Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The relationship between long-term adherence to recommended clinical procedures and health care utilization… Krueger, Hans 2006

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

Item Metadata

Download

Media
831-ubc_2006_fall_krueger_hans.pdf [ 1.28MB ]
Metadata
JSON: 831-1.0092771.json
JSON-LD: 831-1.0092771-ld.json
RDF/XML (Pretty): 831-1.0092771-rdf.xml
RDF/JSON: 831-1.0092771-rdf.json
Turtle: 831-1.0092771-turtle.txt
N-Triples: 831-1.0092771-rdf-ntriples.txt
Original Record: 831-1.0092771-source.json
Full Text
831-1.0092771-fulltext.txt
Citation
831-1.0092771.ris

Full Text

THE RELATIONSHIP BETWEEN LONG-TERM ADHERENCE TO RECOMMENDED CLINICAL PROCEDURES AND HEALTH CARE UTILIZATION FOR ADULTS WITH DIAGNOSED TYPE 2 DIABETES by HANS KRUEGER B.R.S., The Mennonite Brethren Bible College, 1981 B.A. (Hons), The University of Winnipeg, 1982 M.Sc., The University of British Columbia, 1988 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Health Care and Epidemiology)  THE UNIVERSITY OF BRITISH COLUMBIA  © Hans Krueger, 2006  ABSTRACT Background: Diabetes is a common and serious chronic condition. If not well-managed, significant multi-system complications often arise, resulting in increased health care utilization and poor health outcomes. There is considerable evidence that people with diagnosed diabetes are not receiving recommended care. A comprehensive program aimed at improving adherence to recommended care can improve patient outcomes and result in cost-savings. The key aim of this study was to determine whether the long-term receipt of appropriate clinical procedures by patients with type 2 diabetes was associated with higher medical care costs. Methodology: A cohort of 20,288 diagnosed type 2 diabetes patients was identified using physician and hospital records. An analytic file was created by linking information on patient characteristics with utilization of physician and acute care services during a five-year period (1996 to 2001). Adherence to recommended clinical procedures for the assessment of blood glucose, blood pressure and cholesterol levels, as well as retinopathy and nephropathy, were measured during this same five-year period. Subjects were assigned to both a categorical (low, medium and high) and a binary (low and high) adherence group. Physician and acute care resource use was converted to constant 2000 Canadian dollars. Multivariate logistic regression was used to assess the relationship between patient characteristics, including adherence as a categorical variable, and utilization of physician and acute care services. Results: Long-term adherence was suboptimal, with patients receiving just 53% of recommended procedures. Adherence to recommended procedures, however, improved during the five year period. Patient characteristics associated with poor adherence include being male, younger, low socio-economic status, having no diabetes-specific complicating conditions and living in certain geographic areas. Patients with high long-term adherence (receiving 73% of  ii  recommended clinical procedures) were 59% more likely to use a high level of physician resources but 22% less likely to use a high level of acute care resources. On the other hand, patients with low adherence (receiving 31% of procedures) were 28% less likely to use a high level of physician resources but 17% more likely to use a high level of acute care resources. The utilization difference related to adherence was particularly noticeable in older adults with higher levels of morbidity. Elderly patients in this low adherence group were more likely to be hospitalized (64.3% vs. 55.8% over the five-year period) and, when they were hospitalized, tended to stay in hospital for longer periods of time (11.9 vs. 6.7 days) than patients in the high adherence group. Conclusion: Improving long-term adherence may result in the avoidance of $4 in acute care costs for every additional $1 in physician costs. If all patients moved into the high adherence category, as much as $3.1 million in annual costs might be avoided across the study sample. If this analysis is applied to all adults with diagnosed diabetes in the province of British Columbia, the annual costs avoided could reach the level of $34.4 million. Systemic changes are required in the provision of primary care to promote long-term adherence to recommended diabetes care.  iii  TABLE OF CONTENTS ABSTRACT.................................................................................................................................... ii TABLE OF CONTENTS............................................................................................................... iv LIST OF TABLES......................................................................................................................viiii LIST OF FIGURES ..................................................................................................................... xiv LIST OF ABBREVIATIONS..................................................................................................... xvii ACKNOWLEDGEMENTS......................................................................................................xviiii CHAPTER I: INTRODUCTION.................................................................................................... 1 CHAPTER II: BACKGROUND .................................................................................................... 3 2.1 Diabetes................................................................................................................................. 3 2.1.1 Diagnosed vs. Undiagnosed Cases ................................................................................ 6 2.2 The Provision of Medical Care to Diabetic Populations ...................................................... 7 2.2.1 The Burden of Diabetes-Related Morbidity .................................................................. 7 2.2.2 The Effectiveness of Diabetes-Related Medical Care ................................................. 14 2.2.3 The Economic Impact of Diabetes-Related Illness...................................................... 18 2.3 Recommended Clinical Procedures .................................................................................... 24 2.3.1 Clinical Practice Guidelines......................................................................................... 24 2.3.2 Clinical Practice Guidelines for Diabetic Care............................................................ 27 2.3.2.1 Guideline Development in Canada ........................................................................... 27 2.3.2.2 Guideline Development in British Columbia ........................................................... 29 2.3.2.3 Diagnosis of Diabetes ............................................................................................... 29 2.3.2.4 Optimal Targets for the Control of Diabetes ............................................................ 31 2.3.2.5 Recommended Diagnostic Procedures to Assess the Ongoing Control of Diabetes 33 2.3.3 Recommended Clinical Procedures ............................................................................. 37 2.4 Adherence to Recommended Clinical Procedures.............................................................. 42 2.4.1 Compliance versus Adherence..................................................................................... 42 2.4.2 Assessing Adherence to Recommended Clinical Procedures...................................... 43 2.4.3 Why is Adherence Generally Sub-Optimal?................................................................ 49 2.4.3.1 Physician Factors ...................................................................................................... 49 2.4.3.3 Organizational Factors .............................................................................................. 53 2.5 Improving Adherence to Recommended Clinical Procedures............................................ 59 2.6 Potential for Savings Associated with Improved Adherence ............................................. 62 2.7 Summary ............................................................................................................................. 68 CHAPTER III: METHODS.......................................................................................................... 71 3.1 Conceptual Framework....................................................................................................... 71 3.2 Study Design / Overview .................................................................................................... 74 3.3 Specific Aims and Hypothesis ............................................................................................ 75 3.4 Data Sources ....................................................................................................................... 77 3.4.1 British Columbia Linked Health Database .................................................................. 77 3.4.2 British Columbia Medical Services Plan Files ............................................................ 78 3.4.3 British Columbia Hospital Separations Files............................................................... 82 3.4.4 British Columbia Vital Statistics Database.................................................................. 83 3.5 Study Population................................................................................................................. 84 iv  3.5.1 Ascertaining Diabetic Cases ........................................................................................ 84 3.5.2 MSP Exclusions ........................................................................................................... 87 3.5.3 Diagnostic Rule-Outs................................................................................................... 88 3.5.4 Children........................................................................................................................ 88 3.5.5 Gestational Diabetes .................................................................................................... 88 3.5.6 Incident Cases .............................................................................................................. 89 3.5.7 Death ............................................................................................................................ 89 3.5.8 Temporary Residents ................................................................................................... 89 3.5.9 Temporary MSP Registration ...................................................................................... 90 3.5.10 Exclusion of Disease and Age-Specific Sub-Groups................................................. 90 3.5.11 Outliers....................................................................................................................... 91 3.5.12 Summary .................................................................................................................... 94 3.6 Variables and Measures ...................................................................................................... 96 3.6.1 Adherence Variables.................................................................................................... 96 3.6.2 Patient Characteristics.................................................................................................. 99 3.6.3 Resource Use Variables ............................................................................................. 108 3.6.3.1 Acute Care .............................................................................................................. 108 3.6.3.2 General Practitioner ................................................................................................ 112 3.6.3.3 Specialist Physician ................................................................................................ 116 3.6.3.4 Total Acute Care and MSP Costs ........................................................................... 119 3.7 Analytic Methods.............................................................................................................. 122 CHAPTER IV: RESULTS.......................................................................................................... 129 4.1 Description of the Study Population ................................................................................. 129 4.1.1 Overview of Study Population................................................................................... 129 4.1.2 Age and Gender ......................................................................................................... 131 4.1.3 Prevalence and Incidence Rates................................................................................. 133 4.1.4 False Negative Results in Diagnostic Rule-outs........................................................ 136 4.1.5 Comparison of the Generic Morbidity and Disease-Specific Severity Indices ......... 137 4.2 Description of Individual Adherence Variables ............................................................... 142 4.2.1 Overview.................................................................................................................... 142 4.2.2 Trend Analysis ........................................................................................................... 146 4.3 Description of Summary Adherence Variables ................................................................ 148 4.3.1 Adherence as a Continuous Variable......................................................................... 148 4.3.2 Adherence Analyzed as a Categorical Variable (Low, Med, High) .......................... 154 4.3.3 Adherence Analyzed as a Binary Variable (Low, High) ........................................... 155 4.3.4 Adherence by Age, Morbidity and Gender................................................................ 157 4.3.5 Trends in Adherence over Time for Low and High Adherence Groups.................... 159 4.4 Univariate Logistic Regression Models for Adherence................................................... 160 4.5 Multivariate Logistic Regression Model for Adherence .................................................. 163 4.5.1 Development of a Reduced Main Effects Model for Adherence............................... 163 4.5.2 Development of a Final Fitted Model for Adherence................................................ 166 4.5.3 Interpretation of the Final Fitted Model for Adherence............................................. 170 4.6 Description of Resource Use Variables ............................................................................ 172 4.6.1 Overview of Resource Use Variables ........................................................................ 172 4.6.2 Mean Annual per Capita Costs by Age, Morbidity and Gender................................ 176  v  4.7 Univariate Logistic Regression Models for Average Annual Total Physician Costs....... 178 4.8 Multivariate Logistic Regression Model Average Annual Total Physician Costs ........... 180 4.8.1 Development of a Reduced Main Effects Model for Physician Costs....................... 180 4.8.2 Development of a Final Fitted Model for Physician Costs........................................ 183 4.8.3 Interpretation of the Final Fitted Model for Physician Costs .................................... 190 4.10 Multivariate Logistic Regression Model for Average Annual Total Acute Care Costs . 197 4.10.1 Development of a Reduced Main Effects Model for Acute Care Costs .................. 197 4.10.2 Development of a Final Fitted Model for Acute Care Costs ................................... 200 4.10.3 Interpretation of the Final Fitted Model for Acute Care Costs................................ 200 4.11 Univariate Logistic Regression Models for Average Annual Total Costs ..................... 203 4.12 Multivariate Logistic Regression Model for Average Annual Total Costs .................... 206 4.12.1 Development of a Reduced Main Effects Model for Total Costs............................ 206 4.12.2 Development of a Final Fitted Model for Total Costs............................................. 210 4.12.3 Interpretation of the Final Fitted Model for Total Costs.......................................... 215 4.13 Analysis of Mean Annual Per Capita Physician and Acute Care Costs ......................... 218 4.13.1 Comparison of Annual Costs by Adherence and Morbidity.................................... 218 4.13.2 Comparison of Annual Costs By Adherence, Morbidity and Gender ..................... 220 4.13.3 Comparison of Annual Costs By Adherence, Morbidity and Age .......................... 224 4.13.4 Utilization of Acute Care Services By Older Adults with High Morbidity............. 229 4.14 Summary of Key Results ................................................................................................ 232 CHAPTER V: DISCUSSION..................................................................................................... 242 5.1 Study Strengths ................................................................................................................. 242 5.1.1 Administrative Data Set............................................................................................. 242 5.1.2 Inclusion Criteria ....................................................................................................... 246 5.2 Study Limitations.............................................................................................................. 248 5.2.1 Potential Utilization Bias ........................................................................................... 248 5.2.2 Missing Drug Cost Variable ...................................................................................... 249 5.2.3 Issues of External Validity......................................................................................... 250 5.3 Issues for Further Research............................................................................................... 252 5.3.1 Development of Adherence Measures ....................................................................... 252 5.3.2 Provider or System Variables .................................................................................... 256 5.4 The Study Findings in Perspective ................................................................................... 260 5.4.1 Adherence Rates ........................................................................................................ 260 5.4.2 Patient Characteristics Associated with Low Adherence .......................................... 264 5.4.3 Utilization of Health Care Resources ........................................................................ 267 5.4.4 The Relationship between Long-Term Adherence and Use of Health Care Resources ............................................................................................................................................. 269 5.5 Policy Implications ........................................................................................................... 270 BIBLIOGRAPHY....................................................................................................................... 279 Appendix A: Allocation of ACGs Into Morbidity Levels .......................................................... 309 Appendix B: PROSSER’s Algorithm ......................................................................................... 312 Appendix C: Description of Individual Adherence Variables.................................................... 316 Appendix D: Description of Individual Resource Use Variables............................................... 327  vi  Appendix E: Calculation of Change in Acute Care, Physician and Total Costs with Improved Adherence ................................................................................................................................... 342  vii  LIST OF TABLES Table  Page  2.1  Diagnostic Criteria for Diabetes in Non-pregnant Adults Based on the Canadian Diabetes Clinical Practice Guidelines...................................................... 30  2.2  Optimal Targets for the Control of Diabetes Based on the Canadian and British Columbia Diabetes Clinical Practice Guidelines ......................................... 32  2.3  Recommended Diagnostic Procedures to Assess the Ongoing Control of Diabetes Based on the Canadian and British Columbia Diabetes Clinical Practice Guidelines .................................................................................................. 36  2.4  Percent of Patients Receiving Recommended Clinical Procedures......................... 47  2.5  Comparison of Traditional and Collaborative Care in Chronic Illness ................... 57  3.1  Excluded Non-Physician Specialties ....................................................................... 87  3.2  Adherence Variables................................................................................................ 99  3.3  Patient Characteristics............................................................................................ 107  3.4  CHSPR Categorization Matrix for Personnel Funded Through MSP ................... 117  3.5  Price Increases for Specialist Physician Services 1996/97 to 2000/01.................. 119  3.6  Resource Use Variables ......................................................................................... 121  4.1  Study Population Characteristics ........................................................................... 130  4.2  Frequency Distribution For Age (April 1, 1998) and Gender ............................... 132  4.3  People with Diagnosed Diabetes – Fraser Health and Ontario – 1998.................. 134  4.4  People with Newly Diagnosed Diabetes – Fraser Health and Ontario – 1998 ...... 135  4.5  Diagnostic Rule-Outs – False Negatives ............................................................... 136  4.6  Disease Specific Severity Index By Age Group .................................................... 138  4.7  Generic Morbidity Index (ACG) By Age Group................................................... 139  viii  4.8  Comparison of Generic Morbidity Index (ACG) and Disease-Specific Severity Index ........................................................................................................ 141  4.9  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving the Following Recommended Services........................................................................ 143  4.10  Eye Exams by Physician Specialty and Fiscal Year.............................................. 145  4.11  Frequency Distribution For Adherence Scores...................................................... 148  4.12  Adults with Diagnosed Type 2 Diabetes Mean Adherence Scores ....................... 150  4.13  Relationship Between Individual Adherence Measures Fiscal 1998/99................ 152  4.14  Relationship Between Individual Adherence Measures Fiscal 1999/00................ 153  4.15  Relationship Between Individual Adherence Measures Fiscal 2000/01................ 153  4.16  Proportion of Adults with Diagnosed Type 2 Diabetes with Low, Medium or High Adherence................................................................................................. 154  4.17  Proportion of Adults with Diagnosed Type 2 Diabetes with Low or High Adherence .............................................................................................................. 156  4.18  Adults with Diagnosed Type 2 Diabetes Mean Adherence Scores by Age and Level of Morbidity................................................................................................. 158  4.19  Females with Diagnosed Type 2 Diabetes Mean Adherence Scores by Age and Level of Morbidity................................................................................................. 158  4.20  Males with Diagnosed Type 2 Diabetes Mean Adherence Scores by Age and Level of Morbidity................................................................................................. 158  4.21  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving the Following Recommended Services High Adherence Group .................................................. 160  4.22  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving the Following Recommended Services Low Adherence Group ................................................... 160  4.23  Univariate Logistic Regression Models for High vs. Low Adherence.................. 161  4.24  Development and Testing of a Multivariate Logistic Regression Model for High vs. Low Adherence ...................................................................... 164  4.25  Main Effects Multivariate Logistic Regression Model for High vs. Low Adherence ....................................................................................... 165  ix  4.26  ACG by DSSI Cross Tab ....................................................................................... 167  4.27  Final Fitted Multivariate Logistic Regression Model for High vs. Low Adherence .............................................................................................................. 169  4.28  Proportion of Adults with Diagnosed Type 2 Diabetes with High Utilization Of Average Annual GP, Specialist Physician, Acute Care and Total Costs.......... 173  4.29  Per Capita Mean Utilization of Hospital and MSP Services By Adults with Diagnosed Type 2 Diabetes ................................................................................... 175  4.30  Adults with Diagnosed Type 2 Diabetes Mean Annual Per Capita Costs By Age and Level of Morbidity .................................................................................. 177  4.31  Females with Diagnosed Type 2 Diabetes Mean Annual Per Capita Costs By Age and Level of Morbidity............................................................................. 177  4.32  Males with Diagnosed Type 2 Diabetes Mean Annual Per Capita Costs By Age and Level of Morbidity............................................................................. 177  4.33  Univariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Physician Costs ................................................................. 179  4.34  Development and Testing of a Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Physician Costs ...................................................................................................... 181  4.35  Main Effects Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Physician Costs........................................... 182  4.36  Age by Adherence Cross Tab ................................................................................ 184  4.37  Age by DSSI Cross Tab......................................................................................... 186  4.38  Final Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Physician Costs ................................................................. 189  4.39  Proportion of Adults with Diagnosed Type 2 Diabetes – High vs. Low Utilization of Average Annual Physician Costs .................................................... 191  4.40  Univariate Logistic Regression Models for High vs. Low Utilization of Average Annual Acute Care Costs ........................................................................ 196  x  4.41  Development and Testing of a Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Acute Care Costs....................................................................................... 198  4.42  Main Effects Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Costs..................................... 199  4.43  Univariate Logistic Regression Models for High vs. Low Utilization of Average Annual Total Costs ........................................................... 204  4.44  Development and Testing of a Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Costs ................... 207  4.45  Main Effects Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Costs ........................................................... 209  4.46  Final Multivariate Logistic Regression Model for High vs. Low Utilization of Average Annual Total Costs.............................................................................. 214  4.47  Mean Annual Costs per Capita by Level of Adherence and Level of Morbidity ............................................................................................................... 219  4.48  Mean Annual Costs per Capita For Females by Level of Adherence and Level of Morbidity................................................................................................. 221  4.49  Mean Annual Costs per Capita For Males by Level of Adherence and Level of Morbidity................................................................................................. 223  4.50  Mean Annual Costs per Capita For Ages 30 – 59 by Level of Adherence and Level of Morbidity .......................................................................................... 225  4.51  Mean Annual Costs per Capita For Ages 60 – 79 by Level of Adherence and Level of Morbidity .......................................................................................... 227  4.52  Utilization of Acute Care Services By Adults Aged 60 – 79 with Diagnosed Type 2 Diabetes In the High or Very High Morbidity Groups............ 230  4.53  Patient Characteristics Associated with High or Low Adherence ......................... 233  4.54  Summary of Odds Ratios for Adherence and High Average Annual Cost Categories ...................................................................................................... 234  4.55  Patient Characteristics Associated with High or Low Use of Physician Costs ...................................................................................................... 236  xi  4.56  Patient Characteristics Associated with High or Low Use of Acute Care Costs .............................................................................................................. 237  4.57  Patient Characteristics Associated with High or Low Use of Total Costs ............ 239  5.1  Patient Characteristics of BC Health Authority Residents .................................... 251  5.2  Potential Annual Acute Care, Physician and Total Costs Based on Improving Adherence ............................................................................................ 272  5.3  Potential Annual per Capita Change in Acute Care, Physician and Total Costs Based on Improving Adherence................................................................... 274  5.4  Comparison of Traditional and Collaborative Care in Chronic Illness ................. 277  C-1  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving 2 or More HbA1c Tests per Year .................................................................................. 