Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Small-area disability-adjusted life years : a new approach to the spatio-temporal analysis of public… Lima, Viviane Dias 2005

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

Item Metadata

Download

Media
831-ubc_2005-105283.pdf [ 28.9MB ]
Metadata
JSON: 831-1.0092287.json
JSON-LD: 831-1.0092287-ld.json
RDF/XML (Pretty): 831-1.0092287-rdf.xml
RDF/JSON: 831-1.0092287-rdf.json
Turtle: 831-1.0092287-turtle.txt
N-Triples: 831-1.0092287-rdf-ntriples.txt
Original Record: 831-1.0092287-source.json
Full Text
831-1.0092287-fulltext.txt
Citation
831-1.0092287.ris

Full Text

S M A L L - A R E A DISABILITY-AD JUSTED LIFE Y E A R S : A N E W A P P R O A C H TO THE SPATIO-TEMPORAL A N A L Y S I S OF PUBLIC H E A L T H S U R V E I L L A N C E D A T A by VIVIANE DIAS L I M A B.Sc, Escola Nacional de Ciencias Estatisticas, 1996 M . S c , Universidade Federal do Rio de Janeiro, 1997 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES (Health Care and Epidemiology) THE UNIVERSITY OF BRITISH C O L U M B I A June 2005 © Viviane Dias Lima, 2005 11 Abstract When engaged in injury surveillance and regional health planning (e.g., the allocation and distribution of health services and prevention initiatives) it is essential that the health status of regions be monitored over time. The conventional approach to monitoring the health impact of injuries has been to summarize the spatial variation of each region using measures such as mortality and incidence rates. While the use of such measures yields important surveillance information, it does not directly quantify the overall (i.e., fatal and non-fatal) injury burden experienced by individual communities. Although measures of incidence and burden can be derived for small regions, they are often based on a low number of observed injury cases. These low counts are frequently associated with high degrees of chance variation. Recently developed model-based disease mapping methods reduce the influence of chance variation and in so doing yield more reliable estimates of the underlying variation in injury incidence and burden. In this thesis we describe how estimates of disability-adjusted life years (DALY) can be integrated into spatio-temporal Bayesian statistical methods to quantitatively describe how the burden of injuries varies between and within regions of British Columbia (BC) over time. As a demonstration, we examined administrative hospitalization data using the D A L Y methodology to calculate the incidence and burden of 13 external causes of injury at the provincial and health authority level from 1991-2000. To illustrate the small-area analysis of injury burden, we used the same data to describe the impact of road traffic accidents. Lastly, we Ill examined injuries due to adverse events to demonstrate the usefulness of administrative database screening to describe region-specific impacts of these injuries in BC. Obtaining reliable small-area D A L Y estimates has the potential to dramatically improve our understanding of public health problems. As shown in this thesis, this methodology enables researchers to quantify the total number of healthy years lost for leading causes of injury. In addition to scientific gains, this information can be used to inform the setting of regional priorities related to the planning and delivery of health care and quantify the benefits of regional intervention programs in cost-effectiveness analyses. IV Table of Contents Abstract 1 1 List of Tables ix List of Figures xiii Acknowledgements xxi Dedication xxiii Chapter 1. Introduction 1 1.1 Research Objectives 8 1.2 Organization of Dissertation 10 Chapter 2. Literature Review 11 2.1 Measures of Population Health and Disease Mapping Methods 11 2.1.1 Summary Measures of Population Health 11 2.1.1.1 Health Expectancies 13 2.1.1.2 Health Gaps 15 2.1.1.3 WhyUseDALYsandNotQALYs 18 2.1.1.4 How DALYs Have Been Used in the Literature 19 2.1.2 Small-Area Analysis and Disease Mapping 20 2.2 Issues in Medical Adverse Events Research 23 2.3 Validity and Reliability of the ICD9 E-Codes in Identifying External Causes 29 of Injury V 2.3.1 A l l Causes of Injuries Excluding Medical Adverse Events 30 2.3.2 Injuries Due To Medical Adverse Events 32 Chapter 3. Methods 39 3.1 Data 39 3.1.1 Definition of Study Regions 40 3.1.2 Definition of Study Population 43 3.1.3 Categories of External Cause and Nature of Injury 48 3.2 Bayesian Disease Mapping 51 3.2.1 Spatio-Temporal Modeling under a Full Bayesian Approach 55 3.3 Disability-Adjusted Life Years (DALY) 58 3.3.1 Assumptions Made in the Calculation of D A L Y s 60 3.3.1.1 Discounting Future Health Rate (r) 61 3.3.1.2 Age Weighting <j3) 64 3.3.1.3 Comorbidities 66 3.3.1.4 Disability Weights (DW) 67 3.3.1.5 Calculating Duration for Y L D (£>) 76 3.3.1.6 Residual Category of Injury Sequela 82 3.3.1.7 Gender Difference in the Length of Life for Calculating 83 Y L L (L) 3.4 Small-Area Analysis of Disability-Adjusted Life Years (DALY) 85 3.5 Software 87 3.6 Outline of Data Analyses 88 vi 3.6.1 Burden of Injuries in British Columbia and Its Health Authorities, 89 1991-2000 3.6.2 A Small-Area Analysis of the Incidence of Medical Adverse Events 89 in British Columbia, 1991-2000 3.6.3 A Small-Area Analysis of the Burden Due To Road Traffic 90 Accidents in British Columbia, 1991 -2000 Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 91 1991-2000 4.1 The Provincial Level Analysis 92 4.1.1 Conventional Mortality and Hospitalization Rates 92 4.1.2 Years of Life Lost (YLL) 97 4.1.3 Healthy Years Lost Due To Disability (YLD) 102 4.1.4 Disability-Adjusted Life Years (DALY) 107 4.1.5 Comparison of Contributions Associated With Y L D and Y L L (Ten- 111 Year Combined) 4.1.6 Comparison of A l l Health Status Indicators (Ten-Year Combined) 115 4.2 The Health Authority Level Analysis 119 4.2.1 Conventional Mortality Rate 119 4.2.2 Conventional Hospitalization Rate 124 4.2.3 Disability-Adjusted Life Years (DALY) 129 vii Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in 136 British Columbia 5.1 Estimated Mortality and Hospitalization Rates 136 5.1.1 Comparison of the Annual Estimated Rates among Different 140 Geographical Layers in the Provincial Health Care System 5.1.2 Mortality and Hospitalization Rate Ratios 147 5.2 Estimated Disability-Adjusted Life Years 154 Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in 162 British Columbia 6.1 Descriptive Statistics for Medical Adverse Events 162 6.2 Estimated Medical Adverse Event Incidence Rates 164 6.2.1 Comparison of the Annual Estimated Rates among Different 168 Geographical Layers in the Provincial Health Care System 6.2.1.1 Age Group 0-19 Years 170 6.2.1.2 Age Group 65-79 Years 176 Chapter 7. Discussion 182 7.1 Summary of Findings 182 7.2 Injury Viewed as a Public Health Issue 184 7.3 Methodological Considerations, Limitations and Recommendations for 192 Future Research 7.3.1 Data Quality 192 viii 7.3.2 Epidemiological Methodology 194 7.3.3 Statistical Methodology 201 7.3.4 General Issues 202 7.4 Conclusion 205 Bibliography 206 Appendix 229 A Acronyms 230 B Glossary 232 C Important ICD9-CM E-Codes 248 D Example of Disability Weight Spreadsheet Assuming Zero Disability 259 Weight for the Residual Category of Injury Sequela E Example of Duration Spreadsheet Assuming Zero Duration for the Residual 265 Category of Injury Sequela F Sensitivity Analyses 267 G Chapters'General Tables and Figures 291 H Diagnostic Plots 362 ix List of Tables Table 1.1 List of ICD9-CM E-codes for External Causes of Injury (Mathers and 6 Stevenson 1999; Murray and Lopez 1996). Table 1.2 Health Status Section of the Health Indicators Project Framework (Statistics 7 Canada 2004). Table 2.1 Different Studies on Medical Adverse Events (adapted from Baker et al. ' 26 2004). Table 2.2 List of Patient Safety Indicators (AHRQ 2003). 35 Table 2.3 Set of ICD9-CM E- and N-codes Defining the Patient Safety Indicator 37 "Foreign Body Left During Procedure". (AHRQ 2003, pages 98-99) Table 3.1 List of ICD9-CM N-Codes for Nature of Injury (Murray and Lopez 1996). 49 Table 3.2 Disability Weights by Nature of Injury and Age Category (Murray and 68 Lopez 1996). Table 3.3 Disability Weights for Each Age Group and Gender by Injury External 71 Cause. Table 3.4 Disability Weights (All Ages Combined) for Each Injury Sequela by 73 Gender and Injury External Cause. Table 3.5 Duration Assumptions by Type of Injury (Murray and Lopez 1996). 76 Table 3.6 Gender and Age Specific Life Expectancies for British Columbia (BC) in 78 2001. Table 3.7 Durations for Each Age Group and Gender by Injury External Cause. 79 Table 3.8 Durations by Injury Sequela for the External Causes of Injury Drowning 81 and Suffocation and Foreign Bodies. Males 0-19 years. Table 3.9 Japanese Gender and Age Specific Life Expectancies for the Year 1990 Used 84 in the Global Burden of Disease (GBD) Study. Table 4.1 Ten-Year Mortality Rates by Gender, Age Group and External Cause of 93 Injury, 1991-2000. Table 4.2 Ten-Year Conventional Hospitalization Rates by Gender, Age Group and 94 External Cause of Injury, 1991-2000. Table 4.3 Total Years of Life Lost (YLL) by Gender, Age and External Cause of 98 Injury, 1991-2000. Table 4.4 Ten-Year Per Capita Years of Life Lost (or Per Capita Y L L ) per 1,000 100 Population by Gender, Age and External Cause of Injury for 1991-2000. Table 4.5 Total Healthy Years Lost Due To Disability (YLD) by Gender, Age and 103 External Cause of Injury, 1991-2000. Table 4.6 Ten-Year Per Capita Healthy Years Lost Due To Disability (or Per Capita 105 Y L D ) per 1,000 Population by Gender, Age and External Cause of Injury for 1991-2000. Table 4.7 Total Disability Adjusted Life Years (DALY) by Gender, Age and 107 External Cause of Injury, 1991-2000. Table 4.8 Ten-Year Per Capita Disability-Adjusted Life Years (DALY) per 1,000 109 Population by Gender, Age and External Cause of Injury for 1991-2000. X I Table 4.9 Percent Contribution of Y L D to D A L Y (Ten-Year Combined) by Gender, 112 Age and External Cause of Injury, 1991-2000. Table 4.10 Ten-Year Age-Standardized Mortality Rate (ASMR) per 10,000 120 Population by Gender and Health Authority (HA), 1991 -2000 (Canada 2001 Census). Table 4.11 Ten-Year Age-Standardized Hospitalization Rates (ASHR) per 1,000 124 Population by Gender and Health Authority (HA), 1991-2000 (Canada 2001 Census). Table 4.12 Ten-Year Age-Standardized Per Capita Disability-Adjusted Life Years (or 130 age-Standardized Per Capita D A L Y or A S D A L Y R ) per 1,000 Population by Gender and Health Authority, 1991-2000 (Canada 2001 Census). Table 5.1 Regional Age-Specific Mortality Rates per 10,000 Population and 141 Hospitalization Rates per 1,000 Population Due To Road Traffic Accidents in 2000 for Males and Females Aged 20-39 Years. Table 5.2 Regional Per Capita Disability-Adjusted Life Years (DALY) per 1,000 155 Population Due To Road Traffic Accidents in 2000 for Males and Females Aged 20-39 Years. Table 6.1 Estimated Provincial Annual Medical Adverse Event Incidence Rates 165 (Age-Specific and Age-Standardized) per 100 Admissions by Gender for the Year 2000 (Standard Population: Hospital Admission Due To A l l Causes for Both Males and Females). xii Table 6.2 Regional Medical Adverse Event Incidence Rate per 100 Admissions in 169 2000 for Males and Females Aged 0-19 and 65-79 Years. xiii List of Figures Figure 1.1 Relationships among the Different Health Status Indicators Used in the 5 Small-Area Analysis of Disability-Adjusted Life Years (DALY) Methodology. Figure 2.1 Graphical Calculation of Quality-Adjusted Life Years (QALYs) (Gold, 15 Stevenson, and Fryback 2002). Figure 2.2 Graphical Calculation of Disability-Adjusted Life Years (DALYs) (Gold, 18 Stevenson, and Fryback 2002). Figure 3.1 Map of the Health Service Delivery Areas and Health Authorities of 42 British Columbia. Figure 3.2 Mid-Period Population Size by Health Service Delivery Areas during 1991 - 44 2000. Males and Females - Al l Age Groups. Figure 3.3 Mid-Period Number of Admissions by Health Service Delivery Areas 45 during 1991 -2000. Males and Females - A l l Age Groups. Figure 3.4 The Distribution of the Ten-Year Conventional Mortality Rate at the 47 Provincial Level, by Gender and Age Group, and for A l l External Causes of Injury Combined. Figure 3.5 Ten-Year Conventional Mortality Rate per 10,000 population 51 Attributable To Road Traffic Accidents. Males, 20-39 Years, 1991-2000. xiv Figure 3.6 Annual number of deaths, estimated population and conventional 53 mortality rate per 10,000 population attributable to road traffic accidents. Males, 20-39 years, 1991-2000. Figure 3.7 Preference Values Intrinsic in the Calculation of Disability-Adjusted 60 Life Years (DALY) . Figure 3.8 Effect of Time, in Years, on the Discrete Discounting Function. 62 Figure 3.9 Shape of Age Weighting Function for fi in the Set {0.03,0.04,0.05}. 65 Figure 3.10 Example of How to Calculate the Weighted Life Expectancy. 78 Figure 3.11 Hypothetical Mortality Rate Posterior Distribution Density of Different 87 Health Services Delivery Areas (HSDA). Figure 4.1 Categories of external causes of injury. 91 Figure 4.2 Annual Age-Standardized Mortality Rate (ASMR) per 10,000 96 Population for Males and Females, 1991-2000 (Canada 2001 Census). Figure 4.3 Annual Age-Standardized Hospitalization Rate (ASHR) per 1,000 97 Population for Males and Females, 1991-2000 (Canada 2001 Census). Figure 4.4 Annual Age-Standardized Per Capita Y L L (ASYLLR) per 1,000 101 Population for Males and Females, 1991-2000 (Canada 2001 Census). Figure 4.5 Annual Age-Standardized Per Capita Healthy Years Lost Due To 106 Disability (ASYLDR) per 1,000 population for Males and Females, 1991-2000 (Canada 2001 Census). Figure 4.6 Annual Age-Standardized Per Capita D A L Y (ASDALYR) per 1,000 110 Population for Males and Females, 1991-2000 (Canada 2001 Census). XV Figure 4.7 Percent Contribution of Y L D to D A L Y (Ten-Year Combined) by Age 113 and External Cause of Injury for Males, 1991-2000. Figure 4.8 Percent Contribution of Y L D to D A L Y (Ten-Year Combined) by Age 114 and External Cause of Injury for Females, 1991-2000. Figure 4.9 Ranking of External Causes of Injury within Each Age-Standardized Health 117 Status Indicator for A l l Ten Years of Data Combined. Males, 1991-2000 (Canada 2001 Census). Figure 4.10 Ranking of External Causes of Injury within Each Age-Standardized 118 Health Status Indicator for A l l Ten Years of Data Combined. Females, 1991-2000 (Canada 2001 Census). Figure 4.11 Annual Age-Standardized Mortality Rate (ASMR) per 10,000 122 Population by Injury and Health Authority for Males, 1991-2000 (Canada 2001 Census). Figure 4.12 Annual Age-Standardized Mortality Rate (ASMR) per 10,000 123 Population by Injury and Health Authority for Females, 1991-2000 (Canada 2001 Census). Figure 4.13 Annual Age-Standardized Hospitalization Rate (ASHR) per 1,000 127 Population by Injury and Health Authority for Males, 1991-2000 (Canada 2001 Census). Figure 4.14 Annual Age-Standardized Hospitalization Rate (ASHR) per 1,000 128 Population by Injury and Health Authority for Females, 1991-2000 (Canada 2001 Census). xvi Figure 4.15 Annual Age-Standardized Per Capita D A L Y (ASDALYR) per 1,000 133 Population by Injury and Health Authority for Males, 1991-2000 (Canada 2001 Census). Figure 4.16 Annual Age-Standardized Per Capita D A L Y (ASDALYR) per 1,000 134 Population by Injury and Health Authority for females, 1991-2000 (Canada 2001 Census). Figure 5.1 Conventional and Estimated Annual Mortality Rates per 10,000 138 Population by Health Services Delivery Area for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. Figure 5.2 Conventional and Estimated Annual Hospitalization Rates per 1,000 139 Population by Health Services Delivery Area for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. Figure 5.3 Annual Road Traffic Accident Mortality Rate per 10,000 Population for 142 Males Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 5.4 Annual Road Traffic Accident Hospitalization Rate per 1,000 143 Population for Males Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. xvii Figure 5.5 Annual Road Traffic Accident Mortality Rate per 10,000 Population for 145 Females Aged 20-39 Years, at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 5.6 Annual Road Traffic Accident Hospitalization Rate per 1,000 146 Population for Females Aged 20-39 Years, at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 5.7 Annual Ratio between the Mortality Rates of Each Health Services 148 Delivery Area and British Columbia for Road Traffic Accidents in Males and Females Aged 20-39 Years, 1991-2000. Figure 5.8 Map of the Annual Ratio between the Mortality Rates of Each Health 149 Services Delivery Area and British Columbia for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. Figure 5.9 Annual Ratio between the Hospitalization Rates for each Health 152 Services Delivery Area and British Columbia for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. Figure 5.10 Map of the Annual Ratio between the Hospitalization Rates of Each 153 Health Services Delivery Area and British Columbia for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. XV111 Figure 5.11 Annual Per Capita Disability-Adjusted Life Years per 1,000 Population 156 for Road Traffic Accidents Experienced by Males Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 5.12 Annual Per Capita Disability-Adjusted Life Years per 1,000 Population 157 for Road Traffic Accidents Experienced by Females Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 5.13 Annual Ratio between the Per Capita Disability-Adjusted Life Years of 158 Each Health Services Delivery Area and British Columbia for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years. Figure 5.14 Map of the Annual Ratio between the Per Capita Disability-Adjusted 161 Life Years of Each Health Services Delivery Area and British Columbia for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years. Figure 6.1 Annual Age-Specific Hospital Admission Rates per 1,000 Population 163 for Both Genders Combined, 1991-2000. Figure 6.2 Conventional and Estimated Annual Medical Adverse Event Incidence 166 Rates per 100 Admissions by Health Services Delivery Area for Boys and Girls Aged 0-19 Years, 1991-2000. XIX Figure 6.3 Conventional and Estimated Annual Medical Adverse Event Incidence 167 Rates per 100 Admissions by Health Services Delivery Area for Males and Females Aged 65-79 Years, 1991-2000. Figure 6.4 Annual Medical Adverse Event Incidence Rates per 100 Admissions 171 For Boys Aged 0-19 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 6.5 Annual Medical Adverse Event Incidence Rates per 100 Admissions 172 for girls Aged 0-19 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Figure 6.6 Annual Ratio between the Medical Adverse Event Incidence Rate of 173 Each Health Services Delivery Area and British Columbia for Boys and Girls Aged 0-19 Years, 1991-2000. Figure 6.7 Map of Annual Ratio between the Medical Adverse Event Incidence 175 Rate of Each Health Services Delivery Area and British Columbia for Boys and Girls Aged 0-19 Years, 1991-2000. Figure 6.8 Annual Medical Adverse Event Incidence Rates per 100 Admissions 177 for Males Aged 65-79 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991 -2000. Figure 6.9 Annual Medical Adverse Event Incidence Rates per 100 Admissions 178 for Females Aged 65-79 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. XX Figure 6.10 Annual Ratio between the Medical Adverse Event Incidence Rate of 180 Each Health Services Delivery Area and British Columbia for Males and Females Aged 65-79 Years, 1991-2000. Figure 6.11 Map of Annual Ratio between the Medical Adverse Event Incidence 181 Rate of Each Health Services Delivery Area and British Columbia for Males and Females Aged 65-79 Years, 1991-2000. xxi A c k n o w l e d g e m e n t s I have so many people to thank... But, I will try to make it short. I hope you all enjoyed this journey as I did. I am extremely thankful for the support, patience and guidance of my supervisory committee: Dr. Steve Marion, Dr. Ying MacNab, Dr. Jacek Kopec, and Dr. Sam Sheps. And special thanks to the members of my examining committee. I would like to thank Dr. Ying MacNab for the financial support (from October 2002 to the end of my program) from her various research grants (a NSERC discovery grant, a CIHR operating grant, and a MSF new unit research grant). In addition, I would like to thank the Coordenacao de Aperfeicoamento de Pessoal de Nivel Superior (CAPES) of Brazil for the financial support during the earlier years of my PhD program (up to August 2002). I would like also to express my appreciation to all the staff and faculty members at the Department of Health Care and Epidemiology for their help through the past five years. Thanks to the B C Vital Statistics Agency, the Centre for Health Services & Policy Research (CHSPR) for providing the data for this thesis, and in particular to Denise Morettin for her assistance in understanding the injury data. I would like to thank Dr. Kaizo Iwakami Beltrao, Dr. Antonio Marcos Duarte Jr., Dr. Beatriz Vaz de Melo Mendes, and other professors from the Escola Nacional de Ciencias Estatisticas (ENCE) and Universidade Federal do Rio de Janeiro (UFRJ), for their guidance and support in pursuing my PhD in Canada. Dr. John Petkau for making me believe that I XXll should persevere no matter how hard it got. Dr. Richard Mathias and Dr. Mieke Koehoorn for their helpful suggestions and support. The opportunity of meeting new friends from different parts of the world made this experience worthwhile. Thanks to Isabella Ghement, Jeff LudLow, Audrey and Jeroen Steenbeek, Yinshan Zhao, Maiko Shimada, Huiying Sun, Min Xu, Natsuko Montegi and Peter Cheung for their support and laughs when I most needed. To all my friends from St. Mark's discussion group and community, and in particular to Fr. Jim O'Neil, Fr. John McCarthy and Fr. Mark Hoo for their spiritual guidance and friendship. To Chris Richardson for editing my thesis and for his friendship. To my parents, family members and friends in Brazil, who despite the physical distance, have always been so important in my life. XX111 D e d i c a t i o n To my loving parents. " If I speak in human and angelic tongues but do not have love, I am a resounding gong or a clashing cymbal. And if I have the gift ofprophecy and comprehend all mysteries and all knowledge; if I have all faith so as to move mountains but do not have love, I am nothing. If I give away everything I own, and if I hand my body over so that I may boast but do not have love, I gain nothing. Love is patient; love is kind. It is not jealous, (love) is not pompous, it is not inflated, it is not rude, it does not seek its own interests, it is not quick-tempered, it does not brood over injury, it does not rejoice over wrongdoing but rejoices with the truth. It bears all things, believes all things, hopes all things, endures all things. Love never fails. If there are prophecies, they will be brought to nothing; if tongues, they will cease; if knowledge, it will be brought to nothing. For we know partially and we prophesy partially, but when the perfect comes, the partial will pass away. When I was a child, I used to talk as a child, think as a child, reason as a child; when I became a man, I put aside childish things. At present we see indistinctly, as in a mirror, but then face to face. At present I know partially; then I shall know fully, as I am fully known. So faith, hope, love remain, these three; but the greatest of these is love." / Corinthians 13 1 Chapter 1 Introduction One of the important roles of epidemiologists involved in public health surveillance is to monitor trends over time in disease mortality and morbidity rates for a collection of different communities (Schneider 2000). For many years, the conventional approach used to monitor the impact of diseases has been to summarize the spatial variation of each individual community using standardized mortality and incidence ratios. Typically, the individual ratios from each community are then integrated into a geographical map to provide a visual summary of the distribution of disease risk across the different communities. The information displayed in these maps is generally used to guide etiologic research, improve our understanding of the health needs of specific populations, guide future surveillance strategies, highlight areas of high and low risk, and inform policy formulation and resource allocation decisions (Elliott and Wartenberg 2004). When researchers conduct disease mapping studies to assess the health status of communities, the health data are typically extracted from large geographically referenced datasets with aggregate counts of disease occurrence at the regional level. These counts are usually modelled using the Poisson distribution. However, when the focus of these studies is to assess spatial variation of rare diseases in small areas, the observed number of disease cases in any single region may be low, and in many situations, zero. In regions with a low population at risk, the counts are likely to be highly variable, which exposes conventional disease Chapter I. Introduction 2 incidence fates and ratios to larger degrees of chance variation (i.e., these conventional measures of disease risk often fluctuate considerably over space and time) (Volinn et al. 1994; Elliott et al. 2001). One approach to dealing with the high variation often associated with small-area studies involves the use of model-based disease mapping methods. Recent statistical advances in model-based disease mapping strategies have dramatically improved our ability to generate realistic assessments of the geographical variation of small-area disease risk through the use of data smoothing techniques. The application of data smoothing techniques reduces chance variation in estimated measures of disease risk and yields more reliable estimates of the underlying variation in disease risk in individual communities (Lawson et al. 1999). Bayesian statistical methods are often applied to estimate the small-area disease risk in disease mapping studies, and one has the option of taking a full Bayesian approach or an empirical Bayesian approach (Bernardinelli and Montomoli 1992; German et al. 1998; Carlin and Louis 2000). Although the empirical Bayesian approach appropriately estimates the parameters in a model in a timely manner, it does not provide precise estimation of the uncertainty of these parameters. Alternatively, the full Bayesian approach has the advantage of providing both satisfactory estimation of parameters and their uncertainty. However, implementing the full Bayesian approach can be computationally more demanding than the former approach. When engaged in disease surveillance from a public health perspective, it is essential that the health status of communities be monitored repeatedly over time. The Bayesian spatio-temporal approaches to disease mapping is ideally suited to the information that results from such monitoring. Additionally, spatio-temporal Bayesian disease mapping can Chapter I. Introduction 3 accommodate the complex structure typically associated with regional public health data. In view of the apparent benefits of such approaches, the development of statistical methods to detect the spatio-temporal change in disease risk has advanced tremendously. The use of spatio-temporal Bayesian methods for mapping disease risks has been proposed in the small-area analysis literature as an efficient method for capturing spatial and temporal effects, and spatial temporal interaction effects on risk (e.g., see Elliott and Wartenberg 2004; Kleinschmidt et al. 2002; MacNab and Dean 2000, 2002; MacNab 2003a,b, 2004). The main focus of these studies has been to assess the distribution of disease occurrence in different communities using health status indicators such as morbidity and mortality rates and ratios. The development and assessment of community level health status indicators provides information that enables public health analysts to: (1) create and maintain up-to-date community specific health status profiles; (2) generate new hypotheses from information derived from analysis of spatial distributions of health problems; and (3) quantify the health impact of changes put in place to improve the quality of health care (Rothman and Greenland 1998). Although health status indicators, such as morbidity and mortality rates and ratios, are important in public health surveillance, it is also important that the overall disease impact (or burden) on the quality of life of communities be quantified. The World Health Organization (WHO) Global Burden of Disease study (GBD) presents one of the methodologies currently described in the literature that can be used to assess the impact of disease on the quality of life of a population as a whole, or several different communities (Murray and Lopez 1996). This approach combines estimates of mortality and morbidity for populations into a single measure of disease burden called disability-adjusted life years (DALY) (Murray and Lopez Chapter I. Introduction 4 1996, 1997a,b,c,d; Begg and Tornijima 2005). D A L Y s are used as a summary measure of population health that reflects the impact of both mortality and non-fatal health decrements. D A L Y s quantify the gap in years between age at death and some standard age before which death is considered "premature" in addition to time lived in states other than excellent health. D A L Y s are obtained by summing years of life lost (YLL) from premature death and healthy years lost due to disability (YLD). One of the innovative goals of this study is to integrate D A L Y estimates into public health surveillance as a new type of health status indicator at the community level that enables the quantification of the loss of healthy years due to premature death and disability resulting from diseases in a single measure of disease burden. Although previous studies using D A L Y s have been carried out at the national or regional level, D A L Y s have not been appropriately calculated at the small-area level. Spatio-temporal statistical models developed and presented in MacNab and Dean (2001) and MacNab (2003a), together with Markov Chain Monte Carlo (MCMC) methods for full Bayesian inference (implemented via the Gibbs sampling algorithm (Gelman et al. 1998; Carlin and Louis 2000; Spiegelhalter et al. 2003)) will be used to estimate DALYs and other health status indicators at the small-area level. Figure 1.1 illustrates the different health status indicators that will be examined in this study, and the basic sources of information that each of them provides. The solid lines were used to indicate which measures (e.g. counts) were further used to calculate other health status indicators, whereas the broken lines were used to remind the reader that individual measures can also be used individually as health status indicator. Chapter 1. Introduction 5 Disability-Adjusted Life Years (DALY) YLD+YLL Years of Life Lost (YLL) Premature Death • Years lived with disability (YLD) Quality of life lost Mortality rate Number of deaths Morbidity rate Number of hospitalizations Bayesian Disease Mapping Methods Figure 1.1. Relationships among the Different Health Status Indicators Used in the Small-Area Analysis of Disability-Adjusted Life Years ( D A L Y ) Methodology. In this study, the small-area analysis of disability-adjusted life years ( D A L Y ) methodology w i l l be demonstrated in an analysis of injury surveillance data from British Columbia (BC). The results of this analysis w i l l inform stakeholders in the public health arena of the impact (i.e., burden) of injuries in the communities of B C . For example, the results of this analysis could be used to answer the following questions: What are the leading causes o f injury in a specific community?; What are the impacts of injuries sustained on the community's quality o f life?; Which types of injuries result in either long-term or short-term disability?; and What is the incidence of each type of injury over time for a specific community? Chapter 1. Introduction 6 In this study, the classification of nature of injury (e.g., spinal cord injuries, fractures) and external cause of injury (e.g., road traffic accidents, suicide and self-inflicted injuries) will be based on the international classification of disease system (ICD) (CDC, 2003). The ICD Version 9 external codes with clinical modifications (ICD9-CM E-codes), including a description of injury types, can be found in Table 1.1. A l l external causes of injury defined by an E-code in Table 1.1 will be included in the analyses. Table 1.1. List of ICD9-CM E-codes for External Causes of Injury (Mathers, Vos, and Stevenson 1999; Murray and Lopez 1996). External Causes of Injury ICD9-CM E-codes Unintentional Injuries E800-E949 1. Road traffic accidents E 8 1 0 - E 8 1 9 , E826-E829 , E929.0 2. Other transport injuries E800-807 ,E820-E825 ,E830-E848,E929.1 3. Poisoning E 8 5 0 - E 8 6 9 , E929.2 4. Falls E880-E885,E886.9 , E887-E888 , E929.3 5. Burns/fires/scalds E 8 9 0 - E 8 9 9 , E924.0, E924.8, E924.9, E929.4 6. Drowning E910 7. Sports injuries E886.0, E917.0, E927 8. Natural and environmental factors E 9 0 0 - E 9 0 9 , E929.5 9. Machinery injuries E919, E920.0, E920.1, E920.4 10. Suffocation and foreign bodies E 9 1 1 - E 9 1 5 11. Adverse effects o f medical treatment (or medical adverse events) E870-E876 , E878-E879 , E 9 3 0 -E949 a. Surgical/medical misadventure E 8 7 0 - E 8 7 6 b. Surgical and medical procedures as the cause o f abnormal reaction o f patient or later complication, without mention o f misadventure at time o f procedure E 8 7 8 - E 8 7 9 c. Adverse effects o f drugs in therapeutic use E 9 3 0 - E 9 4 9 12. Other unintentional injuries (excluding medical adverse event) E916, E917.1-E917.9 , E918, E920.2, E920.3, E920.5-E920.9 , E 9 2 1 - E 9 2 3 , E 9 2 4 . 