the standard region used for the age and/or sex adjustment; R , R 2 , R 3 etc. are rates for different regions and ^ar{j^J IS th e estimated variance of the rate for each region. For rare events and small populations, this statistic may not be satisfactorily approximated by a chi-square distribution. To overcome such situations, the authors suggested the logarithmic transformation to adjust for skewness in the distribution of the standardized rates. {Carriere 1997} Under this transformation, the estimated variances of the standardized rate is given by: Var(lnRh)=Var(Rh)/R2h 254 Appendix C (continued) Under the Poisson assumption, the estimated variance VardnRh) can be written more explicidy as Var(\n ftj = ^  bljChj/R-l' H e r e > e # i s t n e t o t a ' number of events in the /th stratum in the M i region, and * where y^. and yi are the number of people in the yth stratum in the standard region and in the M i small region respectively. The previous T 2 statistic becomes: 2 «[ln(A)-ln(A)T T = Y i—-WwithH-ldf » M i n A) Z For large n6 or relatively high incidence or procedure rates, both versions of the T 2 statistic will yield similar conclusions. {Carriere 1997} With very rare events and small population sizes, the use of a logarithmic transformation results in T 2 statistic that more accurately approximates the chi-square distribution with degrees of freedom equal to one less the number of areas being compared. 255 Appendix C (continued) Carriere provided two methods for calculating confidence intervals for the standardized rates. However, for comparisons of rates across four or more areas, which is usually the case in health services research, they recommend using the Bonferroni method assuming a normal distribution to correct for multiple comparisons. The 100(l-a)% confidence region for the rates is given by where the m is the total number of comparisons. 256 

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0100710/manifest

Comment

Related Items