317  C-2  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving At Least One Eye Exam per Year ........................................................................................ 318  C-3  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving At Least One Microalbumin Test per Year .......................................................................... 320  C-4  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving At Least One Lipid Test Every Three Years ........................................................................ 321  C-5  Proportion of Adults with Diagnosed Type 2 Diabetes with At Least Four Blood Pressure Measurements per Year ................................................................ 323  C-6  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving All Five Recommended Services ......................................................................................... 325  C-7  Proportion of Adults with Diagnosed Type 2 Diabetes Receiving None of the Five Recommended Services................................................................................. 326  D-1  Adults with Diagnosed Diabetes Utilization of Acute Care Days......................... 328  D-2  Proportion of Adults with Diagnosed Diabetes High Utilization of Average Annual Acute Care Dollars.................................................................................... 329  D-3  Adults with Diagnosed Diabetes Utilization of General Practitioner Visits.......... 331  D-4  Proportion of Adults with Diagnosed Diabetes High Utilization of Average Annual General Practitioner Dollars..................................................................... 332  xii  D-5  Adults with Diagnosed Diabetes Utilization of Specialist Physician Visits.......... 335  D-6  Proportion of Adults with Diagnosed Diabetes High Utilization of Average Annual Specialist Physician Dollars...................................................................... 336  D-7  Adults with Diagnosed Diabetes Mean Total Costs .............................................. 339  D-8  Proportion of Adults with Diagnosed Diabetes with Low or High Utilization of Average Annual Total Costs ........................................................... 340  E-1  Mean Annual Costs per Capita by Level of Adherence and Level of Morbidity ........................................................................................................... 342  E-2  Total Annual Costs by Level of Adherence and Level of Morbidity .................... 342  E-3  Total Annual Costs by Level of Adherence and Level of Morbidity if Low Adherence Groups Move to Medium Adherence.......................................... 343  E-4  Total Annual Costs by Level of Adherence and Level of Morbidity if Medium Adherence Groups Move to High Adherence ......................................... 343  E-5  Total Annual Costs by Level of Adherence and Level of Morbidity if Low Adherence Groups Move to High Adherence ............................................... 344  E-6  Total Annual Costs by Level of Adherence and Level of Morbidity if Low Adherence Groups Move to Medium Adherence and Medium Adherence Groups Move to High Adherence........................................................ 344  E-7  Total Annual Costs by Level of Adherence and Level of Morbidity if Low & Medium Adherence Groups Move to High Adherence............................. 345  xiii  LIST OF FIGURES Figure 3.1  Conceptual Framework for Patients / Physicians with High Adherence to Recommended Medical Care................................................................................... 72  3.2  Study Overview ....................................................................................................... 74  3.3  Identification of Outliers – Acute Care Utilization.................................................. 92  3.4  Identification of Outliers – General Practitioner Utilization ................................... 93  3.5  Identification of Outliers – Specialist Physician Utilization.................................... 93  3.6  Selection of Study Population.................................................................................. 95  4.1  Frequency Distribution By Age (April 1, 1998) and Gender ................................ 133  4.2  Proportion of Adults With Diagnosed Type 2 Diabetes Who Received Recommended Services ......................................................................................... 143  4.3  Proportion of Adults With Diagnosed Type 2 Diabetes Who Received Recommended Services ......................................................................................... 144  4.4  Frequency Distribution for Adherence Scores....................................................... 149  4.5  Age by Sex Interaction........................................................................................... 167  4.6  General By Disease-specific Morbidity Interaction .............................................. 168  4.7  Age by Adherence Interaction ............................................................................... 184  4.8  Age by DSS Index Interaction ............................................................................... 187  4.9  Age by Adherence Interaction ............................................................................... 211  4.10  Morbidity by Adherence Interaction...................................................................... 212  4.11  Socio-economic Status by Gender Interaction....................................................... 213  4.12  Mean Annual Cost per Capita – Low vs. High Adherence.................................... 220  4.13  Mean Annual Cost per Capita – Females – Low vs. High Adherence .................. 222  4.14  Mean Annual Cost per Capita – Males – Low vs. High Adherence...................... 224  xiv  4.15  Mean Annual Cost per Capita for Ages 30 - 59 – Low vs. High Adherence ........ 226  4.16  Mean Annual Cost per Capita for Ages 60 - 79 – Low vs. High Adherence ........ 228  4.17  Adults Aged 60 – 79 with Diagnosed Type 2 Diabetes in the High or Very High Morbidity Groups Proportion Hospitalized and ALOS ....................... 231  5.1  Potential Annual Acute Care and Physician Costs Based on Improving Adherence .............................................................................................................. 273  D-1  Frequency Distribution for Average Annual Acute Care Inpatient Days.............. 327  D-2  Frequency Distribution for Average Annual General Practitioner Visits ............. 330  D-3  Frequency Distribution for Average Annual Specialist Physician Visits.............. 334  D-4  Frequency Distribution for Average Annual Total Costs ...................................... 338  xv  LIST OF ABBREVIATIONS ACG  Adjusted Clinical Groups  ACE Inhibitor Angiotensin-Converting Enzyme (ACE) Inhibitors ARB  Angiotensin Receptor Blockers  ADA  American Diabetes Association  ADG  Aggregated Diagnostic Group  AIDS  Acquired Immune Deficiency Syndrome  AHCPR  Agency for Health Care Policy and Research  AMI  acute myocardial infarction  APP  Alternate Payment Program  BCCDM  British Columbia Chronic Disease Management  BCLHD  British Columbia Linked Health Database  BCMA  British Columbia Medical Association  CDA  Canadian Diabetes Association  CHSPR  Centre for Health Services and Policy Research  CIHI  Canadian Institute for Health Information  CPG  clinical practice guidelines  CPI  Consumer Price Index  CVD  cardiovascular disease  DAD  discharge abstract database  DAN  diabetic autonomic neuropathy  DCCT  Diabetes Control and Complications Trial  DCA  Diabetic Care of America  DDMP  diabetes disease management program  DPN  distal symmetric polyneuropathy  DR  diabetic retinopathy  EA  enumeration area  EDC  expanded diagnostic cluster  ESRD  end-stage renal disease  FFS  fee-for-service  xvi  FHA  Fraser Health Authority  GP  General Practitioner  GPAC  Guidelines and Protocols Advisory Committee  HDL  high-density lipoprotein  HbA1c  glycosylated haemoglobin  HPMG  Health Partners Medical Group  ICD  International Classification of Diseases  IDF  International Diabetes Federation  IDDM  insulin dependent diabetes mellitus  IPPE  income per person-equivalent  LHA  local health area  MRRS  most recently registered specialty  MSC  Medical Services Commission  MSP  Medical Services Plan  NIDDM  non-insulin dependent diabetes mellitus  OOP  out of province  PHCO  Physician Hospital Community Organization  PVD  peripheral vascular disease  QALY  quality adjusted life years  RIW  resource intensity weigh  SES  socio-economic status  SPSS  Statistical Package for the Social Sciences  TOP  type of practice  TRIAD  Translating Research Into Action for Diabetes  UBC  University of British Columbia  UK  United Kingdom  UKPDS  United Kingdom Prospective Diabetes Study  US  United States  VA  Veterans Affairs  VH  Vancouver Hospital  WHO  World Health Organization  xvii  ACKNOWLEDGEMENTS Completion of this dissertation would not have been possible without the support, encouragement and prayers of many faculty members, colleagues, family members and friends. Dr. Charlyn Black, my supervisor and mentor, has been extremely supportive in this process. This project would not have been completed without her generosity of time and resources. My other committee members, Dr. Rob Reid and Dr. Sam Sheps, provided invaluable feedback throughout this process. I am also indebted to Dr. Morris Barer who, during a difficult time early in this process, provided much needed encouragement and support. Dr. Diane Watson was influential in the early stages of study design while Dr. Anne-Marie Broemeling provided support, encouragement and helpful input on her study. Dr. Bob Prosser showed an immense amount of patience in answering my numerous, and at times, unsophisticated statistical questions. Bo Green turned an immense volume of data into meaningful variables. A number of staff at the Centre for Health Services and Policy Research, including Denise Morettin, Kerry Kerluke and Dawn Mooney, generously provided their expertise in accessing and analysing the data from the BC Health Linked Data set. To my colleague, Dan Williams, thank-you for putting in all that extra time keeping the business functioning so that I could finish this dissertation. To friends Ray Saucy, Tom Balke and Ted Andres; we can now go out for coffee! To my father, who through historical circumstances never had the opportunities for education that I have; thank-you for your life-long support. My children, Katrina, Alicia and Joshua grew up while I was working on this dissertation. Thank-you for being such good kids! Finally, I owe an immense debt of gratitude to my wife, Anna, who continued to provide support even when I felt like giving up. Thank-you for your love. October 24, 2006  xviii  CHAPTER I: INTRODUCTION During the last two decades, a series of studies have pointed out the unequivocal relationship between the tight control of blood glucose and blood pressure levels in patients with diabetes and a reduction in both acute and chronic complications. This research has been so compelling that the American Diabetes Association offered the following conclusion: "it is time for all health professionals to treat diabetes aggressively. It is also time for patients to take their diabetes with the utmost seriousness. And it is incumbent upon the health care system to provide the necessary resources for both to be successful. Compromise or acceptance of a disadvantageous and dangerous status quo in people with diabetes should not be tolerated any longer" (Genuth et al., 2003, p. S32). In addition to reducing a patient’s longevity and quality of life, the complications associated with poorly managed diabetes result in significant costs to the health care system. The first step in controlling blood glucose and blood pressure is to determine what the appropriate levels should be and then to execute the necessary diagnostic tests to decide if those levels are being achieved. Information on diagnostic test outcomes is rarely available in administrative data sets. On the other hand, it is often possible to determine whether or not a given test was received by a patient and how often that test was received. In this study we used information on the receipt of five recommended clinical tests (as available in the B.C. Linked Health Data set) to determine if there is a relationship between high adherence to the procedures and the use of health care services. Adherence was measured for the assessment of blood glucose, blood pressure and cholesterol levels, as well as retinopathy and nephropathy, over a five year period. While other studies have assessed adherence to recommended clinical procedures, to our knowledge this is  1  the first study to consider long-term adherence, including repeated patient exposure to the procedure. The primary objectives of this study were: 1) to assess whether adherence to recommended clinical procedures changed over time; 2) to determine which patient-level characteristics were associated with low or high adherence; and 3) to determine whether adults with diagnosed type 2 diabetes with higher adherence to recommended clinical procedures utilize more health care services.  2  CHAPTER II: BACKGROUND 2.1 Diabetes Diabetes mellitus is a chronic disorder of metabolism. It occurs when the body can no longer absorb glucose due to the lack of insulin production or the inability to use the insulin that is produced. Insulin, a hormone produced in the pancreas, is required for glucose to be absorbed from the blood stream into cells, where the glucose is metabolized to produce energy. Without insulin, or without the ability for the body to use insulin appropriately, glucose remains in the blood stream, starving cells of energy; as well, the excess glucose in the blood stream, over time, may result in damage to a variety of body organs and systems. For instance, if there is not enough insulin for the body’s cells to use the available glucose, the body begins to use fat instead, resulting in ketoacidosis; this condition, if left untreated, eventually leads to unconsciousness and death. There are four main types of diabetes. Type 1 diabetes is an autoimmune disease that occurs when the insulin-producing beta cells in the pancreas are damaged or destroyed, causing a reduction in, or the cessation of, insulin production (Atkinson and Maclaren, 1994). The aetiology of type 1 diabetes is not well understood, but the disease is believed to be the result of an individual’s genetic vulnerability together with a possible viral or other infectious trigger; the infection induces an autoimmune response that damages the already vulnerable insulinproducing beta cells in the pancreas (Gavin et al., 2003). The incidence of type 1 diabetes in Canada is highest in children 10-14 years of age (Toth et al., 1997; Blanchard et al., 1997). In Ontario, for example, the incidence rate for female children in the calendar year 2000 ranged from 19.9 per 100,000 among 0-4 year old females to 33.5 among 10-14 year old females (To et al., 2003). Similar results are seen in male children, with the incidence rates ranging from 25.0  3  per 100,000 among 0-4 year old males to 35.9 among 10-14 year old males. Type 1 diabetes is considerably less common than type 2 diabetes, accounting for less than 10% of persons with diabetes. Type 2 diabetes is the most common form of diabetes, occurring in approximately 90% of patients with diabetes. Type 2 diabetes results when the pancreas produces sufficient insulin, but the body cannot use the insulin effectively. This condition, known as insulin resistance, causes the pancreas to secrete additional insulin to maintain normal blood sugar levels. In approximately one-third of people with insulin resistance, either the body’s cells do not respond to the higher levels of insulin or, over time, insulin production decreases, resulting in the high blood glucose levels of type 2 diabetes (DeFronzo et al., 1992). Obesity and physical inactivity aggravate insulin resistance, contributing to the severity of disease. While the incidence of type 1 diabetes is highest in children, type 2 diabetes tends to begin manifesting in adults at mid-life (Engelgau, 2004). It should be noted, however, that the prevalence of type 2 diabetes in children is increasing along side the emerging epidemic of childhood obesity (Ludwig and Ebbeling, 2001). There is a steady increase in incidence rates in the older population. In Ontario in 1999, for example, the incidence of diabetes increased from 0.41 per 100 for women 35-49 years of age to 0.95 per 100 for women 50-64 years of age and 1.28 per 100 for women 65-74 years of age (Hux and Tang, 2003). The rates for men were slightly higher, at 0.51, 1.28 and 1.65, respectively. Twin and family studies have also identified a strong genetic component to type 2 diabetes, with an increased risk among siblings of an individual with diabetes that is at least three times higher than the population at large among individuals with European ancestry (Elbein, 2002; Elbein et al., 2002). The strongest genetic link known at this time is due to variants of the  4  calpain-10 gene, though a number of other genes have been implicated (Elbein et al., 2002; Carlsson et al., 2005). Gestational diabetes is the third main type of diabetes. It occurs in approximately 4% of pregnant women who have not had diabetes before (Engelgau et al., 1988). In Canada, the prevalence of gestational diabetes varies from 3.5% to 3.8% in the non-aboriginal population to 8.0% to 18.0% in aboriginal populations (Harris et al., 1997; Godwin et al., 1999; Rodrigues et al., 1999; Dyck et al., 2002). While the aetiology is not well understood, it is believed that hormones from the placenta block the action of the mother’s insulin in her body, resulting in insulin resistance and the subsequent build-up of blood glucose levels. Gestational diabetes usually disappears with the termination of the pregnancy but there remains an increased risk for the mother of later impaired glucose tolerance and type 2 diabetes (Henry and Beischer, 1991; Ben-Haroush et al., 2003; Albareda et al., 2003). Women with gestational diabetes have a 1763% risk of type 2 diabetes within 5-16 years after their pregnancy (Hanna and Peters, 2002). Finally, the fourth category is diabetes secondary to other conditions. These consist of diabetes associated with genetic defects of beta cell function, genetic defects in insulin action, diseases of the pancreas, endocrinopathies, infections, uncommon forms of immune-mediated diabetes, drug or chemical induced diabetes and other genetic syndromes sometimes associated with diabetes (Canadian Diabetes Association, 2003). This discussion in this study is limited to type 1 and 2 diabetes. Diabetes is usually diagnosed when one or more of a set of common signs and symptoms are exhibited by the person or by screening of high-risk individuals, confirmed by a high level of blood glucose. A positive diagnosis of diabetes is made when an individual’s test results are  5  higher than a preset standard on any of three common tests of plasma glucose. These tests, and the related plasma glucose values, include: •  A fasting plasma glucose value of ≥7.0 mmol/L  •  A casual (any time of day, without regard to the interval since the last meal) plasma glucose value of ≥11.1 mmol/L.  •  An oral glucose tolerance test plasma glucose value of ≥11.1 mmol/L in a blood sample taken two hours after a person has consumed 75 grams of glucose dissolved in water.  A positive result needs to be confirmed by a second positive test on a different day, unless there is unequivocal evidence of hyperglycaemia accompanied by acute metabolic decompensation (Canadian Diabetes Association, 2003). 2.1.1 Diagnosed vs. Undiagnosed Cases While the diagnosis of diabetes is relatively straightforward, there appear to be a significant proportion of the population with undiagnosed type 2 diabetes (Leiter et al., 2001; Worral and Moulton, 1992). A study in Manitoba (Young and Mustard, 2001) found the prevalence of undiagnosed type 2 diabetes to be approximately 2.2% of the adult population in that province, representing approximately one-third of all type 2 diabetes cases. This proportion is similar to the estimated 2.7% of the population aged 20 years and older in the United States with undiagnosed type 2 diabetes (Harris et al., 1998, Wilder et al., 2005). A recent audit in the United Kingdom estimated that 23% of individuals who have type 2 diabetes have not been recorded as having diabetes by their general practitioners (National Diabetes Audit, 2005). The fact that between a quarter and a third of people with type 2 diabetes remain undiagnosed is a public health concern. The onset of type 2 diabetes typically occurs at least 4-7  6  years before clinical diagnosis (Harris et al., 1992). Diabetes-related complications may develop during this time while earlier detection and treatment may reduce the development of these complications (Harris and Eastman, 1996). As presented in the following section, the burden of diabetes-related morbidity is high if the disease is not well-managed. Individuals with diabetes who are undiagnosed are, of course, also untreated. The concern is that preventable diabetesrelated complications can develop prior to diagnosis. 2.2 The Provision of Medical Care to Diabetic Populations 2.2.1 The Burden of Diabetes-Related Morbidity Diabetes is one of the most serious of the chronic diseases, with significant multi-system complications if the disease is not well-managed. Acute complications include diabetic ketoacidosis, hyperosmolar nonketotic coma, and hypoglycaemia, plus a higher susceptibility to common infections (Booth and Fang, 2003). Chronic complications fall into two main categories: microvascular (nephropathy, retinopathy and neuropathy) and macrovascular (ischemic heart disease, stroke and peripheral vascular disease). On the acute side, both ketoacidosis and hyperosmolar coma are characterized by severe elevations in blood glucose levels (hyperglycaemia); emergencies associated with these conditions involve life-threatening metabolic disturbances. Patients with type 1 diabetes are more likely to present with diabetic ketoacidosis than patients with type 2 diabetes. The annual rate of diabetic ketoacidosis is estimated at 46 per 10,000 individuals with diabetes (Faich, et al., 1983; Snorgaard et al., 1989), while hyperosmolar coma occurs less frequently. In a survey of 312 admissions for ketoacidosis and hyperosmolar coma, MacIsaac and co-authors (2002) found that 55% were admitted for ketoacidosis, 15% for a hyperosmolar, hyperglycaemic state, and 30% for a combination of the two. Further, the mortality rate was 7  1.2% for patients with ketoacidosis, 17% for patients presenting with a hyperosmolar, hyperglycaemic state, and 5.3% for patients with a combined state. Similar results were found in a larger study, with a mortality rate of 4.9% for diabetic ketoacidosis and 14.6% for hyperosmolar coma (Hamblin et al., 1989). Among chronic complications, diabetic nephropathy affects 25-45% of patients with diabetes (Jawa et al., 2004). In its earliest stages, diabetic nephropathy presents with low levels of albumin in the urine (microalbuminuria). If the course of diabetic nephropathy progresses, it may eventually lead to chronic or end-stage renal failure (ESRD). In fact, diabetes is the leading cause of ESRD, a condition in which the patient requires a renal transplant or dialysis in order to live. The risk of developing ESRD is up to 13 times higher in persons with diabetes than those without the condition (Brancati et al., 1997; Perneger et al., 1994). Over 40% of patients starting dialysis treatment for renal problems in Canada and the United States have diabetes (Canadian Institute of Health Information, 2001; National Institutes of Health, 2001). Early (non-proliferative) diabetic retinopathy (DR) has a prevalence of at least 70% in persons with type 1 diabetes (Klein et al., 1984a) and 40% in persons with type 2 diabetes (Klein et al., 1984b). Non-proliferative DR may progress to proliferative DR, characterized by the appearance of new retinal blood vessels. If detected early, proliferative DR can be treated with retinal laser photocoagulation to reduce the risk of vision loss (Buhrmann et al., 2003). If left untreated, proliferative DR represents a serious threat to vision, leading to blindness in 50% of patients within 5 years (Caird et al., 1968). Proliferative DR presents in approximately 50% of individuals with type 1 diabetes and 10% of individuals with type 2 diabetes after they have the disease for 20 years (Klein et al., 1984a,b). In individuals under the age of 65, over half of all cases of blindness are caused by diabetes (Jawa et al., 2004).  8  Diabetic neuropathy is among the most common of the long-term complications associated with diabetes, afflicting an estimated 50% of individuals with diabetes (Young et al., 1993; Dyck et al., 1993). There are a variety of types of diabetic neuropathy, with the most common ones being chronic sensorimotor distal symmetric polyneuropathy (DPN) and diabetic autonomic neuropathy (DAN) (Boulton et al., 2005). DAN primarily affects the gastrointestinal, genitourinary and cardiovascular systems. Gastrointestinal disturbances include esophageal enteropathy, gastroparesis, constipation, diarrhoea and fecal incontinence. Genitourinary tract disturbances include bladder and/or sexual dysfunction. In men, it is associated with loss of penile erection and /or retrograde ejaculation. DAN is also associated with reduced cardiovascular autonomic function, resulting in a doubling of the risk of silent myocardial ischemia. Finally, DAN is associated with dry skin, loss of sweating, and the emergence of fissures and cracks that allow micro-organisms to enter, ultimately contributing to the development of ulcers, gangrene and limb loss (Vinik et al., 2003). Pain, especially in the lower limbs, is the most outstanding complaint of people with DPN; the pain is often described as deep and aching or sudden, sharp, and stabbing – like an “electric shock.” Patients may also experience severe weight loss, depression, and, in males, erectile dysfunction. Other symptoms include a constant burning discomfort in the feet, numbness of the feet, and unsteadiness resulting from disturbed proprioception and abnormal muscle sensory function (Boulton et al., 2004). Cardiovascular disease (CVD) accounts for approximately 70% of all deaths among people with diabetes (Gu et al., 1998). Mortality from CVD is two to three times higher in men with diabetes compared to the rest of the male population, and as much as five times higher in women with diabetes (Almdal et al., 2004; Stamler et al., 1993; Kannel and McGee, 1979). On  9  average, individuals with diabetes tend to have an acute myocardial infarction 10-15 years earlier than the general population (Booth et al., 2003). Among persons with diabetes, stroke is 2 to 4 times as common as found among persons without diabetes (Jorgensen et al., 1994; Jamrozik et al., 2000). Diabetes influences the occurrence and experience of stroke in several ways. The diabetic stroke patient is younger on average, recovers more slowly and is at a higher risk of death from a stroke than the non-diabetic stroke patient (Jorgensen et al., 1994). An estimated 37-42% of all strokes are attributable to the effects of diabetes alone or of diabetes in combination with hypertension (Kissela et al., 2005). Persons with diabetes have a two- to four-fold increase in the rate of peripheral vascular disease (PVD), most often affecting the lower leg (Beckman et al., 2002). PVD can result in a significant range of functional impairments. At one end of the spectrum, there is painful walking. More seriously, when it is not possible to restore adequate blood supply to the limbs, amputation may be required. Approximately 40-60% of all lower limb amputations are performed in patients with diabetes (Apelqvist and Larsson, 2000). In addition to the acute and chronic complications associated with diabetes, persons with type 2 diabetes are also at a higher risk of other co-morbidities, including hypertension, depression and ischemic heart disease (Broemeling, et al., 2005). In British Columbia, 31% of persons with diabetes also had hypertension, 11% were diagnosed with depression and 10% with ischemic heart disease. In Saskatchewan (Simpson, et al., 2003), 36% of health care expenditures for people with diabetes are attributable to major co-morbidities. The constellation of possible co-morbidities is consistent with evidence from the United States, which indicates that more than 40% of Americans with a chronic illness have at least one other co-existent chronic condition (Hoffman, et al., 1996). The probability of co-morbidities increases with the age of the individual  10  (Wolff et al., 2002). Evidence from the Netherlands suggests that, while approximately 21% of people with diabetes under the age of 65 have at least one co-morbidity, the proportion increases to 40% for those over the age of 65 (Schellevis et al., 1993). Individuals with type 2 diabetes are also more likely to present with ‘metabolic syndrome.’ The World Health Organization (WHO, 1999) has defined the metabolic syndrome as the presence of at least two of the following criteria in an individual: 1. Central obesity (body mass index > 30 kg/m2 and / or a waist-to-hip ratio > 0.90 m in males, > 0.85 m in females); 2. Dyslipidaemia (triglycerides ≥ 1.7 mmol/l and or HDL < 0.9 mmol/l in males, < 1.0 mmol/l in females or hypolipidemic treatment); 3. Arterial hypertension (≥ 140/90 mmHg or anti-hypertensive treatment); 4. Microalbumineria (30-299 mg/l). The prevalence of individuals with the metabolic syndrome in the United States has increased rapidly during the last two decades in both adults and adolescents. In the year 2000, at least 27% of adults and 9.2% of adolescents were identified as having the metabolic syndrome (Ferranti et al., 2004; Ford et al., 2004). Since the 1999 WHO definition, a number of other groups have attempted to define the metabolic syndrome, leading to substantial confusion and absence of comparability between studies (Alberti et al., 2005). To address this confusion, the International Diabetes Federation (IDF) convened a consensus group in 2004 consisting of all previous organizations involved in generating the previous definitions together with members from all IDF regions. This consensus group have defined the metabolic syndrome as the follows (Alberti et al., 2005):  11  1. The presence of central obesity defined by ethnic specific values for waist circumference; 2. Plus any two of the following: a. Raised triglycerides - > 150 mg/dL (1.7 mmol/L) and / or specific treatment for this lipid abnormality b. Reduced HDL-cholesterol - < 40 mg/dL (1.03 mmol/L) in men; < 50 mg/dL (1.29 mmol/L) in women and / or specific treatment for this lipid abnormality c. Raised blood pressure – Systolic ≥ 130 mm Hg; diastolic ≥ 85 mm/Hg and / or treatment of previously diagnosed hypertension d. Raised fasting plasma glucose – Fasting plasma glucose ≥ 100 mg/dL (5.6 mmol/L) and / or previously diagnosed type 2 diabetes. If above 5.6 mmol/L or 100 mg/dL, oral glucose tolerance test is strongly recommended, but is not necessary to define presence of the syndrome. Despite differences in definitions, the general consensus is that the presence of the metabolic syndrome in patients with type 2 diabetes influences the risk of chronic complications (Isomaa et al., 2001; Hanna and Neary, 2004; Saely et al., 2005; Khunti et al., 2005). Bonora et al. (2003), for example, found that the presence of the metabolic syndrome in patients with type 2 diabetes was independently associated with an almost five-fold increase in cardiovascular disease. Sundstrom, et al. (2006) suggest that the metabolic syndrome can now be added as one of the established risk factors (in addition to smoking, diabetes, hypertension and serum cholesterol) for cardiovascular disease. Due to the high level of both acute and chronic complications, as well as co-morbidities, the diabetes-related risk of mortality is significantly higher than mortality in the general  12  population. In people aged 15-34 years with type 1 diabetes, standardized mortality ratios are approximately 3.5 times higher than the general population (Wibell et al., 2001). Young people who have been admitted to hospital for diabetes have a nine times higher standardized mortality ratio than the general population, which includes a higher risk of death from suicide (Roberts et al., 2004). The age-adjusted relative risk of death from all causes in persons with type 2 diabetes is approximately 2 for men (Lutofo et al., 2001) and 3 for women (Hu et al., 2001), increasing to 5 and 7 respectively if the person with diabetes also has coronary heart disease. In terms of life expectancy, people without diabetes live 12-13 years longer than people with diabetes (Manual and Schultz, 2004). In addition to increased risk of premature mortality, people with diabetes also suffer significant disability, with an estimated 20-50% reporting limitations in their activities (Songer, 1995). The health-related quality of life for people with diabetes has been estimated at between 0.6 - 0.9 on a scale from 0 to 1 with ‘1’ representing perfect health and ‘0’ representing death (Maddigan et al., 2000; Coffey et al., 2002). There is considerable variation depending on who is doing the evaluation (Landy et al., 2002). This scale has also been used to quantify the impact that major complications have on the individual’s health-related quality of life (Clarke et al., 2002). Specifically, researchers estimated the impact of myocardial infarction at -0.055, blindness in one eye at -0.074, ischemic heart disease at -0.090, heart failure at -0.108, stroke at 0.164 and amputation at -0.280. In summary, diabetes is one of the most serious of the chronic diseases, with significant multi-system complications if the disease is not well-managed. As a result, individuals with diabetes tend to have a shorter life expectancy, as well as significantly reduced quality of life compared to the general population.  13  2.2.2 The Effectiveness of Diabetes-Related Medical Care While the morbidity and premature mortality associated with diabetes is significant, there is an important body of evidence which indicates that appropriate management of this chronic condition can delay and / or prevent the related complications. The management of diabetes is aimed at reducing the acute and chronic complications associated with diabetes, primarily by maintaining the patient’s blood glucose, blood pressure and lipid levels as close to normal as possible. This involves a combination of diet, smoking cessation, exercise, social support and drug therapy, the latter consisting of some combination of antihypertensive and cholesterol lowering agents, insulin injections or oral hypoglycaemic agents. In type 1 diabetes, the use of insulin therapy is always required, while normoglycaemia can sometimes be achieved in type 2 diabetes through diet and exercise alone, though concomitant oral hypoglycaemics or insulin are often also required. The Diabetes Control and Complications Trial (DCCT) was a large, comprehensive diabetes clinical study conducted from 1983 to 1993 by the National Institute of Diabetes and Digestive and Kidney Diseases, based in the United States (DCCT Research Group, 1993). It involved 1,441 volunteers with type 1 diabetes from 29 medical centers in the United States and Canada. Volunteers had been diagnosed with diabetes for at least 1 year, but no longer than 15 years. This study compared the effects of two treatment regimens – standard therapy and intensive control – on the incidence of acute and chronic complications of diabetes. Volunteers were randomly assigned to each treatment group. Intensive control involved self-testing blood glucose levels four or more times a day, four daily insulin injections or use of an insulin pump, frequent adjustment of insulin doses according  14  to food intake and exercise, a diet and exercise plan, and monthly visits to a health care team composed of a physician, nurse educator, dietician, and behavioural therapist. This study found that intensive therapy with normalization of blood glucose levels reduced the risk of developing diabetic retinopathy by 76%, prevented the development and slowed the progression of diabetic kidney disease by 50%, and reduced the risk of nerve damage by 60% at 6.5 years of follow-up. Follow-up research on this study population indicates that the beneficial results of this intensive control continued for a period of at least eight years, even though the difference in mean glycated haemoglobin (HbA1c) levels between the two treatment groups diminished over time, so that there was ultimately an average difference of only 0.2% during the follow-up period (DCCT Research Group, 2000, 2002, 2003). The authors concluded that "the current results reaffirm that intensive treatment of type 1 diabetes should be initiated as early as safely possible in order to provide strong and durable protection from the development and progression of diabetic microvascular disease" (DCCT Research Group, 2003, p. 2166). The United Kingdom Prospective Diabetes Study (UKPDS Group, 1998a) followed 1,148 patients with hypertension and type 2 diabetes (enrolled between 1987 and 1991) for an average of nine years. Patients were randomly allocated to a tight control of blood pressure group (with a goal of < 150/85 mm Hg) and a less tight control of blood pressure group (with a goal of <180/105 mm Hg). Patients visited study centres every 3-4 months. At each visit, plasma glucose concentration, blood pressure, and body weight were measured. Treatments to control blood pressure and blood glucose concentration were assessed and adjusted if target values were not met.  15  The results of the UKPDS indicate that the tight blood pressure control group experienced a 32% reduction in death related to diabetes, a 44% reduction in stroke, a 56% reduction in the risk of heart failure, a 37% reduction in microvascular disease, and a 47% reduced risk of deterioration of visual acuity. This study also found a strong relationship between glucose levels and subsequent cardiovascular events. For every 1% reduction in glycated haemoglobin (HbA1c), the authors observed a 14% drop in the incidence of acute myocardial infarction (AMI) and a 16% drop in heart failure rates. After reviewing the new information from the United Kingdom Prospective Diabetes Study, the American Diabetes Association stated that "it is time for all health professionals to treat diabetes aggressively. It is also time for patients to take their diabetes with the utmost seriousness. And it is incumbent upon the health care system to provide the necessary resources for both to be successful. Compromise or acceptance of a disadvantageous and dangerous status quo in people with diabetes should not be tolerated any longer" (Genuth et al., 2003, p. S32). A study in Denmark randomly assigned 80 individuals with type 2 diabetes to a “targeted, intensified, multifactorial intervention” and 80 to receive conventional treatment (Gaede et al., 2003). The intervention group received a stepwise implementation of behaviour modification (diet, exercise, and smoking cessation) and pharmacologic therapy that targeted hyperglycaemia, hypertension, dyslipidemia, and microalbuminuria. The intervention group saw significant improvements (compared to the conventional treatment group) in their glycosylated haemoglobin values, systolic and diastolic blood pressures, serum cholesterol, triglyceride level and urinary albumin excretion rate at the end of the 7.8 year follow-up. At the end of that time period, 44% of patients in the conventional therapy group had one or more cardiovascular events  16  (death from cardiovascular causes, nonfatal myocardial infarction or stroke, coronary- or peripheral-artery revascularization, or amputation as a result of ischemia) compared to only 24% in the intervention group. Solomon (2003), commenting on the study by Gaede et al. (2003) notes that this study “provides the best evidence to date of the magnitude of the benefits that can be derived from instituting several interventions”. But even with the intensive interventions offered in the Denmark study, targeted blood pressure and blood glucose levels were only infrequently achieved. Less than half of the patients in the intensive therapy group achieved target systolic blood pressure levels while less than a fifth achieved targeted glycosylated haemoglobin levels. As noted by Solomon (2003), “although these findings point to the difficulty of achieving the targets in the real world, they also suggest the possibility of even greater benefits if the targets can be met more frequently.” Despite significant strides in the treatment of diabetes, the patients themselves must invest a considerable amount of time, energy and resources in dealing with diabetes, presenting a constant challenge for people with the disease. The 2nd edition of Diabetes in Canada (Health Canada, 2002a) notes that "diabetes exerts a significant effect on the quality of life of those with the disease. The continuous need to monitor intake (in terms of timing, type and amount of food), take medications (whether pills or insulin injections), monitor blood glucose, and anticipate and plan for activities that may affect diabetes control can put a severe strain on daily life" (p. 10). It is perhaps not surprising then that target blood glucose and blood pressure levels are consistently hard to achieve, even in clinical trials. The difficulty in consistently achieving targets noted in the study by Gaede and colleagues (2003) has been confirmed by a number of  17  other studies (Menard, et al., 2005; Rothman and Elasy, 2005; Karter et al., 2005). Karter et al. (2005) found that just 18% of the patients in their study of new antihyperglycaemic therapies reached a target HbA1c of ≤7.0%. Menard et al. (2005) found that even after a year of intensive multitherapy intervention most patients were substantially below the goal of 100% adherence. Just 35% of patients in the intervention arm achieved the goals of HbA1c of ≤7.0%, 64% reached a diastolic blood pressure of < 80 mm Hg, 53% reached a low-density lipoprotein cholesterol level of < 2.5 mmol/L and 44% achieved triglyceride levels of < 1.5 mmol/L. It should be noted that these levels were significantly better than the no intervention control group (8%, 37%, 20% and 14%, respectively for the control group). Just six months after the completion of the research trial and the return to usual care, however, the benefits achieved during the trial had vanished; specifically, as noted in an editorial by Rothman and Elasy (2005), “there were no longer statistically significant differences in haemoglobin A1c concentrations, blood pressure or triglyceride levels between the intervention and control groups.” In summary, the considerable complications associated with diabetes can be delayed and /or minimized with appropriate management. The most successful outcomes are achieved when there is a strong partnership between the patient and the clinical team. When clinical vigilance and patient accountability are relaxed, however, the research suggests that target levels of blood glucose and blood pressure are difficult to maintain. 2.2.3 The Economic Impact of Diabetes-Related Illness The current economic burden of diabetes is substantial, at least partly due to the high level of complications faced by patients whose chronic condition is not well-managed. In 1992, people with diabetes accounted for 15% of total US health care expenditures, even though they constituted only 4.5% of the total population (Rubin et al., 1994). Bagust and colleagues (2001)  18  have estimated the lifetime health care costs for patients with diagnosed type 2 diabetes to be more than twice that for an equivalent non-diabetic population. The cost to health plans in the US increases significantly with every 1% increase in HbA1c levels above 6% (Gilmer et al., 1997). Moss et al. (1999) found that increases in HbA1c levels positively predicted hospitalizations among people with diabetes. In contrast, factors in the diabetic population that were not significantly associated with hospitalizations included age, gender, systolic and diastolic blood pressures, body mass, smoking status, and alcohol consumption. According to a study commissioned by the American Diabetes Association (2003), diabetes was estimated to cost the US economy $132 billion in 2002. Direct medical expenditures are estimated at $91.8 billion, which includes $23.2 (25%) billion for diabetes care, $24.6 (27%) billion for chronic complications attributable to diabetes, and $44.1 (48%) billion for excess prevalence of general medical conditions. Indirect costs, totalling $39.8 billion, were associated with lost workdays, restricted activity days, premature mortality, and permanent disability due to diabetes. The authors note that the $132 billion "likely underestimates the true burden of diabetes because it omits intangibles, such as pain and suffering, care provided by nonpaid caregivers, and several areas of health care spending where people with diabetes probably use services at higher rates than people without diabetes (e.g., dental care, optometry care, and the use of licensed dieticians). In addition, the cost estimate excludes undiagnosed cases of diabetes" (ADA, 2003, pg 917). After adjusting for differences in age, gender, and race/ethnicity, people with diabetes utilized approximately 2.4 times the health care resources of someone in the general population. Estimates of the economic burden of diabetes in Canada are quite variable. A study by Health Canada (2002b) suggests that the economic burden of diabetes in Canada was $1.6 billion  19  (Cdn $) in 1998. Health Canada acknowledges that this is a very conservative estimate, one which does not include physician costs, costs associated with the complications of diabetes, costs borne by patients and costs associated with short-term disability as well as the value of time lost from work and leisure activities by family members or friends who care for the patient. An alternate estimate from Health Canada (2002a), simply based on the relative population sizes of the two countries, suggests that the true economic burden may be as high as 10% of the U.S. figure, or $13.2 billion in 2002. Simpson et al. (2003) estimated the direct health care costs (hospitalizations, physician services and prescription drugs) for the 3.6% of Saskatchewan’s population with diagnosed diabetes to be $143.3 million in 1996, or approximately 15% of total expenditures in these three areas that year. These direct costs averaged $3,524 per person per year. Ohinmaa et al. (2004) used this Saskatchewan data, with the addition of day surgery and outpatient dialysis costs, to estimate the direct health care costs in Canada for people with diagnosed diabetes to be $4.66 billion in 2000. Dawson et al. (2002) used a broader approach in estimating the economic burden of diabetes in Canada, including estimated costs for undiagnosed cases as well as indirect costs (mortality related productivity losses). Their estimate of the total economic burden of diabetes in Canada in 1998 was between US$4.76 and $5.23 billion. Laditka and co-workers (2001) compared the resource use of people with diabetes to those without diabetes in an employer-based population in Ohio in 1996. The commercial health insurer in that state had an enrolment of approximately 828,000 employed individuals, of whom 1.6% had diabetes. This 1.6% of the population generated 9.4% of the insurer's costs. Total annual per capita costs for the non-diabetic population were $909, compared to $5,659 for the diabetic cohort. More specifically, the diabetic population used inpatient resources at a rate 4.8  20  times that of the non-diabetic population, after adjusting for the age and gender differences in the two populations. Rates of resource use were also 2.5 times higher for outpatient facility encounters, 2 times higher for emergency department visits, 2.4 times higher for physician office visits, 3 times higher for physician consultation visits and 2.8 times higher for ancillary services such as laboratory and radiology tests. Several groups within Canada have examined aspects of resource use by persons with diabetes compared to the general population. Research on the level of family physician utilization in Winnipeg, Manitoba indicates that persons with diabetes see their family physician just over two times as often as the general population (Watson et al., 2003). In Ontario, patients with diabetes visited a physician or optometrist 2.2 times more frequently in 2000/01 compared to patients without diabetes (Chan and Harju, 2003). Klarenbach and Jacobs (2003) provide a comparison of health resource utilization in Canada and the US. They found that patients with diabetes in Canada were more likely to have contact with a general physician and an eye specialist, but were less likely to have contact with other medical specialists compared to their American counterparts. Several studies provide information on the average annual medical care costs by people with diabetes. Not surprisingly, hospitalizations create a huge financial burden on society. In a European study, the average annual direct medical cost for people with diabetes was estimated at €2,515 (Jönsson, 2002). Of this amount, 53% was for hospitalization, 24% for ambulatory care and 23% for drugs. Simpson et al. (2003) estimated the health care expenditures for people with diagnosed diabetes in Saskatchewan in 1996. Average costs per capita were $3,524 consisting of $1,889 (53.7% of the total) in hospitalization costs, $836 (23.8%) in prescription drug costs,  21  $583 (16.6%) in physician services costs, $115 (3.3%) in dialysis costs and $96 (2.7%) in day surgery costs. The juvenile burden is similar to that of adults. Hospitalization rates in children with diabetes are approximately 3 to 7 times that of their peers (Aro et al., 1994; Icks et al., 2001). Broemeling and co-authors (2005) examined resource use in adults with diagnosed diabetes in British Columbia in 2000/01. In their study, the authors used Adjusted Clinical Groups (ACGs) to allocate people with diabetes into the following five groups based on comorbidity: •  No co-morbidity  •  Low co-morbidity (2-3 additional conditions ranging from minor acute and timelimited conditions to chronic, medically unstable, psychosocial, and major acute conditions)  •  Medium co-morbidity (4-5 additional conditions)  •  High co-morbidity (6-9 additional conditions)  •  Very high co-morbidity (10+ additional conditions)  Similar to other studies, the authors found that adults with diabetes used 2.4 times the health care resources compared to the general adult population. Their analysis revealed that the level of resources used varied significantly based on the level of co-morbidity. The healthiest group of individuals with diabetes, those with no co-morbidity, used only 0.1 times the level of resources. The group with low co-morbidity used 0.6 times the resources, and those with medium co-morbidity used 1.2 times the resources. Most significantly, the diabetic population with high co-morbidity used 3.9 times the resources and those with very high co-morbidity used  22  11.5 times the resources. Thirty-two percent of the diabetic population were in the high or very high co-morbidity groups. Williams et al. (2002) used disease specific complications when assessing the impact of complications on the costs of type 2 diabetes. They divided people with type 2 diabetes into the following four broad categories of complication status: 1. No complications. 2. One or more microvascular complications only. Microvascular complications include foot ulcer, amputation, retinopathy, photocoagulation, vitrectomy, blindness in one or both eyes, microalbuminuria, nephropathy, dialysis, renal transplant and neuropathy. 3. One or more macrovascular complications only. Macrovascular complications include angina, myocardial infarction, heart failure, percutaneous transluminal coronary angioplasty, coronary artery bypass graft, transient ischemic attack, stroke and peripheral vascular disease. 4. One or more of each of microvascular and macrovascular complications. A patient with no complications cost €1,505 in direct medical costs per year. Compared to a patient with no complications, the presence of microvascular complications added 70% to these costs, macrovascular complications added 100% to these costs and the presence of both micro and macrovacsular complications increased the costs by 330% (€5,226). In summary, the economic costs, both direct and indirect, associated with diabetes are high, largely due to the serious complications associated with the condition if it is not appropriately managed. In both Canada and the United States, average health care resource use for an individual with diabetes is at least two times that of the general population, though this  23  ratio is highly dependent on the presence of co-morbidities or complications. Over half of the average annual direct care costs are due to hospitalizations. 2.3 Recommended Clinical Procedures The complications and costs associated with diabetes are significant, yet both can be substantially reduced if the disease is well-managed, as will be summarized in section 2.5 (Potential for Savings Associated with Improving Planned Management) of this chapter. But can one identify and track ideal management? Clinical practice guidelines (CPGs) are one option for summarizing and distributing best practice information in an accessible format. As knowledge on best practices for a specific disease evolves, CPGs can be modified to take the new information into account. The following section will provide some background on CPGs and review the evolution of CPGs for diabetes in Canada and British Columbia. 2.3.1 Clinical Practice Guidelines Woolf (1990) has defined practice guidelines as “the official statements or policies of major organizations and agencies on the proper indications for performing a procedure or treatment or the proper management for specific clinical problems” (p. 1812). The most commonly used definition of clinical practice guidelines is that provided by Field and Lohr (1992): “(S)ystematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances” (p. 2). Guidelines are developed based on the best available evidence at the time of their development. As noted earlier, as knowledge on best practices for a specific disease evolves, CPGs can be modified to take the new information into account.  24  Guidelines have been used in one form or another for over 50 years in medicine, being described variously as practice standards, recommendations, protocols, policies, parameters, options, care maps, care pathways, and so on (Woolf, 1990, 1995; Woolf et al., 1999). In the United States, the federal government created a new agency in 1989, the Agency for Health Care Policy and Research (AHCPR), whose role, in part, was to develop CPGs. The AHCPR was given the mandate to develop three initial guidelines by January 1, 1991 (Woolf, 1990). Interest in CPGs at that time extended beyond North America to Australia and New Zealand, as well as to countries in Europe and Africa (Woolf et al., 1999). In the Netherlands, for example, guidelines have been worked on since 1987 by the Dutch College of General Practitioners, with the resulting CPGs figuring prominently in Dutch health policy. By 1993, over 1,500 CPGs had been developed in the US alone (Woolf, 1995). The rapid proliferation of CPGs in the late 1980s and 1990s was driven by at least three converging areas of research (Field and Lohr, 1992). The first involves the documentation of unexplained geographic variation in medical practices. Dating back to 1969, studies by Wennberg and others (Lewis, 1969; Wennberg and Gittelsohn, 1973; McPherson et al., 1982; Chassin et al., 1986; Wennberg et al., 1987; Perrin et al., 1989; McMahon et al., 1989) reported large, unexplained variances in the receipt of medical care by patients living in different geographic regions. While the reasons for these variances are not fully known, one suggestion has been that they may be explained in part by clinician uncertainty about the proper indications for procedures (Wennberg et al., 1977). If this is the case, then the provision of CPGs would be a potential approach to reducing this uncertainty (Wennberg, 1984). The second area of research is on the inappropriate use of interventions. The most important work in this area was completed by Chassin and colleagues (1987), in which 5,000  25  medical records were reviewed to measure the appropriateness of three procedures. Their findings suggested that 17% of coronary angiography, 32% of carotid endarterectomy, and 17% of upper gastrointestinal tract endoscopy procedures were inappropriate. Findings such as this have fuelled speculation about the inappropriate use of a broader range of medical procedures. The third area of research led to uncertainty about the health outcomes achieved by the use or non-use of various services and interventions (Eddy, 1984; Eddy and Billings, 1988; Brook, 1989; Brook et al., 1986; Roper et al., 1988). Research on the effectiveness, let alone the cost-effectiveness, of many health care services and procedures did not exist, or was incomplete. In response, Roper and Hackbarth (1988) encouraged the Health Care Financing Agency in the US to “purchase value, the optimal mix of high quality and reasonable cost” (p. 91) and to increase funding for research on effectiveness and the promotion of quality care. The perception that the costs of medical care, particularly in the US, were spiralling out of control in the 1980s, together with the three research streams noted above, led to a powerful motivation for the implementation of CPGs. It was hypothesized that the systematic combination of scientific evidence and clinical judgment would lead to recommendations for appropriate care that would be embraced by physicians and their patients alike, leading to better health outcomes and lower health care costs. In essence, CPGs were an attempt to control costs while improving quality of care, risk management, and patient outcomes (Field and Lohr, 1992). The five major purposes of CPGs as identified by Field and Lohr (1992) are: •  To assist in clinical decision-making by patients and practitioners  •  To educate individuals or groups  •  To assist in assessing and assuring the quality of care  •  To guide the allocation of resources for health care  26  •  To reduce the risk of legal liability for negligent care.  While a myriad of CPGs have been developed, the implementation of their recommendations into every day practice has not been smooth. In 1988, Eddy and Billings (p. 20) warned: To achieve high-quality medical care, we must succeed at three main tasks. First, we must determine just what practices constitute high-quality care. This involves analyzing evidence of the effectiveness, risks, and costs of various medical practices, and designing standards that define appropriate practices. The second task involves monitoring existing practices to compare them against the accepted standards. The third involves changing the behaviour of practitioners to ensure that the care actually delivered meets the standards. Failure at any of these tasks will threaten the quality of care people actually receive. 2.3.2 Clinical Practice Guidelines for Diabetic Care 2.3.2.1 Guideline Development in Canada National clinical practice guidelines for diabetes were first developed and distributed in Canada in September, 1992 (Tan et al., 1992), and then updated in 1998 (Meltzer et al., 1998) and again in 2003 (Canadian Diabetes Association, 2003). The 1992 guideline (Tan et al., 1992) were developed by an expert committee of volunteers consisting of specialist and family physicians, nurse educators, dieticians and a lawyer. This group prepared position papers which were then reviewed by 38 other health professionals prior to preparing a second draft of the guidelines for discussion at a public consensus conference. The final version of this early CPG was published in the September, 1992 issue of the Canadian Medical Association Journal. In 1998, the Canadian Diabetes Association’s Clinical and Scientific Section revised the 1992 CPG, incorporating new research developments (Meltzer et al., 1998). These guidelines were based on a series of recommendations, each recommendation graded according to the level  27  of supporting evidence. The process for developing the 1998 CPG was similar to the one used in 1992 with an expert committee preparing background papers for synthesis into a larger report, which was then reviewed by external experts and discussed at a public forum. The 2003 CPG (Canadian Diabetes Association, 2003) was developed in response to a number of important new research findings (Hu et al., 2001; Tuomilehto et al., 2001; Knowler et al., 2002). The new results included those from the UKPDS (UKPDS Group, 1998) and followup studies on the DCCT population (DCCT Research Group, 2000). The authors of the 2003 CPG generally followed the process established for the earlier versions of the diabetes CPG, with the following key principles (Canadian Diabetes Association, 2003): • •  • • •  Each recommendation had to address a clinically important question related to one or more of the prevention, detection or management of diabetes mellitus and its sequelae. Whenever possible, each recommendation had to be justified by the strongest clinically-relevant empirical evidence that could be identified; the citation(s) reporting this evidence had to be noted adjacent to the relevant guideline. A summary of the strength of this evidence, based on prespecified criteria from the epidemiological literature and other guidelines, had to be noted. The evidence had to be incorporated into a recommendation that was assigned a grade based on the available evidence, evaluating both its strength, and its applicability. Guidelines based on biological or mechanistic reasoning, expert opinion, or consensus had to be explicitly identified and graded as such (p. S4).  