1 . E 9 2 5 -E926, E928, E929.8 Intentional Injuries E 9 5 0 - E 9 7 9 , E990-E999 1. Suicide and self-inflicted injuries E 9 5 0 - E 9 5 9 2. Other intentional injuries: homicide and violence; legal intervention; and injury resulting from operations o f war E 9 6 0 - E 9 7 9 , E990-E999 Chapter 1. Introduction 7 The material presented in this thesis is intended to explain a new methodology, but also to produce useful information about injuries in BC. On a technical note, it is important to mention that all health status indicators examined in this study were previously incorporated in the methodology of the Health Indicators Project, a collaboration that started in 1998 between the Canadian Institute for Health Information (CLHI) and Statistics Canada. The main objective of this initiative was the identification of information requirements that i f addressed could help improve and maintain population health across Canada at the regional, provincial and national levels (CIHI 2004a). The section of the Health Indicators Project framework relevant to this study is displayed in Table 1.2. Table 1.2. Health Status Section of the Health Indicators Project Framework (CIHI 2004a). i Health Status Well-Being Health Conditions Human Function Deaths • Self-rated health • Changes over time in self-rated health • Body mass index (BM1) • Arthritis/rheumatism • Functional health • Two-week disability days • Infant mortality • Perinatal mortality • Self-esteem • Diabetes • Activity limitation • Life expectancy • Asthma • High blood pressure • Chronic pain - affects activities • Conditions causing activity limitation • Disability-free life expectancy • Disability-adjusted life expectancy Mortality crude counts/rates, age-standardized rates: • Total mortality • Circulatory disease deaths • Chronic pain - severity • Depression • Disability-adjusted life years ( D A L Y ) • Health expectancy • Cancer deaths • Respiratory disease deaths • L o w birth weight •Suicide • Unintentional injury deaths • A I D S deaths Cancer incidence, age-standardized rates: • A l l cancer incidence • Lung cancer incidence • Colorectal cancer incidence • Breast-female cancer incidence • Prostate cancer incidence Potential years o f life lost ( P Y L L ) • Total P Y L L • Cancer P Y L L •Circulatory P Y L L • Respiratory P Y L L • Unintentional injuries P Y L L • Suicide P Y L L • Injury hospitalization • Food and waterbome diseases • A I D S P Y L L • Injuries Note: Health status indicators used in this study are highlighted. Chapter 1. Introduction 8 1.1 Research Objectives Injuries have a negative impact on the health of populations. Estimates of the magnitude of this negative impact and its variation between and within jurisdictions can provide useful information for researchers and policy makers. One objective of this study is to present and implement a methodology for estimating the magnitude of the burden of injuries and its variation between and within regions of BC over time. The methodology used tb assess the impact of injuries is based primarily on the calculation of disability-adjusted life years (DALYs) that are then utilized in a Bayesian-based approach to disease mapping and estimation of small-area variation. Two specific examples are presented in this study: road traffic accidents and medical adverse events (or medical AEs or AEs). The former (road traffic accidents) has features that are in many ways typical of unintentional injuries. The latter (medical AEs) was discovered to have a surprisingly large incidence in our study population during initial exploratory analysis; however, the presence of many atypical features complicated the calculation of DALYs (see more details in Chapter 7). As such, the D A L Y measure was not utilized in the analysis of injuries due to medical AEs. The main research objective of this study is therefore to apply Bayesian disease mapping methods to a measure of disability-adjusted life years (DALY) . • The methodology is illustrated by exarniriing patterns, variations, and trends in small-area disability-adjusted life years (DALY) for injuries due to road traffic accidents in British Columbia from 1991 to 2000. Chapter 1. Introduction 9 In addition to the main objective, the secondary research objectives of this study are: (a) To describe the impact of injury at the provincial and health authority levels. • The burden of unintentional and intentional injuries is assessed at the provincial and health authority levels using the disability-adjusted life years (DALY) measure. (b) To assess the regional variation and temporal trends in injury rates due to medical adverse events. • Patterns, variations, and trends in small-area incidence of injuries due to medical adverse events in British Columbia from 1991-2000 are presented and discussed. In addition, this study also explores the following methodological issues involving the measurement of disability-adjusted life years (DALY) : (a) The impact of different values of the discount rate for future health on the calculation of chsability-adjusted life years (DALY). (b) The robustness of disability-adjusted Ufe years (DALY) when different choices of disability weights (weights reflect the severity of injury) are used for those injury sequelae in which no weights have been published previously. Chapter 1. Introduction 10 1.2 Organization of Dissertation The organization of this study is as follows. In Chapter 2 we present a systematic review of all the pertinent literature that was used to develop the epidemiological and statistical methods demonstrated in this study. Chapter 3 discusses in detail the epidemiological aspects inherent in the calculation of disability-adjusted life years (DALY) and introduces the spatio-temporal model as a tool to analyse injury surveillance data via a full Bayesian approach. The provincial and regional analyses, based on conventional estimates of injury incidence and burden, are presented in Chapter 4. In Chapter 5, we present the results for the small-area analysis of injury burden attributable to road traffic injuries. The small-area analysis of injuries due to medical adverse events is shown in Chapter 6. Chapter 7 contains the conclusion of the study and includes a discussion of the results and associated limitations, identification of areas of future research, and suggestions to stakeholders on the possible uses of our results in public health injury prevention. 11 Chapter 2 Literature Review In this chapter we present a systematic review of pertinent literature that was used to develop the epidemiological and statistical methods described in this study. Additionally, the International Classification of Disease System (ICD) Version 9 coding system, used in the literature to identify injuries at the population level, will also be described. 2.1 Measures of Population Health and Disease Mapping Methods This section presents an overview of well-known methods for measuring population health. These include methods of health-adjusted life years (HALYs) such as quality-adjusted life years (QALYs) and disability-adjusted life years (DALYs). In addition, we review the up-to-date work in disease mapping methods that has been previously reported in the small-area analysis literature. 2.1.1 Summary Measures of Population Health For many years, one of the main objectives in several public health studies has been to describe the distribution of disease occurrence in the population using measures of disease frequency (e.g., incidence rates, standardized mortality rates.). The main goal of these studies 12 has been to provide summary information about the population health necessary to public health planners and other stakeholders responsible for managing and allocating health care resources. However, in this thesis, the term summary measures of population health (SMPH) is employed to define those measures that combine information on both mortality and non-fatal health outcomes to represent the health of a particular population in a single number that is expressed in time units (Begg and Tomijima 2005; Gold, Stevenson, and Fryback 2002; Field and Gold 1998; Murray et al. 2002; Murray, Salomon, and Mathers 2000; Iburg and Kamper-Jorgense 2002). There are several scenarios where a SMPH can be useful. For instance: • To compare the health of populations across different regions at different points in time; • To identify and quantify health inequalities within or between populations; • To provide an estimate of the effect of fatal and non-fatal health outcomes on the health of a population; • To provide information to guide the setting of priorities in the planning and delivery of health care for a given population; and • To quantify the benefits of health interventions in cost-effectiveness analyses (Murray et al. 2002). Two classes of summary measures of population health have been developed: health expectancies and health gaps. Both classes of measures use time (lived in health states, or time lost as a result of premature death) as a common metric for measuring the impact of mortality and non-fatal health outcomes (Gold, Stevenson, and Fryback 2002). Chapter 2. Literature Review 13 2.1.1.1 Health Expectancies Health expectancies are population indicators that estimate the average time (in years) that a person could expect to live in a defined health state. Following the work by Gold, Stevenson, and Fryback (2002), we also refer to measures based on health expectancies that incorporate estimates of mortality and morbidity as quality-adjusted life years (QALYs). Economists, operational researchers and psychologists developed QALYs in the late 1960s primarily for use in cost-effectiveness analysis. QALYs are constructed using health-related quality of Ufe (HRQL) weights that are attached to individual descriptions of health. These HRQL weights are not linked to any particular disease/condition/disability; rather, they are derived from values eUcited from individuals regarding either their own health state, or the health states of others. The universe of health states that individuals experience is vast, and the challenge of any health status measure is to capture the complexity of these states, in a valid and consistent manner, across different individuals. Descriptive health status measures that have been applied in studies using QALYs include the Health Utility Index (HUT) (Manuel, Schultz, and Kopec 2002; Torrance et al. 1995; Torrance 1986; Torrance, Boyle, and Horwood 1982), the QuaUty of Well-Being Scale (QWB) (Kaplan 1988), EQ-5D (EuroQoL) (EuroQol Group 1990), and the Health and Activity Limitation Index (HALex) (Erickson 1998). Since the health attributes described in each of these instruments vary, each instrument portrays a different picture of health status, making comparisons complicated across conditions or populations. By convention, since QALYs are a measure of health expectancy, the maximal value 1.0 represents full health for one year, and the minimal value 0 represents death. Note that some of these health expectancy measures are constructed in a way that allows for negative values that Chapter 2. Literature Review 14 reflect health states worse than death. The most commonly used methods to generate preferences or values for HRQL in Q A L Y studies include the standard gamble (SG), rating scale (RS) and time trade-off (TTO) (Torrance et al. 1995; Torrance 1986, 1987; Torrance, Boyle, and Horwood 1982). These techniques ask respondents to value health states by making explicit what they would be willing to sacrifice (in terms of time, or risk of death) in order to return from the health state being described (or experienced) to perfect health. In order to have a better understanding of the calculation of QALYs, we refer to Figure 2.1 (Gold, Stevenson, and Fryback 2002). In this figure, the trajectory of an individual HRQL from "now" to that person's death is shown by the solid line. The life path declines for some years, followed by a brief recovery. Next, the person suffers a major event from which he/she briefly recovers some HRQL, however, the recovery does not last long and it is followed by an irregular decline until an abrupt drop representing death. The area under the curve represents the QALYs accumulated by the person over this portion of his/her life. The area of each rectangle is the product of an HRQL weight and the time in which the person is assumed to experience that HRQL level. Chapter 2. Literature Review 15 1.0 4-Quality of life 0.0 now death years Figure 2.1. Graphical Calculation of Quality-Adjusted Life Years (QALYs) (Gold, Stevenson, and Fryback 2002). 2.1.1.2 Health Gaps Measures of potential years of life lost due to premature mortality have been used for many years to measure the mortality burden of various causes of death. These measures calculate the gap, in years, between age at death and some standard age before which death is considered "premature" (usually 75 years). Health gaps extend the notion of mortality gaps to include time lived in states other than excellent health. The use of time-based health gap measures provides a common metric for population health research that can be used with different outcomes of interest. Additionally, a common metric approach facilitates the integration of economic evaluations of interventions with the monitoring of health system outcomes and assessments of the overall health burden of diseases/injuries/health determinants in the population. The most widely used health gap measure is the disability-adjusted life years (DALY) (Murray and Lopez 1996, 1997a,b,c,d). Chapter 2. Literature Review 16 In the early 1990's, researchers at Harvard University and at the World Health Organization conducted the first global burden of disease (GBD) study using DALYs to quantify the global burden of premature death and functional disability caused by disease/injury. There were four main objectives associated with the original GBD study (Murray and Lopez 1996, 1): • 'To facilitate the inclusion of non-fatal health outcomes in debates on international health policy, which were all too often focused on mortality in children under 5 years of age"; • "To decouple epidemiological assessment from advocacy so that estimates of the mortality or disability from a condition are developed as objectively as possible"; • "To allow estimates of health impact to be mapped to causes, whether in terms of disease and injury, or risk factors and broader social determinants"; • "To quantify the burden of disease using a measure that could also be used for cost-effectiveness analysis". In the first GBD study, Murray and colleagues relied on secondary data and expert opinion to identify different diseases/injuries, and to describe the long-range of disability associated with these diseases/injuries. The D A L Y valuation employed in the GBD study was based on a person trade-off (PTO) health state valuation technique that explicitly addresses trade-offs between life and H R Q L for people with different diseases/injuries. As mentioned previously, there are other health state valuation techniques besides PTO that have been extensively used in the literature, for example: standard gamble (SG), rating scale (RS) and time trade-off (TTO) (Torrance et al. 1995; Torrance 1986, 1987; Torrance, Boyle, and Horwood 1982). PTO has been used in the literature as a method of health state valuation that focuses on social (as opposed to individual) preferences for different health care interventions. In Chapter 2. Literature Review 17 contrast to other methods, the respondents in studies using PTO are asked to make decisions about the health state preferences of other people instead of themselves. Generally speaking, "the PTO basically consists of asking how many outcomes of one kind (e.g., outcome B) respondents consider to be equivalent in social value to X outcomes of another kind (e.g., outcome A). For example, 'If there are x people in adverse health situation A and y people in adverse health situation B, and if you can only help (cure) one group (for example, owing to limited time or limited resources), which group would you choose to help?' " (Green 2001, 234). PTO is a complex technique when compared to the other health state valuation methods, and in addition, little is known about its reliability and validity. In spite of these limitations, the researchers involved in the GBD study identified the PTO assessment as the best technique to address their study objectives. D A L Y is a measure of health gap, where a value of 1.0 represents full disability (or death) and a value of 0 represents no disability (or full health). The computation of D A L Y s is shown in Figure 2.2 (Gold, Stevenson, and Fryback 2002). Note that there is a point called max that represents the ideal gender-specific life expectancy. To compute D A L Y s , the area lost from the ideal lifetime, living to the maximum life expectancy in full health, is approximated by summing the areas of the rectangles, or by calculating the area above the life path extended to the ideal life expectancy. The work in this study follows the convention of using gender-specific life expectancies. Chapter 2. Literature Review 18 0.0 now death max years Figure 2.2. Graphical Calculation of Disability-Adjusted Life Years (DALYs) (Gold, Stevenson, and Fryback 2002). 2.1.1.3 Why Use DALYs and Not QALYs To study the burden of injury we can either use prevalence or incidence measures. To estimate the fatal burden of injury {i.e., the years of life lost (YLL)) we used the mortality rate and the number of years lost due to premature death. When calculating Y L L , it is therefore necessary to adopt an incidence perspective. However, to calculate the non-fatal burden of injury (i.e., the healthy years lost due to disability (YLD)) we can use either incidence or prevalence measurements. In order to use a prevalence measure to calculate Y L D , we need to estimate the prevalence of different sequelae at a point in time (we may have to adjust this prevalence for any seasonal variation i f present) and multiply it by one year, or alternatively use the incidence of each sequela and multiply it by the duration of disability (if the assumption of a steady-state population holds). Chapter 2. Literature Review 19 In this burden of injury study, an incidence perspective has been chosen since: (a) Incidence-based databases are available for both hospitalization (a morbidity measure) and mortality; (b) The method to calculate time lived with disability is consistent with the method for calculating time lost due to premature mortality (i.e., the sum of two incidence measures); (c) Using an incidence measure allows us to capture any epidemiological trend in spatial clusters of new cases; and (d) Statistical models based on incidence information wil l be used to develop a surveillance system to capture change, identify clusters and flag abrupt jumps of new cases of injury. 2.1.1.4 How DALYs Have Been Used in the Literature DALYs have been assessed and used in various burden of disease/injury studies, starting with the collaboration between researchers at Harvard University and the WHO for the World Bank's 1993 "World Development Report" (World Bank 1993), which throughout this thesis we refer to as the global burden of disease (GBD) study (Murray and Lopez 1996, 1997a,b,c,d). Following this study, there have been other important burden of disease studies undertaken in Australia (Mathers et al. 2000, 2001), Victoria (in Australia) (Taylor 2001; Victorian Department of Human Services 1999a,b), New Zealand (Government of New Zealand 2001) and the Netherlands (Melse et al. 2000; Stouthard, Essink-Bot, and Bonsel 2000; Stouthard et al. 1997). Some of the most recent studies include the WHO Global Burden of Disease project estimates for the years 2000,2001 and 2002 (WHO 2004c). In addition to reports and papers describing the burden of disease/injury studies, in recent years, researchers have published papers specifically exannning the strengths and limitations Chapter 2. Literature Review 20 associated with the use of DALYs (Murray and Acharya 1997; Bonneux 2002; Ezzati et al. 2002; Hollinghurst, Bevan, and Bowie 2000; Jelsma, De Weerdt, and De Cock 2002; Williams 1999; Murray and Lopez 2000; Sayers and Fliedner 1997; Arnesen and Kapiriri 2004). The strengths of the D A L Y approach to measuring burden of disease were described in previous sections of this study. However, there are some limitations related to methodological assumptions to carry out the D A L Y calculations. Two of these limitations, related to the age and time of illness adjustment, are further discussed in the Methods chapter of this study. The disability weights associated with the extent of loss due to physical fimctioning are also a concern when applying this method. These weights were developed for the GBD study using expert opinion and secondary data. The question is: How universal are these weights? A study from the Netherlands went one step further than the original GBD study, and changed the way the GBD weights were calculated to identify different levels of severity within each condition. Unfortunately, the Netherlands study did not obtain disability weights for as many categories of injury sequelae as the GBD study, because not all sequela described in the GBD study were considered important to public health in terms of morbidity, mortality and cost. As we are interested in the same categories of injury sequelae as the GBD study, we elected to use the GBD weights from the WHO report on the Global Burden of Injury in 2000 (Mathers et al. 2001) for the analyses presented in this study. 2.1.2 Small-Area Analysis and Disease Mapping For marty years researchers have recognized that disease occurrence is not randomly distributed in time and space. Additionally, the practice of mapping diseases at the small-area Chapter 2. Literature Review 21 level has increased tremendously over the years (Diehr et al. 1992; Howe 1989; Elliott et al. 2001; Volinn et al. 1994). However these maps are usually unreliable if based on conventional rates because areas with a small population at risk are more likely to have extreme rates due to chance variation, especially when exanrining rare diseases. One of the main reasons for using disease mapping methods is to smooth and remove the random variation in conventional rates to provide more statistically reliable assessments of the geographical variation of these rates. Bayesian statistical methods applied to disease mapping have advanced tremendously in the past decade, with several important papers having been published on the use of these methods (Elliott et al. 2001; Clayton, Bernardinelli, and Montomoli 1993; Mollie 1996; Besag, York, and Mollie 1991; MacNab and Dean 2000, 2001, 2002; Manton et al. 1989; MacNab 2002, 2003a,b, 2004; Lawson et al. 1999; Richardson 1992; Kleinschmidt et al. 2002). These papers have used Bayesian disease mapping methods for descriptive purposes, for surveillance arid identification of high (jisease/injury incidence areas, in ecological analyses, and to inform decisions involving health policy and resource allocation. However, it is important to mention that small-area analysis and disease mapping methods have not yet been applied to the study of medical adverse events, or incorporated into the calculation of disability-adjusted Ufe years. Kleinschmidt et al. (2002) used Bayesian disease mapping methods to study spatial and temporal variations in small-area malaria incidence rates from mid-1986 to mid-1999 in two districts in northern KwaZulu Natal, South Africa. By modeling the spatial variation of time trends in incidence rates, the authors explored whether the additional cases of malaria were coming mainly from areas that always had the highest transmission levels, or whether they were originating from previously low-transmission or malaria-free sub-regions. The authors Chapter 2. Literature Review 22 showed that the malaria incidence increased in the period of study, and based on the geographical distribution of the malaria incidence, the number of areas of high incidence increased as well. In addition, it was shown that the spatial distribution of the rise in malaria incidence was uneven, and the additional cases of malaria were coming from the geographic expansion of high-malaria-risk areas. Sun et al. (2000) used a Bayesian hierarchical rhodel in disease mapping to study the spatio-temporal distribution of lung cancer mortality rates in the state of Missouri from 1973 to 1992, for males in the age groups 45-54, 55-64, 65-74 and 75 and older. The main results were: (1) The lung cancer mortality rates increased over time for the two older age groups; (2) There were geographical clusters of lower rates in the north and higher rates in the southeast of Missouri for the two older age groups; and (3) When controlled for age, the lung cancer mortality rates increased more rapidly in the southeast and south central regions of Missouri. MacNab (2004) applied Bayesian spatial and ecological regression models to analyse small-area variation in hospitalization rates due to motor vehicle accident injury among male children and youth (aged 0-24 years), from 1990 to 1999, in 83 local health areas in British Columbia. The author demonstrated how Bayesian modeling techniques could be implemented to assess potential risk factors measured at the regional level. Some of the risk factors at the regional level were neighbourhood socio-economic disadvantage score, medical services availability and utilisation, crime rates, excess speeding charges and seatbelt violation rate. The results of this study indicated that significant spatial variation in motor vehicle accident (MVA) hospitalization rate existed after adjusting for several risk factors. In addition, the author found that: (1) M V A hospitalization rate was inversely associated with Chapter 2. Literature Review 23 socio-economic status (SES) across all age groups; higher in communities with lower SES; (2) For the age group 1-14 years, crime rates were associated with a higher M V A hospitalization rate; and (3) For the age group 20-24 years, seatbelt violations and excess speeding were associated with a higher M V A hospitalization rate. In another study, MacNab (2003a) showed that Bayesian hierarchical Poisson random effects spatio-temporal models could be used in injury surveillance and prevention programs. Information derived from such models could then be used by high-risk regions to help develop and evaluate prevention programs, and to assist in health planning and resource allocation. For example, MacNab (2003a) examined hospitalization rates due to motor vehicle accident (MVA) injury from 1987 to 1996 among male children and youth aged 0-24, in 20 health regions in British Columbia. The fitted model captured inequalities in M V A hospitalization rates, and showed that during the ten years of study, the spatial distribution of areas with higher M V A hospitalization rates shifted from the northeast to the central-interior and southwest of BC. 2.2 Issues in Medical Adverse Events Research Before proceeding with the literature review of studies examining the validity and reliability of the ICD9 E-Codes used in this thesis, it is important to familiarize the reader with critical issues in medical adverse events (or medical A E or AE) research. First, it is often difficult to define, classify, identify, describe and analyse AEs. As an example of the variety of published definitions, Walshe (2000) summarized a selection of definitions of medical A E Chapter 2. Literature Review 24 reported in the literature (see Walshe (2000) for referenced literature). The definitions included: 1 "Any response to medical care in the hospital that is unintended, undesirable and harmful to the patient." (McLamb and Huntley 1967) 2 "A potentially compensable event is a disability caused by healthcare management: • Disability—is a temporary or permanent impairment of physical or mental function (including disfigurement) or economic loss. • Causation—is established when the disability is more probably than not attributable to health care management. • Healthcare management—includes both actions and inactions of any health care provider or attendant." (Mills 1978) 3 "Adverse patient occurrences (APOs) ... refer to untoward patient events which, under optimal conditions, are not a natural consequence of the patient's disease or treatment. The common thread of all APOs is that they are events which health professionals agree are not desirable outcomes of medical management." (Craddick and Bader 1983) 4 "An unintended injury caused by medical management rather than by the disease process. The injury is sufficiently serious to lead to prolongation of hospitalisation or temporary or permanent impairment or disability in the patient." (Harvard Medical Practice Study 1990) 5 "An unintended injury or complication, which results in disability, death or Chapter 2. Literature Review 25 prolonged hospital stay and is caused by health care management." (Wilson, Runciman and Gibberd 1995) 6 "An untoward or undesirable occurrence in the healthcare process which has or potentially has some negative impact on a patient or patients and results or may result from some part of the healthcare process." (Walshe 1998) Additionally, a medical A E has been defined as an injury that is caused by medical management, which prolongs hospitalization or produces disability (Karson and Bates 1999; Brennan et al. 1991; Wilson et al. 1995; Michel et al. 2004, Samore et al. 2004). Table 2.1 illustrates several other important studies using variants of this A E definition. Second, although methods used to estimate the incidence of medical adverse events have developed tremendously, most AEs are still not detected (Baker et al. 2004; CIHI 2004b; Weingart et al. 2000; Kohn, Corrigan, and Donaldson 2000). There are 3 methods to assess the magnitude of AEs: (1) voluntary reporting; (2) direct observation of health care personnel during routine clinical care; and (3) computerized screening algorithms utilized by nurses combined with retrospective physician chart review. Although each of these methods has advantages and disadvantages, method (3) has been applied in most studies on adverse events. For example, the studies described in Table 2.1 all utilized variants of method (3) to estimate A E incidence in in-hospital patients. Chapter 2. Literature Review 26 Table 2.1 Different Studies on Medical Adverse Events (adapted from Baker et al. 2004). Study Satiny (\L\in 1. \clllMOIl III" ll)W-li.sk palienls A l : definition Window ol" scrulinv before index admission Window of semliri} alter index admission %of patients with ~- 1 A l : Baker et al. 2004 n = 3,745 20 Canadian hospitals (2000) Yes Unintended injury or complication that resulted in disability, death or prolonged hospital stay and was caused by health care management rather than by the underlying disease process 12 months AE must have occurred during index admission, but it could be detected up to 12 months afterward 7.5 Thomas et al. 2000 n= 14,700 28 hospitals in Utah and Colorado (1992) No Injury caused by medical management rather than by the disease process and resulted in prolonged length of stay or disability at discharge 6 months if patient <65 yr; 12 months if >=65 yr None: AE must have been detected during index admission 2.9 (3.2$) Wilson et al. 1995 n= 14,179 28 hospitals in New South Wales and South Australia (1992) Partial (did not exclude obstetrics patients) Unintended injuries or complications resulting in death, disability or prolonged hospital stay that arise from health care management 6 months if patient <65 yr; 12 months if >=65 yr AE must have occurred during index admission but could be detected afterward 16.6(10.6$) Brennan et al. 1991; Leapeetal 1991 n = 30,195 51 hospitals in New York (1984) No Unintended injury that was caused by medical management and that resulted in measurable disability Unlimited Unknown 3.7 Vincent, Neale, and Woloshynowych 2001 n= 1,014 2 hospitals in London, England (1999-2000) No Unintended injury caused by medical management rather than by disease process Unlimited AE must have occurred during index admission but could be detected afterward 10.8 Davis et al. 2002, 2003(a,b) n = 6,579 13 hospitals in New Zealand (1998) Partial (did not exclude obstetrics patients) Unintended injuries or complications resulting in death, disability or prolonged hospital stay that arise from health care management Unlimited AE must have occurred during index admission but could be detected afterward 12.9H $ From Thomas et al.In this study, American and Australian investigators harmonized the inclusion criteria and AE definitions between the 2 studies and then re-analyzed the Australian data.This yielded an adjusted AE rate of 10.6%, as compared with 16.6% using the original Australian study methods. \ Represents an unweighted estimate of prevalence. Chapter 2. Literature Review 27 Note that Table 2.1 shows a high variability across studies in the percentage of hospital separations in which an A E occurred, ranging from 2.9% (Utah and Colorado study, UTCOS) to 16.6% in a study in Australia (Quality in Australian Health Care Study, QAHCS). Next, we present in more detail the results of several recent studies on adverse events conducted in Canada. In a recent national study conducted in Canada, 20 hospitals were selected in the 5 most populous provinces (British Columbia, Alberta, Ontario, Quebec and Nova Scotia) (Baker et al. 2004). The definition of a medical A E in this study was any unintended injury or complication resulting in death, disability or prolonged hospital stay that arose from health care management, as opposed to original patient illness. The criteria to identify these medical AEs involved a two-stage process. The first stage consisted of nurses and health records professionals screening each selected hospital chart for the presence of an A E (total of 1,527 out of 3,745 charts). The second stage consisted of physicians screening the selected charts identified in the first stage to confirm the A E . The overall A E rate was 7.5% (95% CI [5.7%, 9.3%]); the risk of A E in men and women was not statistically different; teaching hospitals had higher rates of A E ; older patients were more likely to experience an A E ; the longer mean length of stay was associated with those patients who experienced an A E ; surgical procedures were associated with higher rates of A E ; and 36.9% (95% CI [32.0%, 41.8%]) of AEs were considered preventable. Forster, Asmis et al. (2004) were interested in evaluating characteristics of adverse events affecting patients admitted to a teaching hospital in Ottawa to determine the incidence, preventability, severity, type and timing of adverse events. The definition of a medical A E in this study was poor patient outcomes due to medical care. The incidence of A E was 12.7% Chapter 2. Literature Review 28 (95% CI [10.1%, 16.0%]); the rate of preventable A E was 4.8% (95% CI [3.2%, 7.0%]); 61% (95% CI [49%, 72%]) of patients experienced the adverse event before the index hospitalization; most AEs were due to drug treatment, operative complications and nosocomial infections; overall risk of A E was significantly associated with patient age, admitting service (medicine, surgery obstetrics or gynecology) and admission status (elective or emergency). When only in-hospital AEs were considered, the risk remained significantly associated with age but was no longer associated with admission status, and it was less strongly associated with admitting service. Forster, Clark et al. (2004) conducted a study at the Ottawa Hospital (a multi-site tertiary-care teaching hospital) to determine the incidence, severity, preventability, and ameliorability of AEs in patients discharged from the general internal medicine service, during a 14-week interval in 2002. The definition of a medical A E in this study was adverse outcomes that were caused by medical care. After discharge, 23% (95% CI [19%, 28%]) of patients experienced at least 1 A E ; the A E severity ranged from symptoms only (68% of the AEs) or symptoms associated with a non-permanent disability (25%), to permanent disability (3%) or death (3%); 12% (95% CI [9%, 16%]) of AEs were considered either preventable or ameliorable; and most AEs were due to adverse drug events (72%), therapeutic errors (16%) and nosocomial infections (11%). Wanzel et al. (2000) prospectively monitored complications for all patients admitted to the general surgery service at the Wellesley Central Hospital (Toronto), over a 2-month period in 1996. They were interested in determining the incidence and nature of complications on a general surgery service, and comparing these results with pre-existing institutional recording and reporting methods. They found that 39% of the 192 patients Chapter 2. Literature Review 29 suffered a total of 144 complications with 1% of complications being fatal, 7% life threatening, 63% of moderate severity, and 29% trivial. Approximately 18% of the 144 complications were considered potentially attributable to error. Seventy eight percent of the complications occurred during or after a surgical operation and were related either directly or indirectly to the surgical operation. Lastly, 6% of complications were not documented in the progress notes of the patients' charts. The third important issue in A E research involves the financial expense and time required to conduct an investigation of A E incidence using these commonly used methods. Recently, several methodological studies were published that explored the use of computerized database screening to detect medical AEs. The interest in using computerized database screening has been increasing because it is inexpensive, rapid and it can be applied, for example, to screen medical billings and patient records in hospitals. Section 2.3 presents a review of such studies. 