The authors acknowledged that CPGs have often fallen short of their intended goals. In order to improve participation, they developed a dissemination strategy which included a searchable web-based version of the guidelines, summary articles published in a variety of professional journals, and messages targeted to people with diabetes and the general public.  28  2.3.2.2 Guideline Development in British Columbia In British Columbia, CPGs are developed under the direction of the Guidelines and Protocols Advisory Committee (GPAC), jointly sponsored by the B.C. Medical Association (BCMA) and the Ministry of Health. The GPAC is co-chaired by the BCMA and the Medical Services Plan (MSP). GPAC chooses topics, approves draft guidelines for external review, and approves final guidelines for submission to the Medical Services Commission (MSC) for review and adoption in B.C. It also coordinates strategies to implement and evaluate guidelines. The decision about which guidelines to develop is based on high volume and / or costs of the specific medical condition, high variability in practice patterns, opportunities for improvement in practice, and the support and interest of physicians. The guideline development process in B.C. involves a literature search, draft guidelines, multiple consultations with experts and external reviews. Final approval of the guidelines is made by the B.C. Medical Association and the Medical Services Commission. Published guidelines are reviewed every two years, or even earlier if new evidence warrants. The latest version of the CPG for diabetes care was published in January, 2004. 2.3.2.3 Diagnosis of Diabetes The diagnostic criteria used in determining whether an individual has diabetes are summarized in section 2.1 above. A comparison of the CPGs developed for diabetes in Canada reveal that these criteria have changed since 1992 due to the availability of new research evidence (see Table 2-1 below).  29  Table 2-1: Diagnostic Criteria for Diabetes in Nonpregnant Adults Based on the Canadian Diabetes Clinical Practice Guidelines 1998 (5)  1992 FPG (mmol/L)  >7.8 mmol/L  Casual (1) PG (mmol/L)  >11.1 mmol/L (3) ≥11.1 mmol/L (6) ≥11.1 mmol/L (8)  2hPG in a 75-g OGTT (mmol/L)  >11.1 mmol/L (4) ≥11.1 mmol/L  (2)  ≥7.0 mmol/L  2003 (7) ≥7.0 mmol/L  ≥11.1 mmol/L  Notes (1) Casual = any time of the day without regard to the interval since the last meal (2) On at least two occasions (3) Plus symptoms and signs of diabetes (increased thirst, polydipsia, polyuria, polyphagia, weight loss, fatigue, blurred vision, etc.) (4) Observed on at least two occasions in addition to a FPG value of > 7.8 mmol/L. (5) A confirmatory test must be done on another day in all cases in the absence of unequivocal hyperglycemia accompanied by acute metabolic decompensation. (6) Plus symptoms of diabetes. The classic symptoms of diabetes include fatigue, polyuria, polydipsia and unexplained weight loss. (7) A confirmatory laboratory glucose test (an FPG, casual PG or a 2hPG in 75-g OGTT) must be done in all cases on another day in the absence of unequivocal hyperglycemia accompanied by acute metabolic decompensation (8) Plus symptoms of the disease. Classic symptoms of the disease = polyuria, polydipsia and unexplained weight loss. Abbreviations FPG = fasting plasma glucose, fasting is defined as no caloric intake for at least 8 hours PG = plasma glucose 2hPG = 2-hour plasma glucose OGTT = oral glucose tolerance test  The medical community has become increasingly aware of the dangers of high blood glucose levels, which is reflected in the changes to the CPGs. In 1992, an individual with a fasting plasma glucose level >7.8 mmol/L was considered to have diabetes. The threshold was lowered to ≥7.0 mmol/L in the 1998 guidelines. It is important to note that while the 2003 version of the guidelines maintains the 7.0 mmol/L level, they have added that blood glucose levels below this threshold also have clinical consequences. The authors of the 2003 guideline suggest the term “impaired fasting glucose” for any individual with a fasting plasma glucose level between 6.1 and 6.9 mmol/L (Canadian Diabetes Association, 2003). 30  2.3.2.4 Optimal Targets for the Control of Diabetes A comparison of all the three versions of the Canadian and the current British Columbia CPGs indicates that important targets for the control of blood glucose, blood pressure and cholesterol levels have all changed since 1992 (see Table 2-2 below). The pattern has been toward lowering the test criterion to enable earlier detection and broader inclusion.  31  Table 2-2: Optimal Targets for the Control of Diabetes Based on the Canadian and British Columbia Diabetes Clinical Practice Guidelines  Plasma glucose level (mmol/L) Pre-meal Post-meal Glycosylated hemoglobin level (% of upper limit) Equivalent HbA1c assay Cholesterol level LDL (mmol/L) Diabetes and either CAD or 3 or more other risk factors Diabetes and 2 other risk factors (6) Diabetes and 1 other risk factor Diabetes and no other risk factors High risk (7) Moderate Risk (8) Total (mmol/L) HDL (mmol/L) Total:HDL cholesterol ratio (mmol/L) Diabetes and either CAD or 3 or more other risk factors Diabetes and 2 other risk factors Diabetes and 1 other risk factor Diabetes and no other risk factors High risk (7) Moderate Risk (8)  1992  1998  2003  BC 2004  4.0 - 7.0 5.0 - 10.0  4.0 - 7.0 5.0 - 11.0  4.0 - 7.0 5.0 - 10.0  4.0 - 7.0 5.0 - 10.0  < 110  ≤ 115 < 7.0%  < 7.0%  < 7.0%  < 2.5 < 3.5  < 2.5 < 3.5  < 4.0 < 5.0  < 4.0 < 5.0  < 3.4  (1)  < 2.5 < 3.5 < 4.0 < 5.0  < 5.2 >1.1  < 1.7  BMI (under 65 years of age)  < 25  (5)  (2) (3) (4) (5)  (1)  (9)  "healthy"  < 140/90 (10) < 130/85  Microalbumin / Creatinine Ratio (ACR) Males Females Estimated Glomerular Filtration Rate (eGFR)  (4)  (1)  < 2.0 < 2.0 < 2.0 < 3.0  Blood pressure (mm Hg)  (3)  (1)  < 4.0 < 5.0 < 6.0 < 7.0  Triglyceride level (mmol/L) Diabetes and either CAD or 3 or more other risk factors Diabetes and 2 other risk factors Diabetes and 1 other risk factor Diabetes and no other risk factors  (2)  < 2.0 < 2.8  (2) (3) (4) (5)  18.5 - 24.9  18.5 - 24.9  < 130/80  < 130/80  < 2.0 < 2.0 (11) (11) < 2.8 < 2.8 ≥ 90 mL/min ≥ 90 mL/min  Notes (1) Should be adjusted for other risk factors. Less strict targets may be appropriate for older patients with limited life expectancy (2-5) All 3 target values (LDL, Total:HDL and triglyceride level) must be achieved. (6) Major risk factors include family history of premature CAD, smoking, hypertension, low HDL (≤0.9 mmol/L) and age over 30 years in both men and women (7) Most patients with diabetes (8) Younger age and shorter duration of diabetes and no other complications of diabetes and no other risk factors for vascular disease (9) "There are very little clinical data to support recommendations on TG levels…it is uncommon for a patient to have a significant elevation in serum TGs with LDL-C and TC:HDL-C at target levels. Thus, in order to simplify the lipid targets, a specific TG target level is not provided" (pg. S59). (10) "For patients with diabetic nephropathy, it has been suggested that optimal blood pressure should be 130/80 to 135/85 mm Hg; however, conclusive evidence is lacking" (pg. 700) (11) A patient is considered to have nephropathy at these levels if confirmed on at least 2 out of 3 tests taken at least one week apart. Abbreviations LDL = low-density lipoprotein HDL = high-density lipoprotein BMI = body mass index CAD = coronary artery disease TG = triglyceride Hg = Hemoglobin  32  In the 1992 guidelines, a glycated haemoglobin level of <110 (% of upper limit) was considered appropriate. Six years later this was modified to ≤115 (% of upper limit), or the equivalent haemoglobin A1c of <7%. Recommended cholesterol levels were initially set at a total cholesterol level of <5.2 mmol/L and a high-density lipoprotein (HDL) level of >1.1 mmol/L. By 2003, the recommended Total:HDL level for high risk patients was <4.0 and the level for moderate risk patients was <5.0. Finally, the recommended blood pressure was steadily adjusted downward from <140/90 in 1992 to <130/80 in 2003. These targets have been well supported by research such as that by the DCCT (1993, 1995, 2000, 2002, 2003) and the UKPDS (1998b). The medical community is in agreement that tight control of blood glucose, cholesterol and blood pressure levels results in clinically important reductions in the complications associated with diabetes. The 2003 guidelines recommend even more aggressive targets for some patients. Specifically, a haemoglobin A1c level of ≤ 6% is recommended for patients in whom it can be safely achieved. Unfortunately, the risk of severe hypoglycaemia was three times higher among participants receiving intensive therapy. Therefore, “normoglycemia may not be an appropriate goal in individuals with either type 1 diabetes or type 2 diabetes who are at risk for severe hypoglycaemia” (DCCT, 1993; p. S18). 2.3.2.5 Recommended Diagnostic Procedures to Assess the Ongoing Control of Diabetes In order to determine whether the optimal levels of control are being achieved, the guidelines outline a set of recommended diagnostic tests to be performed on a periodic basis. These include an HbA1c assay, which determines how much glucose has bound to the A1c form of haemoglobin in a process called glycosylation. The identified value reflects how much  33  glucose has been in the blood during the past average 120-day lifespan of red cells. The three month time period for this test was chosen to reflect the continuous monitoring of blood glucose levels given the average life-span of a red blood cell (Saudek, et al., 2006). A lipid profile was recommended every six months in 1992, with the frequency reduced to every 1-3 years in subsequent CPG. The 2003 Canadian Guidelines note that “a fasting lipid profile (TC, HDL-C, TG and calculated LDL-C) should be conducted at the time of diagnosis of diabetes and then every 1 to 3 years as clinically indicated….More frequent testing should be done if treatment for dyslipidemia in initiated” (p.S60). The 2003 Canadian Guidelines suggested that most patients with diabetes fell into the high risk of a vascular event category and that treatment should be initiated if LDL-C <2.5 and TC:HDL-C <4.0. For patients with a moderate risk (e.g. “younger age and shorter duration of diabetes and no other complications of diabetes and no other risk factors for vascular disease”), treatment should be initiated if LDL-C <3.5 and TC:HDL-C<5.0. The recommended treatment approach, in addition to lifestyle modifications, is to prescribe a statin with the possibility of prescribing a fibrate for higher risk patients. The use of statins as the drug of choice to lower cholesterol levels was based on the results of the Heart Protection Study (Heart Protection Study Collaborative Group, 2002; Yusuf, 2002) which found that adding a statin (simvistatin) to existing treatments produced a substantial reduction in vascular events in high-risk patients, regardless of their initial cholesterol concentrations. Finally, the Canadian Guidelines suggest that “when monotherapy fails to achieve lipid targets, the addition of a second drug from another class should be considered” (p. S60). The measurement of blood pressure was supposed to take place at least every six months in the 1992 guidelines, but has subsequently been recommended to take place at every physician  34  visit. The patient care plan included in the 2003 Canadian Guidelines indicates that blood pressure should be measured at least four times per year (p.S123). An eye exam is recommended at least once every two years, more frequently if the patient has retinopathy. The 2003 Canadian Guidelines note that “in people with type 1 diabetes, screening and evaluation for retinopathy by an experienced professional should be performed annually 5 years after the onset of diabetes in individuals ≥15 years of age” (p.S77). Furthermore, “in people with type 2 diabetes, screening and evaluation for retinopathy by an experienced professional should be performed at the time of diagnosis. The interval for followup assessments should be tailored to the severity of the retinopathy. In those with no or minimal retinopathy, the recommended interval is 1 to 2 years” (p. S77). Measurement of urinary microalbumin should take place at diagnosis and at least annually thereafter (see Table 2-3 below). Microalbuminuria is associated with a level of 30 – 299 µg/mg creatinine while macroalbuminuria is associated with levels of 300 or higher (American Diabetes Association, 2005). Because of the variability in urinary albumin excretion, two or three specimens collected over a three to six month period should show abnormal results before a patient is considered to have crossed one of these diagnostic thresholds. The 2003 Canadian Guidelines note that in people with type 2 diabetes who have been diagnosed with nephropathy, the preferred treatment is an ACE inhibitor or ARB if the creatinine clearance is >60 mL/min or an ARB if the creatinine clearance is ≤60 ml/min. These guidelines suggest that serum creatinine and potassium levels be checked “within 2 weeks of initiation of therapy and periodically thereafter” (p. S69). The American Diabetes Association (2005) notes that “the role of annual microalbumin assessment is less clear after diagnosis of microalbuminuria and institution of ACE inhibitor or ARB therapy and blood pressure control.  35  Most experts, however, recommend continued surveillance to assess both response to therapy and progression of disease.” Table 2-3: Recommended Diagnostic Procedures to Assess the Ongoing Control of Diabetes Based on the Canadian and British Columbia Diabetes Clinical Practice Guidelines 1992  1998  2003  BC 2004  at least every 6 months  every 3-4 months if on insulin, otherwise every 6 months  approximately every 3 months  every 3 months  essential for patients taking insulin  essential for all type 1 diabetes, pregnant women with diabetes and insulin-treated type 2 diabetes  type 1 diabetes at least 3 times per day, most with type 2 - at least once per day  at least every 6 months  at every visit  at every visit  at every visit  Lipid Profile  at least every 6 months, if dyslipidemia then every 4 months  every 1-3 years  every 1-3 years  every 1-3 years  Foot Examination  at least annually  at least annually, more often for those at high risk  at least annually, more often for those at high risk  at least annually  annually for all type 2 and type 1 with five + years duration  at least annually  annually if no albuminuria, otherwise at least every six months  at least annually  Blood Glucose Control Over Time HbA1c assay  Patient Blood Glucose Monitoring  Blood Pressure  regular as appropriate  Nephropathy Screen for macroscopic protein & non-renal diseases with dipstick If protein-negative, measure ACR  Measure SCr (currently lab will report eGFR)  Neuropathy Retinopathy Retinal eye exam  annually for all type 2 and type 1 with five + years duration  at least annually, 5+ years diabetic with proteinuria measure urinary protein excretion, without proteinuria measure microalbuminuria.  if no albuminuria, annually for people over 15 with a 5 year historyof type 1 and all type 2  check annually for symptoms  check annually for symptoms  check annually for symptoms  check annually for symptoms  if patient has retinopathy, at least annually  based on the severity of the retinopathy, if no retinopathy then every 2 years  based on the severity of the retinopathy, if no retinopathy then every 2 years  every 1-2 years  Abbreviations ACR = Albumin to creatinine ratio SCr = Serum creatinine eGFR = Estimated Glomerular Filtration Rate  36  In addition to conducting diagnostic procedures, physicians are encouraged to assess and discuss self-management challenges, encourage patient blood glucose monitoring, provide counsel on issues of smoking, weight control and exercise, ensure the patient is vaccinated for influenza and pneumonia, and prescribe medications as required. The 2004 BC CPG includes the following practice points for the care of a patient with diabetes: • • • • •  Minimizing symptomatic hyperglycaemia or hypoglycaemia overrides the stated target levels for A1c. More frequent lipid measurement is required for patients receiving treatment for dyslipidemia. Most adults with diabetes are at high risk of cardiovascular disease. Rigorous control of blood pressure has been shown to reduce the risk of complications and mortality rates. Co-existing depression is common in patients with diabetes. Diagnose and treat as appropriate.  2.3.3 Recommended Clinical Procedures As previously discussed, achieving ideal blood glucose and pressure levels is difficult in an optimal research environment and even more difficult in a usual care setting. One organization that has had some success in the United States is the Veteran’s Health Administration (VHA). Patients from the VHA score significantly higher than a national sample for overall quality of care, chronic disease care and preventative care (Asch et al. 2004; Greenfield and Kaplan, 2004). The VHA has developed more achievable targets that are strongly linked to health outcomes and then rewarded physicians for process measures (e.g. testing for HbA1c, cholesterol and blood pressure levels) with less focus on whether or not these levels are achieved. They established a series of priorities for the medical care of high risk patients. Their thresholds for blood glucose (HbA1c <9.0-9.5%), blood pressure (<135-140/80-85 mmHg) and  37  LDL cholesterol (<130-140mg/dl) were less stringent than the levels recommended in CPGs (Hayward et al., 2004). The VHA program tends to focus on “‘tightly linked’ quality measures in which the clinical interventions or process is strongly and directly linked to patient outcomes, an actionable process is measured, and a high-risk population is targeted” (Hayward et al., 2004). By focussing on ‘tightly linked’ clinical interventions, providers receive credit for appropriate care “regardless of the severity of their patient population” (Hayward et al., 2004). Included in these ‘tightly linked’ clinical interventions for diabetes are the measurements of blood sugar, blood pressure and cholesterol levels. The Organization for Economic Co-operation and Development (OECD) has taken a systems approach in determining which indicators to use in assessing the quality of diabetes care in OECD countries (Greenfield, et al., 2004). Their ultimate recommendation of the indicators to track was determined by: 1) whether the indicator captured an important performance aspect; 2) was scientifically sound and; 3) its tracking was potentially feasible. The importance of an indicator was further broken down into three dimensions: impact on health, policy importance and susceptibility to being influenced by the health care system. The scientific soundness of the indicator was based on both face and content validity. Finally, the feasibility was based on current data availability and potential reporting burden. The review by Greenfeld and colleagues (2004) led to a recommendation of the following nine indicators: •  •  Processes of diabetic care o Annual HbA1c testing o Annual LDL cholesterol testing o Annual screening for nephropathy o Annual eye exam Proximal outcomes  38  •  o HbA1c control o LDL cholesterol control Distal outcomes o Lower extremity amputation rates o Kidney disease in person with diabetes o Cardiovascular mortality in patients with diabetes  With respect to measuring blood pressure, Greenfeld et al. (2004) note that “measuring and reporting blood pressure control in a comparable fashion would be more challenging. The protocols for measuring and reporting of blood pressure would have to be standardized across countries and data collection would require dedicated reporting or Electronic Medical Records, whose implementation lags substantially the implementation of electronic laboratory systems.” In British Columbia, the Chronic Disease Management (CDM) program in the Ministry of Health has recommended tracking the following process measures for diabetes care (BCCDM, 2003): • • • •  Two or more HbA1c tests during each fiscal year At least one eye exam during each fiscal year At least one microalbumin test during each fiscal year At least one lipid test every three years  The recommended clinical process procedures by the BCCDM program are identical to those of the OECD, with the exception of the frequency of testing. Rather than an annual HbA1c testing, the BCCDM recommends a minimum of two annual HbA1c tests. This concurs with recommendations by the American Diabetes Association (2005). The ADA recommends at least two HbA1c tests per year in patients who are meeting glycaemic treatment goals and four times per year in those who are not. On the other hand, the BCCDM recommends at least one lipid test every three years rather than an annual LDL cholesterol test. In this study we will essentially follow the recommendations of the BCCDM in tracking adherence to recommended clinical procedures. In addition to these four procedures, a fifth  39  variable will be assessed as a proxy for blood pressure measurement, namely, whether or not the patient visited a general practitioner (GP) at least four times in a given fiscal year. The assumption behind this proxy measure is that the patient’s blood pressure should be measured at every GP visit. A key assumption in using process measures is that actions necessary to address, for example, poor blood pressure and glucose levels will be more likely when the practitionerpatient partnership adheres to the recommended type and frequency of clinical procedures than if the recommended clinical procedures are not performed. In other words, the assumption is that clinical vigilance and patient activity are synergistic, resulting in more favourable patient outcomes. In a recent editorial, Williams (2005) noted that “we don’t know much about the relationship between the process of care and patient outcomes in the real world…most real world quality improvement efforts measure the process of care because it is easier than measuring outcomes. Therefore, most quality improvement efforts assume that a better process of care will lead to better patient outcomes.” Is this actually the case? Larsen et al. (1990) assessed whether routine measurements of HbA1c resulted in improved metabolic control in people with type 1 diabetes. Patients were randomly assigned to a study group who received HbA1c tests every three months and a control group who did not receive HbA1c tests. After a year, the mean HbA1c level fell significantly in the study group (from 10.1 to 9.5 percent) but not in the control group (from 10.0 to 10.1 percent). In addition, only 30% of individuals in the study group were in the poor control group (HbA1c levels above 10.0 percent) after a year compared to 50% in the control group. Finally, the study group were also hospitalized less often than the control group. Similar changes occurred in the control group when routine measurements of HbA1c were provided for this group the following year. The  40  researchers conclude that “regular measurements of haemoglobin A1c lead to changes in diabetes treatment and improvement of metabolic control, indicated by a lowering of haemoglobin A1c values.” Kahn et al. (1990) found that better processes of care were related to lower 30-day mortality for older patients hospitalized with congestive heart failure, myocardial infarction, pneumonia, and cerebrovascular accident. A number of researchers (Jencks et al., 1988; Park et al., 1990; Thomas et al., 1993) have studied the correlation between medical care and patient outcomes by assessing the validity of the annual hospital mortality statistics released in the United States by the Health Care Financing Administration (1987). The researchers concluded that single year mortality rates were not a good measure of the quality of care provided by the hospital. At best, they found a weak relationship between the two. Park et al. (1990), for example, noted that 56 to 82% of the excess mortality in hospitals with unexpectedly high mortality could result from purely random variation. More recent research on medical errors indicated that a portion of this excess mortality may be due to medical errors rather than simply random variation (Institute of Medicine, 1999; Stelfax et al., 2006). A recent study by Higashi and colleagues (2005), however, found a much stronger relationship between performance as measured by process quality indicators and survival in vulnerable older adults living in the community. The researchers assessed the quality of care received by measuring a set of indicators covering 22 conditions. After adjusting for gender, health status and health service use, a higher quality of medical care was associated with lower mortality after 500 days. A key difference in this study compared to the three studies noted above is the longer follow-up period (three years vs. one year). One of the earlier research groups (Park et al., 1990) mentioned the one-year follow-up period as a significant limitation in its  41  work. This seems to be borne out by the fact that the correlation found by Higashi et al. only emerged after 500 days of follow-up. In commenting on the findings of Higashi and co-authors, Williams (2005) notes that the findings “are important because they provide evidence that quality improvement efforts that focus on the process of care improve patient outcomes.” While such results need to be replicated, they do provide some support for the oft-assumed idea that improving the process of care leads to improved patient outcomes. 2.4 Adherence to Recommended Clinical Procedures 2.4.1 Compliance versus Adherence Haynes et al. (1979) have defined compliance as “the extent to which a person’s behaviour (in terms of medications, diets, or life-style changes) coincides with medical or health advice.” They, as well as others (e.g., Kurtz, 1990; Johnson, 1992; Golin et al., 1996), suggested that the term adherence was interchangeable with compliance. In contrast, Luftey and Wishner (1999) maintained that there are important ideological differences between the two terms. “The term compliance suggests a restricted medical-centred model of behaviour, while the alternative adherence implies that patients have more autonomy in defining and following their medical regimens.” This distinction is particularly important when considering chronic diseases, such as diabetes, that involve complex care requirements. Compliance is a term which indicates that patients are largely responsible for their daily care requirements, while surrendering most of the decision-making to caregivers. Using adherence reflects a shift from an “authoritative practitioner-submissive patient model” (Luftey and Wishner, 1999) to one in which the patient carries an equal role in determining treatment protocol.  42  Anderson and colleagues (Glasgow and Anderson, 1999; Anderson and Funnell, 2000) suggest that the semantic shift recommended by Luftey and Wisher (1999) does not go far enough. They underscore that, in diabetes care, the patient is “fully responsible for the selfmanagement of their illness. This responsibility is non-negotiable and inescapable” (Glasgow and Anderson, 1999, p. 2091). They suggest terms, such as ‘self-care’ and ‘self-management,’ that more appropriately represent “the cluster of daily activities that patients perform to manage their diabetes.” In response to these assertions from Anderson and colleagues, Luftey and Wishner (2000, p.1035) contend: “It is important to remember that even though patients are indeed responsible for their diabetes management, practitioners are also inescapably invested in these processes in ways that will not change with changes in terminology. They are responsible for prescribing regimens that patients can safely execute, and, moreover, for overseeing this self-management in a way that maximizes glucose control while protecting themselves and patients from liability and the negative consequences of uncontrolled diabetes.” In the following section we examine adherence to recommended clinical procedures, keeping in mind the role of both patients and physicians in co-managing diabetes. For the purposes of this discussion, adherence is defined as the degree to which actual practice coincides with recommendations as identified in diabetes CPGs. 2.4.2 Assessing Adherence to Recommended Clinical Procedures People with diabetes who have a regular health care provider visit their provider more often and are more likely to receive recommended clinical procedures than those without a regular provider (O’Conner et al., 1998). A minimum of four annual visits to a primary care physician are required to receive the recommended clinical procedures noted earlier.  43  Harris (1990) estimated that people with diabetes in the United States visited their primary care physician an average of 2.7 times per year in 1985 for the ongoing care of their diabetes. In a further study based on a survey sample in 1989 in the United States, Harris (1996) found that approximately 10% of individuals with diabetes did not have a regular physician, 32% made fewer than four visits to their physician per year, 33% made four to six visits per year, and 26% made more than six visits per year. Also in the United States, Peters et al. (1996) found that patients with diabetes averaged 4.5 visits to their primary care physician in 1993, but 21% had one or fewer visits. Further, Hiss (1996) found that patients who are taking insulin tended to visit their primary care physician more often (i.e., 4.6 visits) compared to those who are not taking insulin (i.e., 3.2 visits). In Canada, Watson et al. (2003) established that residents of Winnipeg, Manitoba who had diabetes tended to see a family practitioner an average of just over seven times in a year. This rate did not change appreciably between 1992 and 2001. In Ontario, researchers found that the average number of visits to a family physician was somewhat higher, at just under ten per year (Chan and Harju, 2003). In the Vancouver area of British Columbia, younger adults with diabetes visited their general practitioner an average of 5.6 times per year in 2000/01. Adults over 65 visited their general practitioner 8.7 times per year (Broemeling et al., 2004). The literature indicated that a substantial proportion of people with diagnosed diabetes were not receiving the recommended clinical procedures. Rubin and co-workers (1998) noted that only 34% of people with diabetes in their study population had at least one HbA1c test per year, only 23% had an annual eye exam, and only 39% received a yearly cholesterol screening. McGlynn et al. (2003) found that only 24% of adults with diabetes had received three or more glycosylated haemoglobin tests over a two-year period, and a bare 14% had an annual eye exam.  44  They noted that 58% had their total serum cholesterol and HDL cholesterol tests documented. Overall, the study by McGlynn and colleagues revealed that people with diabetes received just 45% of the processes recommended for basic care of their chronic condition. The United States Agency for Healthcare Research and Quality published a report (i.e., National Healthcare Quality Report, 2003) which noted that 55% of adults with diabetes reported receiving an influenza vaccination in the previous year, 66% reported having a foot exam, 67% reported having a retinal eye exam, 90% reported having an HbA1c measurement at least once in the past year, and 94% reported receiving a lipid profile in the past two years. Only 21% of patients, however, reported having all five major tests done in the past two years. In Winnipeg, Manitoba, Katz et al. (2004) found that 54% of patients with diabetes had a cholesterol screening test and 37% had an eye exam during a one year period. In Ontario, Harris et al. (2003) found that 84% of their random sample of patients with diabetes had at least one HbA1c test ordered in the previous year, 28% were tested for microalbuminuria, 15% were examined for diabetes-related foot problems, 88% had their blood pressure measured, and 48% had their lipid profiles documented in their chart. The British Columbia Chronic Disease Management (BCCDM, 2003) department of the BC Ministry of Health Services accessed administrative data to determine whether patients throughout the province were receiving the series of services recommended in current clinical practice guidelines. In the fiscal year 2002/03, only 39% of people with diagnosed diabetes in the province had two or more HbA1c tests. A higher number (i.e., 43%) had an eye exam, but only 34% had a microalbumin test. The most encouraging finding was that 78% had at least one lipid test in the three years from 2000/01 to 2002/03 (BCCDM, 2003).  45  Comparing the results of these different studies is problematic in that the definitions of the study populations differed. Furthermore, the results were derived variously from self-report, chart abstracts or administrative data. Some studies used a one year window while others used a two year window. Finally, the diagnostic tests were not identical across settings. Nevertheless, it was apparent that a large portion of patients with diagnosed diabetes were not receiving the recommended clinical procedures. Some studies have uncovered patient characteristics that are associated with variability in adherence. For instance, place of residence appeared to make a difference. Jencks and coauthors (2000), for example, found that adherence varied significantly by US state for annual HbA1c testing, biennial eye exams and biennial lipid profiles. The median rate for HbA1c testing was 71% with a range from 52 to 85%. Similarly, it was 69% for eye exams (with a range from 56 to 80%) and 57% lipid profiles (with a range from 39 to 73%). In a follow-up study, Arday et al. (2002) adjusted for patient characteristics such as age, gender, race and socio-economic status. These adjustments reduced the variance between US states in HbA1c tests, eye examinations and lipid profiles by 30, 23 and 27%. The authors noted that “while the variation explained by person-level characteristics (one-fourth to one-third of the variance among states) is considerable, a majority of the variation among states remains unexplained.” This study also highlighted the potential relationship between patient characteristics and varying levels of adherence. Lower rates of adherence for the three diagnostic tests was observed in patients under the age of 65, blacks, those living in a community with a lower socio-economic status and those with five or fewer outpatient physician visits during the two year study period.  46  Woodward et al. (2006) assessed the frequency and outcomes of HbA1c tests over a period of one year among eastern Ontario patients with diabetes. Only 58% of the study population had at least one HbA1c test during the study year. They found that older individuals were more likely to be tested than younger individuals. In addition, males were more likely to be tested than females and those individuals who visited a physician (i.e., GP or specialist) more often were also more likely to be tested. The literature revealed that while overall results suggest suboptimal adherence, there has been some improvement in adherence over time, at least in certain jurisdictions. Using a retrospective chart review, Stolar et al. (1995) gauged the impact of the ADA’s 1988 clinical practice guidelines for patients with type 2 diabetes. They assessed adherence to recommended clinical procedures for the three years before and after the 1988 guidelines were published. They found significant improvements in a number of areas. Selected results are summarized on table 2-4. Table 2-4: Percent of Patients Receiving Recommended Clinical Procedures Initial Visit 46.3% 59.1% 60.5% 17.0% 15.2% 43.9% 67.7% 7.0%  Eye exam HbA1c Cholesterol High-density lipoproteins Low-density lipoproteins Triglycerides Urinalysis 24-hr creatinine clearance Based on Stolar et al. (1995) * significant change from previous visit  Midpoint Visit 57.3%* 73.9%* 67.0%* 30.3%* 27.7%* 50.2%* 64.0% 9.4%  Final Visit 67.8%* 82.2%* 76.7%* 50.6%* 48.9%* 67.0%* 67.1% 14.9%*  Jencks et al. (2003) assessed changes in the proportion of Medicare beneficiaries with diabetes who received at least one HbA1c test per year, at least one eye exam every two years  47  and at least one lipid profile every two years. Between 1998/99 and 2000/01, these proportions increased from 70% to 78%, 69% to 70% and 58% to 74% respectively. In England, Campbell and colleagues (2005) used medical records data to assess changes in the receipt of recommended care between 1998 and 2003. They found a significant improvement in the quality of care provided to individuals with type 2 diabetes during that time. More specifically, the proportion of individuals who had at least one HbA1c tests during the previous 15 months increased from 87.6% to 92.7%. Similar increases were observed for the measurement of serum creatinine levels (from 79.8% to 89.5%) and serum cholesterol levels (from 74.9% to 97.6%). Increases for eye exams (from 70.8% to 71.8%) and blood pressure (from 92.8% to 94.6%) were more modest. Beerstecher (2005) has cautioned, however, that these results might not reflect an increase in the quality of the care provided but simply an improvement in data recording. The move toward better care has been evident in British Columbia. In this province, the proportion of people with diagnosed diabetes who received two or more HbA1c tests per year increased from 31% in 1999/00 to 39% in 2002/03. Similarly, the proportion with at least one microalbumin test per year increased from 22% to 34%. The proportion with at least one lipid tests in three years increased from 61% to 78%. Unfortunately, the proportion of people with at least one eye exam per year decreased from 47% to 43% (BCCDM, 2004). The problem of sub-optimal adherence to recommended clinical procedures is not isolated to patients with diabetes. A study by McGlynn et al. (2003) found that patients with a wide variety of chronic conditions received just 56% of care as recommended in CPGs. Nor is it isolated to specific countries. A study by Schoen et al. (2005) compared the receipt of recommended services for adult patients with diabetes in six countries (Australia, Canada, New  48  Zealand, United Kingdom, United States and Germany). They discovered that the countries all fell into a range from 38 to 58%, for evidence that patients had received all of the following services: an HbA1c test in the last six months; a foot or eye exam in the past year; and, a cholesterol check in the past year. 2.4.3 Why is Adherence Generally Sub-Optimal? 2.4.3.1 Physician Factors Early research on the reasons for poor adherence to clinical practice guidelines tended to focus on the physician. Cabana et al. (1999) identified three broad potential physician-based barriers to CPG adherence: •  Physician knowledge (lack of awareness or lack of familiarity with CPG).  •  Physician attitudes (lack of agreement, lack of self-efficacy, lack of outcome expectancy, or the basic inertia of previous practice).  •  Physician behaviour (including external barriers).  Weinberger et al. (1984) assessed whether physicians who were more successful at controlling their diabetic patient’s blood glucose levels could be distinguished from less successful physicians: was their success in managing diabetic patients tied to their knowledge? Was it tied to their attitudes? The researchers found that differing levels of knowledge alone did not distinguish between the two groups. It was intentions and beliefs that distinguished the two groups. Other research points to knowledge as a factor. While CPGs are usually widely disseminated, Wolff et al. (1998) found that only 27% of family physicians in the US knew where to find a CPG. Furthermore, approximately one-third of physicians were unfamiliar with the content of specific guidelines (Ward et al., 2002; Wolfe et al., 2004). Targeting individual  49  physicians for education (i.e., academic detailing) has been shown to be successful (Goldberg et al., 1998), but is very labour-intensive and expensive (Greco and Eisenberg, 1993). Many researchers have focused on physician attitudes toward CPGs and the relationship between those attitudes and behaviour (Anderson et al., 1991; Halm et al., 1999). The majority of physicians tend to agree that CPGs are good educational tools and convenient sources of advice, and that they are developed to improve the quality of health care (Lomas et al., 1989; Tunis et al., 1994; Weingarten et al., 1995; Siriwardena, 1995; Gupta et al., 1997; Hayward et al., 1997; James et al., 1997; Wolff et al., 1998; Farquhar et al., 2002). This generally positive attitude toward CPGs, however, was not found in all studies. Interestingly, approximately one quarter of physicians view CPGs negatively, describing them as oversimplified or “cookbook” medicine, too rigid to apply to individual patients and a challenge to physician autonomy (Tunis et al., 1994; Weingarten et al., 1995; Siriwardena, 1995; Hayward et al., 1997; James et al., 1997; Wolff et al., 1998; Costantini et al., 1999; Chasuk et al., 2001; Farquhar et al., 2002; Boyd et al., 2005). When researchers examined the implementation of CPGs, they discovered that few physicians reported making changes to their clinical practice based on published guidelines (Tunis et al., 1994; Hayward et al., 1997; James et al., 1997; Wolff et al., 1998). Indeed, CPGs ranked well below other sources of information, including continuing medical education, discussions with colleagues, and review articles, in influencing physician practice patterns (Tunis et al., 1994; Gupta et al., 1997; Hayward et al., 1997). Weingarten et al. (1995) found no significant association between physicians’ attitudes toward CPGs and the implementation of guidelines.  50  The Canadian study by Hayward et al. (1997) noted that overall, “it seems that the challenge is not so much to overcome negative attitudes about guidelines but more to develop strategies that will influence physicians to read, remember and use them.” Two key conclusions from their study were: 1) for physicians to adopt guidelines, they may require an authoritative endorsement; and 2) CPGs should be presented to doctors in a format that promotes their use, such as short pamphlets, official manuals summarizing a number of guidelines, journal articles summarizing new guidelines, and pocket cards. The concern about endorsement was reflected in the results of other studies, where key factors influencing the uptake of CPGs by physicians included whether the guidelines had been endorsed by appropriate professional organizations and / or physician opinion leaders, as well as whether the guidelines were based on evidence such as systematic reviews (Lomas et al., 1991; Gupta et al., 1997; James et al., 1997; Hayward et al., 1997). Finally, another potential reason for poor adherence to CPGs for diabetes is that officebased management of diabetes has simply increased in complexity during the 1990s (Grant et al., 2004). 2.4.3.2 Patient Factors While the initial research into reasons behind poor adherence to CPGs focused on the physician more recent research has focussed on the role of the patient and the physician-patient relationship. The literature on adherence to diabetes treatments suggests that patients follow regimens more readily given the following conditions: 1) if the treatments involve medications rather than lifestyle changes (Anderson et al., 1993; Glasgow et al., 1987); 2) if the perceived severity of the disease is high (Kurtz, 1990) and there is a recognized direct connection between symptoms and disease (Peyrot et al., 1987); 3) if medications alleviate uncomfortable symptoms  51  and minimize the risk of hypoglycaemia (Kurtz, 1990; Peyrot et al., 1987); 4) if they believe the recommended treatment will enable them to delay or avoid complications (Peyrot et al., 1987; Bobrow et al., 1987); and, 5) if the regimen is simple rather than complex (Ary et al., 1986). Anderson et al. (1993) divided diabetic patients into low and high adherence groups. The largest differences were in the more difficult adherence areas (i.e., in following recommendations for diet and exercise). Larme and Pugh (1998) studied the attitudes of primary care providers towards diabetes using qualitative research methods. Among their results is the conclusion that diabetes is harder for providers to treat than other chronic conditions because its successful management relies to a great extent on lifestyle change which is largely outside of provider control; further, treatment is complex and requires close coordination between patients and physicians. In a US study, James et al. (1998) found that patient-specific factors were associated with a physician’s decision to adhere to guideline recommendations. For instance, physicians were more lax with patients that had difficulties in affording health care, a reduced quality of life because of a co-morbidity, or a desire to stay in their community with family even when that meant limited access to specialty services. The authors assessed whether a “physician’s attention to providing quality interpersonal care may conflict with providing quality technical care”. Their conclusion expands on this phenomenon: “This study suggests that medical decisions in primary care are affected by patient preferences distinct from biomedical aspects of disease. This insight, although not new to family physicians, is extremely important for those who would measure health care quality through measures of physician adherence to disease-specific guidelines”. Wagner (2001) notes that many chronically ill people have socioeconomic factors, disabilities, and co-morbid conditions that make it harder for practitioners and practice systems  52  to help them. In particular, the co-occurrence of mental disorders and other chronic diseases negatively impacts on the interaction between the care provider and the patient (Simon, 2001; Osborn, 2001; Piette et al., 2004; Selby et al., 2004; Frayne et al., 2005). A number of studies have assessed the relationship between the patient’s race or ethnicity and physician or patient adherence to prevention guidelines (e.g. Martin et al., 1995; Harris, 2001; Heisler et al., 2003; Mainous et al., 2004). In a recent review of this literature, Lanting et al. (2005) found that diabetic patients from minority groups had higher mortality rates and were at a higher risk of the complications associated with diabetes. After adjusting for risk factors such as smoking, socioeconomic status, income, years of education, and body mass index, however, ethnic differences tended to disappear. Ethnic differences in process of care (e.g. the receipt of recommended services such as HbA1c tests, blood pressure testing, etc.) were observed only in blacks (compared to whites) in the United States. While blacks in the United States had an increased risk of mortality and diabetes complications compared to whites, the opposite association was found in the United Kingdom. This suggests that differences observed in the United States were probably not due to genetics but most likely due to differences in the health care systems in the two countries (i.e., universal access in the UK). Indeed, differences in the organization of health care services have been posited as a major factor in the level of adherence to CPGs. It is to these organizational factors that we now turn our attention. 2.4.3.3 Organizational Factors Recent research has begun to examine organizational factors beyond the individual provider or patient in assessing the adherence to CPGs (The TRIAD Study Group, 2002). For  53  instance, Curry (2000) has noted that benefit and reimbursement policies have an impact as do investments in clinical information systems. A systematic review of studies comparing community-based and hospital-based care for people with diabetes found that the community setting was as good as or better than the hospital, provided that the community-based system included a computerized central recall to prompt both providers and their patients. Otherwise, unstructured care in the community was associated with poorer follow-up, greater mortality and worse glycaemia control than hospital-based care (Griffen and Kinmouth, 2004). A number of studies (Harrold et al., 1999; Donohoe, 1998) have suggested that specialists are more knowledgeable about the management of conditions associated with their specialty, and that they are more likely to adhere to CPGs. Rothman and Wagner (2003), however, underscore that it is the design of the health care system, rather than the specialty of the physician, that is the primary determinant of chronic care quality. This is an issue that we will return to in the next section. A number of other studies (Renders et al., 2001; Renders et al., 2004; Shiffman et al., 2004) also identify systematic arrangements for patient follow-up as important in improving process outcomes in diabetic care. Literature reviews by Renders et al. (2001, 2004) also revealed that the effective use of nurses in the care process and enhanced patient education were important success factors. Others (Michie and Johnston, 2004) noted the importance of clear, concise behavioural recommendations, and suggested that rewriting guidelines to increase behavioural specificity may be the simplest, most effective method of increasing their implementation.  54  In summary, the literature identified a variety of physician, patient and organizational challenges which interfered with adherence to recommended clinical procedures. The most important recommendations for improvement were organizational, including appropriate financial incentives, a computerized central recall system, the involvement of nurses in the care process, and enhanced patient education. 2.4.4 Factors that Promote Adherence In addition to the challenges that need to be addressed, research has also revealed positive factors that promote adherence. National guidelines for family practice have been developed and disseminated in a rigorous, structured manner since 1987 in the Netherlands (Grol et al., 1995). The result among family practitioners has been an average adherence rate of 67% (Grol, 2001). Key factors predicting adherence include recommendations that define the desired performance very concretely, that are compatible with existing values, and that do not have major consequences for the organization of health care. Lessons learned from what is described, in the opinion of the authors, as the “most comprehensive programs for evidence-based guideline development and implementation in the world” (Grol, 2001, p. II-52) include: • • •  Rigorous development of clinical guidelines at a national level is both feasible and well-accepted by the target group when it is ‘owned and operated’ by the profession itself. A comprehensive strategy to disseminate the guidelines via various channels, both written and personal, appears to be very important. A program to implement a guideline should be well-designed, well-prepared, and preferably pilot-tested before use.  A systematic review by Grimshaw et al. (2001) of interventions designed to change provider behaviour noted that multifaceted interventions targeting different barriers to care were likely to be more effective than single interventions. The most successful programs addressed: 1) clinician behaviour; 2) changes to the organization of practice; 3) information system  55  enhancement; 4) and educational or supportive programs aimed at patients (Renders et al., 2001; Rothman and Wagner, 2003). Renders et al. (2004) noted that key aspects of successful CPG implementation in the care of diabetic patients included: 1) organizational interventions that improved regular prompted recall and review of patients; 2) a stress on patient-oriented interventions; and, 3) the utilization of nurses in patient education and the facilitation of adherence to treatment. Wagner and colleagues in Seattle have developed a chronic care model which includes: 1) key linkages with community resources; 2) active leadership support; 3) more consistent and collaborative self-management support; 4) system redesign to include non-physician personnel in practice teams; 5) clinical information system enhancements to include reminders and feedback on the care provided; and, 6) attention to co-morbid conditions (Wagner et al., 1996; Wagner et al., 2001; Bodenheimer et al., 2002a, 2002b, 2002c; Heisler and Wagner, 2004). An important aspect of the chronic care model is the concept of patient self-management. “Patients with chronic conditions make day-to-day decisions about – self-manage – their illnesses. This reality introduces a new chronic disease paradigm: the patient-professional partnership, involving collaborative care and self-management education” (Bodenheimer et al., 2002c). Bodenheimer et al. (2002c) provide the following comparison of traditional and collaborative care in chronic illness.  56  Table 2-5 Comparison of Traditional and Collaborative Care in Chronic Illness Issue  Traditional Care  Collaborative Care  What is the relationship between patient and health professional?  Professionals are the experts who tell patients what to do. Patients are passive.  Shared expertise with active patients. Professionals are experts about the disease and patients are experts about their lives.  Who is the principal caregiver and problem solver? Who is responsible for outcomes?  The professional.  The patient and professional are the principal caregivers; they share responsibility for solving problems and for outcomes.  What is the goal?  Compliance with instructions. Noncompliance is a personal deficit of the patient.  The patient sets goals and the professional helps the patient make informed choices. Lack of goal achievement is a problem to be solved by modifying strategies.  How is behaviour changed?  External motivation.  Internal motivation. Patients gain understanding and confidence to accomplish new behaviours.  How are problems identified?  By the professional, e.g. changing unhealthy behaviours.  By the patient, e.g., pain or inability to function; and by the professional.  How are problems solved?  Professional solve problems for patients.  Professionals teach problem-solving skills and help patients in solving problems.  Early results, in terms of both adherence and outcomes, from the implementation of this chronic care model in a variety of settings are very positive (Bodenheimer et al., 2002), though more formal evaluations need to carried out, including assessment of longer term patient outcomes (Narayan et al., 2004). A review of 177 Veteran’s Affairs clinics in the United States found that programs “associated with better diabetes control simultaneously have teams that actively involve physicians in quality improvement, use electronic health information systems, have authority to respond to staffing and programmatic issues, and engage patients in care.” (Jackson et al., 2005, p. 225) 57  Financial incentives for improvements in the quality of care have been applied in the United Kingdom (McElduff et al., 2004; Roland, 2004), Australia (Practice Incentives Program, 2001) and British Columbia (Full Service Family Practice Incentive Program, 2003). The most extensive application of this model is found in the UK, where an estimated 30-50% of a general practitioner’s income may be dependent on meeting specific quality targets. This approach is still in its infancy and has not yet been fully evaluated. Early research (Dudley, 2005; Rosenthal, et al., 2005), however, has suggested that “paying clinicians to reach a common, fixed performance target may produce little gain in quality for the money spent and will largely reward those with higher performance at baseline” (p. 1788). In conclusion, while adherence to recommended clinical procedures for diabetes is generally poor, a number of organizational factors which can improve adherence have been identified in the literature. These include: •  active leadership support  •  guidelines that are owned and operated by the profession  •  enhanced patient education and self-management support  •  the inclusion of non-physician personnel in practice teams  •  information systems designed to provide timely reminders and feedback  •  attention to co-morbid conditions  •  comprehensive dissemination strategy  58  2.5 Improving Adherence to Recommended Clinical Procedures In this section we survey a number of specific programs that have been successful in improving adherence. Rubin and co-authors (1998) reviewed the Diabetic Care of America's (DCA) Diabetes NetCareSM program. Within a year of this program’s implementation, 76% of patients received at least one HbA1c test per year (compared with 34% prior to the program). Furthermore the percentage of patients receiving an annual eye exam rose from 23% to 40%, those receiving an annual foot exam rose from 2% to 25%, and those receiving an annual cholesterol screening rose from 39% to 63%. The average HbA1c level dropped from 8.9% to 8.5%. Sperl-Hillen et al. (2000) assessed patients 12 months after the implementation of a comprehensive diabetes management program in which the specific goals were to improve glycemic control and reduce cardiovascular risk in all adult diabetes patients. They found a significant improvement in mean HbA1c levels, including an increased proportion of patients with an HbA1c level below 8%, as well as a significant increase in the fraction of patients with acceptable lipid control. Berg and Wadhwa (2002) assessed patients six months after enrolment to the McKesson Health Solutions Diabetes CareEnhance program. The number of patients having an HbA1c test increased from 56.1% at program intake to 81.3% at 6 months. There was a decrease in symptoms of hyperglycaemia (i.e., during the past two weeks) from 28% at intake to 13.1% at six months. They also found a significant reduction in hospitalizations from a pre-enrolment annualized utilization rate of 1,110 per 1,000 population to 847 per 1,000 population. Bodenheimer et al. (2002a) cite a number of unpublished results from groups that have implemented a variation of the chronic care model developed by Wagner and colleagues. In a  59  program in Ohio, the proportion of patients with HbA1c levels below 7% had increased from 42% to 70% after the implementation of the chronic care model. In Minneapolis, the proportion of diabetic patients with an HbA1c level of less than 8% increased from 60.5% to 68.3%. In Colorado, the average HbA1c level dropped from 10.5% to 8.6% between October of 1998 and March of 2000. The number of patients receiving at least two HbA1c tests within a year increased from 11% to 71% during that time period. A recent study (Dorr et al., 2005) evaluated the impact of implementing the chronic care model at Intermountain Health Care in Salt Lake City. Key elements of the model included care managers who were placed in clinics with the role of facilitating team collaboration and patient education as well as leveraging existing information technology to allow the primary care teams to adopt numerous different care guidelines at once. Using this approach, they found that the odds of being overdue for HbA1c testing decreased by 21%. For those people who were overdue the odds of being tested increased by 49%. Similarly, the odds of having an HbA1c level < 7.0 percent increased by 19%. Beginning in 1995, the Department of Veterans Affairs (VA) in the US embarked on a nation-wide effort to reengineer its services with a view towards improving both efficiency and effectiveness of delivery. Kerr et al. (2004) compared the outpatient care received by patients with diabetes in the VA system to that found in commercial managed-care systems. The results showed that patients in the VA system were more likely to receive HbA1c testing, counselling about aspirin use, and eye and foot examinations. They also had better lipid control. Sperl-Hillen and O’Conner (2005) reported on 10 year trends in glycemic and lipid control in adults with type 2 diabetes enrolled in the HealthPartners Medical Group (HPMG) in Minnesota. The HPMG identified diabetes as a priority area in 1995 and implemented a  60  multifaceted improvement program. Between 1994 and 2003, median A1c levels in the patient population decreased from 8.3 to 6.9. During the same time period, mean LDL (mg/dL) decreased from 132 to 97. Among Sperl-Hillen and O’Conner’s conclusions were the following: •  Primary care clinics can successfully improve diabetes care in the absence of carve-out disease management. Primary care physician continuity of care is significantly related to better diabetes care.  •  The final common pathway to A1c and LDL improvement is intensification of pharmacotherapy.  •  Certain groups of patients have had less improvement in A1c and LDL than other groups. Those with the most difficulty included younger adults and those with a current or former diagnosis of depression  •  Financial accountability and performance incentives for diabetes performance may facilitate improvement.  In a systematic review of diabetes disease management programs, Knight et al. (2005) note that disease management programs on average have a modest but clinically and statistically significant effect on glycaemic control (0.5 percentage point reduction). This compares to a two percentage point reduction in patients with type 1 diabetes in the DCCT and a 0.9 percentage point reduction in patients with type 2 diabetes in the UKPDS. As noted earlier, both of these studies involved very intensive interventions that would be hard to maintain in real world situations. These examples indicate that a concerted effort to improve adherence can indeed improve the care that diabetic patients receive, often within a very short time period after program implementation.  