2.3 Validity and Reliability of the ICD9 E-Codes in Identifying External Causes of Injury The success of the proposed statistical framework depends on the quality of data available for analysis. For this reason, it is imperative to review the injury research literature in which the validity and reliability of ICD9 E-codes have been assessed. ICD9 E-codes (see Table 1.1) have occasionally been used in studies conducted at the population level to identify all external causes of injury, excluding medical adverse events (AEs). However, studies done at Chapter 2. Literature Review 30 the population level to identify injuries due to AEs using ICD9 E-codes appear to be very rare. For clarity, the first part of this section of the study concentrates on studies excluding AEs, and the second part focus on studies specifically examining AEs. In section 2.3.2 we review the few studies in which computerized database screening was employed to examine outcomes related to adverse events. 2.3.1 All Causes of Injury Excluding Medical Adverse Events The number of studies using computerized screening in conjunction with the International Classification of Diseases (ICD) coding system as a method to determine the incidence of selected injuries (excluding AEs) has increased over the years (Barss et al. 1998; Christoffel and Gallagher 1999). Despite the rise in popularity of this methodological approach, investigations assessing the quality of data in these types of studies have been limited. Studies have been conducted in several countries, including Canada, Australia and the United States, that assessed the reliability of injury ICD9 E-codes (excluding AEs) in death certificates and in hospital separation records using the kappa statistic (Landis and Koch 1977). The kappa statistic in these studies varied from 0.40 (fair agreement) to 0.98 (almost perfect agreement) (LeMier, Cummings, and West 2001; Ribbeck et al. 1992; Owen, Bolenbaucher, and Moore 1999; Maclntyre, Ackland, and Chandraraj 1997; Schwartz et al. 1995; Bota, Therrien, and Rowe 1997; Lu, Lee, and Chou 2000; to name a few). For example, Bota, Therrien, and Rowe (1997) assessed the reliability of cause-of-injury information using the ICD9 E-code system to track injuries in an emergency department in Chapter 2. Literature Review 31 Ontario (Canada) using codes assigned by several different health professionals (nurses, physicians and nosologists). The values for the kappa statistic in this study showed almost perfect agreement and ranged from 0.88 to 0.92, depending on the examining health professional. As mentioned previously, the kappa statistic varied substantially from one reliability study to another. The main differences between these studies included: the familiarity and experience of coders; how committed the study specific hospital institutions and vital statistics agencies were in maintaining the quality of ICD9 E-code systems used in database production; and most importantly, how detailed these E-codes were. The use of more detailed E-codes was related to lower reliability (e.g., detailed E-codes such as E810.1 as opposed to non-specific E-codes as E810). In our study, however, we are only working with non-specific ICD9 E-codes, with the exception of a small proportion of records using complete E-codes to identify the late effects of injury (E929), injuries resulted from striking against or struck accidentally by objects or persons (E917), and injuries caused by cutting and piercing instruments or objects (E920). In several other studies, the validity of using the injury ICD9 E-codes (excluding AEs) in hospital separation records or in death certificates was assessed (Maclntyre, Ackland, and Chandrafaj 1997; Winn, Agran, and Anderson 1995; Nelson et al. 1993; LeMier, Cummings, and West 2001; Sniezek, Finklea, and Graitcer 1989; to name a few). These studies indicated that non-specific E-codes provided high sensitivity and positive predicted value. In LeMier, Cummings, and West (2001), the authors assessed the validity of using ICD9 E-codes to ascertain the mechanism of injury in hospital discharge records. For non-specific E-codes, the sensitivity ranged from 91.6% (for firearm injuries) to 98.9% (for Chapter 2. Literature Review 32 poisoning injuries), and the positive predicted value ranged from 87.6% (for motor vehicle accident) to 94.2% (for poisoning injuries). Moyer, Boyle, and Pollock (1989) studied the validity of injury ICD9 E-codes for motor vehicle crash deaths, suicides and homicides in death certificates. The sensitivity and specificity for non-specific E-codes were both higher than 90%. 2.3.2 Injuries Due To Medical Adverse Events Recently, alternative approaches combining computerized database screening and the ICD coding system have been used in studies to identify injuries due to medical adverse events. These studies have mainly focused on using ICD N-codes {i.e., ICD9 codes describing the actual bodily harm caused by the external cause of injury) in contrast to using ICD E-codes and we could find very few studies in which the ICD9 E-codes were used to identify medical adverse events (Karson and Bates 1999; Agency for Healthcare Research and Quality (AHRQ) 2003; Kalish et al. 1995; Zhan and Miller 2003a,b; to name a few). In the study by Samore et al. (2004), the definition of an adverse event was linked to "any patient harm caused by device-related medical or surgical management rather than the patient's illness." (Samore et al. 2004, 326) and was referred to as an A M D E . One of the methods used to detect a medical A E was the use of ICD9-CM E-codes. The ICD9-CM E-codes used were E996.0-E996.7, E878.1, E878.2 and E879.4 (a subset of codes defining a medical A E in this dissertation. See page 6). The authors screened all records for at least one of these ICD9-CM E-codes, and then randomly sampled 141 of these records and found that 72% had a confirmed A M D E (i.e., the presence of 1 or more of these ICD9 E-codes). In the Chapter 2. Literature Review 33 records that failed to confirm an A M D E , a procedure-associated A E that was not device related was identified. This study demonstrated the feasibility of using ICD9-CM E-codes for surveillance in terms of identifying this type of A E and suggests that the method be further explored as a means of increasing patient safety. In the paper by Hunter and Bains (1999), the authors were interested in answering the question "has the quality of health care been compromised by system-wide changes in Ontario's health care system?" (Hunter and Bains 1999, 1). To accomplish this goal, they quantified the rates of AEs among Ontario residents admitted to hospital or undergoing day surgery from 1992 to 1997 using data processed by the Canadian Institute for Health Information (CLHI) and maintained by the Ontario Health Planning Database. The definition of adverse events used in this paper was the same as the one used in this study (i.e., A E was defined as an adverse event, a postoperative complication, or a medical/surgical misadventure defined by the ICD9 E-codes E870-E876, E878-E879 and E930-E949). They found that the rates of AE ' s in the hospital admitted and day surgery populations increased from 1992 to 1997. The rates were higher among the hospital admitted population. More specifically, from 1992 to 1997 the A E rates in the hospital admitted population due to misadventures (E870-E876) increased from 0.18 to 0.30 per 100 admissions, complications (E878-E879) increased from 3.30 to 5.00 per 100 admissions, and adverse drug reactions (E930-E949) increased from 1.04 to 1.62 per 100 admissions. The authors concluded that either there was no substantial impact on the quality of care as measured by the rate of adverse events or coding practices may have changed. However, there was a worrying increase in the trend in hospital complication rates, and they suggested that further study was necessary to understand these trends. Chapter 2. Literature Review 34 Next we review several studies that investigated the validity and reliability of using of these computerized screening systems to flag injuries due to medical adverse events mentioned in the AHRQ's report on patient safety indicators (AHRQ 2003). The Agency for Healthcare Research arid Quality (AHRQ) recently developed and released patient safety indicators (PSIs) that can be derived from hospital inpatient discharge data. According to the AHRQ, these indicators represent an important new tool available to researchers conducting research aimed at improving patient safety (AHRQ 2003). As Zhan and Miller (2003a) mentioned, there have been few studies that used PSIs to screen for patient safety problems (Iezzoni et al. 1992; Miller et al. 2001). The PSIs developed by A H R Q consist of 20 hospital-level and 6 area-level indicators, and five of these indicators use ICD9-CM E-codes in combination with N-codes as part of their formulation (see Table 2.2). Even though patient safety indicators are a positive step towards the broader use of hospital inpatient discharge data to flag patient safety problems and monitor the quality of medical care, there are still problems regarding the validity and reliability of these indicators in terms of detecting AEs. In particular, the PSIs that use E- and N-codes have been shown to have substantial reliability (average kappa was 0.67), but low construct and coding validity (AHRQ 2003). We believe that the lack of validity and reliability of these PSIs is because they are based on both N - and E-codes instead of just E-codes. The N-codes in this situation offer a very broad characterization of the A E , whereas E-codes, which are very specific, can offer a narrower characterization of the same problem (see Table 2.3 for an example of how unnecessary mixing E- and N-codes is in identifying AEs). Chapter 2. Literature Review 35 Table 2.2. List of Patient Safety Indicators (AHRQ 2003). Patient Safety Indicator (PSI) Hospital- Area- Use of ICD9-CM Summary level level Accidental puncture or laceration Yes Yes Yes Required for numerator and denominator definitions This indicator is intended to flag cases of complications that arise due to technical difficulties in medical care -specifically, those involving an accidental puncture or laceration. 2 Postoperative pulmonary embolism (PE) or deep vein thrombosis (DVT) Yes 3 Complications of anesthesia Yes Yes Required for numerator definition This indicator is intended to capture cases flagged by E-codes and complications codes for adverse effects from the administration of therapeutic drugs, as well as the overdose of anesthetic agents used primarily in therapeutic settings. 4 Postoperative respiratory failure Yes 5 Death in low-mortality diagnosis-related groups (DRGs) Yes 6 Postoperative sepsis Yes 7 Decubitus ulcer Yes 8 Postoperative wound dehiscence Yes Yes 9 Failure to rescue Yes 10 Selected infections due to medical care Yes Yes Foreign body left during procedure Yes Yes Yes Required for numerator definition This indicator is intended to flag cases of a foreign body accidentally left in a patient during a procedure. This indicator is defined on both a hospital level (by restricting cases to those flagged by a secondary diagnosis or procedure code) and an area level (by including all cases). 12 Transfusion reaction Yes Yes Yes Required for numerator definition This indicator is intended to flag cases of major reactions due to transfusions (ABO and Rh). This indicator is defined both on a hospital level (by including cases based on secondary diagnosis associated with the same hospitalization) and on an area level (by including all cases of transfusion reactions). Chapter 2. Literature Review 36 Table 2.2. (continued) List of Patient Safety Indicators (AHRQ 2003). P a l ien ( Sa fe ty I n d i c a t o r (J'SI) H o s p i l i i l -k-vcl A r e a -level I s e o f I C D 9 - C M S u m m a r y 13 Iatrogenic pneumothorax Yes Yes 14 Birth trauma—injury to neonate Yes 15 Postoperative hemorrhage or hematoma Yes 16 Obstetric trauma—Cesarean delivery Yes 17 Postoperative hip fracture Yes Yes Required as an exclusion criteria in the denominator This indicator is intended to capture cases of in-hospital fracture specifically, hip fractures. This indicator limits diagnosis codes to secondary diagnosis codes to eliminate fractures that were present on admission. 18 Obstetric trauma—vaginal delivery with instrument Yes 19 Postoperative physiologic and metabolic derangement Yes 20 Obstetric trauma—vaginal delivery without instrument Yes 37 Table 2.3. Set of ICD9-CM E- and N-codes Defining the Patient Safety Indicator "Foreign Body Left During Procedure". (AHRQ 2003, pages 98-99) Foreign Body Left During Procedure Numerator: Discharges with ICD-9-CM codes for foreign body left in during procedure in any secondary diagnosis field per 1,000 surgical discharges. ICD-9-CM diagnosis codes (ICD9-CMN-codes) 998.4 Foreign body accidentally left during a procedure 998.7 Acute reactions to foreign substance accidentally left during a procedure ICD-9-CM external codes (ICD9-CME-codes) Foreign body left in during: E871.0 Surgical operation E871.1 Infusion or transfusion E871.2 Kidney dialysis or other perfusion E871.3 Injection or vaccination E871.4 Endoscopic examination E871.5 Aspiration of fluid or tissue, puncture, and catheterization E871.6 Heart catheterization E871.7 Removal of catheter or packing E871.8 Other specified procedures E871.9 Unspecified procedure Denominator: A l l medical and surgical discharges defined by specific diagnosis related groups (DRGs) (see denominators for decubitus ulcer for medical discharges and complications of anaesthesia for surgical discharges). 38 Zhan and Miller (2003a) conducted a literature review to identify studies oh patient safety research that used administrative data, the authors concluded that since 1970 few studies have used administrative data in patient safety research. For example: a number of studies examined small-area variation in health care and clinical practice patterns; some studies in outcome research were carried out using Medicare data; and some studies reported using administrative data to assess quality and patient safety. More recently, these authors reported that investigators had also utilized patient safety indicators (PSIs) developed by the Agency for Healthcare Research and Quality (AHRQ). In a separate paper, Zhan and Miller (2003b) used patient safety indicators (PSIs) and administrative data to assess excessive length of stay (LOS), charges, and deaths attributed to medical injuries during hospitalization. In their study, they used PSIs to identify medical injuries in 7.45 million hospital discharge abstracts from 994 acute-care hospitals across 28 states in the year 2000. They identified a large number of cases that they considered likely to be medical injuries resulting from failures in the process of care in hospitals and found that the rates for the PSIs were relatively lower than typically reported in other studies. They concluded that PSIs represent a promising tool for evaluating patient safety, however, additional analyses are required to fully understand the potential benefits and problems associated with the use of these indicators. 39 C h a p t e r 3 M e t h o d s In this chapter we: (1) Describe the data analysed in the demonstration of the analytical methodology presented in this study; (2) Describe the spatio-temporal hierarchical statistical model and Bayesian disease mapping theory underlying the data analysis; (3) Review the statistical assumptions associated with the small-area analysis, including assumptions used in estimating disability-adjusted life years (DALYs); and (4) Document the procedure developed to incorporate spatio-temporal Bayesian disease mapping methods into the calculation of disability-adjusted life years (DALYs). 3.1 D a t a To calculate the health status indicators shown in Figure 1.1, injury mortality data, including population estimates, were obtained from the British Columbia (BC) Vital Statistics Agency. In addition, injury hospitalization data and number of admissions were provided by the University of British Columbia's Centre for Health Services and Policy Research (CHSPR). Chapter 3. Methods 40 3.1.1 Definition of Study Regions The analyses presented in this study were conducted using the following geographical boundaries defined by the BC Ministry of Health: Sixteen health service delivery areas (HSDA) nested within five health authority areas (HA) (see Figure 3.1). The level of health authority (HA) represents the primary layer in the provincial health care system that is responsible for "identifying regional health needs; planning appropriate programs and services, and ensuring programs and services are properly funded and managed" (Government of British Columbia 2004). The B C geographical health authorities are the: Interior health authority, Fraser health authority, Vancouver Coastal health authority, Vancouver Island health authority and Northern health authority. The next layer in the provincial health administration system is the health service delivery area (HSDA). The HSDAs are responsible for "managing the delivery of health services in their respective areas, meeting performance objectives set by health authorities, ensuring community input into health service planning and evaluation, and evaluating health status and plan performance" (Government of British Columbia 2004). These geographically-based HSDAs were used to define place of residence of deceased or hospitalized individuals. It is important to recognize that in some areas of British Columbia in which there are not many hospitals or health care professionals qualified to perform certain specialized procedures, some patients travel to Vancouver (in the Vancouver Coastal health authority) or Alberta (neighbouring province to the east of British Columbia) for treatment. The immediate consequence is that the place of residence may not be the same as the place of care (or place of injury for that matter). If the analyses presented in this study Chapter 3. Methods 41 were re-run using place of injury or care, then it is possible that the values and interpretation of the health status indicators would be different from the analyses presented in this study. Since place of residence was used to group individuals into levels of health administration areas, the health status indicators in this study should be interpreted as the impact of injury on the H A or HSD A of residence. Chapter 3. Methods 42 Health Service Delivery Areas 51 Northwest 53 Northeast North Shore/Coast Garibaldi North Vancouver Island Thompson Cariboo Shuswap 14 Fraser East M 13 Health Authorities ; Fraser i Interior i Northern | Vancouver Coastal I Vancouver Island Koolenay/Boundary East Kootenay Central Vancouver Island South Vancouver Island 12 Okanagan 0 100 200 400 Kilometers I I I l I l I 1 I inset 1 0 5 10 20 Kilometers Health Authority Health Service Delivery Area 01 Interior 11 East Kootenay 12 Kootenay Boundary 13 Okanagan 14 Thompson Cariboo 21 Fraser East 22 Fraser North 23 Fraser South 03 Vancouver Coastal 31 Richmond 32 Vancouver 33 North Shore/Coast Garibaldi 41 South Vancouver Island 42 Central Vancouver Island 43 North Vancouver Island 51 Northwest 52 Northern Interior 53 Northeast 02 Fraser 04 Vancouver Island 05 Northern Figure 3.1. Map of the Health Service Delivery Areas and Health Authorities of British Columbia. Chapter 3. Methods 43 3.1.2 Definition of Study Population Hospitalization Data. Individuals were included in this investigation i f they were hospitalized in BC during the period 1991-2000, and had a hospitalization diagnosis included in the ICD9-CM E-codes classification system listed in Table 1.1. The data on hospitalization included all hospital discharge data from any hospital in BC whose level of care was either acute care or day surgery. The data obtained for hospitalization consisted mainly of BC residents (the average percentage of residents was 99.9% in the 10 years of study). The remaining patients (0.1%) included in the hospital data were people not registered to the Medical Services Plan of BC (MSP) and visitors to BC that were hospitalized during 1991-2000. Mortality Data. This set of data included any person who died during the period 1991-2000, and had as cause of death any of the ICD9-CM E-codes listed in Table 1.1. The data obtained for mortality only included residents of British Columbia in the period 1991-2000. Population Data. To calculate the annual incidence rates for all injuries (excluding medical AEs), we used the annual BC population size estimates during 1991-2000 as a denominator. Figure 3.2 maps the mid-period population sizes of each HSD A during 1991-2000. Most of the areas in the southwest of BC are those with the population higher than the median population. Chapter 3. Methods 44 Percentiles Figure 3.2. Mid-Period Population Size by Health Service Delivery Areas during 1991-2000. Males and Females - A l l Age Groups. Admission Data. The number of admissions to any hospital in BC was used as the denominator in calculations of incidence rates of injuries due to medical AEs, whose level of care was either acute care or day surgery, in the period 1991-2000. Figure 3.3 maps the mid-period number of admissions of each HSD A during 1991-2000. Most of the areas in the central and southwest of BC are those with the number of admissions higher than the median number of admissions. Chapter 3. Methods 45 Percentiles • < 2.5th [12,611- 12,626) |—1 2 5th - 50th [12,626 - 44,814) mm 50th - 97 5th [44,814 - 94.143) MM >= 97.5 [94,143-96,891] Mean: 47,859 Standard Deviation. 28.546 Minimum: 12,611 Maximum: 96,891 Figure 3.3. Mid-Period Number of Admissions by Health Service Delivery Areas during 1991-2000. Males and Females - A l l Age Groups. Case or Event Definition. The unit of analysis was person incident (or event). The definition for an injury was an event that led to death or that was non-fatal but severe enough to require hospital treatment. In order to organize the hospitalization data we adopted the following rules to define different events and the date of each event: • If a patient at different admission dates had the same ICD9-CM code in his/her chart, we considered these codes as separate events. This rule allows for multiple events per person. Chapter 3. Methods 46 • If a patient dvrring the same admission had the same ICD9-CM code repeated in his/her chart, we counted as one event the first occurrence of these codes. • If a patient was transferred between hospitals, the admission date was the date of first admission. Age Groups. We performed the analysis separately for males and females. In addition, because we did not have sufficient cases for certain age groups in the hospitalization and mortality data, we used the overall BC mortality rate by age group for all external causes of injury combined to decide on the best way to create appropriate age groups (Figure 3.4). Given that we need a sufficient number of deaths and hospitalizations in each injury category in order to use the proposed methodology, and based on the distribution of observed mortality rates by age group as shown in Figure 3.4, we decided that the new age groupings would be 0-19, 20-39, 40-64, 65-79 and 80 or more years of age. Chapter 3. Methods 47 Males Females 60 50 40 o o o o" i_ a Q) O CL ^ -5 30 TO a. !r o Q- 20 ro o 10 0 -I <>•* •3- 0" T-t C> ^ cn o LO cr> TJ- cn CM CN CO CO i i i i O LO O LO CM CN CO 00 T i i i i i i i O L O O L O O L O O L O -I •vf + 00 LO - i oo o w 00 age group Figure 3.4. The Distribution of the Ten-Year Conventional Mortality Rate at the Provincial Level, by Gender and Age Group, and for A l l External Causes of Injury Combined. Chapter 3. Methods 48 3.1.3 Categories of External Cause and Nature of Injury The International Classification of Diseases, Ninth Revision, Clinical Modification, External coding scheme (i.e., ICD9-CM E-codes) was used in the hospital separation data for all years during the period 1991-2000, and mortality data for all years of study except the year 2000. For the year 2000, the ICD tenth revision coding system was used by the Vital Statistics Agency of British Columbia. The translation from ICD 10 into ICD9 was provided by the Vital Statistics Agency of British Columbia. The specific external causes of injury considered in the analyses at the provincial and H A levels are listed in Table 1.1. A detailed explanation of the ICD9-CM E-codes in Table 1.1 can be found in the Glossary (see Appendix B). The Category of Other Unintentional Injuries (Excluding Medical AE) in Table 1.1. An explanation of the unintentional injuries included in this category can be found in the Glossary (see Appendix B). In this study, we ran the analyses at the provincial and H A levels including other unintentional injuries (excluding medical AE). However, it is difficult to interpret the results of this category of injury because it involves an array of different unintentional injuries. E-Code Categories E850-E858. The ICD9 E-codes E850-E858 are E-codes classified in Table 1.1 as unintentional poisoning. These E-codes are defined by accidental poisoning by drugs, medicinal substances, and biologjcals. It includes accidental drug overdose; wrong drug given or taken in error; drug taken unintentionally; and accidents in the use of drugs and biologicals in medical and surgical procedures (see Appendix B for more detail). These E-codes were excluded from the Chapter 3. Methods 49 category injury due to adverse effects of drugs in therapeutic use mainly because the difficulty in identifying, for example, whether the accidental poisoning was caused by a wrong prescribed dosage of the substance or not. The frequency of E-codes in this category in both data sets (mortality and hospitalization) can be found in Table G. 1 in Appendix G. Nature of Injury. Nature of injury is defined by ICD9-CM N-codes describing the actual bodily harm caused by the external cause of injury, such as an intracranial injury. The ICD9-C M N-codes for nature of injuries are presented in Table 3.1. Table 3.1. List of ICD9-CM N-Codes for Nature of Injury (Murray and Lopez 1996). Injury Nature ICD9-CM N-Codes Fractured skull 800 - 801 Fractured face bones 802 Fractured vertebral column 805 Injured spinal cord 806 and 952 Fractured rib or sternum 807 Fractured pelvis 808 Fractured clavicle, scapula or humerus 810-812 Fractured radius or ulna 813 Fractured hand bones 814-817 Fractured femur 820 - 821 Fractured patella, tibia or fibula 822 - 823 Fractured ankle 824 Fractured foot bones 825 - 826 Other dislocation 830, 833 - 834, 836 - 839 Dislocated shoulder, elbow or hip 831,832, 835 Sprains 840 - 848 Injury Nature Ii D9-CM N-Codes Intracranial injury 850 - 854 Internal injuries 860 - 869 Open wound 870, 872 - 884, 890 - 894 Injury to eyes 871,950 Amputated thumb 885 Amputated finger 886 Amputated arm 887 Amputated toe 895 Amputated foot 896, 897.0-897.1 Amputated leg 897.2 - 897.3 Crushing 925 - 929 Burns < 20% 940-947, 948.0-948.1 Burns > 20% and < 60% 948.2- 948.5 Burns > 60% 948.6-948.9 Injured nerves 951,953 -957 Poisoning 960 - 979, 980 - 989 Residual 900 - 924, 930 - 939 Following the recommendations of Murray and Lopez (1996), a number of other N -codes for ill-defined or minor categories have been distributed across the nature of injury categories in Table 3.1 according to the following rules: Chapter 3. Methods 50 (1) The N-codes 803 (other and unqualified skull fractures) and 804 (multiple fractures involving skull or face with other bones) should be proportionately redistributed across the N-codes 801 and 802; (2) The N-code 809 (ill-defined fractures of trunk) should be proportionately redistributed across the N-codes 807 and 808; (3) The N-codes 818 (ill-defined fractures of upper limb) and 819 (multiple fractures involving both upper limbs, and upper limb with rib(s) and sternum) should be proportionately redistributed across the N-codes 810-817; (4) The N-codes 827 (other, multiple and ill-defined fractures of lower limb) and 828 (multiple fractures involving both lower limbs, lower with upper limb, and lower limb(s) with rib(s) and sternum) should be proportionately redistributed across the N -codes 822 - 826; (5) The N-codes 897.4 - 897.7 (traumatic amputation of leg(s) - unilateral, level not specified, and bilateral (any level)) should be proportionately redistributed across the N-codes 895, 896 and 897.0 - 897.3; (6) The N-code 949 (burn, unspecified) should be proportionately redistributed across the N-codes 940-948; and (7) The N-codes 958 - 959 (certain traumatic complications and unspecified injuries) and 990 - 999 (other and unspecified effects of external causes) should be proportionately redistributed across all N-code categories after the previous steps. Chapter 3. Methods 51 3.2 Bayesian Disease Mapping The analysis of geographical variation in disease risk has been one of the main roles of epidemiologists engaged in public health surveillance. In several studies to date, the ultimate goal of the analyses has been to produce disease maps to: (1) identify areas of high and low disease risk; (2) detect spatial clusters of disease incidence; and (3) provide etiological clues based on the distribution of disease risk. For example, Figure 3.5 contains a typical example of disease map that can be found in the literature illustrating the spatial distribution of disease incidence. 22 Figure 3.5. Ten-Year Conventional Mortality Rate per 10,000 population Attributable To Road Traffic Accidents. Males, 20-39 Years, 1991-2000. Chapter 3. Methods 52 Two important limitations are associated with the map displayed in Figure 3.5. First, the map is based on the conventional mortality rate. Although many studies have used conventional rates and ratios to produce maps such as the one in Figure 3.5, mapping conventional measures of risk can lead to erroneous conclusions, especially in the case of rare diseases and a small population at risk. In such situations, it becomes difficult to distinguish whether differences observed among the spatial units are due to chance, or whether they are indeed true differences. Second, this map shows the conventional mortality rate spatial pattern for all years combined during the period 1991-2000. Consequently, the information on disease risk that could have been obtained annually is lost. Historically, in most disease mapping studies, the approach used to model temporal spatial data has been to either analyse the data cross-sectionally, or to initially pull the temporal data together, and then proceed with the analysis. Consider now, examining over time the same data presented in Figure 3.5. More specifically, Figure 3.6 illustrates the temporal variability in the conventional mortality rate, along with other temporal data on road traffic accidents that occurred between 1991 and 2000 in 3 health services delivery areas in B C - Northeast (with a low population at risk), Simon Fraser (with a median population at risk) and Vancouver (with a high population at risk). Chapter 3. Methods 53 Population at risk Number of road traffic accidents deaths Conventional mortality rate Northeast (53) 1992 1996 2000 1992 1996 2000 1992 1996 2000 year year year Vancouver(32) o o 1992 1996 2000 1992 1996 2000 1992 1996 2000 year year year Figure 3.6. Annual number of deaths, estimated population and conventional mortality rate per 10,000 population attributable to road traffic accidents. Males, 20-39 years, 1991-2000. In Figure 3.6, note that the scales of the population trend and conventional mortality rate plots are very different for small, medium and large regions. There appears to be greater variation in the conventional mortality rate in the HSDA with a low population at risk (Northeast); to see the variability for each HSDA'rate, look at the minimum and maximum values for the rate. Additionally, when we compare the sizes of the population at risk Chapter 3. Methods 54 between the three HSD As, it appears that as the population at risk increases, the variability associated with the mortality rate decreases. As in the spatial context, the annual conventional rates are also subject to chance variation. To conclude, the mortality trends of the geographical units described in Figure 3.6 appear to be quite different. This difference is due to a subtle source of variation induced by unobserved risk factors, called systematic variation, which influences the spatio-temporal pattern of injury incidence among geographical units. In view of the previous modelling concerns, to analyse spatio-temporal data appropriately, it is important to apply a disease mapping methodology that simultaneously accounts for the spatial variability, temporal variability and the interaction between space and time in the statistical model. To date there have been a small number of developments in the spatio-temporal modeling of disease rates, such as the models shown in Bernardinelli et al. (1995), Knorr-Held and Besag (1998), Sun et al. (2000), Bohning, Dietz and Schlattaman (2000), and MacNab (2003a,b,2004). It is also important to chose a disease mapping methodology that is capable of separately modeling the systematic variability induced by unmeasured covariates (or risk factors) not included in the analysis, and the chance variability produced by rare events in regions with a small population at risk. In this study, the chosen spatio-temporal model is based on the Bayesian hierarchical model presented in MacNab (2003a), where the systematic and chance variations are modelled via random spatial and temporal effects. Chapter 3. Methods 55 3.2.1 Spatio-Temporal Modeling under a Full Bayesian Approach The model described in this section was developed by MacNab and Dean (2001) in a small-area study of infant mortality rates in the province of British Columbia, Canada. The model was later extended by MacNab (2003a) to adjust for potential age confounder. In this thesis, the spatio-temporal modelling was carried out separately for each gender x age group combination. Let Yit represent the number of (injured or diseased) cases in HSDA i (/=1,...,16) and year t (t=l991,...,2000). We assume that conditioned on the random effects the number of (injured or diseased) cases has a Poisson distribution with intensity given by p.it. Thus, in the first level of the Bayesian hierarchical (mixed effects) Poisson model, one assumes that given the vector of random effects b, 7,/s are (conditionally) independent, i.e., Yit\b ~ Poisson , with fiit representing the expectancy of Yit given the vector of random effects b. To analyse the incidence rates, one assumes: \og{pit)^\og(nit) + \og{m) + S0{t-t) + (Pi +RSi(t-t) (3.1) 1 0 where t = 7 1 is the average time; log(« 1 () is an offset, i.e., a covariate having a known coefficient; m is a fixed effect representing the provincial mean rate; 4 So([-t) = ^<2o,A(/ ~*)' where aok (AF=1,2,3,4) represent fixed spline coefficients, is a "global" fixed cubic B-spline representing the provincial rate trend; cpi is a random spatial effect that allows for spatially structured variation in rates that arise from unobserved risk Chapter 3. Methods 56 factors or latent effects; and RS((t-t) = ^ PikBk(t-t) , where f3ik (7=1,...,16 and k=\ &=1,2,3,4) , is a family of random cubic B-splines representing the temporal trend for the ith H S D A relative to the "global" spline. Here {Bk (t - t)}Ak=1 represents a set of basis functions of an order three B-spline basis function generated without intercept, and with one inner knot at the mid-period (MacNab and Dean 2001). The motivation for using splines to model the B C and the local temporal trends is the possibility of modelling the non-linear temporal effects without making strong assumptions about the parametric form of the trends (MacNab and Dean 2001; MacNab 2003a). Furthermore, by modelling the local trends using a family of random splines, we allow the shape of the temporal trends to vary among different HSD As. As a result, we gain more precision and stability when estimating each region's rate over time because the model pools information across time and space to fit local trends. In this study, we assume independence priors for all random effects: (i) <p = (<Pi,...,<pl6)' ~MVN(o,o-pI]6), z'=l,..., 16, 7i6 is a 16x16 identity matrix, and cr^  represents the overall variability/dispersion between regions; (ii) /?A ~A^(o,cr^/]6), k=l,2,3,4, and/i6 is a 16x16 identity matrix; and (iii) p =covariance (/3ik, (pt) (i.e., since we assume that (pt and f3i k are not independent) In MacNab (2003a), inferences for the spatio-temporal model (3.1) were made via an empirical Bayesian approach using the penalized quasi-likelihood (PQL) estimation method (Breslow and Clayton 1993; MacNab and Dean 2000). While the empirical Bayesian approach appropriately estimates the model parameters in a timely basis, this approach does not provide precise estimation of the uncertainty of these parameters, requiring the use of Chapter 3. Methods 57 methods such as a parametric bootstrap for proper inference on random effects (MacNab et al. 2004). In this study, inferences for model (3.1) are based on a full Bayesian approach using Markov Chain Monte Carlo (MCMC) simulation procedures via Gibbs sampling algorithms (Beraardinelli and Montomoli 1992; Gelman et al. 1998; Carlin and Louis 2000, MacNab 2003b; MacNab et al. 2004). Additionally, goodness-of-fit plots are used to assess how well the systematic and random variations were modelled (see Appendix G). For full Bayesian analysis the following hyperpriors that are commonly seen in disease mapping literature are used in this study (MacNab et al. 2004): aok ~N(0,0.000l),k =1,2,3,4 m ~ Uniform{0, l) p ~ Uniform^ 1, l) T = \ ~ Gamma(0.0001,0.0001), and r k = \ ~ Gamma(0.0001,0.0001), k =1,2,3,4. °~k In this study, therefore, the following important quantities are obtained from the full Bayesian analysis of the spatio-temporal model (3.1): (i) The estimated non-linear trend of the provincial level rate, exp(log(/w) + S0 (t)); and (ii) The estimated non-linear trend for the ratio between each HSDA's rate and the provincial rate, exp(#>. + RS;(t)). In summary, there are several advantages associated with the full Bayesian approach used in this study. In particular, this approach has (i) the advantage of providing both satisfactory estimation of parameters and their uncertainty; (ii) it allows each estimated Chapter 3. Methods 58 parameter in model (3.1), as well as each estimated measure (e.g., rates, ratios and counts) to have a posterior statistical distribution. Inferences for model (3.1) are then based on posterior summary statistics such as the: posterior mean, posterior standard deviation, posterior median, posterior mode and 0.025 and 0.975 posterior quantiles; and (iii) based on the posterior distribution of injury (mortality or hospitalization) counts, it is possible to calculate the posterior distribution of D A L Y s , as is shown later in this chapter. 