61  2.6 Potential for Savings Associated with Improved Adherence As noted earlier, there is a substantial amount of evidence indicating that the appropriate control of blood glucose and blood pressure results in the avoidance or minimization of the serious complications associated with diabetes. To a lesser extent, there is also evidence that an improvement of process measures (i.e., getting the recommended tests) is associated with improved outcomes. As the examples in the previous section demonstrated, implementation of a comprehensive management program can result in significant and sometimes rapid improvement in adherence to recommended clinical procedures. What is less clear is whether these improvements resulted in lower health care costs either immediately or over the long term. Initial work assessing the outcomes and costs associated with the implementation of comprehensive planned management programs for people with diabetes tended to be based on extrapolation from cost-effectiveness models. With the publication of longitudinal results from the DCCT and the UKPDS, actual results concerning long-term complication rates became available for use in cost-effectiveness studies. The early models assumed, and the results of the DCCT and UKPDS confirmed, that cost savings would result from a reduction in downstream complications, but that these savings would take from 5 to 10 years to become apparent. Funding agencies, however, are not always willing to wait for long-term results. To address the short-term economic imperative, a number of more recent studies have begun to assess the shortterm financial impact of improving blood pressure and glycaemic control (Killilea, 2002; White, 2002; Clouse et al., 2002; Zhang et al., 2004). Siegel et al. (1992) modelled the impact of screening for microalbuminuria or proteinuria followed by early treatment with angiotensin-converting enzyme inhibitors. They suggested that  62  screening for microalbuminuria would cost from $7,000 to $16,500 per year of life saved, while screening to detect proteinuria would actually be cost saving. Javitt et al. (1991) demonstrated that the timely detection and treatment of retinopathy in patients with type 1 diabetes should result in considerable savings to the US federal government while at the same time increasing person-years of sight. Further work by Javitt et al. (1994) showed similar results for patients with type 2 diabetes. Based on computer modelling, they estimated that screening and treatment for eye disease in patients with type 2 diabetes would generate annual savings of $247.9 million to the federal budget. In addition, 53,986 person-years of sight would be saved annually. This estimate was based on a participation rate of 60%. With a 100% participation rate, these estimates increased to an annual savings of $472.1 million and 94,304 person-years of sight. In a follow-up study, Javitt and Aiello (1996) approached the timely detection and treatment of retinopathy from a societal perspective and estimated that the cost per quality adjusted life year (QALY) saved was $3,190. This was broken down into $1,996 for those with type 1 diabetes, $2,933 for those with type 2 diabetes who used insulin for glycaemic control, and $3,530 for those with type 2 diabetes who did not use insulin. Gilmer et al. (1997) modelled the relationship between baseline HbA1c levels in type 2 diabetic patients and health care costs over the following three years. For every 1% increase in HbA1c, their model found a statistically significant increase in health care costs over three years, ranging from $400 to $4,000 per patient. It is likely therefore that decreases in HbA1c levels would result in cost savings. Eastman et al. (1997) developed a model comparing standard treatment (i.e., $890 per patient per year) with comprehensive treatment (i.e., $2,873 per person per year). Their model  63  predicted that comprehensive treatment maintaining HbA1c levels at 7.2% would reduce the cumulative incidence of blindness, end-stage renal disease and lower-extremity amputation by 72%, 87% and 67%, respectively, at a cost of $16,000 per quality adjusted life year saved. With the completion of the DCCT in the United States, actual long-term results were available for the first time to assess the cost-effectiveness of intensive versus conventional control of blood glucose levels in patients with type 1 diabetes. The study by the Diabetes Control & Complications Group (1996) estimated that implementing intensive rather than conventional therapy for every person with type 1 diabetes in the United States in 1994 would cost $4.0 billion over the lifetime of that population, or $28,661 per year of life saved. In addition, intensive therapy would produce the following results per patient compared to conventional approaches: •  7.7 years of additional sight,  •  5.8 additional years free from ESRD,  •  5.6 additional years free from lower extremity amputation,  •  15.3 additional years free from any significant microvascular or neurologic complications.  A major criticism of the DCCT study is that participants were highly motivated and thus the results may not be transferable to general practice settings (Rubin et al., 1998). Furthermore, the additional medical resources used in the intensively treated population of $4,000 to $5,800 per participant (DCCT, 1995) would make the protocol prohibitively expensive to replicate in general practice settings. In the UKPDS (1998b), the cost of treatment for patients in the tight control of blood pressure group increased by 21%, primarily due to a doubling of the cost of antihypertensive  64  drugs compared to the less tight control of blood pressure group. This increase, however, was more than offset by the 17% reduction in the cost of complications. The overall undiscounted cost of treatment and complications was 2.3% less for the tight control of blood pressure group. The authors conclude that “tight control of blood pressure in hypertensive patients with type 2 diabetes substantially reduced the cost of complications, increased the interval without complications and survival, and had a cost effectiveness ratio that compares favourably with many accepted healthcare programmes.” Similar conclusions were drawn by Gray et al. (2000) for intensive blood glucose control in patients with type 2 diabetes. Intensive glucose control increased costs by £695 per patient but reduced the cost of complications by £957 compared with conventional management. Most of the savings were due to a reduction in hospital-based costs in the intensive blood glucose control group. Rubin et al. (1998) studied the short term experience with of a group of 7,000 people with diabetes enrolled in the Diabetic Care of America's (DCTA) Diabetes NetCareSM program. They found that the initiation of a comprehensive diabetes management program resulted in overall savings to the health care system of 12.3% within the first year. These savings were primarily related to a reduction in both acute care admissions (by 18%) and in the length of stay once admitted. This study also noted a 1.8% decrease in overall pharmacy costs (i.e., after adjusting for general drug price increases), but an increase in physician costs of 2.1%, associated with the implementation of a comprehensive diabetes management program. The actual costs of implementing the program remained proprietary, but the authors did state that there would be a breakeven point at 1,265 diabetic patients.  65  Testa and Simonson (1998) studied the very short-term effect (i.e., within 15 weeks) of active hypoglycaemic therapy versus placebo in a randomized controlled trial. They examined symptoms, quality of life, work productivity and health care use. Patients receiving active hypoglycaemic therapy reported better health and work productivity and less use of ambulatory care. Wagner et al. (2001) compared the health care costs of two diabetic cohorts over a five year period (1992 – 1997). They examined a group in whom glycaemic control improved and a group in whom glycaemic control did not improve. Differences in the number of primary care visits reached statistical significance within two years. Differences in total costs reached statistical significance within three years. Mean total health care costs were between $685 and $950 less per year for patients in whom glycaemic control improved. Thus this study provided a robust example of short-term cost savings. Menzin et al. (2001) also investigated the short-term economic benefits of improved glycaemic control. Using a retrospective chart review design, they assigned patients with diabetes to three groups based on HbA1c levels: good control (<8%), fair control (8% - 10%) and poor control (>10%). They assessed differences in inpatient admissions for short-term complications, including hyperglycaemia, hypoglycaemia, selected infections, and electrolyte imbalance. Over a three year period, the adjusted rate of inpatient treatment ranged from 13 admissions per 100 patients for the good control group, to 16 per 100 for the fair control group, to 31 per 100 for the poor control group. Average costs per diabetic patient in the three groups were $970, $1,380 and $3,040, respectively. Among patients with chronic diabetic complications, the difference in results between the good and poor glycaemic control scenarios were even more marked.  66  Berg and Wadhwa (2002) assessed patients six months after enrolment to the McKesson Health Solutions Diabetes CareEnhance program. The number of patients having an HbA1c test increased by 44.9%. Six months after enrolment, patients showed a 53.2% decrease in symptoms of hyperglycaemia. The researchers also found a significant reduction in inpatient admissions, resulting in a calculated return on investment of more than four to one (i.e., 4.34:1). Villagra and Ahmed (2004) studied the short-term outcomes associated with the implementation of diabetes disease management programs (DDMPs) in ten US States, involving over 43,000 individuals with diabetes. Within a year after the implementation of the DDMPs, significant improvements were observed in the provision of dilated retinal exams, microalbumin testing, lipid testing and tobacco use. A positive trend was observed for HbA1c testing and prescriptions for angiotensin-converting enzyme inhibitors or angiotensin receptor blockers. Overall costs decreased by 8.1%, driven largely by a 22-30% reduction in hospitalization costs. However, pharmacy costs were higher with the DDMPs in place. Not all studies of disease management programs have found cost savings. Fireman et al. (2004) reviewed the Kaiser Permanente experience in Northern California with disease management programs in diabetes, asthma, heart failure and coronary artery disease between 1996 and 2002. They found evidence of substantial improvements in the quality of care associated with the implementation of disease management programs, but not cost savings. Crosson and Madvig (2004) explained that this study used a very restrictive definition of cost savings and that it was carried out in a mature delivery system known for its efficiency. They refuted the criticism of Fireman and colleagues, asserting that much of the “low-hanging fruit, in terms of cost management” may have been harvested prior to the implementation of their disease management programs.  67  In summary, the majority of studies assessing the planned management of care for people with diabetes show, in the immediate and longer term, an improved adherence to recommended clinical procedures, the avoidance or minimization of the serious complications associated with diabetes and a reduction in costs. 2.7 Summary Diabetes is a common and serious chronic condition. If it is not well-managed, significant multi-system complications often arise resulting in an increase in health care utilization. There is considerable evidence which indicates that people with diagnosed diabetes are not receiving the recommended care. A comprehensive program aimed at improving adherence to recommended care can improve the treatment that these patients receive, often within a very short time period. Furthermore, the implementation of a diabetes disease management program results in better patient outcomes which translated into reductions in health care utilization and concomitant costsavings. The primary objective of this study was to determine whether adults with diagnosed type 2 diabetes who have a higher adherence to recommended clinical procedures utilized a higher volume of physician services and a reduced volume of acute care services over a given five year period. A further objective was to assess whether adherence for this study population changed over time. For this analysis, routinely collected administrative data for the five year period from April 1, 1996 to March 31, 2001 was accessed. The population selected for analysis consisted of patients with diagnosed type 2 diabetes living within the geographic boundary of the Fraser Health Authority, a large region in south western British Columbia with a population of 1.4 million.  68  More specifically, the questions addressed will be: 1. Has the level of adherence to recommended clinical procedures by adults with diagnosed type 2 diabetes living within the geographic boundaries of the Fraser Health Authority in British Columbia changed during the five years from April 1, 1996, to March 31, 2001? 2. Which patient characteristics (e.g. age, gender, socio-economic status, general level of morbidity, disease-specific severity level and geographic area of residence) are associated with improved long-term adherence to recommended clinical procedures? 3. What is the relationship between improved long-term adherence and the average annual utilization of physician, acute care and total costs? A number of studies have assessed various aspects of the relationship between adherence and patient characteristics. These studies tend to assess adherence based on a single snapshot of the receipt of recommended services over a one or two year period. In the current study we will develop an adherence measure based on a clinical protocol that spans a five year time period. To our knowledge, this is the first study to assess the relationship between patient characteristics and long-term adherence to recommended clinical procedures. The individual patient’s adherence level will be included as an additional patient characteristic. We will then assess the relationship between patient characteristics and the utilization of physician, acute care and total health care costs, paying special attention to the role of the patient’s adherence level in this relationship. A number of studies have assessed the impact of the overall improvement of adherence on health care costs. To our knowledge, this is the first study to investigate the relationship between long-term adherence to recommended  69  clinical procedures, patient characteristics and of health care costs in patients with diagnosed type 2 diabetes.  70  CHAPTER III: METHODS 3.1 Conceptual Framework In patients with diagnosed type 2 diabetes, appropriate control of blood glucose, blood pressure and cholesterol levels results in a reduced risk of both acute and chronic complications. The process of controlling blood glucose, blood pressure and cholesterol levels requires a number of key steps: 1. Knowledge of ideal target levels 2. Monitoring of actual results 3. Implementation and maintenance of lifestyle changes and, if appropriate, use of medications 4. Ongoing monitoring and changes to arrive as close as possible to the ideal target levels over time. This process requires an increased utilization of physician resources, diagnostic tests, behaviour change counselling and medications compared to a situation in which monitoring and subsequent changes do not take place. On the other hand, improved control of blood glucose, blood pressure and cholesterol levels leads to a long-term reduction in both acute and chronic complications and the acute care services required to address these complications. Ultimately, the process of improved control of blood glucose, blood pressure and cholesterol levels leads to a reduction in premature mortality and an improved health-related quality of life. This information is summarized in the conceptual framework shown in figure 3-1.  71  Adult with Diagnosed Type 2 Diabetes  Adherence to Recommended Medical Care  Health Care System Impacts  Figure 3-1: Conceptual Framework  b) Macro vascular  Monitor Risk Level: 1) Nephropathy assessment (at least 1 microalbumin test per year) 2) At least 1 dilated eye exam per year 3) Annual foot examination 4) Monitor for neuropathy annually  i) ischemic heart disease ii) stroke iii) peripheral vascular disease  i) nephropathy ii) neuropathy iii) retinopathy  Intermediate Outcomes Reduced Risk of: 1) Acute Complications 2) Chronic Complications a) Micro vascular  For Patients / Physicians with High Adherence to Recommended Medical Care Proximal Outcomes  1) Physical activity 2) Heart healthy diet 3) Weight loss  Target Values Blood Glucose A1c of 7 or below Blood Pressure 130/80 4) Smoking cessation 5) Appropriate use of medications  Target Values Achieved By:  Cholesterol LDL below 2.5 Total:HDL below 4  Process Monitor Target Values 1) At least 2 HbA1c tests per year 2) At least 4 blood pressure measurements per year 3) At least 1 lipid test every 3 years  1) Increased physician visits 2) Increased use of diagnostic services  1) Increased physician visits 2) Increased use of medications 3) Increased use of behaviour change counseling  1) Increased physician visits 2) Increased use of diagnostic services 3) Decreased use of emergency services 4) Decreased use of acute care services Surgical interventions (e.g. CABG, PCTA), renal transplantation, cataract removal, etc. 5) Decreased use of home help / living assistance / long term care 6) Decreased use of rehabilitation services  1) HbA1c and lipid tests from 1) Physician visits from 1) Microalbumin and eye exam from MSP MSP billing information MSP billing information billing information 2) Physician visits from MSP 2) Physician visits from MSP billing Data Sources billing information information Used In 3) Blood pressure measurement 3) Visits to an emergency department from Study Assumed to occur at each MSP billing information GP visit 4) Use of acute care services from hospital separation files Notes: Recommended clinical procedures and health care system impacts tracked in this study are highlighted in bold/italics print.  Distal Outcomes  1) Reduced early mortality 2) Improvement in health-related  quality of life  72  Figure 3-1 provides a conceptual framework for patients who receive recommended clinical procedures. As noted previously, a significant number of patients with type 2 diabetes do not receive recommended clinical procedures. Without appropriate monitoring, neither patients nor physicians will be aware of any variance between ideal and actual blood glucose, blood pressure and cholesterol levels. In turn, it is also less likely that they will address the lifestyle changes or the medication adjustments that are needed in order to reduce the variance between ideal and actual levels. When recommended clinical procedures are not provided and appropriate lifestyle/medication adjustments are not made, patients are at a higher risk of both acute and chronic long-term complications necessitating an increased use of emergency, acute care, long term care and rehabilitation services. In addition, the patient is at an increased risk of early mortality and deterioration in health-related quality of life. Specifically, this study used routinely-collected administrative data to determine whether adults with diagnosed diabetes with higher long-term adherence to recommended clinical procedures utilized a higher volume of physician services and a reduced volume of acute care services during the five years from April 1, 1996, to March 31, 2001, after adjusting for patient age, gender, socio-economic status, location of residence and levels of morbidity. The objective of this study is to determine if the delivery of appropriate diagnostic tests for diabetes is associated with elevated costs for physician visits and reduced costs for acute care. The recommended clinical procedures for which routinely collected administrative data are available in British Columbia are highlighted in Figure 3-1 in bold/italicized print. This includes the tests for monitoring blood glucose, blood pressure, and cholesterol levels, and the tests used to assess for nephropathy and retinopathy. Furthermore, measures of health care utilization which are routinely available in administrative data in British Columbia are also  73  highlighted in bold/italicized print. The data sources used in this study are also summarized in figure 3-1. A key assumption of this study is that actions necessary to bring blood pressure and blood glucose closer to ideal levels will be more likely if the patient undergoes the recommended type and frequency of tests than if the recommended tests are not performed. That is, on average, physicians providing the recommended tests are more likely to address unfavourable results than physicians not providing the recommended tests. In the same vein, it is assumed that patients who obtain unfavourable test results are more open to making changes than patients who are uninformed. 3.2 Study Design / Overview This study is an observational, cross-sectional study with the individual as the unit of analysis. An overview of the study is shown in figure 3-2 below, with the various components discussed in the following sections.  Figure 3-2: Study Overview Remove incident cases using 1996/97and 1997/98 as a washout period  1996/97  Fiscal Year 1997/98 1998/99  1999/00  2000/01  Y/N Y/N Y/N Y/N  Y/N Y/N Y/N Y/N  Identify Cohort of Prevalent Cases People with diagnosed type 2 diabetes living in the Fraser Health Authority  Assess Adherence Two or more HbA1c tests during a year At least one eye exam per year At least one microalbumin test per year At least four BP measurements per year  Y/N Y/N Y/N Y/N  Y/N Y/N Y/N Y/N  Y/N Y/N Y/N Y/N  Y/N  At least one lipid test every three years  Y/N Y/N  Health Care Utilization Acute care discharges / days (adjusted by Resource Intensity Weight (RIW)) Surgical day care cases (adjusted by RIW) Medical Service Plan (MSP) general practitioner services MSP specialist physician services  74  Per capita annual utilization adjusted for age, sex, SES and level of morbidity differences in the relevant patient populations  3.3 Specific Aims and Hypothesis Hypothesis: That the long-term receipt of appropriate clinical procedures for patients with diagnosed type 2 diabetes is associated with an increase in physician costs but a decrease in acute care costs. Aim 1: To determine whether adherence to recommended clinical procedures for adults with diagnosed type 2 diabetes living within the geographic boundaries of the Fraser Health Authority has changed during the five years from April 1, 1996 to March 31, 2001. Question 1.1 - Has the receipt of the following recommended services by adults with diagnosed type 2 diabetes living within the geographic boundaries of the Fraser Health Authority changed during the five years from April 1, 1996 to March 31, 2001? •  Two or more HbA1c tests during each fiscal year  •  At least one eye exam during each fiscal year  •  At least one microalbumin test during each fiscal year  •  At least one lipid test every three years  •  At least four blood pressure measurements each year  Question 1.2 – What proportion of adults with diagnosed type 2 diabetes living within the geographic boundaries of the Fraser Health Authority have received each of the five recommended services in each fiscal year? Question 1.3 – What proportion of adults with diagnosed type 2 diabetes living within the geographic boundaries of the Fraser Health Authority received all five of the recommended services in each year during the entire five year time period (i.e., showed the highest adherence)?  75  Question 1.4 - What proportion of adults with diagnosed type 2 diabetes living within the geographic boundaries of the Fraser Health Authority received none of the five recommended services during the entire five year time period (i.e., showed the lowest adherence)? The previous two questions represent the extremes of potential adherence. Aim 2: To determine which patient characteristics (e.g. age, gender, socio-economic status, general level of morbidity, disease-specific severity level and geographic area of residence) are associated with improved long-term adherence to recommended clinical procedures. Aim 3: To determine whether the utilization of physician and acute care services has changed during the five year period. Aim 4: To assess the relationship between patient characteristics and the utilization of physician, acute care and total health care costs. Aim 5: To determine whether adults with better adherence utilize varying levels of different types of health care, after adjusting for the patient’s age, gender, socio-economic status, location of residence and levels of morbidity. Question 5.1 – Do adults with better adherence use more physician services than those with poor adherence? Question 5.2 – Do adults with better adherence use fewer acute care services than those with poor adherence? Question 5.3 – Do adults with better adherence use fewer total services (physician and acute care) than those with poor adherence?  76  3.4 Data Sources 3.4.1 British Columbia Linked Health Database Data for this study are drawn from the British Columbia Linked Health Database (BCLHD) housed by the Centre for Health Services and Policy Research (CHSPR) at the University of British Columbia. This is a linked longitudinal database including all residents of British Columbia who are registered with the Medical Services Plan. The BCLHD is an extensive data resource for applied heath services and population health research. Data from the BCLHD have been used for over one hundred health care and health services research projects since 1996. The BCLHD includes data files containing: •  individual-level information on health care service use;  •  claims made to the B.C. Workers' Compensation Board;  •  basic information about the location and background of select health care service providers;  •  surveys that provide a deeper level of understanding for (small) groups of B.C. residents; and  •  descriptive information about neighbourhoods and communities derived from census data.  Where possible, the information from each data file in the BCLHD is linkable at the level of the individual. Data are linked using deterministic and probabilistic linkage methodology and linkage rates in excess of 95% have been achieved (Chamberlayne et al., 1998). These data can be used to explore population-based trends, allowing researchers to trace the experiences of groups of individuals across various arenas of the health system.  77  All data are anonymized to ensure protection of privacy and confidentiality. Specific data files used in this study are from the Medical Services Plan Registration and Premium Billing files, Medical Service Plan Payment Information Master files, the Hospital Separations files and the Vital Statistics - deaths files. 3.4.2 British Columbia Medical Services Plan Files There are a number of potential sources of data for information on physician services and expenditures on physicians in British Columbia. These include the Medical Service Plan (MSP) database files, the MSP Out of Province (OOP) database, the Alternative Payment Program (APP) and the Primary Health Care Organization (PHCO) Encounter Data Set. The MSP database files include fee-for-service (FFS) payments to B.C. physicians for services to B.C. and non-B.C. residents. It also contains payments made on behalf of B.C. residents who obtained services in Quebec, the U.S. and other countries (as these jurisdictions are not covered through reciprocal billing agreements). The MSP Out of Province (OOP) database includes FFS payments made to out-ofprovince physicians who provided services to B.C. residents. When B.C. residents receive services in other provinces (except for Quebec), those other provinces bill back to B.C. manually – i.e., not through the standard claim submission process. This process is used as a result of the reciprocal agreements set up between and among the provinces. Thus, these payments are not included in the main MSP database, but are in this OOP database. The Alternative Payment Program (APP) database includes salary, sessional payment, and service agreement data Note that salary and sessional data are physician-specific, but service agreement data are not.  78  The PHCO Encounter Data Set includes data from primary care demonstration sites established in the 1999/00 fiscal year. Approximately 80% of B.C. physicians are self-employed professionals working on a fee-for-service (FFS) basis. In the FFS system, the MSP pays physicians an established fee for each service provided to each patient. Fees compensate physicians for their professional services and pay for overhead including staff salaries, medical equipment, supplies, rent, continuing education, insurance, business licenses, and other costs associated with running a business. Physicians billing FFS must submit claims to MSP in a computer-readable format within 90 days of the service date. Claims can be submitted via Teleplan or by contracting with a service bureau equipped to make the submissions. Teleplan is a web-based telecommunications system used by practitioners to securely submit claims, notes and eligibility requests to MSP, and receive payment statements, rejected claims and patient eligibility data from MSP through an encrypted Internet connection. A second group of physicians are paid through salaried or sessional arrangements. Salaried physicians are typically on staff at hospitals, private corporations, government agencies or universities. For example, medical directors of health authorities and physicians employed by the B.C. Cancer Agency, Riverview Hospital or Centre for Disease Control, and regional and provincial medical health officers, are salaried. Sessional payments are used primarily for physicians working in mental health and palliative care. The sessional payment is based on time, rather than service provided, with one session equalling 3.5 hours. This type of payment allows physicians to bill the MSP for the actual time spent with patients instead of the type of service or treatment provided.  79  The Alternative Payments Program (APP) administers the salaried or sessional payments to physicians. The Alternate Payments Program is in turn administered from within the MSP. The APP salaried payments are made through the employing health care agency. These payments are based on submission of a “Claim for Reimbursement of Shareable Expenditure.” Conditions for salaried physician payment are negotiated on a province-wide basis and set out in an agreement between the government of BC and the BCMA with respect to salaried physicians in government service. The APP sessional payment provides funding to an agency, which in turn enters into a “Personal Sessional Contract” with a physician for the delivery of services. The agency is required to pay the physician directly for his/her time (based on a completed Certificate of Services form summarizing the amount of time required to deliver care services), and completion of a “Claim for Reimbursement of Shareable Expenditure” made to APP. Payment to the physician is made based on a proration of the standard 3.5-hour session rounded down to the nearest quarter hour. The sessional payment scale is negotiated between the BCMA and the Medical Services Commission (MSC). Service agreements are used for physicians working solely in a publicly funded health care facility. For example, physicians working in emergency rooms or pathologists are typically paid through service agreements. Under a service agreement, regional health authorities and other government-funded agencies contract with physicians to deliver agreed-upon services / deliverables. In addition to the major sources of physician payment (i.e., fee-for-service paid through MSP and salaried or sessional paid through APP), two minor sources include PHCO Encounter Data and BCMA-related funds. PHCO encounter funding is an encounter-based system that is  80  related to primary care practices, but does not contain physician-specific payments (i.e., payments go to practices that decide on dispersement). The fact that a service occurred, the date of the service and the diagnosis codes relating to that service, however, are available at the patient level in the PHCO Encounter Data. Finally, some funds are available to physicians through the BCMA (e.g., call reimbursement, Canadian Medical Physician Association fee repayment). In the provision of primary health care in British Columbia, the vast majority of services are provided through MSP FFS payments. In 2000/01 spending on primary care physicians in the province totalled an estimated $664.0 million, of which $649.1 million (97.6%) was through the FFS system. In an analysis for residents living within the geographic boundaries of the FHA, this proportion of funding through FFS was 99.2% (Watson et al., 2004). For specialist physician care a higher proportion of total expenditures are paid through the APP system than for primary care providers, particularly for certain specialities. Approximately 24% of psychiatry, 12% of oncology and 10% of paediatric services, for example, are funded through APP (Auditor General of British Columbia, 2003). In 2000/01, an estimated1 23% of total expenditures for specialist physicians was paid through the APP with the remainder (77%) paid through MSP. The primary source for physician utilization and costs for this study is from the MSP files. This source provides comprehensive information on over 99% of all payments to primary care physicians and an estimated 77% of all payments to specialist physicians.  81  The Medical Services Plan Registration and Premium Billing files include information on the patient’s gender, birth date, postal code, enrolment to and cancellation date from MSP. Medical Service Plan Payment Information Master files include additional information on the physician, the type and date of services billed as well as the amount paid for the services. 3.4.3 British Columbia Hospital Separations Files All hospitals in British Columbia submit information on acute care and same day surgery separations to the Canadian Institute for Health Information (CIHI). Upon discharge from hospital, the patient’s medical record is coded and abstracted based on criteria determined by CIHI. The resulting Discharge Abstract Database (DAD) abstract is submitted to CIHI where the data are edited for quality and additional information added (e.g., case-mix grouping, resource intensity weight, etc.). Hospital-specific reports are then produced and returned to the hospital for further review and corrections, prior to being used in the production of CIHI reports and distribution to the provinces (CIHI, 2004). In recognition of the importance of data accuracy and quality, CIHI has instituted a variety of data control measures. These include the use of abstracting software, educational programs for abstractors, a production system editing and correction process as well as special studies assessing the quality of the data (CIHI, 2004). These special studies usually involve accessing the original data sources (i.e., medical records) and re-abstracting information to  1  This estimate was developed in the following manner. PURRFECT Version 9.0 MSP Referral Patterns indicates that $1,766.1 million was spent on physician services through MSP in 2000/01. Of this amount, $649.1 was spent on GP services, leaving $1,117 million for ‘other’ services. From this amount we removed $130.2 million for non-medical practitioners and $359.4 for diagnostic specialists, leaving $627.4 million for specialist services. Total expenditures through APP totaled approximately $200 million in 2000/01 (Auditor General of BC, 2003).Of these funds, $11.9 million is allocated to primary care services (Watson et al., 2004). We have assumed that the remaining APP expenditures of $188.1 million are for specialist physician services. Based on these assumptions, an estimated 23% ($188.1 of $815.5 million) of specialist physician expenditures are paid through the APP.  82  compare with the original abstracted information (Long et al., 2001; Mitchell and Brown, 2002; Brown and Richards, 2002). Despite this caution, data quality issues do surface. A review of the 1999/00 DAD indicated a discrepancy rate between the original and the re-abstracted information of 13.4% with respect to the most responsible diagnosis field, 10.0% for the principal procedure field, 9.0% for the postal code field, 6.5% for the entry code field and so on (Richards et al., 2001). The most common reasons for these discrepancies included the original coder missing information that was included in the medical chart and differences in the interpretation of the documentation (Richards et al., 2001). 3.4.4 British Columbia Vital Statistics Database The British Columbia Vital Statistics Agency is responsible for the ascertainment, registration, and certification of vital events through the administration of the Vital Statistics Act, Marriage Act and Name Act (British Columbia Vital Statistics Agency, 2003). The Vital Statistics database was used to ascertain whether a patient in the study had died and the date of that death. If a person dies in British Columbia, the death must be registered with the Vital Statistics Agency. The process is as follows. A medical practitioner or coroner will complete and sign a medical certificate within 48 hours after the death. The medical certificate will be forwarded to a funeral director. On the request of the funeral director, the particulars of the death are provided by an appropriate person, e.g., the nearest living relative present at the death or latest illness. The funeral director then registers the death and provides a death certificate and a burial permit for the deceased.  83  The British Columbia Vital Statistics Agency database does not include deaths for BC residents that occurred outside of the province (British Columbia Vital Statistics Agency, 2003).2 The BCLHD includes options to access Vital Statistics data on deaths, births and clinical information. For this study, we accessed data on deaths only. The date of death was identified by the year and month in which the individual died. 3.5 Study Population The study patient cohort consists of all adults with diagnosed type 2 diabetes who lived continuously within the geographic boundaries of the Fraser Health Authority during the five year time period from April 1, 1996 to March 31, 2001 and who were not identified as incident cases. Additional exclusions from this cohort are noted below. 3.5.1 Ascertaining Diabetic Cases While diabetes can be clinically defined based on plasma glucose values, determining the total number of people with diabetes in a population varies somewhat based on how cases are ascertained and whether or not estimates are included for undiagnosed cases. Accurate, population based estimates are important for both policy-makers and planners who influence the provision of resources and care for patients with diabetes. The prevalence of diabetes has been ascertained using various methodologies such as: •  National population-based surveys (Tan and MacLean, 1995; Harris and Robbins, 1994)  2  The fact of a death in another Canadian province of person borne in BC is reported electronically to BC. No other particulars are currently sent, though a system is under development. This process is intended to deter identify theft. Lorne Verhulst, personal communication, February, 2006.  84  •  Registries (LaPorte et al., 1985), cohort studies in highly selected populations (Leibson et al., 1997)  •  Medical record reviews (Martin et al., 2000)  •  Standardized telephone surveys (Mokdad et al., 2000, 2001)  •  Routinely collected administrative data (Blanchard et al., 1996; Hux et al., 2002).  Population based surveys in Canada such as the National Population Health Survey and the more recent Canadian Community Health Survey have facilitated population based estimates of prevalence using patient self-reporting. There is evidence, however, that self-reporting tends to underestimate the true prevalence of diagnosed diabetes (Manuel and Schultz, 2004; Hux et al., 2002; Mackenbach et al., 1996). Another disadvantage is that the cost of primary data collection at a national level can be high. The use of routinely collected administrative data to ascertain the prevalence of diabetes is an approach that is considerably less resource intensive than other approaches and thus appropriate for ongoing surveillance. Blanchard and co-workers (1996) in Manitoba used comprehensive databases of physician service claims and hospital discharge abstracts to identify individuals diagnosed with diabetes in that province. Their algorithm specified that any patient with two physician service claims or one hospitalization with a diagnostic code of diabetes within a two year period would be identified as having diabetes. More recent work (Watson et al., 2003) in Manitoba used a similar algorithm but used two physician service claims or one hospitalization bearing a diagnosis of diabetes within a three year period. Hux and colleagues (2002) in Ontario tested two algorithms for assessing the prevalence of diabetes using administrative data bases. They used the algorithm developed by Blanchard and co-workers (1996) as well as one which required only one physician service claim. They also  85  excluded gestational diabetes by identifying any record bearing a diabetes diagnostic code followed by an obstetrical event within 5 months after the date of the diabetes diagnostic code. The results of these algorithms were validated by comparing them to the National Population Health Survey in which a random sample of the population was asked whether they had diabetes that had been diagnosed by a physician. In addition, primary chart abstraction was conducted on a random sample of 3,317 patients from 520 physicians. They concluded that using only a single physician service claim resulted in an unacceptable level of false positive identifications, possibly due to cases where diabetes was suspected but subsequent laboratory tests did not confirm this suspicion. In British Columbia, the Chronic Disease Management group of the B.C. Ministry of Health Services has used the following case definition in determining the number of individuals in the province with diagnosed diabetes: •  At least one hospital discharge coded as ICD-9 250 since April 1, 1992, based on three levels of care (acute, rehabilitation, and day care), and all 16 hospital diagnoses; or  •  At least two Medical Service Plan services (on different dates) coded as ICD-9 250 within a moving 365 day period; or  •  At least one pharmacare service: glucose testing strips (98995003), oral hypoglycaemics (682020012-682092002 and 682092021-682092043), or insulin (682008).  •  Remove gestational diabetes cases: delete hospital, MSP, and pharmacare/pharmanet records that occur five months prior and three months after delivery (admission) dates.  86  For this study, the initial study population was identified based on Medical Service Plan claimants who had resided within the geographic boundaries of the Fraser Health Authority at any time between April 1, 1996 and March 31, 2001. These patients met the following criteria: •  had at least one hospital discharge coded as ICD 9-250 in any of the 16 diagnostic fields between April 1, 1998 and March 31, 2000 , or,  •  had at least one MSP service coded as ICD 9-250 between April 1, 1998 and March 31, 2000.  This initial draw resulted in a population of 64,020. 3.5.2 MSP Exclusions MSP claims associated with non-physician specialties were excluded (see Table 3.1). Table 3-1: Excluded Non-Physician Specialties 30 – Chiropractor 32 – Physiotherapist 35 – Orthotics 38 – Podiatry 40 – Dental Surgery 42 – Orthodontics 80 – Midwife 82 – Nutritionist / Dietician 84 – Educator 86 – Medical Office Assistant 88 – Respiratory Therapist  31 – Naturopath 34 – Osteopath 37 – Oral Surgery 39 – Optometry 41 – Oral Medicine 43 – Massage Therapy 81 – Registered Nurse 83 – Counsellor 85 – Licensed Practical Nurse 87 – Nurse Practitioner 89 – Home Support  In addition, unknown fee item codes, mileage fee item codes, anaesthesia and dentistry fee item codes, records with no payment information and records for out-of province patients were excluded. These MSP exclusions resulted in the reduction of 168 individuals from the initial population of 64,020.  87  3.5.3 Diagnostic Rule-Outs A sub-group of individuals identified included those who only had one MSP service coded as ICD 9-250 between April 1, 1998 and March 31, 2000. These were individuals who fell outside of the ‘standard’ algorithm for identifying a person with diabetes using administrative databases. As noted earlier, Hux and colleagues (2002) tested algorithms which included either one or two physician service claims identified as ICD 9-250 over a two year period. They found that using only a single physician service claim resulted in an unacceptable level of false positive identifications, possibly due to cases where diabetes was suspected but subsequent laboratory tests did not confirm this suspicion. This process resulted in the exclusion of 18,054 individuals from the initial population of 64,020. 3.5.4 Children Children, particularly adolescents, with diabetes tend to have different care patterns than adults. Adolescents, for example, experience poor glycaemic control and acute complications more frequently than adults (DCCT, 1994; Svoren et al., 2003). Potential reasons for this include changing physiology (pubertal growth and development) as well as behavioural and adherence issues (Amiel et al., 1986; Wysocki et al., 1996; Rydall et al., 1997; Wolfsdorf, 1999). Individuals who were under the age of 20 years on April 1, 1996 were identified as children and excluded from the initial population. This process resulted in the exclusion of 800 individuals from the initial population of 64,020. 3.5.5 Gestational Diabetes To exclude individuals with gestational diabetes, all females who had an obstetrical event within five months after a hospital or MSP ICD 9-250 code were identified. An obstetrical event  88  was identified using ICD 9 codes 630.0 to 676.9 based on a search of both hospitalization and MSP data. Only individuals for whom this was the only time that the ICD 9-250 code appeared were excluded. This process resulted in the exclusion of 1,106 individuals from the initial population of 64,020. 3.5.6 Incident Cases To exclude incident cases (i.e., newly diagnosed cases), any individual without at least one hospital discharge coded as ICD 9-250 in any of the 16 diagnostic fields during the two year period between April 1, 1996 and March 31, 1998, or, at least one MSP service coded as ICD 9250 during the two year period between April 1, 1996 and March 31, 1998 was identified and removed from the initial population. This process resulted in the exclusion of 12,569 individuals from the initial population of 64,020. 3.5.7 Death Individuals who died were identified to take into account an increase in health care utilization during the time period prior to death. Between April 1, 1996 and March 31, 2001, a total of 3,268 individuals died. 3.5.8 Temporary Residents Individuals who did not live within the geographic boundaries of the Fraser Health Authority for the entire five years from April 1, 1996 to March 31, 2001 were also identified. This was done because one of the control variables in the study was location of residence within the Fraser Health Authority. Of the 28,055 individuals remaining in the study population, 3,584 residents spent at least part of the five year time period living outside of the geographic boundary of the FHA, so they were removed from the analysis.  89  3.5.9 Temporary MSP Registration A further sub-group of individuals identified was those who were not registered with MSP during the entire five years from April 1, 1996 to March 31, 2001. One of the key outcome variables in the study was utilization of health care services, particularly those identified through MSP and hospitalization data, during the entire five year time period. Consequently, a total of 1,782 individuals with temporary MSP registration were removed. 3.5.10 Exclusion of Disease and Age-Specific Sub-Groups There were a number of patient sub-groups included in the remaining population of 22,689 for whom adherence to recommended clinical procedures might have been unusual for clinical reasons. Arday and co-authors (2002), for example, found that the End Stage Renal Disease (ESRD) subpopulation in their study had much lower rates of adherence to recommended clinical procedures than the general population with diagnosed diabetes. The authors noted that clinicians were dealing with “an elderly or disabled population with competing co-morbidities that may influence diabetes care decisions.” For individuals with high impact cancers and Acquired Immune Deficiency Syndrome (AIDS), the potentially life-saving medical care associated with these diseases would likely usurp the imperative, for example, to have blood glucose levels monitored on a regular basis. We identified and excluded patients with type I diabetes, high impact cancers, AIDS, ESRD and the very old. Type 1 and type 2 diabetes are distinct clinical entities, as noted earlier. Unfortunately, there is no reliable way to identify these two populations from the administrative data without the availability of two-digit suffix coding (i.e. ICD9-250.01, ‘type 1 diabetes mellitus’). In the absence of this detailed coding information, researchers made the distinction using age 30 as a cut off point because type I diabetes is usually diagnosed early in life whereas type 2 diabetes  90  typically develops later in life. A study by Laakso and Pyorala (1985) indicated that the cumulative proportion of prevalent cases of type 1 diabetes in Finland was 84% at age 30 and 95% at age 50. More recent research (Fagot-Campagna and Narayan, 2001; Duncan, 2006) pointing to the increasing prevalence of early onset type 2 diabetes makes the 30 year cut-off point less reliable. Nevertheless, we have used the age 30 cut-off point as an exclusion criterion. Patients with AIDS were identified based on the presence of the ICD9 code 042 in either the MSP or hospitalization data between April 1, 1998 and March 31, 2000. Patients with high impact cancers were identified based on the presence of the expanded diagnostic cluster (EDC) MAL03 code between April 1, 1998 and March 31, 2000. Patients with end stage renal disease (ESRD) were identified based on the presence of the EDC REN01 code between April 1, 1998 and March 31, 2000 (Weiner, et al, 2005). The very old were identified as individuals age 80 or older on April 1, 1998. This age cut-off for the very old in diabetes research follows the criterion established by the Manitoba Center for Health Policy Research (Fransoo et al., 2005). This process resulted in the exclusion of the following number of patients: •  Age 80 or older – 1,542  •  Younger than age 30 – 290  •  Individuals diagnosed with high impact cancers – 330  •  Individuals diagnosed with ESRD – 282  •  Individuals diagnosed with AIDS – 3  The resulting study population was thus further reduced from 22,689 to 20,242. 3.5.11 Outliers Fourteen individuals were removed from the analysis based on their high resource use over the entire five year period. This included one individual with an average of 194.6 acute care  91  days each year over the five year period (the next highest value was an average of 88.8 acute care days each year, see Figure 3-3), ten individuals whose average annual utilization of GP visits ranged from 75.8 to 119.0 for each year of the five year study period (see Figure 3-4) and three individuals whose average annual utilization of specialist visits ranged from 47.2 to 52.2 (see Figure 3-5). On visual inspection of the data, these fourteen subjects were considered to be outliers and removed from further analysis. Figure 3-3: Identification of Outliers Acute Care Utilization  400  Number of Individuals  350  300  250  200  150  100  Outlier 50  0 0  20  40  60  80  100  120  Average Annual Days in Acute Care  92  140  160  180  200  Figure 3-4: Identification of Outliers General Practitioner Utilization  400  Number of Individuals  350  300  250  200  150  100  Outliers 50  0 0  20  40  60  80  100  120  Average Annual GP Visits  Figure 3-5: Identification of Outliers Specialist Physician Utilization 1400  1200  Number of Individuals  1000  800  600  400  Outliers  200  0 0  10  20  30  40  Average Annual Specialist Physician Visits  93  50  60  3.5.12 Summary The final study population consists of 20,228 individuals. A summary of the study selection process and results is shown on Figure 3-6.  94  Figure 3-6: Selection of Study Population Exclusions Include all MSP claimants who have resided within the geographic boundary of the Fraser Health Authority (FHA) between April 1, 1996 to March 31, 2001 (and have at least one hospital discharge coded as ICD 9-250 in any of 16 diagnostic fields, or, at least one MSP service coded as ICD 9-250 between April 1, 1998 and March 31, 2000).  Target Population Number % of Total Remaining Number 64,020 100%  Exclude individuals with an ICD-9 250 code only in the following records: nonphysician specialty codes, non-physician fee item codes, unknown fee item codes, mileage fee item codes, anesthesia and dentistry fee item codes, records with no payment information and records for out-of-province patients.  MSP Exclusions  168  0.3%  63,852  Exclude individuals with only one MSP service coded as ICD 9-250 between April 1, 1998 and March 31, 2000.  Diagnostic Rule-Outs  18,054  28.2%  45,798  800  1.2%  44,998  1,106  1.7%  43,892  Exclude individuals under the age of 20 as of April 1, 1996.  Exclude females with an obstetrical event within 5 months after a hospital or MSP ICD 9-250 code, if this is the only time that the ICD 9-250 code appears.  Children  Gestational Diabetes  Exclude individuals without at least one hospital discharge coded as ICD 9-250 in any of 16 diagnostic fields, or, at least one MSP service coded as ICD 9-250 between April 1, 1996 and March 31, 1998.  Incident Cases  12,569  19.6%  31,323  Exclude individuals who died during the time period from April 1, 1996 to March 31, 2001.  Death  3,268  5.1%  28,055  Exclude individuals who did not live in the geographic boundary of the FHA during the entire time period from April 1, 1996 to March 31, 2001.  Temporary Residents  3,584  5.6%  24,471  Exclude individuals who were not registered with MSP during the entire time period from April 1, 1996 to March 31, 2001.  Temporary Registration  1,782  2.8%  22,689  Exclude individuals 80 years of age and older as of April 1, 1998.  Elderly  1,542  2.4%  21,147  Exclude individuals under the age of 30 as of April 1, 1998.  Type I Diabetes  290  0.5%  20,857  Exclude individuals with high impact cancers (n=330), ESRD (n=282) and AIDS (n=3).  High Impact Diseases  615  1.0%  20,242  Outliers  14  0.0%  20,228  20,228  31.6%  Exclude statistical outliers based on high resource use during entire five year period.  Study Population  95  3.6 Variables and Measures 3.6.1 Adherence Variables The key independent variable is the level of adherence to a set of recommended tests and procedures. The set of variables that will be measured, and their operational definitions, are as follows: •  HbA1c – Whether or not the person received two or more HbA1c tests during the fiscal year as identified by the MSP fee item 91745 (haemoglobin A1C). There will be five observations, one for each fiscal year from April 1, 1996 to March 31, 2001.  •  Microalbumin – Whether or not the person received one urinary microalbumin test during the fiscal year as identified by MSP fee items 92396 (microalbumin, semiquantitative) or 91985 (microalbumin). There will be five observations, one for each fiscal year from April 1, 1996 to March 31, 2001.  •  Lipid – Whether or not the person had one lipid test over a three year period as measured by MSP fee items 91375 (cholesterol, total), 91780 (HDL cholesterol) or 92350 (triglycerides, serum/plasma). There will be three observations, one for the three year period from April 1, 1996 to March 31, 1999, one for the three year period from April 1, 1997 to March 31, 2000 and one for the three year period from April 1, 1998 to March 31, 2001.  •  Eye exam – Whether or not the person had one eye exam during the fiscal year as measured by MSP fee items 2010 (consultation – ophthalmology), 2015 (eye examination), 2039 (fundus photography), 2040 (retinoscopy under general anesthetic), 2898 (re-examination or minor exam) or 2899 (full optometric  96  diagnostic). There will be five observations, one for each fiscal year from April 1, 1996 to March 31, 2001 •  Blood pressure measurements – Information on whether or not an individual with diagnosed diabetes had a blood pressure measurement is not directly available in the administrative data. The Canadian and British Columbia CPGs have indicated that blood pressure should be measured at every physician visit and that this should occur at least four times per year. In this study, we have used the presence of at least four general practitioner (GP) visits in a fiscal year as a proxy for direct information on the appropriate number of annual blood pressure measurements. There will be five observations, one for each fiscal year from April 1, 1996 to March 31, 2001.  Information on the receipt of these tests/procedures over the five year period from April 1, 1996 to March 31, 2001 was combined to develop a measure of adherence. Each individual in the study population of 20,228 had 23 possible opportunities for these tests. Specifically, if all the recommended tests were received at appropriate intervals, the patient could earn 23 points during the five year period. Adherence points were assigned in the following manner: •  1 point for two or more HbA1c tests per year, to a total of 5 points  •  1 point for at least one urinary test each year, to a total of 5 points  •  1 point for an eye exam each year, to a total of 5 points  •  1 point for at least four blood pressure measurements per year, to a total of 5 points  •  1 point for a lipid test during 1996-1999; 1 point for a lipid test during 1997-2000; and 1 point for a lipid test during 1998-2001, to a total of 3 points  97  Thus, an individual with perfect adherence on all five measures would be assigned a score of 23. On the other hand, an individual who received none of these tests or procedures over the five year period was assigned a value of 0. All others received a score between 0 and 23, depending on their receipt of the recommend tests or procedures over the five year period. Weighting all tests equally follows the convention of the majority of research in this area, including the landmark research by McGlynn and colleagues (McGlynn et al., 2003; Asch et al., 2004; Asch et al., 2006). These researchers, for example, used 439 indicators to assess the quality of care provided for 30 different conditions. Each of the indicators received an equal weight, thus allowing the researchers to identify the proportion of recommended procedures received overall as well as for each specific condition. The adherence score assigned to each individual was utilized to create three summary adherence variables; first, an adherence variable with continuous values from 0 to 23, second, an adherence variable with the categorical values of low, medium and high adherence and third, a binary adherence variable with the values of low or high adherence. To determine the best clustering of these values into categorical (low, medium and high adherence) and binary variables (low or high adherence), we used the Jenks optimization algorithm (Environmental Systems Research Institute, Inc., ArcView GIS 3.3, 1992-2002). This method is based on an algorithm developed by Fisher (1958) and belongs to a class of clustering procedures designated as methods of partition by exact optimization (Hartigan, 1975). The Jenks method is used to create a grouping or partition of N objects into K non-intersecting subsets – P(N,K) – in such a way that an error function – e[P(N,K)] – is minimized. The method guarantees a partition with the smallest possible within-group variance for a given K value. The Jenks optimization algorithm indicated a clustering of scores from 0 to 9 (low adherence), 10 to  98  14 (medium adherence) and 15 to 23 (high adherence). The clustering for the binary variable was from 0 to12 (low adherence) and 13 to 23 (high adherence). The adherence variables, and their data source, are identified on table 3-2. Table 3-2 Adherence Variables Variable  Description  Values  Properties  Data Source  Status  MSP Payment Information Master Files MSP Payment Information Master Files MSP Payment Information Master Files MSP Payment Information Master Files  Modify  MSP Payment Information Master Files  Modify  Individual Adherence Variables HbA1c Test  MSP fee items 91745  Microalbumin Test  Either MSP fee item 92396 or 91985  Lipid Test  Any of MSP fee items 91375, 91780, or 92350 Any of MSP fee items 2010, 2015, 2039, 2040, 2898, or 2899 General Practitioner Visits  Eye Exam  Blood Pressure Measurements  2 or more tests per fiscal year 1 per fiscal year  Binary (yes/no)  1 every three fiscal years  Binary (yes/no)  1 per fiscal year  Binary (yes/no)  4 or more GP visits per fiscal year  Binary (yes/no)  Binary (yes /no)  Modify  Modify  Modify  Summary Adherence Variables Adherence Score I  Adherence Score II  Adherence Score III  Receipt of the 5 process variables over the five year period Receipt of the 5 process variables over the five year period Receipt of the 5 process variables over the five year period  0 to 23  Continuous  Created  Derived  Low (0-9), Medium (1014), High (1523) Low (0-12), High (13-23)  Ordinal  Created  Derived  Binary  Created  Derived  3.6.2 Patient Characteristics There are a number of known patient characteristics which have an independent influence on the utilization of health care services and may influence adherence to recommended clinical 99  procedures. These include age, gender, socio-economic status, location of the patient residence, general level of co-morbidity and diseases-specific severity. The calculation of age was based on the individual’s age on the first day (April 1, 1996) of the five-year study period. An additional year was added to the individual’s age for each subsequent fiscal year. Age was included as both a continuous and a categorical variable. The categorical variable was developed by grouping individuals into the following age categories for each fiscal year: • • • • •  Ages 30 to 39 Ages 40 to 49 Ages 50 to 59 Ages 60 to 69 Ages 70 to 79  Income quintiles by neighbourhood were used as a proxy for socio-economic status. The Centre for Health Services and Policy Research uses a methodology developed by Ng et al. (1997). Ng and co-authors developed an Income Per Person-Equivalent (IPPE) which “takes into consideration the economies of scale possible when two or more people share a household” (p.22). Enumeration area (EA) income information is available from census data including the average household income (total EA income divided by the number of private households in that EA) and average personal income (total EA income divided by the population aged 15 and over in the EA).These calculations, however, do not take into account the number of persons per household. Two people sharing a residence do not require twice the income of a person living alone to maintain the same standard of living. Thus, an EA with relatively low average personal income, but many multi-person households, may have a standard of living similar to an EA with relatively high average personal income but with many one-person households. The calculation of IPPE adjusts average household income for the bias introduced by the unequal distribution of household sizes across EAs. (p. 22)  100  IPPE is calculated as follows: •  IPPE = total household income in an EA / person-equivalents  •  Where person-equivalents = 1.00 (number of one-person households) + 1.36 (number of two-person households) + 1.72 (number of three-person households) + 1.98 (number of four-person households) + 2.30 (number of five- or more person households).  