3.3 Disability-Adjusted Life Years (DALY) The D A L Y methodology extends the concept of potential years of life lost due to premature death to include equivalent years of "healthy" life lost by being in states other than good health. One D A L Y represents one lost year of healthy life. Thus the D A L Y represents the number of years of life lost (YLL) summed with a measure of healthy years lost due to disability (YLD) based on the severity and duration of the sequelae, age and gender of the injured person. The parameters to obtain D A L Y are explained in more detail in the next section. Here, we only present the simplified formulae to calculate Y L D , Y L L and D A L Y . The D A L Y for each external cause of injury is calculated as: DALY= YLD + YLL (3.2) Y L D is the morbidity component of D A L Y , also known as the non-fatal burden component, and the simplified formula to calculate Y L D is given by: Without discoimting rate: YLD - Ix DW x D Chapter 3. Methods 59 With discounting rate: YLD - IxDWx-x\l-e {rxD) (3-3) r where / = number of hospitalizations, DW= disability weights, D = average duration of disability, and r = discount of future health rate. Y L L is the mortality component of D A L Y , or the fatal burden component. The simplified formula to calculate Y L L is given by: where// = number of deaths, L - life expectancy (at age of death), and r = discount of future health rate. For more information on general formulae to calculate Y L D and Y L L please see Begg and Tomijima (2005) or Murray and Lopez (1996). The next section explains the assumptions made to calculate DALYs in this study. Without discounting rate: YLL = N x L (3.4) Chapter 3. Methods 60 3.3.1 Assumptions Made in the Calculation of DALYs To calculate D A L Y s , there are several parametric assumptions that must be made in order to obtain Y L L and Y L D (see Figure 3.7). YLD Y L L Number of hospitalizations Residual category Discounting rate (r) Duration (£>) Comorbidity Disability weight (DW) Age weighting tf>) Discounting rate (r) Life expectancy (L) Number of deaths (AO Age weighting (/?) ' Incidence Parameters Assumptions DALY Figure 3.7. Parameters Intrinsic in the Calculation of Disability-Adjusted Life Years (DALY). Chapter 3. Methods 61 To obtain Y L L values for the discounting rate, life expectancy and age weights must be chosen. For calculating Y L D , values for the discounting rate, age weights, disability weights and durations have to be chosen. In addition to making choices concerning parameters intrinsic to D A L Y calculations, decisions must be made regarding the control for comorbid conditions as well as assumptions about the unknown disability weights and durations for the residual category of injury sequela. The following sections discuss these choices in detail. 3.3.1.1 Discounting Future Health Rate (r) Discounting refers to the practice of valuing the same thing in the future as less valuable than in the present. For a better justification of using discount rates in this study, we should consider several aspects inherent in their use. One aspect is the economic perception involved in discounting. For example, suppose that you have to choose between receiving a guaranteed amount in dollars today or the possibility of the same amount in a year from now, which would you choose? Discounting says that most people would choose to receive the amount today. Suppose that instead of receiving the same amount in one year, you will receive 50% more because of the interest rates, what would you choose? Some people might prefer to receive the amount in one year. There are several factors involved in the decision of an individual who prefers to receive money today as opposed to in the future. These factors include uncertainty of what the future holds in terms of economy; monetary conditions; or the possibility of not being alive or fully capable of utilizing this money in the future. In this study, however, we are not focusing on an individual's discount rates for future health, but Chapter 3. Methods 62 rather on societal discount rates for future health, where the preference affects decisions for a society as a whole as compared to preferences affecting each individual. The subject of discounting is complex and there are several papers describing arguments in favour and against its use (Murray and Lopez 1996; Murray and Acharya 1997; Williams 1999). In this study we decided to include discounting in the calculation of D A L Y s by age, gender, external cause of injury, and region. The next question that must be addressed is how much discounting we should use. It is necessary to find a balance between the sacrifices of future generations and gains of current generations. In this study as well as in others, discounting is best described as an exponential decay function over time. For example, Figure 3.8 illustrates the ' 1 V effect of time on discounting using a discrete form of discovmting function as \ + r where t denotes time in years, and the discounting rates (r) assume the values 3%, 6% and 10%. Ll - O — .52 o -o 1.0 0.8 0.6 0.4 0.2 0.0 V 20 40 60 Year 80 100 r = 3% • r = 6% r= 10% Figure 3.8. Effect of Time, in Years, on the Discrete Discounting Function. Chapter 3. Methods 63 Figure 3.8 demonstrates that a year of healthy life gained in 10 year's time is worth 26% less than one gained now i f we use a discounting rate of 3%. However, i f we adopted a discounting rate of 10%, a year of healthy life gained in 10 year's time is worth 61% less than one gained now. Although researchers in the G B D study used a discount rate of 3%, they also advised other researchers to look at higher values. However, it is important to mention that there is no solid theoretical reasoning to use a 3% discount rate. The effect of discounting on the value of Y L L , Y L D and D A L Y is illustrated in the Appendix F under "Sensitivity Analysis for Discounting of Future Health Rate". The relevance of performing this sensitivity analysis was to emphasize the importance of understanding the impact of each different discounting rate on Y L L , Y L D and D A L Y for injuries in BC. The discount rates used in this sensitivity analysis were 0%, 3%, 6% and 10%. If we use no (r = 0%) or a high value (r = 10%) for discounting, in the first choice, the future is valued exactly as the present, and the second choice says that healthy life gained in future is much less worthy than i f we gained it now. The decision then should be made between a discounting value of 3% or 6%. Although the value of 3% for discount is somewhat arbitrary, it has been extensively used in several economic evaluations and cost-effectiveness analyses. In this study we therefore assumed a discounting of future health rate equal to 3% after considering what value of discounting is conventionally used in the literature (Murray and Acharya 1997; Murray and Lopez 1996; World Bank 1993); and after reviewing the results obtained in the sensitivity analysis presented in Appendix F. Chapter 3. Methods 64 3.3.1.2 Age Weighting (/?) The GBD study weighted a year of healthy Ufe lived by a young adult more highly than a year lived by a young child or an elderly person (Murray and Lopez 1996). Although age weighting and discounting appear similar in nature, the processes are distinct and care should be taken not to confuse them. The development of age-dependent weights is based on the belief that there is a social preference to value life differently depending on age. For example, one justification for valuing the life of a young adult more highly than an older adult is because there are limited resources (e.g., monetary and human resources) associated with the health care system. Or possibly, because society is willing to invest more resources on preventing death at young ages than preventing a death at any later age; or because young adults have been educated and are ready to contribute economically and productively to society for (potentially) a greater amount of time (Murray and Lopez 1996). The implementation of age weighting is one of the most controversial elements in the calculation of DALYs. There are several arguments in favour (e.g., years of productive adult life should receive greater importance) and in opposition to these weights (e.g., on equity and empirical grounds). In Murray et al. (2002) and Anand and Hanson (1997), the authors point out that the methodology to obtain age weights is still in an experimental stage, and that the weights published to date should be treated with caution. Given the current controversy involving the use of age-based weights, it is worth examining how the use of age weights can influence the final results in a burden of disease study. For the GBD study, Murray and Lopez (1996) assumed an arbitrary continuous age-weighting function g(C, age, 0) for the weights at each age of the form: Chapter 3. Methods 65 g{C, age, fi)=Cxagex e^™86 (3.5) where J3 determines the importance of age weights (i.e., the preference of valuing the lives of young adults higher than the lives of children and elderly); and C is an adjustment constant. As pointed out in the hterature, values for fi are purely arbitrary and values of C differ according to different study populations. In the GBD study, f3 was equal to 0.04 and C equal to 0.1658, and it was emphasized that i f the age weighting function changed, the constant C would have to change as well. The general formula for the function (3.5) for varying values for fim {0.03,0.04,0.05} without the influence of Cis shown in Figure 3.9. 14 -, 0 -f 1 1 - i • 1 1—; 1 0 20 40 60 80 100 120 age Figure 3.9. Shape of Age Weighting Function for fi in the Set {0.03,0.04,0.05}. Chapter 3. Methods 66 For example, for /?=0.04, the value of a year of life lived by a 30-year old person is 2.21 times higher than a 5-year old, 1.01 times higher than a 20-year old, 1.12 times higher than a 40-year old, and 1.66 times higher than a 60-year old. However, i f a lower /? is used, say /?=0.03, then the value of an year of life lived by a 30-year old person is 2.83 times higher than a 5-year old, 1.11 times higher than a 20-year old, 1.01 times higher than a 40-year old, and 1.23 higher times than a 60-year old. If age weighting is used, then the formulas (3.3 and 3.4) should be slightly modified, and the D A L Y s should be calculated as the sum of two continuous functions (/ , and f2) involving both Y L D and Y L L and the age weighting function g(C,age,0), such as (see Murray and Lopez (1996) for more details): DALY = fx (g(C, age, /?), YLD) + f2 (g(C, age, /?), YLL) (3.6) Because of g{C,age,0), the shape of the D A L Y function (3.6) for different age weights is very similar to shape of formula (3.5) as shown in Figure 3.9. The main concern involving the use of age weights is which age weight value (ft) should be used in this study context. Until the choice of specific age weights can be based on more solid theoretical grounds, we have elected to adopt uniform age weights across all age groups. 3.3.1.3 Comorbidities Studies to date have not explicitly attempted to adjust disability weights for comorbidities with the exception of the Australian and Victorian Burden of Disease Study (Mathers et al. 2000; Victorian Department of Human Services 1999b). In addition, it is important to mention that no Chapter 3. Methods 67 study has mentioned adjustment for durations of comorbidities. In this study, even after recognizing the importance of controlling for prevalent comorbid conditions, no attempt was made to adjust for comorbidities because the available data did not provide information on prevalent diseases for individuals in the study. However, we did adjust for coexisting comorbidities due to injury in the same individual. The details of how we implemented this procedure is presented in the sections on disabihty weights and durations (see sections 3.3.1.4 and 3.3.1.5). 3.3.1.4 Disability Weights (DW) Population burden due to a particular injury is calculated using disability weights. Disability weights used in D A L Y s are developed via the application of a specific and deliberative form of the person trade-off (PTO) method for measuring social preferences for different health state choices. The final weights used in the G B D study were based on the application of the PTO protocol to a particular reference group of experts convened at the World Health Organization with representatives from each region in the world. The resulting disability weights quantify societal preferences for health states in relation to the societal "ideal" of optimal health defined by this group of experts. Separate weights have been developed for treated and untreated forms of each disability. The disability weights for the G B D study can be found in Murray and Lopez (1996). Two assumptions intrinsic to the calculation of the GBD disability weights had an effect on the results of this study: (1) the weights were stationary over time; and (2) the same disability weight was assigned to everyone who had the same injury sequela, regardless of differences in the level of disability. Table 3.2 presents the disability weights for nature of Chapter 3. Methods 68 injury for treated cases as every individual in the hospitalization data was treated. Also in this table, it should be noted that for certain sequelae there are weights for short-term and long-term disabilities. Table 3.2. Disability Weights by Nature of Injury and Age Category (Murray and Lopez 1996). Nature of injury ICD9-CM N-Codcs Disability weights Treated G B D assumptions Fractured skull - short term 8uu-801 0.431 85% of incident cases Fractured skull - life long 800-801 0.410 For ages 60+ the D W is 0.404. 15% of incident cases Fractured face bones 802 0.223 100% of incident cases Fractured vertebral column -short term 805 0.266 100% of incident cases Injured spinal cord - life long 806 and 952 0.725 100% of incident cases Fractured rib or sternum -short-term 807 0.199 100% of incident cases Fractured pelvis - short-term 808 0.247 100% of incident cases Fractured clavicle, scapula or humerus - short-term 810-812 0.153 For ages greater than 15 years theDW is 0.136 100% of incident cases Fractured radius or ulna -short-term 813 0.180 100% of incident cases Fractured hand bones 814-817 0.100 100%) of incident cases Fractured femur - short-term 820-821 0.372 95% of incident cases Fractured femur - life long 820-821 0.272 5% of incident cases Fractured patella, tibia or fibula - short-term 822 - 823 0.271 100% of incident cases Fractured ankle - short-term 824 0.196 100% of incident cases Fractured foot bones 825 - 826 0.077 100%) of incident cases Other dislocation 830, 833 - 834, 836 - 839 0.074 100% of incident cases Dislocated shoulder, elbow or hip - short-term 831,832, 835 0.074 100% of incident cases Sprains 840 - 848 0.064 100% of incident cases Intracranial injury - short-term 8 5 0 - 854 0.359 95% of incident cases Intracranial injury - life long 8 5 0 - 854 0.350 For ages 60+ the D W is 0.404. 5% of incident cases. Internal injuries - short-term 860 - 869 0.208 100% of incident cases Open wound 870, 872 - 884, 890 - 894 0.108 100%) of incident cases Injury to eyes - life long 871,950 0.301 For ages 5-14 D W is 0.300 and for ages 15+ the D W is 0.298. It 100% of incident cases Amputated thumb - life long 885 0.165 100% of incident cases Amputated finger - life long 886 0.102 100% of incident cases Amputated arm - life long 887 0.257 100% of incident cases Amputated toe - life long 895 0.102 100% of incident cases Amputated foot - life long " 896, 897.0-897.1 0.300 100% of incident cases Chapter 3. Methods 69 Table 3.2. (continued) Disability Weights by Nature of Injury and Age Category (Murray and Lopez 1996). Nature of injury |( iv>-r\1 N-Codes Disability weiuhls Trailed G B D assumptions Amputated leg - life long 897.2 0.300 100% of incident cases Crushing - short-term 925 - 929 0.218 100% of incident cases Burns < 20% - short-term 940-947, 948.0-948.1 0.158 0% of incident cases Burns < 20% - lifelong 940-947, 948.0-948.1 0.001 100% of incident cases Burns > 20% and < 60% - short term 948.2 - 948.5 0.441 0% of incident cases Burns > 20% and < 60% - life long 948.2-948.5 0.255 100% of incident cases Burns > 60% - short-term 948.6-948.9 0.441 0% of incident cases Burns > 60% - life long 948.6-948.9 0.255 100% of incident cases Injured nerves - life long 951,953-957 0.064 100% of incident cases Poisoning - short-term 960-979, 980-989 0.611 For ages 15+ the D W is equal to 0.608 100% of incident cases Residual 900 - 924, 930 - 939 - > Need to be accessed As an example, in Table 3.2, the weight for poisoning of 0.611 means that, on average, society judges a year of disability arising from an open wound (weight of 0.108) to be preferable to a year of disability due to poisoning. Similarly, a year with a disability resulting from poisoning (weight of 0.611) is considered preferable to a year with disability from a spinal cord injury (weight of 0.725). Note that there are some unexpected disability weights in Table 3.2, such as the extremely low weight of 0.001 for Burns <20% - life long; or the high weight of 0.611 for poisoning. It is hard to imagine a situation of a person that a year of disability due to poisoning been considered much worse than living an entire year with disability due to burns (note that all disability weights for burns were less than the disability weight for poisoning). Although it is beyond the scope of this study to develop new disability weights for all categories of nature of injury in Table 3.2, we will proceed with the D A L Y calculation based on the disability weights obtained for the GBD study, and in the discussion of this study we will propose other methods to obtain alternative disability weights. Chapter 3. Methods 70 It was mentioned previously that adjustment of the disability weights for comorbidities was not carried out in this study. However, the data obtained for this study does allow for the identification of individuals with more than one nature of injury (or injury sequela). Therefore, a multiplicative functional form was used to calculate the weights for comorbid injury sequelae (Torrance et al. 1995; Torrance 1986; Torrance, Boyle, and Horwood 1982). Consider for example, a person with two injuries (e.g., fractured skull and open wound), with the DWs for these sequelae being DW\ and DW2. The overall disability weight for this individual is calculated as DW\2=\-(l-DW\)*(\-bW2). If instead we used an additive functional form, ignoring the dependence between sequelae, the combined disability weight is obtained as DW\2=DW\+DW2, which clearly is a very strong assumption about the level of disability for that person. In addition, the additive functional form is not desirable because it allows for disability weights greater than one (e.g., when a person has several injury sequelae). The overall disability weight for each specific external cause of injury by gender is given in Table 3.3. Note that the disability weights in Table 3.3 are based on a weighted average of the adjusted disability weights for comorbid injury sequelae. A hypothetical example of the worksheet to calculate the disability weights for any external cause of injury is given in the Appendix D. Chapter 3. Methods 71 Table 3.3 Weighted Average Disability Weights for Each Age Group and Gender by Injury External Cause. Description Disability wci aht bv ace group - Males . 1 - |W 20 - 39 40-64 65 - 79 8()i Road Traffic Accidents 0.202 0.247 0.244 0.260 0 264 Other transport injuries 0.233 0.218 0.218 0.245 0.262 Poisoning 0.625 0.642 0.636 0.620 0.624 Falls 0.224 0.203 0.223 0.270 0.300 Burns/Fires/Scalds 0.022 0.039 0.049 0.053 0.053 Drowning .0.289 0.277 0.216 0.366 0.170 Sports Injuries 0.195 0.137 0.115 0.131 0.158 Natural and environmental factors 0.253 0.249 0.257 0.290 0.324 Machinery injuries 0.144 0.145 0.142 0.134 0.129 Suffocation and foreign bodies 0.218 0.269 0.283 0.198 0.247 Other unintentional injuries1 0.154 0.147 0.165 0.179 0.200 Suicide and self-inflicted injuries 0.546 0.556 0.598 0.559 0.492 Other intentional injuries 0.249 0.258 0.303 0.334 0.387 Description Disa .)- |u lilily \\eig it by ago group - Females 20 - 39 40-64 65 - 79 80+ Road Traffic Accidents 0.262. 0.242 0.241 0.247 0.252 Other transport injuries 0.225 0.220 0.224 0.261 0.279 Poisoning .0.631 0.652 0.644 0.625 0.615 Falls 0.220 0.197 0.220 0.272 0.308 Burns/Fires/ S calds 0.025 0.050 0.046 0.060 0.048 Drowning 0.258 0.165 0.185 0.161 0.172 Sports Injuries 0.165 0.130 0.123 0.160 0.197 Natural and environmental factors 0.240 0.264 0.253 0.280 0.215 Machinery injuries 0.126 0.127 0.120 0.161 0.190 Suffocation and foreign bodies 0.282 0.260 0.212 0.245 0.207 Other unintentional injuries1 0.167 0.124 0.129 0.169 0.227 Suicide and self-inflicted injuries 0.624 0.624 0.648 0.620 0.554 Other intentional injuries 0.401 0.365 0.414 0.419 0.420 Excludes medical adverse events Based on Table 3.3, note how variable the disability weight is for each combination of gender x age group x external cause of injury. For example, consider the range of disability weights for poisoning and bums/fires/scalds; note that the disability weight for poisoning is in average more Chapter 3. Methods 72 than 11 times higher than the average disability weight for bums/fires/scalds, hi order to understand the source of such difference, we present Table 3.4, where the highlighted cells contain 'the percentage distribution of people in the study population classified according to each combination of gender x injury sequelax disability weight (for nature of injury; see Table 3.2 page 68), and that contributed to more than 50% of the overall disability weight of each external cause of injury. For example, based on Table 3.4, note that more than 90% of males that had poisoning as external cause of injury received a disability weight equal to 0.608 (primarily due to the injury sequela poisoning), and more than 90% of males that had burns/fires/scalds as external cause of injury received a disability weight equal to 0.001 (primarily due to the injury sequela burns < 20%). Thus, after adjusting for comorbid injury sequelae in the same individual, the overall disability weight for the external cause of injury poisoning in males was 0.633 and for bums/fires/scalds in males was 0.037. In addition, note that for those external causes of injury with similar distribution of injury sequelae for males and females (e.g., natural and environmental factors), the resulting DWs are very similar; whereas for external causes of injury in which the distribution of injury sequela is very different between males and females (e.g., other intentional injuries), the resulting DWs are apparently different. Chapter 3. Methods 73 Table 3.4. Disability Weights (All Ages Combined) for Each Injury Nature by Gender and Injury External Cause. l\temal Cause oflniiirv I ri] lit > Sequelae Averaye (OBI)) Road 1 ral'lic Accidents Other transport inj Poisoning Falls M r M 1 M I- M 1 Fractured skull 0.419 Fractured face bones 0.223 Fractured vertebral column 0.266 Injured spinal cord 0.725 Fractured rib or sternum 0.199 Mmm -.(, Fractured pelvis 0.247 Fractured clavicle, scapula or humerus 0.139 i i . : Fractured radius or ulna 0.180 E E s 6.6 MEM 8.5 15.0 Fractured hand bones 0.100 Fractured femur 0.322 S.I • m 29.7 Fractured patella, tibia or fibula 0.271 m 7.1 F^a CZ Fractured ankle 0.196 7.1 |_"j0£L" 9.9 Fractured foot bones 0.077 Other dislocation 0.074 Dislocated shoulder, elbow or hip 0.074 Sprains 0.064 Intracranial injury 0.367 MkfM 12.6 7:5* 8.3 Internal injuries 0.208 8.8 7.6 i i Open wound 0.108 14.9 12.8 11.5 7.1 Injury to eyes 0.298 Amputated thumb 0.165 Amputated finger 0.102 Amputated arm 0.257 Amputated toe 0.102 Amputated foot 0.300 Amputated leg 0.300 Crushing 0.218 Burns < 20% 0.001 Burns > 20% and < 60% 0.255 Burns > 60% 0.255 Injured nerves 0.064 Poisoning 0.608 97.5 Weighted Average DW (adjusted to injury sequela) 0.251 0.248 0.224 0.231 | 0.633 | 0.637 | 0.237 0.263 Chapter 3. Methods 74 Table 3.4. (continued) Disability Weights (All Ages Combined) for Each Injury Sequela by Gender and Injury External Cause. Injury Sequelae Average D W I G U D J lixlcrmil Cause of Injurv Hums Fires. Scald-. Drowning Spoil* Injuries Naiural unJ environmental Itl . 'IOIS M 1 M . F . M „,.v r M Fractured skull 0.419 Fractured face bones 0.223 Fractured vertebral column 0.266 Injured spinal cord 0.725 10.8 Fractured rib or sternum 0.199 Fractured pelvis 0.247 Fractured clavicle, scapula or humerus 0.139 Fractured radius or ulna 0.180 9.3 Fractured hand bones 0.100 Fractured femur 0.322 Fractured patella, tibia or fibula 0.271 ma Fractured ankle 0.196 11.8 Fractured foot bones 0.077 Other dislocation . 0.074 19.3 E H Dislocated shoulder, elbow or hip 0.074 Sprains 0.064 34.4 Intracranial injury 0.367 Internal injuries 0.208 1 1 1 Open wound 0.108 17.1 mm.: 41.1 B Injury to eyes 0.298 Amputated thumb 0.165 Amputated finger 0.102 Amputated arm 0.257 Amputated toe 0.102 Amputated foot 0.300 Amputated leg 0.300 Crushing 0.218 Burns < 20% 0.001 K 1 U 90.8 Burns > 20% and < 60% 0.255 Burns > 60% 0.255 Injured nerves 0.064 Poisoning 0.608 W9E1 0.258 14.2 Weighted Average DW (adjusted to injury sequela) 0.037 0.041 0.275 0.221 0.149 0.143 0.252 Chapter 3. Methods 75 Table 3.4. (continued) Disability Weights (All Ages Combined) for Each Injury Sequela by Gender and Injury External Cause. I'xtcinal C aiisc of lniur\ Injun ScqucliiL--Ueiage .DW (GBD)' ' ^Machincr\ iiiiuric . Suffocation,, and foreign bodict Oilici unintentional iniurkN Suicide and sclt-mlliLicd miunct Other inu.-iuioii.il injuries I'- M F M I M 1 M • Fractured skull 0.419 Fractured face bones 0.223 27.0 Fractured vertebral column 0.266 Injured spinal cord 0.725 Fractured rib or sternum 0.199 Fractured pelvis 0.247 Fractured clavicle, scapula or humerus 0.139 Fractured radius or ulna 0.180 Fractured hand bones o.ioo m i l l •= 11.4 Fractured femur 0.322 Fractured patella, tibia or fibula 0.271 Fractured ankle 0.196 Fractured foot bones 0.077 Other dislocation 0.074 Dislocated shoulder, elbow or hip 0.074 Sprains 0.064 Intracranial injury 0.367 10.0 Internal injuries 0.208 3"\8 1 . 1 Open wound 0.108 H i M S M I 22.3 20.5 mil 3').8 20.5 14... Injury to eyes 0.298 48.6 Amputated thumb 0.165 Amputated finger 0.102 Amputated arm 0.257 Amputated toe 0.102 Amputated foot 0.300 Amputated leg 0.300 Crushing 0.218 Burns < 20% 0.001 Burns > 20% and < 60% 0.255 Burns > 60% 0.255 Injured nerves 0.064 9.0 14.0 Poisoning 0.608 75.5 g E W l .rr Weighted Average D W (adjusted to injury sequela) 0.143 0.129 0.262 0.243 0.155 0.146 | 0.566 | 0.629 0.268 0.388 Chapter 3. Methods 76 3.3.1.5 Calculating Duration for Y L D (D) Duration in this section refers to the average duration of disability measured in years. To calculate the durations required to determine the values of Y L D , it is necessary to first distinguish short-term duration from long-term duration (or life long duration - LL) of disability. Table 3.5 displays both types of durations for each of injury considered in the original GBD study (Murray and Lopez 1996). To calculate the life long durations, we used the B C life expectancy of 2001 (see Appendix G). A hypothetical example of a worksheet to calculate the durations for any external cause of injury is given in the Appendix E. Table 3.5. Duration Assumptions by Type of Injury (Murray and Lopez 1996). Nature of injury ICD9-CM N-Codes Duration (years) "Iicated G B D assumptions Fractured skull - short-term 800 - 801 0.107 85% of incident cases Fractured skull - life long 800 - 801 L L 15% of incident cases Fractured face bones 802 0.118 100% of incident cases' Fractured vertebral column - short-term 805 0.140 100% of incident cases Injured spinal cord - life long 806 and 952 L L 100% of incident cases Fractured rib or sternum - short-term 807 0.115 100% of incident cases Fractured pelvis - short-term 808 0.126 100% of incident cases Fractured clavicle, scapula or humerus - short-term 810-812 0.112 100% of incident cases Fractured radius or ulna - short-term 813 0.112 100% of incident cases Fractured hand bones 814-817 0.070 100% of incident cases Fractured femur - short-term 820-821 0.139 95% of incident cases Fractured femur - life long 820- 821 L L 5% of incident cases Fractured patella, tibia or fibula -short-term 822 - 823 0.090 100% of incident cases Fractured ankle - short-term 824 0.096 100% of incident cases Fractured foot bones 825 - 826 0.073 100% of incident cases Other dislocation 830, 833 - 834, 836 -839 0.019 100% of incident cases Dislocated shoulder, elbow or hip -short-term 831,832,835 0.035 100% of incident cases Chapter 3. Methods 77 Table 3.5. (continued) Duration Assumptions by Type of Injury (Murray and Lopez 1996). Nature of injury ICD9-CM N-Codes Duration (vcars) 'I realed G B D assumptions Sprains 840 - N4S 0.038 100% of incident cases Intracranial injury - short-term 850 - 854 0.067 95% of incident cases Intracranial injury - life long 850 - 854 L L 5% of incident cases. Internal injuries - short-term 860 - 869 0.042 100%> of incident cases Open wOund 870, 872 - 884, 890 - 894 0.024 100% of incident cases Injury to eyes - life long 871,950 L L 100% of incident cases Amputated thumb - life long 885 L L 100% of incident cases Amputated finger - life long 886 L L 100% of incident cases Amputated arm - life long 887 L L 100% of incident cases Amputated toe - life long 895 L L 100% of incident cases Amputated foot - life long 896, 897.0- 897.1 L L 100% of incident cases Amputated leg - life long 897.2- 897.3 L L 100% of incident cases Crushing - short-term 925 - 929 0.094 100% of incident cases Burns < 20% - short-term 940-947, 948.0-948.1 0.083 0% of incident cases Burns < 20% - lifelong 940-947, 948.0-948.1 L L 100% of incident cases Burns > 20% and < 60% - short-term 948.2 - 948.5 0.279 0% of incident cases Burns > 20% and < 60% - life long 948.2 - 948.5 L L 100% of incident cases Burns > 60% - short-term 948.6 - 948.9 0.279 0% of incident cases Burns > 60% - life long 948.6 - 948.9 L L 100% of incident cases Injured nerves - life long 951,953 -957 L L 100% of incident cases Poisoning - short-term 960- 979, 980-989 0.008 100% of incident cases Residual 900 - 924, 930 - 939 - Need to be accessed Note 1: The durations are for all age groups or otherwise indicated. Note 2: L L stands for life long duration, and it will be calculated based on the life expectancy for British Columbia in 2001, and it is specific to age and gender. The calculation of BC life expectancies was primarily based on a life table model called the Coale and Denieny West Level 26 (Coale and Guo 1989). First, an abridged life table was calculated using the age groups 0, 1-4, 5-9,..., 80-84, 85+ (see Table G.2 in Appendix G). Life expectancies for the age groups 0-19, 20-39, 40-64, 65-79 and 80+ were then generated by calculating a weighted average of the life expectancies in the abridged life tables. The process is shown in Figure 3.10 and the final life expectancies for BC in 2001 are given in Table 3.6. Chapter 3. Methods 78 Age 1 ile Number ol group expectancy deaths <1 L, N , 1-4 L 2 N 2 5-9 U N 3 10-14 U N 4 80-84 Ll8 N , 8 85+ L19 N 1 9 Age Life Number ol deaths group expectanc} illlNliilMllliifHW 0-19 A-11,new 20-39 1 L2,«ew 40-64 L3,new 65-79 / 80+/ ^5,new _Nl*Ll+... + N5*L5 l.new — N{+... + N5 Figure 3.10. Example of How to Calculate the Weighted Life Expectancy. Table 3.6. Gender and Age Specific Life Expectancies for British Columbia (BC) in 2001. Age group Life ex UC leclancies - 2001 Males Females 0-19 71.62 78.15 20-39 50.21 54.95 40-64 28.75 32.89 65-79 13.89 16.86 80+ 7.42 9.00 The final weighted average duration for each external cause of injury is given in Table 3.7. Chapter 3. Methods 79 Table 3.7. Weighted Average Durations for Each Age Group and Gender by Injury External Cause. Description Duration bv auc urou •) - Males 0- 19 20 - 39 40 - 64 65 - 79 80( Road traffic accidents 1.743 1.832 0.973 0.483 0.258 Other transport injuries 3.332 3.669 2.565 0.620 0.317 Poisoning 0.972 0.666 0.298 0.060 0.039 Falls 1.020 1.400 0.637 0.311 0.178 Burns/fires/scalds 70.415 47.294 26.520 12.869 6.911 Drowning 9.306 7.245 1.049 3.546 0.092 Sports injuries 1.543 0.572 0.359 0.204 0.175 Natural and environmental factors 2.560 2.296 1.611 0.798 0.038 Machinery injuries 16.785 11.819 7.066 4.045 2.628 Suffocation and foreign bodies 25.931 28.652 15.815 1.889 0.899 Other unintentional injuries1 16.331 11.989 6.900 3.311 1.525 Suicide and self-inflicted injuries 1.212 1.163 0.565 0.213 0.061 Other intentional injuries 1.766 1.653 0.972 0.445 0.443 Description Duration by age group - Females 0-19 20 - 39 40 - 64 65 - 79 80) Road traffic accidents 1.712 1.887 0.946 0.392 0.208 Other transport injuries 6.491 3.195 1.297 0.690 0.437 Poisoning 0.725 0.450 0.103 0.010 0.011 Falls 0.889 0.964 0.380 0.232 0.166 Burns/fires/scalds 76.263 51.097 30.469 15.343 8.466 Drowning 6.567 0.072 0.078 0.084 0.060 Sports injuries 1.065 0.515 0.327 0.330 0.188 Natural and environmental factors 3.245 1.508 1.109 0.390 0.233 Machinery injuries 15.894 15.488 9.407 4.779 1.018 Suffocation and foreign bodies 21.923 13.107 6.781 0.702 1.354 Other unintentional injuries1 13.970 13.821 8.566 2.682 0.695 Suicide and self-inflicted injuries 0.479 0.637 0.398 0.258 0.146 Other intentional injuries 2.650 1.443 1.018 0.354 0.242 Excludes medical adverse events Note how different the durations are for each combination of gender x age group x duration (for nature of injury; see Table 3.3 page 76). Observe that the durations for burns/fires/scalds are much higher than the duration of other injuries. The reason for this is because several sequelae under this external cause of injury ended up being classified as life long. Another Chapter 3. Methods 80 example is related to the long average durations (adjusted for comorbid injury sequelae) for males aged 0-19 years for drowning (9.306 years) and suffocation and foreign bodies (25.931 years). To understand the source of such difference we present Table 3.8. This table shows how different the distribution of injury sequelae is for these two external causes of injury; and pay special attention to the proportion of individuals with Ufe long injury sequelae, which explains the differences in the final durations for these two external causes of injury. Chapter 3. Methods 81 Table 3.8. Durations by Injury Sequela for the External Causes of Injury Drowning and Suffocation and Foreign Bodies. Males 0-19 years. Iiiiurs SL'IIIIUIJO lAkrn . i l Ctu-c ol Iniun. Drowning Sulloialiun .ind toieign bodies Distribution i%) I ' lllldjUsled Dilution K i H U ) Distribution C.'ui I. nadiusiLd Duration ( d U D i Fractured skull - short-term Fractured skull - life longs Fractured face bones 3.2 0 . 