The income quintile categorization provides an ecologic measure of socioeconomic status for individuals residing in Fraser Health Authority neighbourhoods. Location of a person’s residence is based on linking the first three digits of the individual’s postal code to the geographic area designated by local health areas (LHAs). The Fraser Health Area consists of 13 distinct LHAs. Information on the LHA of residence was calculated for each of the five fiscal (April 1 – March 31) years. If an individual moved from one LHA to another they were allocated to the geographic region in which they lived for the majority of the fiscal year. Information on an individual’s residence is included as earlier work by H. Krueger & Associates Inc. (2003) indicated a significant variance in terms of the diagnostic care services received by people with diabetes living within the various FHA LHAs. Likewise, research in the United States has identified significant variation between residents of different states in the receipt of recommended services even after adjusting for patient-level characteristics (Jencks et al., 2000; Arday et al. 2002).  101  Adjusted Clinical Groups (ACGs) were used as a proxy for the level of morbidity. ACGs, were originally developed in the United States, as a measure of the burden of morbidity in populations. ACGs have been validated in the Canadian setting (Reid et al., 1999, 2002) and in particular, in British Columbia (Reid et al., 2001). These Canadian studies have found that ACGs explain about 50 per cent of same year physician costs and about 40 per cent of same year total medical and hospital costs. The ACG case-mix system assigns the over 14,000 International Classification of Diseases (ICD) codes into 32 clinically similar aggregated diagnostic groups (ADGs) based on the following criteria:3 •  Expected duration of illness (e.g., acute, chronic, or recurrent)  •  Disease severity (i.e., expected prognosis with respect to disability or longevity)  •  Diagnostic certainty (e.g., sign and symptoms versus well defined conditions)  •  Etiology (e.g., infections, neoplasms, psychosocial conditions)  •  Expected need for specialist care or hospitalization  This information is then combined with the patient’s age and gender to assign each patient to one of 82 mutually exclusive adjusted clinical groups (ACGs). There are a number of key advantages of using the ACG system as opposed to other methods to quantity the burden of morbidity in populations (e.g., self-reported health status, age and gender, measures of social deprivation, premature mortality rates, etc.). First, the ACG system “does not rely on only the most important or most common diagnosis, but instead identifies common combinations of morbidities (related and unrelated) that build upon each other, both additively and multiplicatively, to determine an individual’s overall need for health 3  The Johns Hopkins ACG® Case-Mix System, Version 5.0: Software Release Notes, Chapter 5, pages 41-43. Johns Hopkins University Bloomberg School of Public Health.  102  services” (Reid et al., 1999). Second, the main data elements required for the system – age, gender and diagnosis – are often routinely collected in administrative data systems and thus are available for total populations. Third, the ACG system assigns individuals to illness categories based on all of the diagnoses they receive over an extended period of time (e.g., one year) from multiple providers (e.g., hospitalizations, physician visits, ambulatory care procedures). And finally, because it “uses only diagnosis – not procedures or hospitalizations – to define illness levels, it does not reward practices that elect to hospitalize patients more readily or perform more procedures” (Verhulst et al., 2001). A recent study by Broemeling and colleagues (Broemeling et al., 2005) used ACGs to identify the existence of co-morbidities in individuals with a confirmed chronic condition and then allocated individuals with the chronic condition to five categories depending on the number of co-morbidities experienced. The authors defined a co-morbidity as “the co-occurrence of additional conditions among individuals with an index condition (Broemeling et al., 2005)”. One of the difficulties encountered was the distinction between complications and co-morbidities. Complications can be defined as “the existence of a second disease when the occurrence of an index disease is required (Gijsen et al., 2001)”. Thus, diabetic retinopathy is a complication associated with diabetes while hypertension and depression are co-morbidities. While this is a fairly clear example of the distinction between complications and co-morbidities, the distinction between the two is often less clear. In the Broemeling et al. study (2005), individuals with an index condition of diabetes were further assigned to five groups based on their co-morbidity level. The presence of comorbidities was identified using ACGs. The five categories are as follows (see Appendix A for a detailed listing of the ACGs in each of the five categories):  103  1. Level 1 – Very low co-morbidity 2. Level 2 – Low co-morbidity (2 or 3 types of conditions) 3. Level 3 – Medium co-morbidity (4 or 5 types of conditions) 4. Level 4 – High co-morbidity (6 to 9 types of conditions) 5. Level 5 – Very high co-morbidity (10+ types of conditions) For people with diagnosed diabetes, the proportion of this adult population in British Columbia in Levels 1 to 5 was 9%, 29%, 30%, 25% and 7% respectively. We have used the same approach as Broemeling et al. (2005) in assigning patients with diabetes to these five levels of morbidity. A measure of disease-severity used in this study was the diabetes-specific disease severity index developed by Reid (1998). This methodology uses diabetes-related complications and pre-existing conditions which exacerbate diabetes management to group patients into the following five groups: 1. No complicating conditions 2. ≥ one minor complicating conditions 3. ≥ one intermediate complicating condition 4. One major complicating condition 5. ≥ two major complicating conditions The group of minor complicating conditions were defined by Reid (1998) as hypertension, lipid disorders and chronic psychiatric disorders. Patients were classified as having a physician-diagnosed complicating condition if they had two or more claims with the relevant diagnosis. Diagnosis of the minor complicating conditions was based on the following ICD9 codes:  104  1. Hypertension • • • • •  401 – Primary hypertension 402.4 – Hypertensive renal or heart disease 405 – Secondary hypertension 437.2 – Hypertensive encephalopathy 796.2 - Elevated blood pressure  2. Disorders of Lipid Metabolism •  272 – Lipid disorders  3. Major Psychiatric Disorders • • • • •  295 – Schizophrenia 296 – Major affective disorder 297.9 – Other psychoses 303.0 – Alcohol abuse 304.0 – Drug dependence  The group of intermediate complicating conditions were defined as diabetic eye disease (i.e., including retinopathy, glaucoma, and cataract), neuropathy, and peripheral vascular disease. Patients were classified as having a physician-diagnosed complicating condition if they had two or more claims with the relevant diagnosis. Diagnosis of the intermediate complicating conditions was based on the following ICD9 codes: 1. Eye Disease • • • • • •  250.5 – Diabetic retinopathy 262 – Retinopathy 365 – Glaucoma 366 – Cataract 379.3 – Lens Aphakia 743.3 – Congenital cataract  2. Neuropathy • • • •  250.6 – Diabetic neuropathy 350.7 – Mononeuropathy or polyneuropathy 377.1 – Autonomic neuropathy 729.2 – Unspecified neuropathy  105  •  723 – Other neuropathy  3. Peripheral Vascular Disease • • •  440 – Atherosclerosis 443 – Unspecified peripheral vascular disease 785.4 – Gangrene  The group of major complicating conditions included two acute conditions (i.e., diabetic ketoacidosis and hyperosmolar non-ketotic coma) and two chronic conditions (i.e., kidney disease and ischemic heart disease). Patients were classified as having a physician-diagnosed complicating condition if they had two or more claims with the relevant diagnosis. Diagnosis of the major complicating conditions was based on the following ICD9 codes: 1. Acute Coma • • •  250.1 – Diabetic ketoacidosis 250.2 – Hyperosmolar non-ketotic coma 250.3 – Other coma  2. Renal Disease • • • • • • • •  250.4 – Diabetic renal disease 581 – Nephrotic syndrome 582 – Glomerulonephritis 583 – Other nephritis 584 – Acute renal failure 585 – Chronic renal failure 586 – Unspecified renal failure V56 – Dialysis care  3. Ischemic Heart Disease • • • • •  410 – Acute myocardial infarction 411 – Subacute myocardial infarction 412 – Old myocardial infarction 413 – Angina pectoris 414 – Coronary atherosclerosis  106  The algorithm developed by Reid (1998) was applied to the current study population to develop a diabetes-specific disease severity index variable. The diabetes-specific disease severity index provided a measure of the level and severity of diabetes-specific complications while the general measure of morbidity using ACGs provided a gauge of the level of co-morbidities experienced by the patient, regardless of whether these comorbidities were directly associated with the patient’s diabetes. These patient characteristics, and their data sources, are identified on table 3-3. Table 3-3 Patient Characteristics Variable  Description  Values  Properties  Data Source  Status  Age in Years  Age as at April 1 of each year  Age 30-79, unknown  Continuous, Categorical  MSP Registration and Premium Billing Files  Existing  Male, female, unknown  Binary (Male, Female)  Existing  Assign to SES quintile each year based on residence, unknown Level 1 (very low morbidity), 2 (low morbidity), 3 (medium morbidity), 4 (high morbidity), 5 (very high morbidity) and 6 (pregnancy-related) Level 1 (no complicating conditions), 2 (≥ one minor complicating conditions), 3 (≥ one intermediate complicating condition), 4 (one major complicating condition), 5 (≥ two major complicating conditions) Assign to one of 13 FHA LHAs each year  Ordinal  MSP Registration and Premium Billing Files CHSPR derived  Ordinal  MSP Payment Information Master Files and Hospital Separations Files  Derived  Ordinal  MSP Payment Information Master Files and Hospital Separations Files  Derived  Categorical  MSP Registration and Premium Billing Files  Existing  Gender  Socioeconomic status Level of Morbidity  Assign individuals to SES quintile  Diseasespecific Severity Index  Use diseasespecific complications to assign individual patients to one of five levels  LHA of patient residence  Local health area of residence in the FHA  Use co-morbidities to assign individual patients to one of 5 levels  107  Existing  3.6.3 Resource Use Variables Previous research indicates that the utilization of health care services, particularly acute care inpatient services, can change within the first few years after the implementation of a comprehensive diabetes management program. We will examine the use of acute care and Medical Service Plan based health care utilization. The hypothesis is that acute care services, namely, acute care discharges / days (adjusted by Resource Intensity Weight {RIW}) and surgical day care cases (adjusted by RIW), will be lower in those patients with good diabetes management, as defined by better adherence to the recommended tests. On the other hand, general practitioner and specialist physician services (i.e., which are MSP based services) will be higher in those patients with more appropriate diabetes management. 3.6.3.1 Acute Care In the hospital separations files used for this project, all separations (i.e. discharges and deaths) are identified as either acute, extended, rehabilitative care or surgical day care. In addition, some patients may be discharged from a discharge planning unit which means they have been receiving long-term care while in an acute care bed (sometimes identified as alternate level of care). For acute care separations, we used separations identified as acute care only. That is, all records coded as ‘A’ (for acute) in the level of care field in the hospital separations file were included. Separations and the attendant patient days were allocated to each fiscal year based on the date of the patient’s release from hospital. That is, if a patient was admitted on March 20, 1997 and discharged on April 10, 1997, then their discharge (and all patient days associated with the discharge) would be allocated to the 1997/98 fiscal year rather than the 1996/97 fiscal year.  108  Because of this assumption, it was possible for an individual patient who remained in hospital for more than a year to have 365+ patient days allocated to the year of their discharge. Information on acute care inpatient days was used to create both a continuous and a binary variable. The binary variable was created by assigning an individual’s annual acute care inpatient day utilization to either a low or high utilization category based on an 80/20 rule. That is, the 80% of patients with the lowest utilization of acute care days in the year were assigned to the low utilization category while the 20% of patients with the highest utilization of acute care days in the year were assigned to the high utilization category. This process was used for each of the five fiscal years. The mean annual utilization was calculated using the individual’s utilization history during the entire five years. An additional outcome variable was the calculation of acute care costs. In calculating these costs we combined information on both acute care inpatient services and surgical day care services received by individuals in the study. For surgical day care procedures, all records coded as ‘S’ (for surgical day care) in the level of care field in the Hospital Separations file were included. Surgical day care “is a surgical service provided to patients who do not require inpatient services, are admitted and discharged on the same calendar day and are usually discharged between one and six hours following the procedure” (PURRFECT 10.1)4 In addition to information on the number of discharges and days for acute care inpatients, information on the resource intensity weighting assigned to each acute care inpatient discharge by the Canadian Institute for Health Information (CIHI) was gleaned from the hospital separations files. Similar information was available for each surgical day care procedure.  4  PURRFECT (Population Utilization Rates and Referrals For Easy Comparative Tables) is an electronic database updated annually by the BC MoH and distributed to interested parties in British Columbia.  109  Resource intensity weights or RIW values are calculated by CIHI in the following manner.5 As noted earlier, each hospital in British Columbia prepares a discharge abstract database (DAD) abstract for every discharge and submits this to CIHI. Based on information in the DAD abstract, the discharge is assigned to a case mix group (if they received acute inpatient care) or to a day procedure group (if they received a surgical day care procedure) based on the patient’s most responsible diagnosis. Three additional elements are assigned to each acute care inpatient discharge (Hicks and Zhang, 2003). First, information on co-morbid conditions present either at the time of admission or realized during the inpatient stay are used to assign each discharge to a complexity level. Cases are assigned to one of four levels. Level 1 denotes the absence of co-morbid conditions, while Level 4 denotes the presence of co-morbid conditions that may be potentially life threatening. Second, the expected length of stay is calculated based on the length of stay of similar discharges across Canada. Finally, cases are assigned to a typical and atypical category. Atypical cases include all deaths while in hospital, individuals who sign themselves out against a physician’s advice, those transferred from one hospital to another and long-stay outliers. All other cases are considered to be typical. Information on the assigned case-mix group, complexity level, expected length of stay, typical/atypical status and the patient’s age (i.e., three age categories [0–17, 18–69, 70+]) were used in assigning an RIW to each discharge case. “RIW are used to standardize the expression of hospital case volumes, recognizing that not all patients require the same health care resources. Volume is then expressed as weighted cases".6  5 6  See http://secure.cihi.ca/cihiweb/dispPage.jsp?cw_page=casemix_riw_e (Accessed October 2005) for more information. Ibid.  110  Each acute care inpatient and surgical day care procedure in British Columbia is thus assigned a resource intensity weight indicating the expected level of relative resources used in caring for the patient compared to other patients. The B.C. Ministry of Health Services uses CIHI’s methodology for calculating a cost per weighted case for each hospital in the province.7 In essence, this involves teasing out costs associated with inpatient acute care services and then dividing these costs by the volume of weighted acute care cases treated at the hospital level. This generates a hospital specific cost per weighted case. In 2000/01, the calculated cost per weighted case in hospitals located within the Fraser Health Authority ranged from $2,150 to $3,520. The provincial average that year was $3,440.8 Since we used all acute care discharges and surgical day care procedures in this study, regardless of the hospital in which they were performed, we used the provincial average of $3,440 in estimating the cost of providing acute and surgical day care services to the patient population in this study. To standardize costs to the 2000/01 fiscal year, we multiplied the weighted case value for each patient’s use of acute inpatient or surgical day care services in each of the five years by the $3,440. There are a number of ways to estimate patient-specific acute care costs. Perhaps the most crude is to multiply patient days by the hospitals average cost per patient day. This does not take into account differences in the complexity of care provided to patient groups. Arguably the most precise manner is to generate patient specific case costs based on actual resources used by individual patients multiplied by the cost per unit of resource use. Several hospitals within B.C. have moved in this direction by implementing case costing methodologies. Two of these hospitals have published results of a comparison of using RIW in estimating costs compared to 7  Stephen Lee, Information Consultant, Information Resource Management, B.C. Ministry of Health, Personal communication, September 29, 2005.  111  their case costing methodology. The two hospitals are located in Victoria (Poole et al., 1998) and Vancouver (Borsa and Anis, 2005). For the Victoria hospital the results of the comparison suggested a relatively close approximation of costs calculated using these two methodologies (Poole et al., 1998) whereas greater variability was found for the Vancouver hospital (Borsa and Anis, 2005). In the absence of case costing methodologies at all of the hospitals utilized by the sample of patients in this study, we have used a cost per weighted case as a closer approximation of actual costs than using a cost per patient day. Information on acute care costs was used to create both a continuous and a binary variable. The binary variable was developed by assigning an individual’s annual acute care costs to either a low or high utilization category based on an 80/20 rule. That is, the 80% of patients with the lowest utilization of acute care costs in the year were assigned to the low utilization category while the 20% of patients with the highest utilization of acute care costs in the year were assigned to the high utilization category. This process was used for each of the five fiscal years. A mean annual utilization was calculated using the individual’s utilization history during the entire five years. 3.6.3.2 General Practitioner Physician specialties can be identified in a number of ways. The most familiar methodology is by most recent registered specialty (MRRS) as designated by the physician’s most recent specialty registration with the Medical Services Plan. This is a self-reported measure of each physician’s licensure and registration status.  8  Ibid.  112  In contrast, type of practice (TOP) is a methodology that uses each physician’s billing information to categorize the physician based on the way they actually practice. For example, a physician could report their MRRS as a family physician, but may actually have a billing pattern more closely representing emergency medicine. He or she would be identified as an emergency medicine physician by type of practice, despite their registered status as a family physician. By comparison, a physician could report their MRRS as pediatrics, but may actually have a billing pattern more closely representing a family physician. He or she would be identified as a family physician by type of practice, despite their registered status as a pediatrician. BC’s Medical Services Plan uses the TOP methodology for publishing practitioner profiles (Verhulst and Starr, 2003). Fee item specialty is a third type of methodology that identifies different types of services. All services covered under the Medical Services Plan are identified by particular fee items in the MSC payment schedule used by fee-for-service physicians. These fee items are grouped into broad categories, and different types of physicians are said to “own” a section for the purposes of fee negotiations. For example, fee item 0532 electrocardiogram and interpretation for children under 2 years of age is owned by specialty 14-paediatrics. Another such category is called “general practice” and specialty 00-family physicians own fee items in this area. This ownership, however, does not imply exclusive billing: any practitioner billing under the Medical Services Plan can bill any applicable fee item. The use of fee item specialty methodology, therefore, captures all billings in a particular category of the MSC payment schedule used by fee-for-service physicians, regardless of the type of physician who provided those services. PURRFECT, a BC Ministry of Health Services database uses fee item specialty  113  methodology to identify services provided by general practitioners, even though these services may be delivered by physicians who have other MRRS specialty designations. In this study we use the most recent registered specialty (MRRS). For general practice this was ‘0 - General Practice’. The MSP payment information master files used to calculate the number of GP visits (i.e., and charges associated with those visits) required ‘cleaning’ in order to generate an accurate determination of costs and counts of the number of visits. This was due to the following issues associated with the raw data in the MSP payment information master files: •  Claims represented by multiple records  •  No charge referral records  •  Claims that were never paid  •  Retroactive adjustments  The process of cleaning or netting the claims combines claim records and amounts that are determined to pertain to the same service on the same date from the same provider to the same patient. In addition to cleaning the costs paid out to physicians, this process also allows visits to be counted accurately. Information on GP visits was used to create both a continuous and a binary variable. The binary variable was created by assigning an individual’s annual GP visits to either a low or high utilization category based on an 80/20 rule. That is, the 80% of patients with the lowest utilization of GP visits in the year were assigned to the low utilization category while the 20% of patients with the highest utilization of GP visits in the year were assigned to the high utilization category.  114  This process was used for each of the five fiscal years as well as for a mean annual utilization calculated using the individual’s utilization history during the entire five years. Information on both the number of patient-specific visits and the payments associated with those visits were generated from the raw data in the MSP payment information master files. In addition to using GP visits to create a continuous and a binary variable, we also used the information on payments to create a cost variable. Payment information in the MSP payment information master files is based on the year in which the payment was made. For comparative purposes with respect to differences in resource use, we needed to adjust for price increases. Adjusting for price increases or inflation is based on actual fee item increases received by GPs and specialist physicians between 1996/97 and 2000/01 as calculated by the British Columbia Medical Association (BCMA).9 Price increases for GP services between 1996/97 and 2000/01 were as follows: • • • •  2.72% change to adjust 1996/97 data to 2000/01 levels 4.98% change to adjust 1997/98 data to 2000/01 levels 3.48% change to adjust 1998/99 data to 2000/01 levels 1.27% change to adjust 1999/00 data to 2000/01 levels  In 1997/98, there was a decrease in fee item prices of 2.15% followed by modest increases each of the following three years (1.45% in 1998/99, 2.18% in 1999/00 and 1.27% in 2000/01). The appropriate price increases were applied at a patient-specific level each of the five years so that all GP costs were adjusted to reflect 2000/01 prices. Information on GP payments was used to create both a continuous and a binary variable. The binary variable was created by assigning an individual’s annual GP costs to either a low or high utilization category based on an 80/20 rule. That is, the 80% of patients with the lowest utilization of GP costs in the year were assigned to the low utilization category while the 20% of  115  patients with the highest utilization of GP costs in the year were assigned to the high utilization category. This process was used for each of the five fiscal years. A mean annual utilization was calculated using the individual’s utilization history during the entire five years. 3.6.3.3 Specialist Physician As noted earlier, we used the most recent registered specialty (MRRS) in identifying the specialty of physicians in this study. Table 3-4 follows a new categorization matrix established for the primary care project currently being completed by CHSPR in which all personnel potentially paid with MSP funds were grouped into five categories. In defining specialist physicians for this study, we combined categories II and III in Table 3.4.  9  Mr. Jim Aikman, Director, Economics Department, BCMA, Personal communication, November 10, 2005.  116  Table 3-4 CHSPR Categorization Matrix for Personnel Funded Through MSP I. General Practice 00 - Family Practitioner II. Primary Care Related Specialists 05 - Obstetrics and Gynaecology 15 - Internal Medicine 28 - Emergency Medicine  14 - Pediatrics 24 - Geriatric Medicine  III. Non-primary Care Related Specialists 01 - Dermatology 02 - Neurology 03 - Psychiatry 04 - Neuropsychiatry 06 - Ophthalmology 07 - Otolaryngology 08 - General Surgery 09 - Neurosurgery 10 - Orthopedic Surgery 11 - Plastic Surgery 12 - Cardio & Thoracic Surgery 13 - Urology 16 - Radiology 17 - Pathology 18 - Anesthesia 19 - Pediatric Cardiology 20 - Physical Medicine and Rehabilitation 21 - Public Health 23 - Occupational Medicine 29 - Medical Microbiology 33 - Nuclear Medicine 44 - Rheumatology 45 - Clinical Immunization and Allergy 46 - Medical Genetics 47 - Vascular Surgery 48 - Thoracic Surgery IV. Non-physician Providers, Possibly Primary Care Related 80 - Midwife 81 - Registered Nurse 82 - Nutritionist/Dietitian 83 - Counselor 84 - Educator 85 - Licensed Practical Nurse 86 - Medical Office Assistant 87- Nurse Practitioner 88 - Respiratory Therapy 89 - Home Support 91 - Pharmacy V. Non-physician Providers, Other 30 - Chiropractics 32 - Physical Therapy 37 - Oral Surgery 39 - Optometry 41 - Oral Medicine 45 - Clinical Immunization and Allergy  31 - Naturopathy 34 - Osteopathy 38 - Podiatry 40 - Dental Surgery 42 - Orthodontia  Information on both the number of patient-specific visits to a specialist physician and the payments associated with those visits were generated from the raw data in the MSP payment  117  information master files. The raw data for specialist physicians had to be cleaned in the same manner as the raw data for GPs. Information on visits to a specialist was used to create both a continuous and a bi