1 1 8 Fractured vertebral column - short-term Injured spinal cord - life long Fractured rib or sternum - short-term Fractured pelvis - short-term Fractured clavicle, scapula or humerus - short-term 3.2 0 . 1 1 2 Fractured radius or ulna - short-term Fractured hand bones I.I 0 . 0 7 0 Fractured femur - short-term Fractured femur - life long Fractured patella, tibia or fibula - short-term I.I O.O'JO Fractured ankle - short-term I.I 0 . 0 % Fractured foot bones Other dislocation 2.1 0.01'J Dislocated shoulder, elbow or hip - short-term 1.1 0 . 0 3 5 Sprains Intracranial injury - short-term 1.0 0.0(. •> Intracranial injury - life long 0.1 7 1 . 6 2 0 Internal injuries - short-term 1 1 - 0 .042 Open wound 3 ~ 2 0 . 0 2 4 Injury to eyes - life long 36 .2 7 1 . 6 2 0 Amputated thumb - life long Amputated finger - life long Amputated arm - life long Amputated toe - life long Amputated foot - life long Amputated leg - life long Crushing - short-term Burns < 20% - lifelong Burns > 20% and < 60% - life long Burns > 60% - life long . . Injured nerves - life long Poisoning - short-term I.I 0 . 0 0 8 Weighted Average Duration (unadjusted to injury sequela) 10.229 25.969 Weighted Average Duration (adjusted to injury sequela) 9.306 25.931 Chapter 3. Methods 82 3.3.1.6 Residual Category of Injury Sequela Residual categories are natures of injury with no published disability weights or durations. The residual categories presented in Tables 3.1 are based on the ICD9 N-codes 900-924 (injury to blood vessels; late effects of injuries, poisonings, toxic effects, and other external causes; superficial injury; contusion with intact skin surface), 930-939 (effects of foreign body entering through orifice), 958-959 (certain traumatic complications and unspecified injuries) and 990-999 (other and unspecified effects of external causes). Note that the injury ICD9 N-codes included in the residual category are considered in the ICD9 classification system as of secondary importance when compared to the ICD9 N-codes not included in this category. The recommendation of the W H O regarding the cases of injuries classified according to the categories 900-924 and 930-939 is to distribute these cases proportionately among all other nature of injuries codes for which disability weights or durations have been developed, and then proceed with the calculation of the disability weights and durations for each external cause of injury. For the nature of injury categories 958-959 and 990-999 there was no clear recommendation from the W H O as to what disability weights to use. For this reason, a sensitivity analysis was conducted to understand how different disability weights assigned to these N-codes influenced the D A L Y s across different external causes of injury (see Appendix F - "Sensitivity Analysis for Disability Weights for the Residual Category o f Injury Sequela"). The disability weights for the residual category of injury were 0, 0.25 and 0.50. This sensitivity analysis demonstrated that using different weights for the ICD9 N-codes Chapter 3. Methods 83 codes 958-959 and 990-999 did not affect the ranking (of importance) of different external causes of injury used in this study. For this reason, zero disability weights for these N-codes were used in this study to calculate the overall disability weight for each external cause of injury. 3.3.1.7 Gender Difference in the Length of Life for Calculating YLL (L) When used to estimate burden of disease, DALYs represent a measure of the gap between a population's health status and some ideal life expectancy. The ideal life expectancy used in this study as a reference when assessing the length of life lost was a life expectancy of 82.5 years at birth for women and a life expectancy of 80 years at birth for men (see Table G.3 in Appendix G). These life expectancies were used in the GBD study and are based on the Japanese population, which in 1990 had the highest Ufe expectancy in the world. The main reason for using the same Ufe expectancy as in the GBD study is to facilitate comparisons across different burden of disease/injury studies. The same process exemplified in the Figure 3.9 was appUed in the calculation of Japanese life expectancies for the age groups 0-19, 20-39, 40-64, 65-79 and 80+. The final life expectancies are reported in Table 3.9. Chapter 3. Methods 84 Table 3.9. Japanese Gender and Age Specific Life Expectancies for the Year 1990 Used in the Global Burden of Disease (GBD) Study. Age group Life expectancies Japan -Males Females 0-19 72.16 76.93 20-39 49.49 51.43 40-64 25.52 28.90 65-79 11.53 13.47 80+ 4.95 5.55 In summary, the key assumptions and decisions made with respect to estimation of the D A L Y parameters throughout section 3.3.1 were: a discounting future health rate (r) equal to 3%; age weighting was not utilized; coexisting injury sequelae were adjusted for in the calculation of disability weights (DW) and durations (D); short-term durations were obtained from the original GBD study (Murray and Lopez 1996); life long durations were based on the gender-specific life expectancies for British Columbians in 2001; and the gender-specific life expectancies (L) used in the calculation of years of life lost (YLL) were obtained from the original GBD study (Murray and Lopez 1996). Note that these choices apply to all chapters of this study that deal with D A L Y s . Chapter 3. Methods 85 3.4 Small-Area Analysis of Disability-Adjusted Life Years (DALY) This section incorporated Bayesian disease mapping into the calculation of disability-adjusted life years (DALY) . From the mortality and hospitalization rates estimated using the linear spatio-temporal Bayesian disease mapping model (3.1) in section 3.2, we calculated the fitted number of deaths (./V/() and hospitalizations (/„) separately for the age groups 0-19, 20-39, 40-64, 65-79 and 80+ (j=l,...,5). After combining the results for the estimated number of hospitalizations {liJt) and deaths (A^.Jfo'r each age group j (/=1,...,5), the corresponding Y L D estimate for each external cause of injury, for the ith HSDA, the/* age group, and the tlh year, was derived as: Y L D , , = / * , x i ) ^ x ^ x ( l - e x p ( - 0 . 0 3 x D y ) ) (3.7) where DWj represents the disability weight and Dj represents the average duration of disability (measured in years) for the j'h age group. The corresponding Y L L estimate for each external cause of injury, for the i,h HSDA, the/* age group, and the t'h year, was derived as: Y L L , , = Nijtx x (l - exp(-0.03 x L].)) (3.8) where Lj represents the life expectancy (at age of death) for the j"1 age group. Chapter 3. Methods 86 Therefore the D A L Y estimate for each external cause of injury, for the i HSDA, the jth age group, and the t'h year, was derived as: D A L Y , , = Y L L , , + Y L D , , (3.9) Comparison and inference were based on the following sample posterior summary statistics: posterior mean, posterior standard deviation, posterior median, posterior mode and 0.025 and 0.975 quantiles. In this study, confidence intervals1 were derived for the estimates of all health status indicators in Figure 1.1- mortality and hospitalization rates, Y L L , Y L D and D A L Y . Note that Through the use of posterior inference, it is possible to test, for example, whether the indicators of different HSDAs are statistically different; or whether the indicators of a specific H S D A are different than its HA's indicators; or whether the indicators of each H A are statistically different than the provincial indicators; and so on. The confidence intervals for these indicators were based on their respective posterior distributions (see a hypothetical example in Figure 3.11). Thus, the confidence intervals of the indicators that we were interested in comparing were visually inspected to determine i f they overlapped; i f they did not overlap then the indicator estimates were considered statistically different. Note that multiple-comparison corrections to the level of 5% should be applied when several statistical tests are being performed simultaneously (Rothman and Greenland 1998). 1 In this thesis, Bayesian confidence sets (or credible intervals) calculated based on the 0.025 and 0.975 quantiles were referred to as the frequentist 95% confidence intervals. We adopted this rule since most readers of this thesis will be familiar with the frequentist notion of confidence intervals. Chapter 3. Methods 87 0.2 0.5 0.8 0.2 0.3 0.4 Figure 3.11. Hypothetical Mortality Rate Posterior Distribution Density of Different Health Services Delivery Areas (HSDA). 3.5 Software We used the software SAS version 8.0 (SAS Institute Inc. 1999), Microsoft Excel 2000 and Microsoft Access 2000 to read the original ASCII data files, clean, explore, manipulate, organize and store the data for all the subsequent analyses. We used the software WinBUGS version 1.4 (Spiegelhalter et al. 2003) to run Markov Chain Monte Carlo (MCMC) methods Chapter 3. Methods 88 (using in particular the Gibbs sampling algorithm) to estimate the model parameters in Section 3.2. Convergence of each parameter estimate was assessed using the Gelman-Rubin statistic (Gelman et al. 1998; Carlin and Louis 2000) and by plotting the density of each of the M C M C chains of each parameter estimate. Diagnostic plots for the residuals and other plots for the health status indicators were done using the software Splus 2000 (Venables and Ripley 1997). Patterns of the health status indicators at the HSD A level were explored using maps produced by the Geographical Information System (GIS) using ESRI ArcView's ArcMap 8.3 (Environmental Systems Research Institute Inc. 2003). 3.6 Outline of Data Analyses In this section we explain in more detail how we addressed the main and secondary objectives of this study. The small-area analyses performed in this study controlled for age by analysing the data separately for each age group. The reason we decided to proceed in this way was the difficulty in achieving adequate convergence using M C M C methods (using in particular the Gibbs sampling algorithm) in the software WinBUGS adjusting for the age effect in the statistical model. We believe this problem is due to a slow mixing of our Markov chains largely because of the high-dimension parametric space and high posterior correlation between our different parameters (Bernardinelli and Montomoli 1992). Per capita measures were obtained using mid-period (1995) population estimates; and age-standardized measures were obtained by direct standardization using as standard Chapter 3. Methods 89 population the 2001 Canadian Census population. For more details on age-standardization and definition of per capita quantities see Glossary in Appendix B. 3.6.1 Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 Chapter 4 of this study contains the results for all health status indicators (Figure 1.1) at the provincial and health authority levels, during the period 1991-2000 specific to gender, age group, and external cause of injury. The results of each indicator are shown for the 10 years of data aggregated and over time. A detailed explanation of which ICD9-CM E-code categories contributed the most to the D A L Y of each external cause of injury at the provincial level are summarized in Appendix C. 3.6.2 A Small-Area Analysis of the Incidence of Medical Adverse Events in British Columbia, 1991-2000 Chapter 5 contains the results of the analysis of small-area variation in injury incidence due to medical adverse events using the spatio-temporal Bayesian disease mapping model as described in section 3.2. For illustrational purposes, we present the results of the small-area analysis for the age groups 0-19 years and 65-79 years. It is novel to the literature on medical adverse events to assess the impact of injuries due to AEs among children and youth (ages between 0 and 19 years), at the population level, and over an extensive period of time. In Chapter 3. Methods 90 addition, we present the results for the age group 65-79 years, because this age group had the highest impact of injuries due to A E during the entire study period. The results for the other age groups are presented in Appendix G. 3.6.3 A Small-Area Analysis of the Burden Due To Road Traffic Accidents in British Columbia, 1991-2000 In Chapter 6, we illustrate the small-area analysis of D A L Y . As opposed to previous sections, we focused our examination on the age group and external cause of injury that contributed most to provincial and community level years of healthy life lost as a result of disability and death; i.e., road traffic accidents, in males and females, between the ages 20 and 39 years. 91 Chapter 4 Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 In this chapter, the health status indicators presented in Figure 1.1 were used to assess the burden of injuries from 1991 to 2000 in the province df British Columbia as a whole as well as in its individual health authorities (HA). The injury categories examined in this chapter are shown in Figure 4.1. For a detailed explanation of which ICD9-CM E-code categories contributed the most to the D A L Y of each external cause of injury please refer to Appendix C. For a description of the method used for standardization of rates and for the calculation of per capita values, please refer to the Glossary in Appendix B. r Road traffic accidents Other transport injuries Poisoning Falls Burns/fires/scalds Natural and environmental factors Other intentional injury Drowning Sports injuries Machinery injuries Suffocation and foreign bodies Other unintentional injuries (excluding AE) Suicide and self-inflicted injury Figure 4.1. Categories of external causes of injury. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 92 4.1 The Provincial Level Analysis 4.1.1 Conventional Mortality and Hospitalization Rates From 1991 to 2000, injuries, excluding medical AEs, killed nearly 20,500 people in British Columbia; 70% of whom were males. For males, the categories of external cause of injury contributing more than 60% of the total number of deaths were suicide and self-inflicted injuries, road traffic accidents and poisoning (mainly due to drug abuse). For females, falls, road traffic accidents and suicide and self-inflicted injuries contributed to more than 60% of the total number of deaths. Table 4.1 shows the ten-year mortality rate associated with each external cause of injury examined in this study. The injury related mortality rates were remarkably different for males and females and across all age groups. The age group 80+ had the highest injury mortality rate over the 10-year period while the age group 20-39 for males and the age group 65-79 for females had the second highest mortality rates. Observe that for males between the ages 20 and 64 years, the mortality rate due to poisoning during 1991-2000 was much higher than the other age groups. The main mechanism associated with these later deaths was related to drug abuse. Also of note in Table 4.1 is the instability in mortality rates associated with rare events for several external causes of injury such as sports injuries (males and females), natural and environmental factors, other transport injuries, machinery injuries and other external injuries (excluding AE) (females). For example, because the total number o f deaths associated with sports injuries (29 deaths for both genders over 10 years) was small, the mortality rate varied considerably depending on the size of the population at risk in each age group. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 93 Table 4.1. Ten-Year Mortality Rates by Gender, Age Group and External Cause of Injury, 1991-2000. Mortality rate per 10.000 population Description Males • ' \ - 0- 19 20 - 39 4 ' i - M 65 - 79 80) Road traffic accidents 10.93 22.58 12.88 14.29 31.51 Other transport injuries 1.40 3.92 3.65 2.39 3.78 Poisoning 0.97 24.44 20.60 4.42 2.52 Falls 0.85 2.59 4.39 23.62 202.65 Burns/fires/scalds 0.87 0.83 . 1-33 2.51 4.28 Drowning 1.64 2.77 1.77 2.27 3.53 Sports injuries 0.08 0.20 0.07 0.12 0.25 Natural and environmental factors 0.14 0.98 1.15 1.02 2.52 Machinery injuries 0.24 1.11 1.38 1.08 0.76 Suffocation and foreign bodies 0.77 0.98 1.38 2.99 13.11 Other unintentional injuries* 0.47 1.94 2.22 1.20 2.52 Suicide and self-inflicted injuries 4.17 25.72 26.08 22.84 39.32 Other intentional injuries 1.78 7.16 4.76 2.63 2.77 Total 24.30 95.21 81.67 81.38 309.52 Description Mortality rate per 10.000 population (1991-2000) 0 - 19 20 - 39 40 - 64 io - ~l» 80+ Road traffic accidents 5.57 7.20 5.68 10.84 16.35 Other transport injuries 0.40 0.52 0.40 0.50 0.58 Poisoning 0.56 7.18 5.39 2.50 2.63 Falls 0.29 0.44 1.56 15.53 198.05 Burns/fires/scalds 0.48 0.37 0.58 1.30 3.65 Drowning 0.65 0.28 0.61 0.80 1.46 Sports injuries 0.02 0.03 0.02 0.00 0.29 Natural and environmental factors 0.06 0.27 0.24 0.25 1.02 Machinery injuries 0.02 0.05 0.05 0.10 0.00 Suffocation and foreign bodies 0.33 0.18 0.43 2.30 10.51 Other unintentional injuries1 0.10 0.20 0.13 0.25 0.44 Suicide and self-inflicted injuries 1.44 6.31 8.14 7.14 8.61 Other intentional injuries 1.71 2.93 2.15 1.65 3.21 Total 11.64 25.97 25.37 43.15 246.80 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 94 In 1991-2000, over 614,000 hospitalizations were attributable to injuries in British Columbia, of which 58% occurred in males. Falls plus road traffic accidents accounted for 52% of the total number of hospitalizations for males, and 66% of total number of hospitalizations for females. Table 4.2 shows the ten-year hospitalization rate associated with each external cause of injury. The hospitalization rates were different between males and females and across the age groups. In general, males had higher hospitalization rates than females. The age group 80+ had the highest hospitalization rate over the 10-year period, while the age groups 20-39 for males and 65-79 for females had the second highest hospitalization rate. Table 4.2. Ten-Year Conventional Hospitalization Rates by Gender, Age Group and External Cause of Injury, 1991-2000. Description Hospitalization rate per 1,000 population (1991-2000) 0-19 ' 20- 39 ; 40-64 65 - 79 80 Road traffic accidents 40.5 58.4 34.2 33.2 45.8 Other transport injuries 7.0 9.5 6.0 4.0 4.5 Poisoning 3.5 3.1 2.9 3.9 7.5 Falls 44.5 34.1 45.3 102.4 366.6 Burns/fires/scalds 5.5 3.8 3.3 3.5 5.8 Drowning 0.5 0.1 0.1 0.2 . 0.2 Sports injuries 16.0 20.7 10.8 4.8 4.8 Natural and environmental factors 2.1 1.9 2.1 1.9 2.8 Machinery injuries 3.0 11.5 10.4 7.4 3.2 Suffocation and foreign bodies 4.4 1.6 2.5 5.8 10.2 Other unintentional injuries1 15.6 26.8 18.4 8.2 7.6 Suicide and self-inflicted injuries 4.7 18.1 11.3 5.2 7.8 Other intentional injuries 10.8 28.2 11.9 4.1 3.8 Total 158.0 217.8 159.2 184.6 470.6 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 95 Table 4.2. (continued) Ten-Year Conventional Hospitalization Rates by Gender, Age Group and External Cause of Injury, 1991-2000. Description Hospitalization rate per 1.000 population (1991-2000) Females II - 19 20 - 39 40 - 64 65 - 79 80+ Road traffic accidents 26.8 29.5 22.6 31.3 32.2 Other transport injuries 2.1 2.0 1.5 2.0 3.4 Poisoning 3.0 2.6 2.5 4.4 7.4 Falls 26.2 16.8 37.1 161.2 623.1 Burns/fires/scalds 3.1 1.3 1.5 2.5 5.4 Drowning 0.2 0.0 0.0 0.1 0.1 Sports injuries 5.2 6.7 4.4 4.6 7.9 Natural and environmental factors 1.8 1.1 1.3 1.3 2.3 Machinery injuries 0.7 0.8 0.8 0.5 0.5 Suffocation and foreign bodies 3.5 0.8 1.6 3.5 7.1 Other unintentional injuries* 6.2 6.4 4.5 3.6 6.7 Suicide and self-inflicted injuries 14.8 28.7 16.2 4.5 4.2 Other intentional injuries 4.2 8.3 4.2 2.7 3.4 Total 97.8 104.9 98.1 222.1 703.7 Excludes medical adverse events Next we show the annual age-standardized mortality rate (ASMR) per 10,000 population and the annual age-standardized hospitalization rate (ASHR) per 1,000 population associated with road traffic accidents, poisoning (mainly due to drug abuse), falls, suicide and self-inflicted injuries and other intentional injuries during 1991-2000 (Figures 4.2 and 4.3). These five external categories of injury were chosen because, as mentioned above, they accounted for more than 60% of the total number of hospitalizations and deaths for males and females, thus producing annual A S M R and ASHR susceptible to lower chance variation than the other external causes of injury. The annual ASMRs for the categories falls and other intentional injuries were similar for males and females (see Figure 4.2). The A S M R for falls was slightly higher in females Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 96 compared to males. For males, the A S M R of suicide and self-inflicted injuries was consistently higher than the A S M R of road traffic accidents during the entire study period. Road traffic accidents Poisoning Falls 1992 1994 1996 1998 2000 year Suicide and self-inflicted injury 1992 1994 1996 1998 2000 year Other intentional injury Males Females 1992 1994 1996 1998 2000 1992 1994 1996 1998 2000 year year Figure 4.2. Annual Age-Standardized Mortality Rate (ASMR) per 10,000 Population for Males and Females, 1991-2000 (Canada Census 2001). Over the years, the annual ASHR of suicide and self-inflicted injury, as well as of falls, was higher for females than males (see Figure 4.3). In addition, the annual ASHR for poisoning was very similar for males and females. Note as well that the annual ASHR due to falls was higher than any of the other annual ASHRs. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 97 Road traffic accidents Poisoning Falls 8 * a. x < 1992 1994 1996 1998 2000 year Suicide and self-inflicted injury 1992 1994 1996 1998 2000 year Other intentional injury a. i < i 1992 1994 1996 1998 2000 year 1992 1994 1996 1998 2000 year 1992 1994 1996 1998 2000 year Males Females Figure 4.3. Annual Age-Standardized Hospitalization Rate (ASHR) per 1,000 Population for Males and Females, 1991-2000 (Canada Census 2001). 4.1.2 Years of Life Lost (YLL) Years of life lost or Y L L is the mortality component of D A L Y and is also known as the fatal burden from injury. From 1991 to 2000, fatal injuries were responsible for 287,272 years of life lost in males and 102,737 years of life lost in females, which represents 61% of the D A L Y for all injuries combined over all age groups for males, and 56% of the D A L Y for all Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 98 injuries combined over all age groups for females. The burden from fatal injuries (YLL) was 2.8 times higher for males than for females in 1991-2000. Suicide and self-inflicted injuries had the highest ten-year burden of the fatal injuries (YLL) experienced by males (27% of all Y L L ) , and road traffic injuries the highest ten-year burden for females (28% of all Y L L ) (see Table 4.3). Table 4.3. Total Years of Life Lost (YLL) by Gender, Age and External Cause of Injury, 1991-2000. YLL (1991-2000) Description Males 0 - 19 20 - 39 40-64 65 - 79 80- A l l ages Road traffic accidents 16,318 35,758 12,822 2.329 575 67,802 Other transport injuries 2,095 6,213 3,638 390 69 12,405 Poisoning 1,446 38,697 20,507 721 46 61,418 Falls 1,269 4,099 4,369 3,850 3,697 17,284 Burns/fires/scalds 1,298 1,315 1,320 409 78 4,420 Drowning 2,449 4,383 1,765 370 64 9,032 Sports injuries 118 309 71 19 5 523 Natural and environmental factors 207 1,547 1,141 166 46 3,106 Machinery injuries 354 1,753 1,373 175 14 3,670 Suffocation and foreign bodies 1,151 1,547 1,373 487 239 4,797 Other unintentional injuries1 708 3,068 2,211 195 46 6,228 Suicide and self-inflicted injuries 6,226 40,734 25,964 3,723 717 77,365 Other intentional injuries 2,656 11,344 4,743 429 51 19,222 Total 36,294 150,768 81,299 13,265 5,647 287,272 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 99 Table 4.3. (continued) Total Years of Life Lost (YLL) by Gender, Age and External Cause of Injury, 1991-2000. Y L L (1991-2000) Description Females n - |w 20 - 39 4 0 - M 65 - 79 sn All ages Road traffic accidents 8,015 11.269 6,069 2,404 573 28,330 Other transport injuries 570 812 425 111 20 1,939 Poisoning 810 11,243 5,760 554 92 18,460 Falls 420 681 1,662 3,446 6,941 13,150 Burns/fires/scalds 690 577 619 288 128 2,301 Drowning 931 446 657 177 51 2,262 Sports injuries 30 52 19 0 10 112 Natural and environmental factors 90 419 251 55 36 852 Machinery injuries 30 79 58 22 0 189 Suffocation and foreign bodies 480 288 464 510 368 2,110 Other unintentional injuries1 150 314 135 55 15 671 Suicide and self-inflicted injuries 2,071 9,880 8,698 1,584 302 22,535 Other intentional injuries 2,461 4,586 2,300 366 113 9,826 Total 16,750 40,648 27,118 9,572 8,649 102,737 Excludes medical adverse events Table 4.4 illustrates the ten-year (i.e., 1991-2000) per capita years of life lost (or per capita Y L L ) per 1,000 population for the different categories of external causes by gender and age group. Per capita Y L L was obtained dividing the Y L L by the mid-period (1995) population estimates. For males and females, the age groups 20-39 and 80+, respectively, had the highest ten-year per capita burden from fatal injuries. For falls, the age group 80+ years had a very high ten-year per capita Y L L for both genders. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 100 Table 4.4. Ten-Year Per Capita Years of Life Lost (or Per Capita Y L L ) per 1,000 Population by Gender, Age and External Cause of Injury for 1991-2000. Per capita Y L L per 1,000 population (1991-2000) Description Males 1) - lo 20 - 39 4u - (4 65 - 79 80+ Road traffic accidents 32.24 58.21 22.97 13.93 14.49 Other transport injuries 4.14 10.12 6.52 2.33 1.74 Poisoning 2.86 63.00 36.73 4.31 1.16 Falls 2.51 6.67 7.83 23.02 93.19 Burns/fires/scalds 2.57 2.14 2.36 2.45 1.97 Drowning 4.84 7.14 3.16 2.21 1.62 Sports injuries 0.23 0.50 0.13 0.12 0.12 Natural and environmental factors 0.41 2.52 2.04 0.99 1.16 Machinery injuries 0.70 2.85 2.46 1.05 0.35 Suffocation and foreign bodies 2.27 2.52 2.46 2.91 6.03 Other unintentional injuries* 1.40 4.99 3.96 1.17 1.16 Suicide and self-inflicted injuries 12.30 66.32 46.51 22.26 18.08 Other intentional injuries 5.25 18.47 8.50 2.56 1.27 Total 71.70 245.45 145.63 79.32 142.33 Per capita Y L L per 1.000 population (1991-2000) Description Females 0-19 20 - 39 40 - 64 65 - 79 80-Road traffic accidents 16.72 18.87 10.9S 12.01 8.36 Other transport injuries 1.19 1.36 0.77 0.55 0.30 Poisoning 1.69 18.82 10.42 2.77 1.34 Falls 0.88 1.14 3.01 17.21 101.30 Burns/fires/scalds 1.44 0.97 1.12 1.44 1.87 Drowning 1.94 0.75 1.19 0.89 0.75 Sports injuries 0.06 0.09 0.03 0.00 0.15 Natural and environmental factors 0.19 0.70 0.45 0.28 0.52 Machinery injuries 0.06 0.13 0.10 0.11 0.00 Suffocation and foreign bodies 1.00 0.48 0.84 2.55 5.37 Other unintentional injuries* 0.31 0.53 0.24 0.28 0.22 Suicide and self-inflicted injuries 4.32 16.54 15.73 7.91 4.40 Other intentional injuries 5.13 7.68 4.16 1.83 1.64 Total 34.93 68.05 49.05 47.81 126.23 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 101 Next we show the trend in the annual age-standardized per capita Y L L (ASYLL) per 1,000 population for each external cause of injury during 1991-2000 (Figure 4.4). The categories of external causes of injuries presented in Figure 4.4 are the same as in Figures 4.2 and 4.3. Road traffic accidents Poisoning Falls 1992 1994 1996 1998 2000 year Suicide and self-inflicted injury 1992 1994 1996 1998 2000 year Other intentional injury 1992 1994 1996 1998 2000 year Males Females 1992 1994 1996 1998 2000 year 1992 1994 1996 1998 2000 year Figure 4.4. Annual Age-Standardized Per Capita Y L L (ASYLL) per 1,000 Population for Males and Females, 1991-2000 (Canada Census 2001). The gender difference in the annual A S Y L L for falls and other intentional injuries were relatively small compared to injury the categories suicide and self-inflicted injuries, poisoning and road traffic accidents where the gender difference is more pronounced. It is also worth noting that the A S Y L L for road traffic accidents in males has decreased Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 102 considerably since 1991. In addition, note the bimodality of the A S Y L L trend for poisoning in males and females during 1991-2000. 4.1.3 Healthy Years Lost Due To Disability (YLD) Healthy years lost due to disability or Y L D is the disability component of D A L Y and is also known as the non-fatal burden from injury. From 1991 to 2000, non-fatal injuries were responsible for 182,416 healthy years lost due to disability in males, and 79,260 healthy years lost due to disability in females, which translates into 39% of the D A L Y for all injuries and ages combined for males, and 44% of the D A L Y for all injuries and ages combined for females. The burden from non-fatal injuries (YLD) was 2.3 times higher for males than the Y L D for females in 1991-2000. For males, other unintentional injuries (excluding AE) had the highest ten-year burden from non-fatal injuries (28% of all Y L D ) , followed by road traffic accidents (16.5% of all Y L D ) . For females road traffic accidents had the highest ten-year non-fatal burden (21.4% of all Y L D ) , followed by other unintentional injuries (excluding AE) (17.4% of all Y L D ) (see Table 4.5). It is important to mention that the higher Y L D for the category other unintentional injuries (excluding AE) was primary due to the ICD9 E-codes (a) E916.0: struck accidentally by falling object such as: rock, snowslide (nowhere else specified), stone, tree, object falling from machine not in operation or from stationary vehicle; (b) E917.9: struck accidentally by an air rifle; (c) E918.0: caught accidentally in or between objects; (d) E920.3, E920.8: cutting and piercing instruments or objects mainly: knives, swords and daggers, and other cutting Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 103 and piercing objects (e.g., broken glass, nail, tin can lid, splinter, etc.); and (e) E929.8: late effect of other accidents (nowhere else specified). Table 4.5. Total Healthy Years Lost Due To Disability (YLD) by Gender, Age and External Cause of Injury, 1991-2000. Y L D (1991-2000) Description Males (1 - 10 20 - 39 40 - 64 65 - 79 80 A l l ygcs Road traffic accidents 9,129 15,767 4,468 691 123 30,178 Other transport injuries 2,598 4,403 1,800 100 15 8,916 Poisoning 1,050 802 309 24 7 2,193 Falls 5,051 5,835 3,549 1,429 775 16,639 Burns/fires/scalds 1,799 . 2,305 1,657 333 76 6,170 Drowning 551 139 18 42 0 750 Sports injuries . 2,385 986 247 21 5 3,644 Natural and environmental factors 673 634 467 73 1 1,848 Machinery injuries 2,900 10,177 5,268 634 41 19,021 Suffocation and foreign bodies 8,705 4,977 4,943 351 88 19,064 Other unintentional injuries1 15,751 24,395 10,546 771 89 51,552 Suicide arid self-inflicted injuries 1,548 7,078 2,117 104 9 10,856 Other intentional injuries 2,332 7,193 1,931 102 26 11,584 Total 54,474 84,691 37,318 4,676 1,257 182,416 Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 104 Table 4.5. (continued) Total Healthy Years Lost Due To Disability (YLD) by Gender, Age and External Cause of Injury, 1991-2000. Deseriptio i YI.D (1991-2000) Females 0-19 20 - 39 40 - 64 65 - 79 801- A l l ages Road traffic accidents 5,611 7,807 2,813 603 115 16,949 Other transport injuries 1,322 812 240 73 28 2,475 Poisoning 660 447 92 6 4 1,208 Falls 2,421 1,879 1,709 2,030 2,182 10,221 Burns/fires/scalds 1,112 981 744 365 133 3,335 Drowning 160 0.2 0.3 0.2 0.1 161 Sports injuries 435 264 97 49 20 865 Natural and environmental factors 645 255 196 28 8 1,132 Machinery injuries 523 773 427 73 6 1,802 Suffocation and foreign bodies 7,631 1,323 1,122 120 134 10,330 Other unintentional injuries* 5,657 5,355 2,393 311 72 13,788 Suicide and self-inflicted injuries 2,107 6,751 2,293 142 23 11,317 Other intentional injuries 2,047 2,561 967 79 23 5,677 Total 30,330 29,209 13,095 3,878 2,749 79,260 Excludes medical adverse events Table 4.6 contains the ten-year per capita healthy years lost due to disability (or per capita Y L D ) per 1,000 population for different categories of external causes by gender and age group. For males and females, the age groups 20-39 and 0-19, respectively, had the highest ten-year per capita burden from non-fatal injuries (see Table 4.6). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 105 Table 4.6. Ten-Year Per Capita Healthy Years Lost Due To Disability (or Per Capita Y L D ) per 1,000 Population by Gender, Age and External Cause of Injury for 1991-2000. Per capita Y L D per 1.000 population (1991-20001 Description Males iBlHil 0-19 20 - 39 4 u - M # o - 80-Road traffic accidents 18.04 25.67 8.00 4.13 3.11 Other transport injuries 5.13 7.17 3.22 0.60 0.37 Poisoning 2.07 1.31 0.55 0.14 0.18 Falls 9.98 9.50 6.36 8.54 19.54 Burns/fires/scalds 3.55 3.75 2.97 1.99 1.92 Drowning 1.09 0.23 0.03 0.25 0.00 Sports injuries 4.71 1.60 0.44 0.13 0.13 Natural and environmental factors 1.33 1.03 0.84 0.44 0.03 Machinery injuries 5.73 16.57 9.44 3.79 1.04 Suffocation and foreign bodies 17.20 8.10 8.85 2.10 2.23 Other unintentional injuries1 31.12 39.72 18.89 4.61 2.25 Suicide and self-inflicted injuries 3.06 11.52 3.79 0.62 0.23 Other intentional injuries 4.61 11.71 3.46 0.61 0.65 Total 107.62 137.88 66.85 27.96 31.69 Description Per capita Y L D per 1.000 population (1991-2000) Females 0 - 19 20 - 39 40 - 64 65 - 79 80 Road traffic accidents 11.70 13.07 5.09 3.01 1.68 Other transport injuries 2.76 1.36 0.43 0.36 0.41 Poisoning 1.38 0.75 0.17 0.03 0.05 Falls 5.05 3.15 3.09 10.14 31.84 Burns/fires/scalds 2.32 1.64 1.35 1.82 1.94 Drowning 0.33 0.00 0.00 0.00 0.00 Sports injuries 0.91 0.44 0.18 0.24 0.29 Natural and environmental factors 1.35 0.43 0.35 0.14 0.11 Machinery injuries 1.09 1.29 0.77 0.36 0.09 Suffocation and foreign bodies 15.92 2.22 2.03 0.60 1.95 Other unintentional injuries1 11.80 8.96 4.33 1.56 1.05 Suicide and self-inflicted injuries 4.39 11.30 4.15 0.71 0.34 Other intentional injuries 4.27 4.29 1.75 0.39 0.34 Total 63.26 48.90 23.68 19.37 40.11 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 106 Figure 4.5 shows the annual trends in age-standardized per capita Y L D (ASYLD) per 1,000 population for the same injury categories described in Figures 4.2 through 4.4. Road traffic accidents Poisoning Falls 1992 1994 1996 1998 2000 year Suicide and self-inflicted injury 1992 1994 1996 1998 2000 year Other intentional injury 1992 1994 1996 1998 2000 year Males Females Figure 4.5. Annual Age-Standardized Per Capita Healthy Years Lost Due To Disability (ASYLD) per 1,000 population for Males and Females, 1991-2000 (Canada Census 2001). The annual A S Y L D of suicide and self-inflicted injury for most years in 1991-2000 was slightly higher for females than males. The annual A S Y L D for poisoning (mainly due to drug abuse) was very similar for males and females, and it was much lower than the A S Y L D of the other external causes of injury. In addition, the A S Y L D of road traffic accidents, for both males and females, has been decreasing significantly since 1991. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 107 4.1.4 Disability-Adjusted Life Years (DALY) The total burden due to injury (or lost years of healthy life) was assessed by calculating D A L Y via the formula D A L Y = Y L D + Y L L . From 1991 to 2000, fatal and non-fatal injuries were responsible for 469,688 lost years of healthy life in males and 181,997 lost years of healthy life in females. The total burden from injuries (DALY) was 2.6 times higher for males than the D A L Y for females in 1991-2000. Road traffic accidents represented the highest ten-year total burden from injuries (DALY) for males (21% of all D A L Y ) and females (25% of all D A L Y ) . Suicide and self-inflicted injuries had the second highest ten-year D A L Y for both males and females (see Table 4.7). Table 4.7. Total Disability Adjusted Life Years (DALY) by Gender, Age and External Cause of Injury, 1991-2000. D A L Y (1991-2000) Description Ma ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ 0-19 20 - 39 40 - 64 05 - 79 80+ Al l ages Road traffic accidents 25,447 51,525 17,289 3,020 698 97,980 Other transport injuries 4,693 10,616 5,438 490 84 21,321 Poisoning 2,496 39,500 20,817 745 53 63,611. Falls 6,320 9,934 7,918 5,279 4,472 33,923 Burns/fires/scalds 3,098 3,620 2,976 742 154 10,590 Drowning 3,000 4,522 1,783 412 64 9,782 Sports injuries 2,503 1,295 319 41 10 4,167 Natural and environmental factors 880 2,181 1,608 239 47 4,955 Machinery injuries 3,254 11,930 6,641 810 55 22,690 Suffocation and foreign bodies 9,856 6,524 6,316 838 327 23,861 Other unintentional injuries* 16,459 27,463 12,757 966 135 57,781 Suicide and self-inflicted injuries 7,774 47,812 28,081 3,827 727 88,221 Other intentional injuries 4,988 18,537 6,674 531 76 30,806 Total 90,768 235,459 118,616 17,941 6,904 469,688 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 108 Table 4.7. (continued) Total Disability Adjusted Life Years (DALY) by Gender, Age and External Cause of Injury, 1991-2000. D A L Y (1991-20001 1 ) C M . I iption 0- 19 20 - 39 4ii - M 65 - 79 80+ Al l ages Road traffic accidents 13,625 19,077 8,883 3,007 688 45,280 Other transport injuries 1,892 1,624 666 183 49 4,414 Poisoning 1,471 11,690 5,852 559 96 19,668 Falls 2,841 2,561 3,371 5,476 9,122 23,371 Burns/fires/scalds 1,802 1,557 1,363 653 261 5,636 Drowning 1,091 446 657 177 51 2,423 Sports injuries 465 316 117 49 30 977 Natural and environmental factors 735 674 448 83 44 1,984 Machinery injuries 553 852 485 95 6 1,991 Suffocation and foreign bodies 8,111 1,611 1,586 630 502 12,440 Other unintentional injuries* 5,807 5,669 2,529 367 87 14,459 Suicide and self-inflicted injuries .4,178 16,631 10,991 1,727 325 33,852 Other intentional injuries 4,508 7,148 3,267 444 136 15,503 Total 47,080 69,857 40,213 13,450 11,397 181,997 Excludes medical adverse events Table 4.8 contains the ten-year per capita D A L Y per 1,000 population for each of the different categories of external causes of injury by gender and age group. The age groups 20-39 for males and 80+ for females were found to have the highest ten-year per capita total burden from injuries. For every 1,000 people monitored over 10 years (i.e., 10,000 years of life observed), 383.33 D A L Y s (i.e., approximately 383 lost years of healthy life due to injury) were lost for males in the age group 20-39 years, and 166.34 D A L Y s (i.e., approximately 166 lost years of healthy life due to injury) were lost for females in the age group 80+ years. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 109 Table 4.8. Ten-Year Per Capita Disability-Adjusted Life Years (DALY) per 1,000 Population by Gender, Age and External Cause of Injury for 1991-2000. Per capita D A L Y per 1.000 population (1991-2000) Description Males 0-19 20 - 39 40 - 64 65 - 79 80-Road traffic accidents 50. r 83.88 30.97 18.06 17.60 Other transport injuries 9.27 17.28 9.74 2.93 2.11 Poisoning 4.93 64.31 37.29 4.46 1.34 Falls 12.49 16.17 14.18 31.57 112.73 Burns/fires/scalds 6.12 5.89 5.33 4.44 3.89 Drowning 5.93 7.36 3.19 2.46 1.63 Sports injuries 4.95 2.11 0.57 0.24 0.25 Natural and environmental factors 1.74 3.55 2.88 1.43 1.19 Machinery injuries 6.43 19.42 11.90 4.84 1.39 Suffocation and foreign bodies 19.47 10.62 11.31 5.01 8.25 Other unintentional injuries* . 32.52 44.71. 22.85 5.78 3.41 Suicide and self-inflicted injuries 15.36 77.84 50.30 22.89 18.31 Other intentional injuries 9.85 30.18 11.96 3.18 1.92 Total 179.33 383.33 212.48 107.28 174.02 Per capita D A L Y per 1.000 population (1991-2000) Description Females 0- 19 20 - 39 40 - 64 65 - 79 So Road traffic accidents 28.42 31.94 16.07 15.02 10.04 Other transport injuries 3.95 2.72 1.20 0.92 0.71 Poisoning 3.07 19.57 10.58 2.79 1.40 Falls 5.93 4.29 6.10 27.35 133.14 Burns/fires/scalds 3.76 2.61 2.46 3.26 3.81 Drowning 2.27 0.75 1.19 0.89 0.75 Sports injuries 0.97 0.53 0.21 0.24 0.44 Natural and environmental factors 1.53 1.13 0.81 0.42 0.64 Machinery injuries 1.15 1.43 0.88 0.47 0.09 Suffocation and foreign bodies 16.92 2.70 2.87 3.15 7.33 Other unintentional injuries* 12.11 9.49 4.57 1.83 1.28 Suicide and self-inflicted injuries 8.71 27.84 19.88 8.62 4.75 Other intentional injuries 9.40 11.97 5.91 2.22 1.98 Total 98.19 116.95 72.73 67.17 166.34 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 110 Figure 4.6 (see below) illustrates the annual trend in age-standardized per capita D A L Y (ASDALY) per 1,000 population during 1991-2000 for the same injury categories as those presented in previous sections examining A S M R , ASHR, A S Y L D and A S Y L L . Note that the pattern of the annual A S D A L Y for each injury category was very similar to the pattern of the A S Y L L , and for all these injuries, the annual A S D A L Y was higher in males than in females. The difference in the annual A S D A L Y between males and females was the highest for suicide and self-inflicted injury followed by road traffic accidents (see Figure 4.6). Road traffic accidents Poisoning Falls 1992 1994 1996 1998 2000 year Suicide and self-inflicted injury 1992 1994 1996 1998 2000 year Other intentional injury 1992 1994 1996 1998 2000 year Males Females 1992 1994 1996 1998 2000 year 1992 1994 1996 1998 2000 year Figure 4.6. Annual Age-Standardized Per Capita D A L Y (ASDALY) per 1,000 Population for Males and Females, 1991-2000 (Canada Census 2001). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 111 4.1.5 Comparison of Contributions Associated With YLD and Y L L (Ten-Year Combined) To compare the relative contributions of Y L D and Y L L to D A L Y totals, we calculated the percent contribution of the non-fatal burden to the total burden from injuries ( Y L D ^ z.e.,100x by gender, age group and external cause of injury for ten years of data V D A L Y ) (see Table 4.9). This percentage-based approach is different than that used in the WHO GBD YLD study, which relied on an assessment of the ratio instead. However, because the Y L L YLL for some injuries was equal to zero, it was not possible to adopt the ratio-based approach in this study. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 112 Table 4.9. Percent Contribution of Y L D to D A L Y (Ten-Year Combined) by Gender, Age and External Cause of Injury, 1991-2000. % of Y L D contribution to D A L Y (1991-2mMM Description Males 0 - 19 2 i i - 3 1 ' 40 - 64 65 - 79 80+ Al l ages Road traffic accidents 35.9 30.6 25.S 22.9 17.7 30.8 Other transport injuries 55.4 41.5 33.1 20.4 17.6 41.8 Poisoning 42.1 2.0 1.5 3.2 13.5 3.4 Falls 79.9 58.7 44.8 27.1 17.3 49.0 Burns/fires/scalds 58.1 63.7 55.7 44.8 49.3 58.3 Drowning 18.4 3.1 1.0 10.1 0.2 7.7 Sports injuries 95.3 76.1 77.6 52.4 53.5 87.5 Natural and environmental factors 76.5 29.1 29.0 30.7 2.9 37.3 Machinery injuries 89.1 85.3 79.3 78.3 75.0 83.8 Suffocation and foreign bodies 88.3 76.3 78.3 41.9 27.0 79.9 Other unintentional injuries* 95.7 88.8 82.7 79.8 66.0 . 89.2 Suicide and self-inflicted injuries 19.9 14.8 7.5 2.7 1.3 12.3 Other intentional injuries 46.8 38.8 28.9 19.3 - 33.7 37.6 Total 60.0 36.0 31.5 26.1 18.2 38.8 Description liiiiiiiilllll^ ^ % of Y L D contribution u. 1) M Y 11991-2000) iiiiiiB^ 0- 1') 20 - 39 40 - 64 65 - 79 80 H A l l ages Road traffic accidents 41.2 40.9 31.7 20.0 16.7 37.4 Other transport injuries 69.9 50.0 36.1 39.6 58.1 56.1 Poisoning 44.9 3.8 1.6 1.0 3.7 6.1 Falls 85.2 '. 73.4 50.7 37.1 23.9 43.7 Burns/fires/scalds 61.7 63.0 54.6 55.9 51.0 59.2 Drowning 14.7 0.1 0.1 0.1 0.1 6.6 Sports injuries 93.5 83.4 83.4 100.0 66.2 88.5 Natural and environmental factors 87.8 37.8 43.9 33.4 17.8 57.1 Machinery injuries 94.6 90.8 88.0 76.6 100.0 90.5 Suffocation and foreign bodies 94.1 82.1 70.7 19.1 26.6 83.0 Other unintentional injuries* 97.4 94.5 94.6 84.9 82.4 95.4 Suicide and self-inflicted injuries 50.4 40.6 20.9 8.2 7.2 33.4 Other intentional injuries 45.4 35.8 29.6 17.7 17.1 36.6 Total 64.4 41.8 32.6 28.8 24.1 43.6 Excludes medical adverse events Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 113 The shaded cells in Table 4.9 highlight injury types with percentages of Y L D > Y L L . Note that in page 102 we explained which type of injury belongs to the category other unintentional injuries (excluding AE). Figures 4.7 and 4.8 contain plots of the percent contribution of Y L D to D A L Y (ten-year combined) for each external cause of injury by age group for men and women. The detailed explanation of each external cause of injury in Figures 4.7 and 4.8 is presented in the Glossary in Appendix B. % of contribution of YLD to DALY BC Males Other intentional injuries Suicide and self-inflicted injuries (| Other unintentional injuries*. Suffocation and foreign bodies Machinery injuries Natural and environmental factors Sports injuries Drowning Burns/fires/scalds Falls Poisoning Other transport injuries Road traffic accidents Note: * Excludes medical adverse events ...Q .1.0. 20 30 40 50 60 70 80 90 100 • 80+ • 65-79 • 40-64 20-39 • 0-19 Figure 4.7. Percent Contribution of Y L D to D A L Y (Ten-Year Combined) by Age and External Cause of Injury for Males, 1991-2000. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 114 % of contribution of YLD to DALY BC Females Other intentional injuries Suicide and self-inflicted injuries Other unintentional injuries*-Suffocation and foreign bodies Machinery injuries Natural and environmental factors Sports injuries Drowning Burns/fires/scalds Falls Poisoning Other transport injuries Road traffic accidents 0 10 Note: * Excludes medical adverse events Figure 4.8. Percent Contribution of Y L D to D A L Y (Ten-Year Combined) by Age and External Cause of Injury for Females, 1991-2000. In summary, the preceding results indicated that the burden imposed by fatal and non-fatal injuries had very different age structures, and varied across gender and external cause of injuries. In addition, considering all external causes, the fatal burden of injuries (i.e., Y L L ) contributed the most to the total burden (DALY) from injuries in both males (61.2%) and females (56.4%). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 115 4.1.6 Comparison of All Health Status Indicators (Ten-Year Combined) In this section of the study, the relative rankings of the external causes of injury (i.e., from highest value (1) to lowest value (13)) are presented for each health status indicator by gender for the period 1991-2000 (see Figures 4.9 and 4.10). From Figures 4.9 and 4.10, it can be seen that for males in B C , suicide and self-inflicted injuries and falls had the highest A S M R and ASHR, respectively. For females, falls had both the highest A S M R and ASHR. For both males and females, road traffic accidents had the highest total burden from injuries (ASDALY) , and the second highest A S M R and ASHR. These results indicate that the injury identified as having the largest impact on the health of the B C population changes depending on which health status indicator is examined. The variation in the ranking of different external causes of injury reveals important areas of inequality in health status, since each health status indicator provide different information about the impact of injury in a population. D A L Y quantifies how many years a given population has lost due to premature death and disability resulted from an injury, in addition to provide information on the relative importance of mortality and morbidity in the total burden of each external cause of injury. This information can be useful to identify areas of larger health gains from interventions directed at reducing the overall burden of injury (e.g., when deciding among different interventions for different external causes of injury, which intervention will produce the larger D A L Y gain per dollar spent?). In contrast, rates quantify the scale of the problem by measuring how many people have died or been hospitalized in a given time period due to injury. This information can be useful for service Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 116 planning and evaluate coverage of current services (e.g., are the available number of hospital beds sufficient for the demand?). For example, consider the ranking of falls and road traffic accidents for females according to the five different health status indicators in Figure 4.10. Note that falls had the highest A S M R and ASHR, whereas road traffic accidents had the highest A S Y L L , A S Y L D and A S D A L Y during 1991-2000. It is important to mention that this change in ranking is due to the assumptions made about the quantities life expectancy, disability weights, durations and discounting rate necessary to calculate Y L L , Y L D and D A L Y ; and also to the fact that those individuals who died or were hospitalized due to falls tended to be older than individuals that suffered a road traffic accident (Tables 4.1 and 4.2, pages 93-95), which resulted in a higher number of healthy years lost due to premature mortality and disability attributed to road traffic accidents. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 117 Males • Road traffic accidents 0 ASMR ASHR ASYLL ASYLD ASDALY a Other transport injuries 1 - A A • # A Poisoning 2 - • » • A - Falls 3 - A • A A CD X Burns/fires/scalds CD > 4 - — X X + • Of) CD 5 - X _ — . • Drowning O) x : w 6 - • A X X - Sports injuries ireser 7 - + • A •> O Natural and environmental factors 2> 8 - • • • • + c -9 - 0 X 0 X • + Machinery injuries ankii 10 - X A X _ X o Suffocation and foreign bodies cc 11 - + \ / + A • • Other unintentional injuries 12 - o o o 0 o (excluding AE) 13 - - - • - A Suicide and self-inflicted injuries 1 A X Other intentional injuries I H Note: Ranking (1) was given to the injury with the highest value for a particular health status indicator. ASMR: age-standardized mortality rate per 10,000 population; ASHR: age-standardized hospitalization rate per 1,000 population; A S Y L L : age-standardized years of life lost per 1,000 population; A S Y L D : age-standardized per capita healthy years lost due to disability per 1,000 population; A S D A L Y : age-standardized per capita D A L Y per 1,000 population; Standard population: Canada Census 2001. Figure 4.9. Ranking of External Causes of Injury within Each Age-Standardized Health Status Indicator for A l l Ten Years of Data Combined. Males, 1991-2000. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 118 0 1 2 3 4 5 6 7 8 i 9 I 10 ca 11 12 13 14 ASMR ASHR ASYLL -Females ASYLD ASDALY • • # • • • A A A A __ A • — _ A X - X 0 X V X X X • A • X 0 0 <> • X • X • + Q o • 0 A • • o • o o -1 - • -• Road traffic accidents • Other transport injuries A Poisoning - Falls X Burns/fires/scalds • Drowning - Sports injuries o Natural and environmental factors - f Machinery injuries o Suffocation and foreign bodies • Other unintentional injuries (excluding AE) A Suicide and self-inflicted injuries X Other intentional injuries Note: Ranking (1) was given to the injury with the highest value for a particular health status indicator. ASMR: age-standardized mortality rate per 10,000 population; ASHR: age-standardized hospitalization rate per 1,000 population; A S Y L L : age-standardized years of life lost per 1,000 population; A S Y L D : age-standardized per capita healthy years lost due to disability per 1,000 population; A S D A L Y : age-standardized per capita D A L Y per 1,000 population; Standard population: Canada Census 2001. Figure 4.10. Ranking of External Causes of Injury within Each Age-Standardized Health Status Indicator for A l l Ten Years of Data Combined. Females, 1991-2000. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 119 4.2 The Health Authority Level Analysis 4.2.1 Conventional Mortality Rate The external causes of injury responsible for the highest age-standardized mortality rate (ASMR) per 10,000 population varied considerably across health authorities (HA) (see Table 4.10). The highlighted cells in Table 4.10 identify the external causes of injury for which the A S M R of the H A was higher than the A S M R of BC. In addition, cells with the A S M R in bold text represent the H A with the maximum A S M R for each external cause of injury. For males, the highest ASMRs across different injury categories were concentrated in the Northern HA, whereas for females the highest ASMRs appeared more spread out (see Table 4.10). As in the BC level analysis, there were injury categories associated with rare mortality events (e.g., sports injuries, natural and environmental factors, machinery injuries), which made the associated mortality rates susceptible to substantial regional chance variation. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 120 Table 4.10. Ten-Year Age-Standardized Mortality Rate (ASMR) per 10,000 Population by Gender and Health Authority (HA), 1991-Canada Census 2001. Description Age-standardized mortality rate per 10,000 population (1991-2000) Males Interior Fraser Vancouver Coastal Vancouver Island Northern BC Road traffic accidents 25.824 6.172 9.025 13.141 3 0 . 0 0 5 15.649 Other transport injuries 4.040 0.282 1.847 3.072 8 . 5 1 9 3.009 Poisoning 8.138 3.035 2 3 . 4 9 7 13.413 9.220 14.507 Falls 9.118 7.754 10.041 9.865 1 1 . 1 9 1 8.965 Burns/fires/scalds 1.636 0.703 1.020 1.126 3 . 0 4 8 1.233 Drowning 2.642 0.281 1.731 2.577 3 . 4 4 2 2.104 Sports injuries 0.183 0.019 0.189 0.031 0 . 2 2 6 0.117 Natural and environmental factors 1.879 0.156 0.480 0.549 2 . 6 9 4 0.843 Machinery injuries 1.656 0.014 0.199 0.984 3 . 0 4 3 0.951 Suffocation and foreign bodies 1.412 0.780 1.719 1.461 2 . 3 5 2 1.501 Other unintentional injuries* 2.522 0.048 0.823 1.987 3 . 6 4 1 1.580 Suicide and self-inflicted injuries 2 4 . 4 1 5 5.089 18.798 22.029 23.254 20.035 Other intentional injuries 4.287 2.098 4.749 3.629 6 . 3 3 8 4.402 Total 87.751 26.431 74.118 73.863 1 0 6 . 9 7 2 74.895 Excludes medical adverse events Note: The highlighted cells represent the external causes of injury in which the A S M R of the H A was higher than the A S M R of British Columbia. The cells with A S M R in bold text represent the H A with the maximum A S M R for each external cause of injury. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 121 Table 4.10. (continued) Ten-Year Age-Standardized Mortality Rate (ASMR) per 10,000 Population by Gender and Health Authority (HA), 1991-2000 (Canada Census 2001). Description Age-standardized mortality rate per 10,000 population (1991-2000) Females Interior Fraser Vancouver Coastal Vancouver Island Northern B C Road traffic accidents 11.797 6.113 5.135 5.551 11.642 7.048 Other transport injuries 0.726 0.279 0.404 0.324 0.935 0.450 Poisoning 3.343 3.039 6.792 4.046 3.466 4.278 Falls 11.555 8.019 9.471 13.150 12.784 10.284 Burns/fires/scalds 0.806 0.700 0.301 0.849 1.308 0.694 Drowning 0.834 0.277 0.571 0.918 0.548 0.584 Sports injuries 0.032 0.020 0.042 0.058 0.000 0.032 Natural and environmental factors 0.456 0.155 0.081 0.255 0.637 0.235 Machinery injuries 0.158 0.014 0.000 0.024 0.142 0.047 Suffocation and foreign bodies 0.802 0.780 1.098 1.288 0.269 0.940 Other unintentional injuries* 0.225 0.047 0.172 0.220 0.423 0.167 Suicide and self-inflicted injuries 6.009 5.067 6.449 7.147 4.073 5.880 Other intentional injuries 3.057 2.091 1.900 2.247 2.435 2.249 Total 39.801 26.602 32.414 36.077 38.661 32.888 Excludes medical adverse events Note: The highlighted cells represent the external causes of injury in which the A S M R of the H A was higher than the A S M R of British Columbia. The cells with A S M R in bold text represent the H A with the maximum A S M R for each external cause of injury. The annual A S M R per 10,000 population from 1991 to 2000 by gender for each H A and BC for the same external causes of injury as in the B C level analysis are presented in Figures 4.11 and 4.12. Note that the annual ASMRs for the external causes of injury in these figures appear susceptible to large chance variation, and it was difficult to see any pattern over time for these ASMRs. However, it was clear that for both genders, road traffic accidents in the Northwest and Interior HAs had ASMRs consistently higher than B C and the Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 122 other health regions during 1991-2000. In addition, poisoning (mainly due to drug abuse) afflicted mainly males in this province especially in the Vancouver Coastal HA. Road Traffic Injuries Poisoning - i 1 1 r* 1992 1994 1996 1998 2000 Falls A . • y \ A — i i 1 1 1992 1994 1996 1998 2000 Suicide and Self-inflicted Injury Other Intentional Injuries — i 1 1 1 1992 1994 1996 1998 2000 1998 2000 Interior Fraser Vancouver Coastal — • — Vancouver Island Northwest RC. Figure 4.11. Annual Age-Standardized Mortality Rate (ASMR) per 10,000 Population by. Injury and Health Authority for Males, 1991-2000 (Canada Census 2001). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 123 Figure 4.12. Annual Age-Standardized Mortality Rate (ASMR) per 10,000 Population by Injury and Health Authority for Females, 1991-2000 (Canada Census 2001). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 124 4.2.2 Conventional Hospitalization Rate The external causes of injury responsible for the highest age-standardized hospitalization rate (ASHR) per 1,000 population varied considerably across HAs (Table 4.11). For males and females, the highest ASHRs were concentrated in the Northern H A for all injury categories. The Interior H A also had a high ASHR for sports injuries. Table 4.11. Ten-Year Age-Standardized Hospitalization Rates (ASHR) per 1,000 Population by Gender and Health Authority (HA), 1991-2000 (Canada Census 2001). Description Age-standardized hospitalization rate per 1,000 population (1991-2000) Males Interior Fraser Vancouver Coastal Vancouver Island Northern BC Road traffic accidents 62.88 38.41 28.50 42.58 66.31 42.95 Other transport injuries 12.03 5.05 2.94 7.19 16.28 7.02 Poisoning 3.69 2.69 3.43 3.23 4.68 3.31 Falls 64.98 49.60 44.81 56.51 74.28 54.05 Burns/fires/scalds 4.75 3.18 3.16 4.20 9.41 4.13 Drowning 0.27 0.20 0.17 0.28 0.34 0.23 Sports injuries 18.12 12.99 11.28 16.37 16.88 14.36 Natural and environmental factors 3.49 1.36 1.17 1.93 4.79 2.03 Machinery injuries 12.68 7.25 3.71 8.74 17.26 8.29 Suffocation and foreign bodies 3.93 3.15 2.43 3.17 4.26 3.20 Other unintentional injuries* 28.25 14.79 11.25 20.21 37.00 18.85 Suicide and self-inflicted injuries 11.35 11.02 8.94 12.32 12.29 10.84 Other intentional injuries 15.74 13.20 13.85 15.76 27.38 15.36 Total 242.17 162.89 135.67 192.48 291.18 184.63 Excludes medical adverse events Note: The highlighted cells represent the external causes of injury in which the A S H R of the H A was higher than the A S H R of British Columbia. The cells with ASHR in bold text represent the H A with the maximum A S H R for each external cause of injury. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 125 Table 4.11. (continued) Ten-Year Age-Standardized Hospitalization Rates (ASHR) per 1,000 Population by Gender and Health Authority (HA), 1991-2000 (Canada Census 2001). Description Age-standardized hospitalization rate per 1,000 population (1991-2000) Females Interior Fraser Vancouver Coastal Vancouver Island Northern BC Road traffic accidents 39.38 24.06 18.55 24.80 44.03 27.08 Other transport injuries 3.04 1.34 1.12 2.15 3.69 1.94 Poisoning 3.80 2.53 2.88 2.87 4.69 3.04 Falls 77.19 61.80 54.35 66.89 95.98 62.20 Burns/fires/scalds 2.25 1.93 1.66 1.95 3.93 2.05 Drowning 0.13 0.07 0.06 0.11 0.06 0.09 Sports injuries 8.01 4.25 4.08 6.06 7.62 5.48 Natural and environmental factors 2.27 1.14 0.85 1.37 2.50 1.39 Machinery injuries 0.79 0.59 0.47 0.85 1.77 0.73 Suffocation and foreign bodies 2.47 2.25 1.78 2.40 2.91 2.21 Other unintentional injuries* 7.57 4.43 3.94 5.59 9.40 5.49 Suicide and self-inflicted injuries 18.59 17.96 13.32 19.33 24.05 18.12 Other intentional injuries 5.75 3.94 4.27 4.78 12.74 5.31 Total 171.25 126.29 107.34 139.14 213.38 135.12 Excludes medical adverse events Note: The highlighted cells represent the external causes of injury in which the A S H R of the H A was higher than the A S H R of British Columbia. The cells with A S H R in bold text represent the H A with the maximum A S H R for each external cause of injury. Similar to Figures 4.11 to 4.12, Figures 4.13 to 4.14 present the annual ASHRs for all injuries from 1991 to 2000 for each H A and BC as a whole. In contrast to the plots for A S M R , the larger sample size associated with the number of hospitalizations for each injury made it easier to see the pattern over time of the ASHR for some external causes of injury. It is apparent that for road traffic accidents and falls in both genders, the Northwest and Interior HAs had ASHR higher than both the province as whole and the other health regions for all Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 126 years. The Northwest H A also had a high ASHR due to other intentional injuries for both males and females. In general, for both males and females, the ASHR for road traffic accidents, falls, and other intentional injuries has been decreasing since 1991 in BC and in most HAs. For suicide and self-inflicted injuries, the ASHR for females has been increasing since 1991 in the Northwest HA. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 127 „ . . . • A u i J i ' Other Intentional Injuries Suicide and Self-inflicted Injury —i 1 1 1 H i 1 1 1 r 1 1992 1 994 1 996 1 998 2000 1 992 1994 1 996 1 996 2000 Figure 4.13. Annual Age-Standardized Hospitalization Rate (ASHR) per 1,000 Population by Injury and Health Authority for Males, 1991-2000 (Canada Census 2001). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 128 Suicide and Self-inflicted Injury Q t h e r MenlAo™\ Injuries 1992 1994 1996 1998 2000 1992 1994 1996 1998 2000 Figure 4.14. Annual Age-Standardized Hospitalization Rate (ASHR) per 1,000 Population by Injury and Health Authority for Females, 1991-2000 (Canada Census 2001). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 129 4.2.3 Disability-Adjusted Life Years (DALY) This section of the results presents the age-standardized per capita disability-adjusted life years (ASDALY) per 1,000 population for each H A and B C (see Table 4.12). For males during 1991-2000, for every 1,000 population, 364.60 D A L Y s (or approximately 365 lost years of healthy life due to injury) were lost in the Northern H A due to injuries (primarily due to mortality), followed by 307.82 D A L Y s (or approximately 308 lost years of healthy life due to injury) lost in the Interior H A due to injuries (primarily due to mortality). For females, in 1991-2000, per 1,000 population, 131.95 D A L Y s (or approximately 132 lost years of healthy life due to injury) were lost in the Northern H A due to injuries (primarily due to mortality) and 117.10 D A L Y s (or approximately 117 lost years of healthy life due to injury) were lost in the Interior H A due to injuries (primarily due to mortality). For males and females in the Interior, Fraser and Northern HAs, road traffic accidents had the highest A S D A L Y (see Table 4.12). In the Vancouver Island H A , the A S D A L Y was the highest for suicide and self-inflicted injuries. The Vancouver Costal H A had the highest A S D A L Y for poisoning (mainly due to drug abuse) (for males) and suicide and self-inflicted injuries (for females). In addition, note the high variability in A S D A L Y for some external causes of injury across different HAs. For example, the A S D A L Y for road traffic accidents ranged from 29 to 88 per 1,000 population in males and from 15 to 39 per 1,000 population in females. With regard to poisoning, the A S D A L Y ranged from 19 to 52 per 1,000 population for males and from 7 to 16 per 1,000 population for females. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 130 Table 4.12. Ten-Year Age-Standardized Per Capita Disability-Adjusted Life Years (or age-Standardized Per Capita D A L Y or A S D A L Y ) per 1,000 Population by Gender and Health Authority, 1991-2000 (Canada Census 2001). Description Age-standardized per capita DAI.Y per 1.000 population (1991-2000) Males Interior Fraser Vancouver Coastal Vancouver Island Northern BC: Road traffic accidents 80.65 42.93 29.14 44.95 87.67 49.82 Other transport injuries 16.37 7.52 5.64 11.64 28.16 10.98 Poisoning 19.17 27.91 51.50 30.33 20.56 32.45 Falls 20.13 15.28 18.05 19.64 21.44 18.01 Burns/fires/scalds 7.05 4.01 4.46 5.42 12.58 5.59 Drowning 6.63 3.62 3.75 6.1M 8.32 5.02 Sports injuries 2.8" 1.82 1.92 2.17 2.55 2.16 Natural and environmental factors 5.62 1.25 1.52 2.09 6.56 2.59 Machinery injuries 18.60 9.93 4.84 12.29 25.08 11.69 Suffocation and foreign bodies 15.d' 12.30 10.24 12.99 15.73 12.68 Other unintentional injuries* 44.72 22.74 17.79 32.81 57.19 29.72 Suicide and self-inflicted injuries 55.74 38.18 41.47 51.25 54.50 45.49 Other intentional injuries 15.25 14.02 15.68 13.99 24.26 15.56 Total 307:82 201.50 205.99 245.75 364.60 241.77 Excludes medical adverse events Note: The highlighted cells represent the external causes of injury in which the A S D A L Y of the H A was higher than the A S D A L Y of British Columbia. The cells with A S D A L Y in bold text represent the H A with the maximum A S D A L Y for each external cause of injury. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 131 Table 4.12. (continued) Ten-Year Age-Standardized Per Capita Disability-Adjusted Life Years (or age-Standardized Per Capita D A L Y or A S D A L Y ) per 1,000 Population by Gender arid Health Authority, 1991-2000 (Canada Census 2001). Xye-standardi/ed per capita D A L Y per 1,000 population i i w 1-2001 n Description Females Interior Vancouver Vancouver Northern rraser Coastal Island Road traffic accidents 37.02 19.51 15.47 20.69 39.07 23.22 Other transport injuries 3.6" 1.40 1.49 2.02 5.02 2.26 Poisoning 7.27 7.18 15.71 9.70 8.94 10.04 Falls 14.93 10.70 11.59 14.73 17.42 12.88 Burns/fires/scalds 3.49 2.78 1.66 3.16 5.99 2.96 Drowning 1.80 0.73 1.08 2.09 1.18 1.28 Sports injuries 0.73 0.36 0.47 0.57 0.58 0.50 Natural and environmental factors 1.9n 0.68 0.53 1.00 2.15 1.03 Machinery injuries 1.32 0.77 0.59 1.14 2.55 1.02 Suffocation and foreign bodies 6.50 6.45 5.29 7.64 7.10 6.51 Other unintentional injuries1 10.26 5.76 5.56 7.78 12.78 7.40 Suicide and self-inflicted injuries 17.89 16.23 16.54 20.88 16.81 17.54 Other intentional injuries 10.30 6.71 6.49 7.63 12.34 7.92 Total 117.10 79.28 82.45 99.02 131.95 94.59 Excludes medical adverse events Note: The highlighted cells represent the external causes of injury in which the A S D A L Y of the H A was higher than the A S D A L Y of British Columbia. The cells with A S D A L Y in bold text represent the H A with the maximum A S D A L Y for each external cause of injury. Tables G.10 and G . l l in Appendix G present the Y L L , Y L D and D A L Y by gender and external cause of injury for B C and each H A for the years 1991-2000. Tables G.10 and G . l l also display the percentage contribution of non-fatal burden to the total burden from injuries. For both males and females in all regions, the injuries with Y L D > Y L L were burns/fires/scalds, sports injuries, machinery injuries, suffocation and foreign bodies and Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 132 other unintentional injuries (excluding medical AE). In addition, for females, natural and environmental factors had Y L D greater than Y L L in all regions. Finally, for falls, the Y L D was greater than Y L L in the Interior H A (males), Fraser H A (males) and Northern H A (males and females). Figures 4.15 to 4.16 present annual A S D A L Y from 1991 to 2000 for each H A and BC for the same injuries as in plots for A S M R and ASHR. For both males and females, the A S D A L Y for road traffic accidents tended to be higher for the Northwest and Interior HAs than for BC and the other HAs. Lost years of healthy life due to poisoning (mainly due to drug abuse) were more common in males in this province especially in the Vancouver Coastal HA. Additionally, individuals living in the Northern H A lost years of healthy life due to falls (females) and other intentional injury (mainly due to homicide) (males and females) in all years. For suicide and self-inflicted injury the Interior H A (males), the Northern H A (males), and the Vancouver Island H A (both males and females) had A S D A L Y s above the provincial A S D A L Y over the 10 years of observation. In addition, the A S D A L Y due to road traffic accidents appears to be decreasing for males in all regions and for females in all regions with the exception of the Northern and Interior HAs. Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 133 c • - J ^ o i<. L j . • Other Intentional Injuries Suicide and Self-Inflicted Injury ' 1992 1994 1996 1998 2000 1992 1994 1996 1998 2000 Figure 4.15. Annual Age-Standardized Per Capita D A L Y (ASDALY) per 1,000 Population by Injury and Health Authority for Males, 1991-2000 (Canada Census 2001). Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 134 Road Traffic Injuries Poisoning Falls 1992 1994 1996 1998 2000 year <! / V A / \ ! \ x>\ '• I \\J-1 1 1 1 1992 1994 1996 1998 2000 Suicide and Self-inflicted Injury Other Intentional Injuries year year Figure 4.16. Annual Age-Standardized Per Capita D A L Y (ASDALY) per 1,000 Population by Injury and Health Authority for females, 1991-2000 (Canada Census 2001). In summary, the results of this section of this study show that for both males and females, road traffic accidents represent the injury imposing the greatest impact on the quality of life of BC residents. In other words, BC residents are loosing more years of Chapter 4. Burden of Injuries in British Columbia and Its Health Authorities, 1991-2000 135 healthy life due to disability and premature death from road traffic accidents than from any other single type of injury. In addition, road traffic accidents affect residents of Northern and Interior health authorities to greater extent than residents of other HAs. 136 Chapter 5 Small-area Analysis of the Burden Due to R o a d Traf f ic Accidents in Bri t ish C o l u m b i a Given the relatively high impact of road traffic accidents among males and females between 20 and 39 years of age, this injury category was selected to demonstrate the utility of Bayesian disease mapping methods to examine small-area patterns, variations, and trends in disability-adjusted life years (DALY). More generally, this type of analysis can be used for monitoring the dynamic evolution of patterns in the spatial distribution of injury incidence and burden over time. The results for the estimated parameters for model (3.1) (described in section 3.2.1), based on the hospitalization and death data of males and females, can be found in Appendix G. 5.1 Estimated Mortality and Hospitalization Rates For the 10 years of observation (i.e., 1991-2000), the estimated BC average mortality rate (parameter m in model 3.1, page 55) due to road traffic accidents was 4 per 10,000 population for males and 0.9 per 10,000 population for females. In the same period in B C , the estimated average hospitalization rate (parameter m) due to road traffic accidents was 7.6 per 1,000 population for males and 3.7 per 1,000 population for females. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 137 Columbia To illustrate the model fitting for hospitalization and mortality data, respectively, we plotted the annual conventional and estimated rates by H S D A for males and females during 1991-2000 (Figures 5.1-5.2). In general, the conventional hospitalization and mortality rates fluctuated over the period of observation because of the increased influence of chance variation associated with small case numbers at the HSDA level. However, the estimated hospitalization and mortality rates once smoothed showed a clear trend during the period 1991-2000. In addition, Figure 5.1 shows a decreasing trend in the smoothed estimated mortality rates for most regions from 1991 to 2000, and Figure 5.2 shows moderate curvature in the trend of the smoothed estimated hospitalization rates for most regions from 1991 to 2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 138 Columbia East Koolen ay Kooten, V 1992 19$ ay/Boundary 19S2 19! 1992 19! 1992 1993 1992 1998 year 1992 199B 1992 1996 year 1992 1993 year 1992 1996 year 1992 1993 year 1992 19! year 1992 199S year Northern Interior 1992 1998 year 1992 1998 Males i i 1 * Males rate Con 4 6 K 1 1992 1998 year ft I 1992 1998 year 1992 199S k ' I ij 1992 19! 1992 199B 1992 19! 1992 1998 year North Shore/Coast Gari baldi South Vancouver Island Central Vancouver Island North Vancouver Island its. 1992 1998 1992 19! year 1992 1998 year 1992 1998 year 1992 1998 L 1992 1998 year 1992 1998 year 1992 19! year 1992 1998 I i l | 1992 1998 Figure 5.1. Conventional and Estimated Annual Mortality Rates per 10,000 Population by Health Services Delivery Area for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 139 Columbia East Kocten ey Mates Kooton ay/Boundary Males Okanagan Males Thompson Cariboo Shu swap , Males , Fraser Val lay Males Simon Fraser Males South Fraser Males Richmond Males _ Conventional rate _. 8 10 12 14 16 A _ Conventional rate _, 8 10 12 14 16 _ Conventional rate _. 8 10 12 14 16 r \ _ Conventional rate _ 8 10 12 14 16 \ _ Conventional rate _ 8 10 12 14 16 \ Conventional rate _ 8 10 12 14 16 Conventional rate _ 8 10 12 14 16 Conventional rate _ 8 10 12 14 16 Estimated rate _ 2 4 6 Estimated rate _ 2 4 6 i y Estimated rate. 2 4 6 Vv Estimated rate. 2 4 6 Estimated rate, 2 4 6 Estimated rate. 2 4 6 Estimated rate. 2 4 6 Estimated rate. 2 4 6 1992 1998 1992 1998 1992 1998 1992 1998 1992 1996 1992 1998 1992 1998 1992 1998 year year year year year year year yaar Vancouver North Shore/Coast Gari bald! South Vancouver Island Central Vancouver Island North Vancouver Island Northwest Northern Interior Northeast Males , Males ( Males ( Males { Males , Males ' , Males , Males 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 year year year year year year year year East Kooten ay Kooton ay/Boundary Okanagan Thompson Cariboo Shuswap Fraser Valley Simon Fraser South Fraser Richmond Females . Females . Females . Females . Females . Females . Females . Females 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 year year year year yaar year year year Figure 5.2. Conventional and Estimated Annual Hospitalization Rates per 1,000 Population by Health Services Delivery Area for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 140 Columbia 5.1.1 Comparison of the Annual Estimated Rates among Different Geographical Layers in the Provincial Health Care System Table 5.1 shows the mortality and hospitalization rates in 2000 associated with road traffic accidents for males and females, respectively. The cells in Tables 5.1 with the rates in boldface represent the HAs with rates statistically higher or lower than the province of BC as a whole. Appendix G presents the estimated rates and confidence intervals set at the HSDA, H A and BC levels for all years. Overall, Table 5.1 shows that in 2000, males living in the Thompson Cariboo Shuswap represented the H S D A with the highest hospitalization and mortality rates. The Vancouver and Richmond HSDAs had the lowest mortality and hospitalization rates, respectively. And for females in 2000, the Thompson Cariboo Shuswap H S D A had the highest mortality rate while the Northeast and East Kootenay HSDAs had the highest hospitalization rate. Richmond and Vancouver HSDAs had the lowest mortality and hospitalization rates, respectively. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 141 Columbia Table 5.1. Regional Age-Specific Mortality Rates per 10,000 Population and Hospitalization Rates per 1,000 Population Due To Road Traffic Accidents in 2000 for Males and Females Aged 20-39 Years. HSDA Age-specific roa mortality rale per ' in 2000 J iralTic accident 10,000 population 95%CIJ Age-specilic road traffic accident hospitalization rate per 1,000 population in 2000 [95%CI| Males Females Males Females Interior HA 2.46 |2.22. 2.67| 1.01 |0."7, 1.16| 7.51 |7.32,7.69] 4.41 |4.23. 4.57| East Kootenay 1.76 [1.02, 2.78] 0.93 [0.34, 2.59] 6.11 [4.85,7.45] 7.36 [5.41, 8.38] Kootenay/Boundary 2.49 [1.51, 3.87] 0.90 [0.35,2.18] 8.37 [6.83, 10.0] 2.17 [1.54, 3.02] Okanagan 1.95 [1.24, 2.73] 0.52 [0.15, 0.95] 6.69 [5.93, 7.47] 3.82 [3.22, 4.41] Thompson Cariboo Shuswap 3.35 [2.26, 4.60] 1.68 [0.70, 3.58] 8.78 [7.79, 9.83] 4.68 [4.10, 5.581 Fraser HA 1.39 [1.27, 1.50] 0.42 |0.34, 0.50| 4.01 |3;92, 4.10| 1.86 |1.79,1.93| Fraser Valley 1.81 [1.17,2.63] 0.57 [0.20, 1.06] 4.18 [3.58,4.83] 2.19 [1.86, 2.84] Simon Fraser 1.01 [0.70, 1.40] 0.33 [0.11,0.59] 3.84 [3.47,4.24] 1.25 [1.15, 1.63] South Fraser 1.61 [1.15,2.17] 0.38 [0.13, 0.65] 4.12 [3.72, 4.56] 1.73 [1.61,2.17] Vancouver Coastal HA 0.88 |0.79, 0.96| 0.41 |0.32, 0.47| 3.37 |3.28, 3.46| 1.64 |1.57, 1.71] Richmond 0.97 [0.56, 1.53] 0.31 [0.08,0.68] 2.31 [1.85,2.83] 1.10 [0.87, 1.591 Vancouver 0.69 [0.45, 0.99] 0.33 [0.14, 0.58] 2.80 [2.52, 3.10] 1.06 [1.01, 1.391 North Shore / Coast Garibaldi 1.40 [0.93,2.09] 0.58 [0.24, 1.43] 5.93 [5.21, 6.70] 2.47 [2.07, 3.05] Vancouver Island HA 1.41 [1.27, 1.54] 0.61 [0.49, 0.70] 5.32 [5.17,5.47]. 2.30 |2.17, 2.43] South Vancouver Island 1.14 [0.76, 1.69] 0.41 [0.15,0.74] 4.68 [4.15, 5.25] 1.60 [1.43, 2.06] Central Vancouver Island 1.81 [1.12, 2.62] 0.66 [0.23, 1.22] 6.01 [5.24, 6.86] 2.71 [2.18,3.28] North Vancouver Island 1.75 [0.98,2.79] 0.75 [0.25, 1.83] 7.10 [5.63, 8.73] 1.97 [1.62, 3.50] Northern 11A 2.62 |2.35, 2.87| 0.75 [0.60, 0.86] 6.82 [6.59, 7.04] 4.94 |4.71,5.17| Northwest 2.36 [1.43, 3.48] 0.60 [0.16, 1.21] 6.85 [5.70, 8.10] 4.25 [3.07, 4.921 Northern Interior 2.61 [1.74, 3.65] 0.83 [0.32, 1.56] 5.36 [4.56, 6.23] 3.61 [3.14, 4.55] Northeast 3.00 [1.84,4.56] 0.81 [0.27, 1.81] 10.0 [8.41, 11.9] 7.69 [5.63, 8.83] BC 1.69 [1.20,2.32] 0.66 [0.57, 0.74] 5.47 [4.26,6.91] 3.12 [3.04,3.18] Figures 5.3-5.4 show, respectively, the annual trends for road traffic accident mortality and hospitalization rates for males during 1991-2000. The display offered by these figures provides a graphical summary of the provincial temporal trend, the trend for health authorities with temporal trends over and above the provincial trend, as well as health service delivery areas with temporal trends greater or lower than their respective health authority trend. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 142 Columbia Road Traffic Accidents - Males Interior Health Authority Road Traffic Accidents - Males Fraser Health Authority E q Simon Frosw Soulfi Fraser FIMW HA Figure 5.3. Annual Road Traffic Accident Mortality Rate per 10,000 Population for Males Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 143 Columbia Road Traffic Accidents - Males Interior Health Authority Road Traffic Accidents - Males Fraser Health Authority Frosar VaMy Simon FfOMT South FfnMf Fnwnr HA Road Traffic Accidents - Males Vancouver Coastal Health Authority Road Traffic Accidents - Males Vancouver Island Health Authority Road Traffic Accidents - Males Road Traffic Accidents - Males Northern Health Authority : British Columbia year year Figure 5.4. Annual Road Traffic Accident Hospitalization Rate per 1,000 Population for Males Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 144 Columbia In summary, based on Figures 5.3-5.4, it appears that the overall annual mortality and hospitalization rates for the province and the health authorities decreased in a non-linear fashion over the decade of observation. In 1991, note that the rates for most health service delivery areas and health authorities were further apart, and they became smaller as we approached the year 2000. In addition, though hospitalizations due to road traffic accidents appear to be declining for all health authorities, there are certain health service delivery areas for which this decline is far faster than the health authorities (e.g., the East Kootenay, Fraser Valley, Richmond and Northern Interior HSD As). There are also health service delivery areas with rates increasing during the last three years of study, such as the North Shore/Coast Garibaldi and Northeast HSD As. Next, Figures 5.5-5.6 show, respectively, the annual trend for road traffic accident mortality and hospitalization rates for females during 1991-2000. Similar to males, the overall annual mortality rates in females for both the province as a whole and individual health authorities decreased by means of a non-linear temporal trend over the decade of observation. In contrast to the overall annual mortality rates, the overall annual hospitalization rates had a flat trend, except for the Vancouver Island and Fraser health authorities. Note that in the Interior HA, the hospitalization trend of each HSDA was very distinct. In addition, there are certain health service delivery areas with rates increasing during the last three years of study, such as the North Shore/Coast Garibaldi HSDA (for both mortality and hospitalization rates), and the East Kootenay, Thompson Cariboo Shuswap, Richmond and Northeast HSD As (for only the hospitalization rate). Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 145 Columbia Road Traffic Accidents - Females Interior Health Authority East Kootenay Kootenay/Boundary Okanagan Thompson Canboo Shuswap ~] 1 1 r 1992 1994 1996 1996 2000 Road Traffic Accidents - Females Fraser Health Authority Fraser Valley Simon Fraser South Fraser Franw HA 1 1 1 1 1 ' 1992 1994 1996 1998 2000 Road Traffic Accidents - Females Road Traffic Accidents - Females Vancouver Coastal Health Authority Vancouver Island Health Authority year year Road Traffic Accidents - Females Road Traffic Accidents - Females Northern Health Authority British Columbia 1992 • 1994 1996 1998 2000 1992 1994 1996 1998 2000 year year Figure 5.5. Annual Road Traffic Accident Mortality Rate per 10,000 Population for Females Aged 20-39 Years, at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 146 Columbia Road Traffic Accidents - Females Road Traffic Accidents - Females Vancouver Coastal Health Authority Vancouver Island Health Authority 1992 1994 1996 1998 2000 1992 1994 1996 199B 2000 year year Road Traffic Accidents - Females Road Traffic Accidents - Females Northern Health Authority British Columbia year year Figure 5.6. Annual Road Traffic Accident Hospitalization Rate per 1,000 Population for Females Aged 20-39 Years, at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 147 Columbia 5.1.2 Mortality and Hospitalization Rate Ratios To monitor emerging regional variations, one may plot injury rate ratios during 1991-2000 (Figure 5.7). This figure indicates which HSDAs had rates above (ratio>l) or below (ratio<l) the provincial rate during 1991-2000. For males and females, the ratio between the mortality rates of most HSDAs and BC was flat during 1991-2000. For females, the ratios for the East Kootenay, Kootenay/Boundary and Thompson Cariboo Schuswap HSDAs increased steadily during 1991-2000. And to monitor emerging patterns over time, one may map injury rate ratios during 1991-2000 (Figure 5.8). The resulting maps provide a visual summary of the spatio-temporal pattern of the mortality ratio estimates during the 10-year study period. Overall, for males, the geographical pattern in the year 2000 was very similar to the pattern in 1991, with no abrupt change over the years. However, for females there was a gradual change in the geographical distribution of the mortality rate ratio from 1991 onward. Additionally, in 2000 more HSDAs had mortality rates higher than the provincial rate. Additionally, for males in the year 2000, we observed that the Thompson Cariboo Shuswap (ratio=2.01, 95%CI [1.32,2.92]), Northern Interior (ratio=1.56, 95%CI [1.03,2.31]) and Northeast (ratio=1.80, 95%CI [1.11,2.80]) HSDAs had mortality rate ratios that were significantly higher than 1. The Simon Fraser (ratio=0.61, 95%CI [0.40,0.88]), Richmond (ratio=0.59, 95%CI [0.34,0.93]) and Vancouver (ratio=0.42, 95%CI [0.27,0.63]) HSDAs had mortality rate ratios that were significantly less than 1. For females in the year 2000, the Thompson Cariboo Shuswap HSDA (ratio=3.22, 95%CI [1.33,7.91]) had a mortality rate ratio significantly higher than 1. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 148 Columbia 5 Mortality rate ratio 1-0 1.5 Mortality rate ratio 1.0 1.5 Mortality rate ratio 1.0 1.5 MortaJity rate ratio 1-0 1.5 Mortality rate ratio 1.0 1.5 Mortality rate ratio 1.0 1.5 Mortality rate rabo 1.0 1.5 Mortality rate ratio 1.0 1.5 s s 5 S 5 992 1996 992 1998 992 1998 992 1998 992 1998 992 1998 992 1998 992 1998 1 5 1.5 2.0 1.5 2.0 ratio 1.5 2.0 ratio 1.5 2.0 ratio 1.5 2.0 ratio 1.5 2.0 ratio 1.5 2.0 ratio 1.5 2.0 rate rate rate rate rate rate rate / V. s Morta 0.5 1.t £ Morta 0.5 l.t 1 5 £ Morta 0.5 1.C Morta 0.5 1.t £ 992 1998 992 1998 1992 1998 992 1998 992 1998 1992 1998 992 1998 992 1998 year year year year year year year year 1992 1998 year 1992 1998 1992 1998 year 3 ° 1992 1998 1992 19E 1992 19S North Shore/Coast Gari baldl South Vancouver Island Central Vancouver Island North Vancouver Island 1992 1998 year ate ratio :.0 2.5 3.0 ate ratio .0 2.5 3.0 ate ratio .0 2-5 3.0 ate ratio .0 2.5 3.0 ate ratio 0 2.5 3.0 ate ratio .0 2.5 3.0 ate ratio .0 2.5 3.0 Mortality r 1.0 1.5 2 Mortality r 1.0 1.5 2 Mortality r 1.0 15 2 Mortality r 1.0 1.5 2 Mortality r 10 1.5 2 - ~ «V Mortality r 1.0 1.5 2 Mortality r 1.0 1.5 2 "> 992 1996 992 1998 992 1998 year 992 1998 year 992 1998 year 992 1998 yaar 992 1998 yaar 992 1998 year Figure 5.7. Annual Ratio between the Mortality Rates of Each Health Services Delivery Area and British Columbia for Road Traffic Accidents in Males and Females Aged 20-39 Years, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British Columbia Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 150 Columbia Some results for the analysis of hospitalization data can be found in Figure 5.9. This figure contains annual plots of the ratio between the estimated hospitalization rate of each H S D A and the provincial hospitalization rate due to road traffic accidents for males and females aged 20-39 years. For males and females, the ratios between the hospitalization rates of each H S D A and BC varied significantly between 1991 and 2000, with the exception of a few HSDAs {e.g., the Simon Fraser and Richmond HSDAs). Figure 5.10 contains maps of the annual hospitalization rate ratios for road traffic accidents experienced by males and females. For males, note that in contrast to the relatively unchanging mortality rate ratios, there was a gradual change in the geographical distribution of the hospitalization rate ratios from 1991 onward. For females, there was significant variation in the hospitalization rate ratios across most of the HSDAs between 1991 and 2000. In summary, for males in the year 2000, the HSDAs Kootenay/Boundary (ratio=1.55, 95%CI [1.13,2.10]), Thompson Cariboo Shuswap (ratio=1.63, 95%CI [1.24,2.11]) and Northeast (ratio=1.86, 95%CI [1.38,2.47]) had hospitalization rate ratios significantly higher than 1. The HSDAs Simon Fraser (ratio=0.71, 95%CI [0.55,0.92]), South Fraser (ratio=0.77, 95%CI [0.59,0.99]), Richmond (ratio=0.43, 95%CI [0.31,0.57]) and Vancouver (ratio=0.52, 95%CI [0.40,0.67]) had hospitalization rate ratios significantly less than 1. For females in the year 2000, the HSDAs Kootenay/Boundary (ratio=2.57, 95%CI [1.79,3.60]), Okanagan (ratio=1.42, 95%CI [1.03,1.93]), Thompson Cariboo Shuswap (ratio=1.82, 95%CI [1.30,2.45]), Northwest (ratio=1.48, 95%CI [1.00,2.10]), Northern Interior (ratio=1.44, 95%CI [1.02,1.96]) and Northeast (ratio-2.68, 95%CI [1.87,3.79]) had hospitalization rate ratios that were significantly higher than 1. The HSDAs Simon Fraser Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 151 Columbia (ratio=0.52, 95%CI [0.37,0.72]), South Fraser (ratio=0.71, 95%CI [0.50,0.96]), Richmond (ratio=0.45, 95%CI [0.29,0.66]), Vancouver (ratio=0.45, 95%CI [0.32,0.61]) and South Vancouver Island (ratio=0.65, 95%CI [0.46,0.90]) had hospitalization rate ratios significantly less than 1. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 152 Columbia East Kooteri ay Kootenay/Boundary Okanagan Thompson Cariboo Shuswap FreserValley Simon Fraser South Fraser Richmond Males Males Males Males Males Males Males Males A \ / \ 1 ' 1 1 ' i ~ 1 " 1 " v y \ rate rate rate rate rate rate rate CN ' \ C Of C <N C <N \ 1 1 ' ' I I ' 1 " 1 ** I I 1 ^ \ is 5 o' 1 s 1 3 1 5 \ I s 1 s I X X X X - , 1992 1998 992 1998 1992 1998 1992 1998 1992 1998 1992 199B 992 1998 992 1998 year year year year year year year year Vancouver North Shore/Coast Gari baldl South Vancouver Island Central Vancouver Island North Vancouver Island Northwest Northern Interior Northeast Males Males Males Males Males Males Males Males 1992 1998 1992 1998 1992 1998 1992 1998 1992 1996 1992 1998 1992 1998 1992 1998 year year year year year year year year East Kootenay Kootenay/Boundary Okanagan Thompson Cariboo Shuswap Fraser Valley Simon Fraser South Fraser Richmond Females Females Females Females Females Females Females Females 1992 1998 1992 1998 1992 1998 1992 1998 1992 199B 1992 1998 1992 1998 1992 1998 year year year year year year year year Vancouver North Shore/Coast Gari baldl South Vancouver Island Central Vancouver Island North Vancouver Island Northwest Northern Interior Northeast Females Females Females Females Females Females Females Females 1992 1998 1992 1996 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 year year year year year year year year Figure 5.9. Annual Ratio between the Hospitalization Rates Delivery Area and British Columbia for Road Traffic Accidents Females Aged 20-39 Years, 1991-2000. for each Health Services Experienced by Males and Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British Columbia Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 154 Columbia 5.2 Estimated Disability-Adjusted Life Years Table 5.2 presents the per capita D A L Y per 1,000 population due to road traffic accidents for males and females. The cells in Table 5.2 with the per capita D A L Y s in boldface represent the HSDAs with per capita D A L Y s that were statistically higher or lower than their respective H A and BC. The HAs with the per capita D A L Y s statistically higher or lower than B C are also highlighted using boldface text. Appendix G contains the estimated per capita D A L Y and confidence intervals at the HSDA, H A and B C levels for all years. hi 2000, for males and females, the Thompson Cariboo Shuswap HSDA had the highest per capita D A L Y s , whereas the Vancouver and Richmond HSDAs had the lowest per capita D A L Y s for males and females, respectively. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 155 Columbia Table 5.2. Regional Per Capita Disability-Adjusted Life Years (DALY) per 1,000 Population Due To Road Traffic Accidents in 2000 for Males and Females Aged 20-39 Years. Age-specific road traffic accident per capita 1 > \ l N per 1.000 population in 2000 |95'\. C1J I-'enidles Interior 11.\ 9.65 |9.05. 10.2| 4.61 [3.99,5.02| East Kootenay 7.23 [6.41, 7.90] 5.48 [4.51, 5.88] Kootenay/Boundary 10.1 [8.99, 11.0] 3.35 [2.58, 3.74] Okanagan 7.98 [7.34, 8.59] 3.06 [2.70, 3.39] Thompson Cariboo Shuswap 12.5 [11.4,13.4] 6.56 [5.21, 7.42] Fraser HA 5.36 [5.05, 5.65] 1.95 [1.74, 2.151 Fraser Valley 6.53 [5.86, 7.10] 2.54 [2.15, 2.87] Simon Fraser 4.30 [3.97, 4.61] 1.47 [1.25,1.68] South Fraser 5.97 [5.50, 6.38] 1.83 [1.59,2.06] Vancouver Coastal H A 3.77 [3.53, 3.99| 1.81 [1.58,1.98| Richmond 3.54 [3.09,3.91] 1.36 [1.07, 1.59] Vancouver 3.03 [2.79, 3.25] 1.41 [1.21, 1.59] North Shore / Coast Garibaldi 6.24 [5.69, 6.68] 2.65 [2.14, 2.89] Vancouver Island 1IA 5.98 |5.60. 6.321 2.63 |2.3I,2.87| South Vancouver Island 5.01 [4.58, 5.36] 1.86 [1.57, 2.10] Central Vancouver Island 7.31 [6.63, 7.94] 2.95 [2.50, 3.30] North Vancouver Island 7.64 [6.79, 8.35] 3.08 [2.41, 3.46] Northern HA 9.77 [9.07, 10.4| 4.16 [3.76, 4.49| Northwest 9.12 [8.20, 9.95] 3.32 [2.83,3.73] Northern Interior 9.09 [8.17, 9.90] 3.88 [3.31,4.34] Northeast 12.1 [10.8,13.2] 5.28 [4.58, 5.76] BC 5.96 [5.69. 6.22] 3.13 [2.88, 3.33] Next, Figures 5.11-5.12 show the annual D A L Y trends for road traffic accidents per capita for males and females during 1991-2000. Observe that for males, the mortality trend for each region highly influenced the trend of the regional per capita D A L Y . For females, the trend in per capita D A L Y in each region was influenced by both mortality and hospitalization trends. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British Columbia 156 Road Traffic Accidents - Males Interior Health Authority Road Traffic Accidents - Males Fraser Health Authority Kootoaay/Boundary Okanaoan Thomson Cariboo Shuawap Int.'kv Ha Road Traffic Accidents - Males Road Traffic Accidents - Males Vancouver Coastal Health Authority Vancouver Island Health Authority Road Traffic Accidents - Males Road Traffic Accidents - Males Northern Health Authority British Columbia Figure 5.11. Annual Per Capita Disability-Adjusted Life Years per 1,000 Population for Road Traffic Accidents Experienced by Males Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British Columbia 157 Road traffic Accidents - Females Road Traffic Accidents - Females Interior Health Authority Fraser Health Authority - , • , , , r- 1 ' , , , , r- 1 1992 1994 1996 1998 2000 1992 1994 1996 1996 2000 Road Traffic Accidents - Females Road Traffic Accidents - Females Vancouver Coastal Health Authority Vancouver Island Health Authority —I ' 1 1 1 ! — 1 1 1 I I 1992 1994 1996 1998 2000 1992 1994 1996 1998 2000 Road Traffic Accidents - Females Road Traffic Accidents - Females Northern Health Authority British Columbia NorUrmsl —i 1 1 : 1 1— 1 1 I 1 I 1992 1994 1996 1998 2000 1992 1994 1996 1996 2000 Figure 5.12. Annual Per Capita Disability-Adjusted Life Years per 1,000 Population for Road Traffic Accidents Experienced by Females Aged 20-39 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 158 Columbia Figure 5.13 contains the per capita D A L Y plots for both genders, the ratio between the per capita D A L Y of each HSDA and the provincial per capita D A L Y . Note that for both genders in a few HSDAs, the per capita D A L Y ratio varied significantly during 1991-2000. Kooten ay/Boundary 1992 19. year 1992 1998 1992 1998 1992 1998 1992 1998 year 1992 1998 year Vancouver North Shore/Coast Gari baldi South Vancouver Island Central Vancouver Island North Vancouver Island Northern Interior Males Males Males Males Males Males Males , Males ratio 2.0 ratio 2.0 ratio 2.0 ratio 2.0 ratio 2.0 ratio 2.0 ratio 2.0 > in Per capita Dt 0.5 1.0 Per capita Dt 0.5 1.0 — - Per capita Dt 0.5 1.0 Per capita Dt 0.5 1.0 Per capita Df 0.5 1.0 Per capita Dt 0.5 1.0 Per capita Dt 0.5 1.0 1992 1998 992 1998 1992 1998 1992 1998 1992 1998 992 1998 1992 1998 1992 1998 „ , year year year year year year East Kooten ay Fern ales Kooten ay/Bou ndary Fern ales Okanagan Fem ales Thompson Cariboo Shuswap Females Fraser Val ley Fem ales Simon Fraser Fem ales South Fraser Females Richmond Fem ales • / .Y ratio .5 2.0 .Y ratio .5 2.0 .Y ratio .5 2.0 .Y ratio .5 2.0 .Y ratio 5 2.0 .Y ratio 5 2.0 ,Y ratio 5 2.0 Per capita DAI 0.5 1.0 1 Per capita DAI 0.5 1.0 1 Per capita DAI 0.5 1.0 1 Per capita DAI 05 1.0 1 Per capita DAL 0.5 1.0 1. Per capita DAL 0.5 1.0 1, Per capita DAL 0.5 1.0 1. 1992 19. year 1992 1998 year 1992 1998 year 1992 1998 year 1992 1998 1992 1998 1992 1998 year North Shore/Coast Gari baldl South Vancouver Island Central Vancouver Island North Vancouver Island Northwest Females Females Females Females Females 1992 1998 year 1992 1998 1992 1998 1992 1996 1992 1998 1992 1998 1992 1998 Figure 5.13. Annual Ratio between the PenCapita Disability-Adjusted Life Years of Each Health Services Delivery Area and British Columbia for Road Traffic Accidents Experienced by Males and Females Aged 20-39 Years. Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 159 Columbia To conclude the results presented in this chapter, Figure 5.14 maps the per capita D A L Y ratio during 1991-2000. Observe that these maps have a very different spatial pattern in comparison to the maps of rate ratios for mortality and hospitalization, and gradual changes are observed from 1991 to 2000 in the per capita D A L Y ratios for both males and females. As mentioned in section 3.2, the main goal of using Bayesian disease mapping modeling approach in this study is to remove the chance variation among the spatial units, and at the same time model the systematic variability due to unmeasured risk factors, to then uncover the true underlying variation among different spatial units. The variability in mortality rate among different regions might be explained by risk factors such as differences in mode of transport, road and climate conditions, driver's risk behaviours, number of hours or miles/kilometres travelled, socio-economic factors, and so on. We noticed that even though the mortality rate is decreasing for all regions, the mortality rate ratio showed a consistent pattern over time for most regions, which might be explained by the success of efforts to reduce road traffic accidents in different regions, and thus reducing the case fatality related to road traffic accidents. Note that the hospitalization rates of different regions present a much higher variability over time than the mortality rates (see Tables G.12 and G.13 for the values of o~ ). In addition, observe that in contrast to the mortality rate, the hospitalization rate is not decreasing in all regions, and the hospitalization rate ratio has been constantly changing over time. The risk factors that might have influenced the regional variability in hospitalization rates, beside the ones for mortality, have to do for example, with the annual supply level of Chapter 5. Small-area Analysis of the Burden Due to Road Traffic Accidents in British 160 Columbia health services in different regions, such as the number of medical health professionals, hospital beds, ambulances, and efficiency in the delivery of emergency services. In order to explain the regional variation in D A L Y , remember that D A L Y is a composite measure involving information on both premature mortality and years lived with disability. Note that the per capita D A L Y has been decreasing in all regions, and the per capita D A L Y ratio showed a varying pattern over time for most regions (showing that the D A L Y declining pace has been different among regions). Changes in the spatial distribution of D A L Y over time are largely related to the severity of the accident, type of sequelae caused by the accident, and possibly to the risk factors mentioned above. 162 Chapter 6 A Small -area Analysis of the Incidence of Med ica l Adverse Events in Bri t ish C o l u m b i a In this chapter we examine patterns, variations, and trends in small-area incidence of injuries due to medical adverse events in the 16 health service delivery areas (HSDA) of British Columbia. For illustrational purposes, we present the results of the small-area analysis for the age groups 0-19 years and 65-79 years. It is novel to the literature on medical adverse events to assess the impact of injuries due to AEs among children and youth (ages between 0 and 19 years), at the population level, and over an extensive period of time. In addition, we present the results for the age group 65-79 years, because this age group had the highest impact of injuries due to A E during the entire study period. The results and figures presented in this chapter for the other age groups can be found in Appendix G. In addition, the estimated parameters for model (3.1) (described in section 3.2.1), based on the A E incidence data of males and females for all other age groups, can be found in Appendix G. 6.1 Descriptive Statistics for Medical Adverse Events During 1991-2000 there were 3.1 million hospital admissions (all causes) for males and 4 million hospital admissions (all causes) for females in BC. For the year 2000, the number of hospital admissions (all causes) was 312,832 for males and 399,699 for females. In addition, 163 the hospital admission rates due to all causes decreased for males and females during 1991-2000 (Figure 6.1). 1992 1996 2000 | 1992 1996 2000 | 1992 1996 2000 | 1992 1996 2000 | 1992 1996 2000 year year year year year Figure 6.1. Annual Age-Specific Hospital Admission Rates per 1,000 Population for Both Genders Combined, 1991-2000. In this study, a medical adverse event (or AE) is defined as an adverse event, a postoperative complication, or a medical/surgical misadventure defined by the ICD9 E-codes E870-E876, E878-E879 and E930-E949. Based on this definition there were 161,411 hospital separation records for males and 158,958 hospital separation records for females that contained at least one of the ICD9 E-codes defining an A E between the years 1991 and 2000. Although medical adverse events rarely result in death, during 1991-2000 there were 457 deaths among males and females with AEs as the underlying cause of death. Table G.6-G.9 (see Appendix G) presents the distribution of in-hospital incidents associated with each of the ICD9 E-codes defining an A E during 1991-2000. The majority of medical adverse event incidents were associated with the E-codes E878-E879 (i.e., surgical and medical procedures as the cause of abnormal reaction of patient or later complication, Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 164 Columbia without mention of misadventure at the time of procedure) for both males (88% of all medical AEs that occurred during hospital stay) and females (84% of all medical AEs that occurred during hospital stay). Moreover, the E-codes E878-E879 were also identified as being responsible for the majority of medical adverse events that resulted in death for both males (54% of all injury-related deaths) and females (52% of all injury-related deaths). 6.2 Estimated Medical Adverse Event Incidence Rates Between 1991 and 2000, the estimated BC average A E incidence rate (parameter m) per 100 admissions varied from 1.40 [1.28, 1.53] (age group 0-19 years) to 6.28 [5.55, 7!07] (age group 65-79 years) for males, and from 1.11 [0.99, 1.24] (age group 0-19 years) to 5.87 [5.24, 6.55] (age group 65-79 years) for females. Table 6.1 shows the estimated provincial A E incidence rate per 100 admissions in 2000. Note that for only the age groups 0-19 and 20-39 years, and for all ages combined, was the estimated A E incidence rate statistically different between males and females, with males having a higher rate. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 165 Columbia Table 6.1. Estimated Provincial Annual Medical Adverse Event Incidence Rates (Age-Specific and Age-Standardized) per 100 Admissions by Gender for the Year 2000 (Standard Population: Hospital Admission Due To A l l Causes for Both Males and Females). Provincial estimated A E rate per 100,admissions (2000) \ue ' Mules Females Both 0-19 1.93 [1.75,2.11] 1.52 [1.35, 1.71] 1.70 [1.52, 1.88] 20-39 3.75 [3.27,.4.27] 2.02 [1.82, 2.24] 2.77 [2.45,3.12] 40-64 4.94 [4.26, 5.69] 4.53 [4.04, 5.05] 4.71 [4.14, 5.33] 65-79 6.87 [5.93, 7.88] 5.68 [4.95, 6.53] 6.20 [5.38, 7.12] 80+ 5.79 [4.S5. 6.SS] 4.65 [3.64. 5.86] 5.15 [4.17, 6.30] Age-standardized 4.96 [4.41.5.56] 3.91 |3.49.4.36| 4.37 [3.89,4.88] To illustrate the impact of using a model fitting approach for the analysis of A E data we plotted the annual conventional and estimated A E incidence rates during 1991-2000 for the age groups 0-19 and 65-79 years (Figures 6.2 and 6.3, respectively). Observe that the conventional A E incidence rates fluctuated over the years due to chance variation and that the estimated A E rates show a smooth trend during the ten-year period. In addition, note how different the trends are between males and females, and between the age groups 0-19 and 65-79 years. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 166 Columbia 1992 1996 year 1992 1998 V Okanagan Males Thompson Cariboo Shuswap Males Fraser Vol ley Males Simon Fraser Males 1992 1996 1992 1998 North Shore/Coast Gari baldl South Vancouver Island Central Vancouver Island North Vancouver Island 1992 1998 1992 1998 1992 1996 South Fraser Males 1992 1998 year Northern Interior Males fa* Richmond 1 1992 1998 year T5 -E 1992 1998 1992 1998 1992 1998 year 1992 1998 year 1992 1998 year -1992 1998 yaar Vancouver North Shore/Coast Gari baldl South Vancouver Island Central Vancouver Island North Vancouver Island Northwest Northern Interior Northeast Females Females Females Females Females Females Females Females 1992 1996 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 1992 1998 Figure 6.2. Conventional and Estimated Annual Medical Adverse Event Incidence Rates per 100 Admissions by Health Services Delivery Area for Boys and Girls Aged 0-19 Years, 1991-2000. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 167 Columbia Easl Kooten ay Males 1992 1998 year Vancouver Males E •* fi 1992 1998 year East Kooten ay Fern ales 1992 1998 year Vancouver Fern ales 1992 1998 year Kooten ay/Boundary Males 1992 1998 year Okanagan Males 1992 1998 Thompson Cariboo Shuswap Males Simon Fraser Males South Fraser Males Figure 6.3. Conventional and Estimated Annual Medical Adverse Event Incidence Rates per 100 Admissions by Health Services Delivery Area for Males and Females Aged 65-79 Years, 1991-2000. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 168 Columbia 6.2.1 Comparison of the Annual Estimated Rates among Different Geographical Layers in the Provincial Health Care System Table 6.2 shows the estimated A E incidence rates in 2000 for males and females in the age groups 0-19 and 65-79 years. The cells in Tables 6.2 with the rates in boldface represent the HSDAs that had rates statistically higher or lower than their respective H A and the province of B C as a whole, hi addition, HAs with rates statistically higher or lower than the B C rate are also boldfaced. Considering only the year 2000, Table 6.2 shows that for people between 0 and 19 years of age, the East Kootenay and Fraser Valley HSDAs had the highest A E incidence rates for boys and girls, respectively; and the South Vancouver Island and Vancouver HSDAs had the lowest A E incidence rates for boys and girls, respectively. Also note that the regions with a higher A E rate in 2000 tended to be clustered in the southeast regions of BC (boys and girls). Moreover, Table 6.2 also shows that in 2000, for those aged 65 to 79 years, the South Fraser and Northeast HSDAs had the highest A E incidence rate for males and females, respectively; and the North Vancouver Island and Okanagan HSDAs had the lowest A E incidence rate for males and females, respectively. Observe that the regions with higher rates in 2000 are clustered in the urban health authorities located in the southwest of B C (males and females). Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 169 Columbia Table 6.2. Regional Medical Adverse Event Incidence Rate per 100 Admissions in 2000 for Males and Females Aged 0-19 and 65-79 Years. Medical Al". rates per 100 admissions in 2000 [95% CI] Medical A l : rales per 100 admissions in 2000 | 95%CI | 0-19 \ears 65-79 years Bovs (iids Males Kemales 1 nterior HA 2.02 |l.')5.2.101 1.57 |I.5I. 1.62| 6.13 [6.01,6.24] 4.69 |4.58, 4.79| East Kootenay 2.22 [1.79,2.81] 1.61 [1.28,2.00] 7.08 [6.06, 8.22] 5.56 [4.62, 6.62] Kootenay/Boundary 2.11 [1.70,2.61] 1.77 [1.39, 2.19] 6.45 [5.56, 7.37] 4.84 [4.08, 5.67] Okanagan 1.95 [1.66,2.28] 1.53 [1.29, 1.80] 6.06 [5.59, 6.57] 4.29 [3.87, 4.72] Thompson Cariboo Shuswap 2.00 [1.61, 2.39] 1.52 [1.28, 1.80] 5.76 [5.17, 6.39] 4.99 [4.44, 5.58] 1 rascr 11A 2.04 [1.98,2.10] 1.65 [1.59, 1.70] 8.10 |7.98, 8.211 6.62 [6.52, 6.73] Fraser Valley 2.20 [1.87, 2.55] 2.18 [1.84, 2.52] 6.07 [5.48, 6.67] 4.86 [4.32, 5.44] Simon Fraser 1.90 [1.65, 2.15] 1.51 [1.30, 1.73] 7.97 [7.44, 8.53] 6.86 [6.33, 7.40] South Fraser 2.07 [1.82, 2.34] 1.47 [1.29, 1.68] 9.37 [8.81, 9.96] 7.34 [6.84, 7.86] Vancouver Coastal HA 1.82 [1.75, 1.88] 1.31 [1.26. 1.37| 7.68 [7.54, 7.82] 5.96 [5.84, 6.08] Richmond 1.86 [1.50, 2.25] 1.62 [1.30, 1.99] 7.74 [6.74, 8.81] 7.15 [6.23, 8.15] Vancouver 1.69 [1.46, 1.94] 1.18 [1.01, 1.38] 7.92 [7.37, 8.49] 6.02 [5.53, 6.54] North Shore / Coast Garibaldi 2.03 [1.71,2.45] 1.40 [1.17, 1.67] 7.23 [6.54, 7.96] 5.29 [4.71, 5.91] Vancouver Island HA 1.75 [1.68, 1.81] 1.47 [1.41, 1.53] 7.03 [6.90, 7.17] 5.66 [5.53, 5.78]. South Vancouver Island 1.66 [1.41, 1.92] 1.65 [1.43, 1.90] 7.69 [7.16, 8.24] 5.89 [5.43, 6.39] Central Vancouver Island 1.84 [1.54, 2.18] 1.24 [1.02, 1.47] 6.34 [5.78, 6.93] 5.27 [4.74, 5.83] North Vancouver Island 1.88 [1.44,2.32] 1.30 [1.00, 1.64] 5.70 [4.53, 7.03] 5.85 [4.66, 7.27] Northern HA 1.83 [1.75, 1.91] 1.57 | l 50. 1.63] 6.48 [6.24, 6.71] 5.52 [5.30, 5.74] Northwest 1.85 [1.51,2.25] 1.56 [1.26, 1.88] 7.32 [6.13, 8.64] 5.86 [4.75,7.10] Northern Interior 1.76 [1.45, 2.10] 1.57 [1.31, 1.90] 6.10 [5.21, 7.07] 4.42 [3.69, 5.25] Northeast 1.97 [1.52,2.44] 1.59 [1.26, 1.97] 6.14 [4.82, 7.61] 7.94 [6.34, 9.86] BC 1.93 [1.75,2.11] 1.52 [1.35. 1.7I| 6.87 [5.93, 7.88] 5.68 [4.95, 6.53] In the following two sub-sections we present a few figures comparing the annual trend of the A E incidence rates for males and females separately for the age groups 0-19 and 65-79 years, across the different levels of the geographical hierarchy. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 170 Columbia 6.2.1.1 Age Group 0-19 Years Figures 6.4-6.5 compare the A E incidence trends during 1991-2000 at the HSDA, H A and BC levels for boys and girls, respectively. Overall, the fitted temporal splines capture moderate curvature in the underlying regional rates, as seen in Figures 6.4 and 6.5. Note that the overall annual A E incidence rates for the province, health authorities and most health service delivery areas tended to increase over the decade of observation. Also note that the A E incidence rates of most health service delivery areas followed the trend associated with their respective health authorities. There were however, several health service delivery areas for which the increase in A E incidence rate was far faster than that of their associated health authority {e.g., the Okanagan (boys) and Kootenay/Boundary (girls) HSDAs). Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 171 Columbia Medical Adverse Events - Males Interior Health Authority East Kootenay Kowenay/Boundsfy Okanagan Thompton Cariboo Shutwap Medical Adverse Events - Males Fraser Health Authority \V '•• \ \ \ ' \ / \ \ / / ^ ^ ^ ^ ^ ft/ / 1 : 11/ A Fre»«fVaH«iy ll South Frasar 11 ',/ I Medical Adverse Events - Males Vancouver Coastal Health Authority Medical Adverse Events - Males Vancouver Island Health Authority Medical Adverse Events - Males Medical Adverse Events - Males Northern Health Authority British Columbia Figure 6.4. Annual Medical Adverse Event Incidence Rates per 100 Admissions For Boys Aged 0-19 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 172 Columbia Medical Adverse Events - Females Interior Health Authority Medical Adverse Events - Females Fraser Health Authority § 3 Medical Adverse Events - Females Vancouver Coastal Health Authority Medical Adverse Events - Females Vancouver Island Health Authority Medical Adverse Events - Females Northern Health Authority 1992 1994 19! year 1998 2000 1 -Medical Adverse Events - Females British Columbia Figure 6.5. Annual Medical Adverse Event Incidence Rates per 100 Admissions for girls Aged 0-19 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Next, Figure 6.6 shows the HSDAs whose estimated A E incidence rates were above (ratio>l) or below (ratio<l) the provincial rate during 1991-2000 for boys and girls. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 173 Columbia Figure 6. Services 2000. 6. Annual Ratio between the Medical Adverse Event Incidence Rate of Each Health Delivery Area and British Columbia for Boys and Girls Aged 0-19 Years, 1991-Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 174 Columbia To provide a visual summary of the spatio-temporal pattern of the A E incidence rate estimates during the 10-year study period, Figure 6.7 maps the annual ratio between the medical adverse events incidence rates of each H S D A and B C for boys and girls. The map for boys shows a dramatic change in the geographical distribution of A E rate ratios observed in 1991-1999 and in 2000; the pattern of the A E rate ratio is more clearly seen in Figure 6.6, and the abrupt change in pattern especially seen in the Northern regions is worthy more investigation. Note that in contrast to the maps for boys, there was a gradual change in the geographical distribution of the A E rate ratio among girls. It is also important to note that because of the small-area problem (i.e., small population at-risk) in the age group 0-19 years, it was difficult to find HSDAs in which the A E rate ratio was statistically different from 1. In summary, Figure 6.7 shows that for boys in the year 2000, the East Kootenay (ratio = 1.15, 95%CI [0.94,1.46]), Fraser Valley (ratio = 1.15, 95%CI [0.97,1.34]) and South Fraser (ratio = 1.07, 95%CI [0.93,1.24]) HSDAs had A E rate ratios that were higher than 1 (however not statistically significant). The South Vancouver Island HSDA (ratio = 0.87, 95%CI [0.73,1.01]) had an A E rate ratio lower than 1 (however not statistically significant). For girls in the year 2000, the Fraser Valley HSDA (ratio = 1.43, 95%CI [1.20,1.69]) had A E rate ratio significantly higher than 1. The Kootenay/Boundary H S D A (ratio = 1.16, 95%CI [0.92,1.45]) also had an A E rate ratio that was higher than 1 (however not statistically significant). The Vancouver (ratio = 0.78, 95%CI [0.65,0.92]) and Central Vancouver Island (ratio = 0.81, 95%CI [0.67,0.97]) HSDAs had A E rate ratios that were significantly less than 1. The North Vancouver Island HSDA (ratio = 0.86, 95%CI [0.66,1.07]) had an A E rate ratio lower than 1 (however not statistically significant). Medical Adverse Events Males, 0-19 years Adverse event rate ratio JBH 90 ! ! [0 90 • 1.10) = 1.10 Medical Adverse Events Females, 0-19 years Adverse event rate ratio < D 9 0 j [0 9 0 - 1 . 1 0 ) I >= 1.10 ,0 Q 1 a TO a TO TO S-TO' a TO TO TO Figure 6.7. Map of Annual Ratio between the Medical Adverse Event Incidence Rate of Each Health Services Delivery Area and British Columbia for Boys and Girls Aged 0-19 Years, 1991-2000. to 3 a-Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 176 Columbia 6.2.1.2 Age Group 65-79 Years Figures 6.8-6.9 compare the A E incidence trends during 1991-2000 at the HSDA, H A and BC levels, for males and females respectively. The provincial A E incidence trend was mostly flat for males, and it decreased slightly for females. The annual A E incidence rates decreased during 1991-2000 for most health authorities by means of a non-linear temporal trend, with the exception of the Northern H A (both males and females) and Vancouver Coastal H A (for males), in which the A E incidence rates increased over the ten-year period. Observe that in the Northern health authority, the rates of all health service delivery areas increased more distinctively in males than in females from 1991 to 2000. Also note that in the Interior H A , the A E incidence trend of each HSDA was very distinct, with the rates for most health service delivery areas being further apart in 1991, and becoming more similar as we approached the year 2000. Finally, there were certain health service delivery areas for which the A E incidence rate declined far faster than their associated health authorities {e.g., the Fraser Valley HSDA); and there were health service delivery areas with increasing rates over the last four years of the study (e.g., the East Kootenay HSDA). Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British \11 Columbia Medical Adverse Events - Males Medical Adverse Events - Males Interior Health Authority . Fraser Health Authority 1992 1994 1996 1 998 2000 1992 1994 1996 1998 2000 year year Medical Adverse Events - Males Medical Adverse Events - Males Vancouver Coastal Health Authority Vancouver Island Health Authority —I 1 1 1 1— 1 1 1 1 I— 1992 1994 1996 1998 2000 1992 1994 1996 1908 2000 Medical Adverse Events - Males Medical Adverse Events - Males Northern Health Authority British Columbia ~\ 1 1 r 1992 1994 1996 1998 2000 1992 1994 1996 1998 2000 year year Figure 6.8. Annual Medical Adverse Event Incidence Rates per 100 Admissions for Males Aged 65-79 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 178 Columbia Medical Adverse Events - Females Medical Adverse Events - Females Vancouver Coastal Health Authority Vancouver Island Health Authority 1992 1994 1996 1998 2000 1992 1994 1996 1998 2000 year ' year Medical Adverse Events - Females Medical Adverse Events - Females Northern Health Authority British Columbia 1992 1994 1 996 1998 2000 1992 1994 1996 1996 2000 year year Figure 6.9. Annual Medical Adverse Event Incidence Rates per 100 Admissions for Females Aged 65-79 Years at the Health Services Delivery Area, Health Authority and Provincial Levels, 1991-2000. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 179 Columbia Figure 6.10 shows the annual ratios between the A E incidence rate estimate of each HSDA and the provincial A E incidence rate for males and females in the age group 65-79 years. Once more, note how different these trends are between different HSD As and between males and females within the same HSDA. To conclude this chapter of the study, Figure 6.11 maps the A E rate ratio estimates of each HSDA relative to BC for males and females between 65 and 79 years of age. Note that for males in the year 2000, the South Fraser (ratio = 1.37, 95%CI [1.18,1.60]) and Richmond (ratio = 1.19, 95%CI [1.04,1.34]) HSDAs had A E rate ratios that were significantly higher than 1. The Thompson Cariboo Shuswap HSDA (ratio = 0.84, 95%CI [0.71,0.99]) had an A E rate ratio significantly less than 1. For females in the year 2000, the Simon Fraser (ratio = 1.21, 95%CI [1.04,1.42]), South Fraser (ratio = 1.30, 95%CI [1.11,1.50]), Richmond (ratio = 1.26, 95%CI [1.05,1.52]), and the Northeast (ratio = 1.40, 95%CI [1.10,1.79]) HSDAs had A E rate ratios that were significantly higher than 1. The Okanagan (ratio = 0.76, 95%CI [0.64,0.89]) and the Northern Interior (ratio = 0.78, 95%CI [0.63,0.96]) HSDAs had A E rate ratios that were significantly less than 1. In conclusion, observe that the regions with at higher rate ratio in 2000 for males and females in the age group 65-79 years tend to be clustered in the urban health authorities located in the southwest of BC. Chapter 6. A Small-area Analysis of the Incidence of Medical Adverse Events in British 180 Columbia Figure 6.10. Annual Ratio between the Medical Adverse Event Incidence Rate of Each Health Services Delivery Area and British Columbia for Males and Females Aged 65-79 Years, 1991-2000. Medical Adverse Events Males, 65-79 years Adverse event rate ratio I 1 [0.90 - 1.1Q) Medical Adverse Events Females, 65-79 years Adverse event rate ratio «0.90 [D 90- 1.10) * * 1.10 Figure 6.11. Map of Annual Ratio between the Medical Adverse Event Incidence Rate of Each Health Services Delivery Area and British Columbia for Males and Females Aged 65-79 Years, 1991-2000. P Q o a-a a-S3 _0\ S3 SI S3 S3 a-TO TO S: TO a TO TO TO Br TO S3 «-~. TO TO TO a E? a' to a-oo 182 Chapter 7 Discussion 7.1 Summary of Findings This study was motivated by the opportunity to demonstrate the utility associated with the integration of a recently developed statistical methodology, known as spatio-temporal Bayesian disease mapping, into the quantification of health status indicators derived from existing population health monitoring data collected by the province of British Columbia. In addition to mortality and hospitalization rates and rate ratios, the health status indicators examined in this study included measures associated with the impact on the quality of life of the population such as years of life lost ( Y L L or fatal burden), healthy years lost due to disability ( Y L D or non-fatal burden) and disability-adjusted life years ( D A L Y or total burden). To provide an example particularly relevant to the public health field, we chose to apply this novel methodology to injury surveillance data to show that the injury impact varies according to different segments of the population. As a possible first step towards addressing the problem of injury, this study demonstrated how researchers can effectively utilize existing data available in administrative databases to answer important questions involving the epidemiology and impact of injury on individual communities within BC. In Chapter 4, we provided an overall picture of the impact of different unintentional and intentional injuries in BC using health status indicators calculated at the Chapter 7. Discussion 183 provincial and health authority levels. At the provincial level analysis, we found that over the ten years of study, among all external causes of injury afflicting males in B C , road traffic accidents, poisoning (mainly due to drug abuse) and suicide and self-inflicted injuries ranked the highest; and road traffic accidents, falls and suicide and self-inflicted injuries ranked the highest among all external causes of injury afflicting females in BC. In addition, the fatal burden (i.e., years of life lost due to premature mortality) associated with injuries contributed at least 50% to the overall burden of this health determinant. In general, when all injuries were combined, males experienced a higher burden from injuries than females. Some gender specific age groups, such as 20-39 years (males) and 80+ years (females), had substantially higher ten-year per capita total burdens from injuries compared to other age groups. In addition, the health authority level analysis revealed that different injuries had different geographical distributions. For example, high injury incidence clusters were seen in the Interior and Northern health authorities of BC. The next objective in this study was to investigate the impact of injuries at the health services delivery area level from 1991 to 2000, and to demonstrate the importance of employing statistical approaches over and above mapping conventional rates, ratios and D A L Y s . Investigating health impacts within specific regions (i.e., health service delivery areas), which are much smaller in size than the health authorities, requires the application of Bayesian disease mapping methods to ensure reliable estimates of rates, ratios and D A L Y s . In Chapter 5, the small-area analysis of injuries due to road traffic accidents highlighted some high-risk health service delivery areas in the Northern and Interior health authorities. In addition, distinct geographic distributions of high and low incidence were observed for males and females between 20 and 39 years of age. In Chapter 6, we showed that medical AEs can Chapter 7. Discussion 184 be effectively monitored using ICD9 E-codes identified in administrative data scanning. The results associated with AEs at the provincial level were consistent with other studies, however, it is important to note that this is the first research at the population level that used Bayesian disease mapping methods to assess the true (i.e., clear of change and systematic variation) underlying differences in medical A E rates for all age and gender groups in all health services delivery areas of B C over an extensive period of time (from 1991 to 2000). Our study showed that the medical A E incidence was consistent over time and across different health regions, and contained significant differences across age groups. 7.2 Injury Viewed as a Public Health Issue The relevance of this study to the field of public health can be traced to the Canadian Health Act of 1984 (c.6, s.3) (Department of Justice 2004), which states that: "The primary objective of Canadian health care policy is to protect, promote and restore the physical and mental well-being of residents of Canada and to facilitate reasonable access to health services withoutfinancial or other barriers" Thus, the methods presented in this study can be used to identify regions and sub-populations where health disparities exist and in so doing provide the first step towards promoting equal access to health for all of BC 's residents. Injury has become a major public health concern, with most injured victims being young adults at their most economically productive age. There is a high cost associated with injury: (1) extensive health care resources are expended to treat, care for, and rehabilitate Chapter 7. Discussion 185 injured persons; (2) a high number of productive years of life are lost due to premature death and long-term disability; and (3) substantial physical and emotional costs associated with pain and suffering are experienced by injured persons and their families. It is expected, for example, that road traffic accidents will soon be considered a serious public health problem worldwide, as it has been projected that by 2020, road traffic accidents will be the third leading cause of death and disability in the world, after ischaemic heart disease (first) and unipolar major depression (second). Of note in 2000, road traffic accidents ranked the ninth highest burden among all health problems in the world (Murray and Lopez 1996; WHO 2004a; MacKenzie 2000). Many researchers have thus come to view injuries as a silent pandemic. Given the forecasted trends, the critical question is what can public health officials in BC do to make sure that the health care system, the economy and society at large are able to handle this population health problem? The first two steps in dealing with the problem of injuries in BC are to define and measure the actual burden, intensity, distributional pattern and trend associated with injuries at the macro level (i.e., provincial level), and then at the micro level (i.e., regional level), since the latter is the level at which services are ultimately managed and delivered. Provincial and regional injury surveillance should be an integral component of our strategy to maintain and improve the health of British Columbians. This study has demonstrated that valuable surveillance information can be extracted from data that are readily available to researchers through existing administrative databases. Additionally, this study demonstrated how fundamental knowledge on the impact of a total of 13 unintentional and intentional injuries was obtained by using commonly employed health status indicators such as mortality and hospitalization rates, in conjunction with health status indicators measuring disease burden such as years of Chapter 7. Discussion 186 life lost (YLL) from premature death, healthy years lost due to disability (YLD) and disability-adjusted life years (DALY) . It is important to stress that each health status indicator provides different information in the assessment of the population's health. Rates quantify the scale of the problem, i.e., these measure how many people have died or been hospitalized in a given time period due to injury. However, many injuries are non-fatal and consequently a proportion of those injured are left with short-term or long-term disabilities. To take into account the resulting disability-related reductions in quality of life, it is necessary to use measures such as Y L L , Y L D and D A L Y to quantify how many years a given population has lost due to injury. The appropriate analysis {e.g., using small area analysis methods) of surveillance data has the potential to efficiently produce important information for public health planners. For example, the methods outlined in this study could be used in the following ways: (1) The creation and maintenance of community-specific health profiles that contain up-to-date information about current health status; (2) The generation of etiological hypotheses could be facilitated by information derived from long-term analyses of spatial distributions of health problems; (3) Subgroups within the B C population that are the most/least affected by injuries could be identified; (4) Surveillance information could be used to capture change and flag abrupt jumps in injury incidence; (5) Detailed information on the health profile of communities could be used by local health planners to help ensure that the quality and quantity of available health resources are appropriate for the local needs; Chapter 7. Discussion 187 (6) Data obtained from surveillance systems could be used to detect the impact of changes put in place to improve quality of health care practices, as well as to monitor the effectiveness of future interventions; and (7) Surveillance information could also be used by health care providers, planners and administrators to efficiently plan effective injury prevention and public health intervention strategies. In order to develop more effective preventive strategies and reduce the impact of injuries on the population, we must improve the quality and quantity of available information. Prevention of injuries is a multidisciplinary undertaking involving collaboration among many different disciplines, agencies and government layers. When developing prevention strategies, it is imperative that researchers ask: Who is being injured? Where (e.g., communities, or health regions) are the injury indicators the highest, and alternatively, the lowest? Furthermore, does the present magnitude of the health status indicators differ from previous levels? To answer these questions it is crucial that researchers conduct descriptive epidemiological studies that are focused on describing the distribution of injuries according to person, place and time. In this study, complex statistical methods were used to analyse ten years of injury surveillance data for individual communities in BC. Some readers may ask: "Why it is necessary to use such statistical methodology instead of just using conventional estimates to calculate health status indicators?". This is a legitimate question and we reiterate with emphasis that small-area analyses based on conventional estimates of health status indicators can be misleading because conventional estimates of risk for areas with a small "population at risk" are influenced by chance variation (see for example the comparison between the conventional and fitted mortality rate for Chapter 7. Discussion 188 road traffic accidents in the Northeast HSDA as shown in Figure 5.1, page 138). As demonstrated in this study, Bayesian spatio-temporal analysis of the injury data enabled us to quantify spatial variation and temporal trends in small-area risks arid burden estimates. The small-area analyses performed in this study provide descriptive information, which could be used by health planners to inform the development of future patient safety preventive initiatives. Although the incidence of injuries due to medical adverse events is growing, there is still insufficient information to gauge the full scale of the problem in B C (e.g., which risk factors most influence the regional differences in A E incidence?). With this in mind, one important contribution of this study is the assessment of the magnitude of injuries due to medical adverse events (as defined by existing E-codes) at the community level across gender and age groups over time. In Chapter 6 we identified areas with a high incidence of injuries due to medical adverse events, we identified subgroups of the population more afflicted by these injuries, and we monitored injury clusters, changepoints and trends in different regions over time. For example, in this study we have shown that injuries due to medical adverse events were not uncommon (the age-standardized B C rate for both males and females was 4.37 per 100 admissions), and the incidence was the highest among elderly patients (65-79 years of age). In-hospital management of elderly patients (65-79 years of age) is a challenge due to a higher than normal incidence of comorbidities and polypharmacy, and consequently increased likelihood of adverse events. (Gridelli 2004; Trinka 2003; Thomas and Brennan 2000). In 2000, the estimated elderly population in B C was just over 399,000 people, representing an increase of 15% compared to the year 1991 (Ministry of Management Services 2004). With an aging population, it is plausible to assume that a higher incidence Chapter 7. Discussion 189 will be associated with medical adverse events taking place during hospitalization for this age group (note that this incidence is also likely to increase in out-patient care of seniors). If so, then this raises an important flag in terms of health care safety, and calls for the development of patient safety polices to deal with this particular problem {e.g., implementation of a universal real-time electronic prescription database for seniors that flags inappropriate prescription combinations). One important recommendation of this study is the development of standardized provincial and regional monitoring systems that facilitate the application of methodologies similar to the approaches demonstrated in this study. The application of such methods would improve our ability to detect and estimate rates of injuries due to adverse events at the small-area level. This approach would also minimize the incidence of false alarms associated with chance fluctuations in areas with small population at risk, which might be misinterpreted in a system based on conventional estimates of injury incidence. In addition, for those health services delivery areas with a higher incidence of injuries due to adverse events, our hierarchical model can be modified and extended to include hospital level data, which allows for the identification of problematic hospitals within those areas as well as the factors that increase their likelihood of a poor performance. In Chapter 5 we demonstrated the importance of calculating the small-area burden of injuries using the disability-adjusted life years (DALY) measure for prevention and surveillance of injuries such as road traffic accidents. We found that the burden of this injury is strongly associated with premature mortality, especially among youth and adults. We showed that in most regions the incidence and burden of this external cause of injury is decreasing, however we also identified areas that are not doing so well. In these problematic Chapter 7. Discussion 190 areas, more attention should be dedicated to looking for alternative methods to improve health outcomes. For example, in addition to reinforcing existing strategies already shown to be effective in preventing these injuries, new community-specific educational programs and bylaw regulations could be developed to help lower traffic fatalities in small communities within each health services delivery area. Several other interesting results were obtained performing the provincial and H A level analyses. Our results showed that the mortality rate and the per capita years lost due to premature death attributable to poisoning was not uncommon among males with age between 20 and 39 years old in the Vancouver Coastal H A (for every 1,000 people monitored over 10 years, approximately 63 years were lost due to premature death). This is a significant result, which required a closer examination of the primary ICD9 E-codes associated with this external cause of injury. The ICD9 E-codes contributing to more than 50% of all poisoning related deaths were E850.0, E855.2, E858.8, E858.9, which are codes that refer to drug overdose mainly due to cocaine, heroin and central appetite depressants. Additionally, we observed that burden associated with the category other unintentional injuries (excluding AE), mainly due to years lived with disability, was especially high for males younger than 65 years. It is important to recall that the higher Y L D for this external cause of injury was primary due to ICD9 E-codes (a) E916.0: struck accidentally by falling object such as: rock, snowslide (nowhere else specified), stone, tree, object falling from machine not in operation or from stationary vehicle; (b) E917.9: struck accidentally by an air rifle; (c) E918.0: caught accidentally in or between objects; (d) E920.3, E920.8: cutting and piercing instruments or objects mainly: knives, swords and daggers, and other cutting and piercing objects {e.g., broken glass, nail, tin can lid, splinter, etc.); and (e) Chapter 7. Discussion 191 E929.8: late effect of other accidents (nowhere else specified). We believe that the Y L D for this injury category was higher because of the combination of type of nature of injury (several individuals with life long sequelae) and age distribution of injured individuals (largely young individuals). In conclusion, the extensive results presented throughout Chapters 4 to 6 demonstrate the type of benefits that can be realized by analysing data collected from the long-term monitoring of injuries in BC. We believe that the application of the solid methodology presented in this study to the analysis of such injury surveillance data has the potential to benefit the health system in many ways. Based on the differential regional impact of injuries using all the health status indicators presented in Figure 1.1, our results could be used as criteria to establish targets in promoting population health, as well as to prioritize and guide future research. Over time, important information and interventions will accumulate and fewer health care resources will be required to treat, care for, and rehabilitate injured persons; a lower number of productive years of life will be lost due to premature death and long-term disability; and lower costs will be associated with the pain and suffering experienced by injured persons and their families. Chapter 7. Discussion 192 7.3 Methodological Considerations, Limitations and Recommendations for Future Research 7.3.1 Data Quality The first assessment of data quality was conducted during the initial data cleaning process. During this process, a very small number of records were identified as having a mistake. For example, the same person was coded as both male and female, was listed as living at the same time in two different HSDAs, or the record was missing an ICD9 E- and/or N-code. To deal with this problem, it was decided that the most efficient (i.e., least time consuming) solution was to remove these mistaken records, as they represented a very small percentage of our data (<1%). The second assessment of data quality relates to how valid and reliable standard classification codes for mortality and morbidity are for assessing the health impact of different injuries (e.g., the International Classification of Diseases System Version 9 external codes (ICD9 E-codes)). In this study, problematic classification codes (e.g., coding disparities among different regions) could have directly affected health status indicators such as the mortality and hospitalization rates; and indirectly affected health status indicators such as the years of lost life (YLL) (based on mortality data), healthy years lost due to disability (YLD) (based on hospitalization data), and ultimately disability-adjusted life years (DALY) (based on both mortality and hospitalization data). Chapter 7. Discussion 193 Years of lost life and mortality rates are affected by the classification of each external cause of injury using ICD9 E-codes included in the mortality data obtained from the B C Vital Statistics agency. In order to assess the quality of the mortality data, one needs to examine the reliability and validity of the completion of medical death certificates. One source of such information are annual reports published by the B C Vital Statistics agency which contain explanations of the procedures involved in completing these medical death certificates to assure data quality2. On the other hand, healthy years lost due to disability and hospitalization rates are influenced by the reliability and validity associated with the completion of hospital medical charts, and transcription into medical computerized records. With these problems in mind, we conducted a literature review on this topic and found that the more non-specific an injury E-code is, the more reliable and valid the assignment of the code in both death certificates and hospital records {e.g., detailed E-codes such as E810.1 as opposed to non-specific E-codes as E810) (see for example LeMier, Cummings, and West 2001; Maclntyre, Acklahd, and Chandraraj 1997). Regarding the accuracy and reliability of ICD9 N-codes 800-999 necessary to calculate disability weights and durations, the study by Maclntyre, Ackland and Chandraraj (1997) reported that the more non-specific an N-code in the hospital separation record was, the lower the frequency of mistakes in assigning these codes was. In this thesis, we only worked with non-specific ICD9 N - and E-codes, with the 2 The physician in attendance at the last illness of the deceased person, or the coroner conducting an inquiry into the death of the person is required to complete a Medical Certification of Death. In addition, the registration of death is completed by the informant with assistance from the funeral home. Funeral Directors obtain the Medical Certification of Death, issue the burial permit and submit the Medical Certification of Death and the Registration of Death documents to the Agency to complete the registration (Vital Statistics Agency of British Columbia (2001)). Chapter 7. Discussion 194 exception of a small proportion of records using complete E-codes to identify the late effects of injury (E929), injuries resulted from striking against or struck accidentally by objects or persons (E917), and injuries caused by cutting and piercing instruments or objects (E920). Assessing the reliability and validity of ICD9 E-codes in data originating from hospital records and death certificates in BC is an important topic for future research. One possible approach to this type of investigation involves additional data linkage to other data sources (e.g., Medical Service Plan billing data), which would yield the detailed information at the patient level necessary to compare injury codes between the different sources of data. Additionally, the data quality of electronic hospital medical records could be assessed through the selection of a sample of medical records, and later comparison of these records with the corresponding original medical charts. Ultimately in the latter case, one would be verifying whether what is described in the medical charts is properly transcribed in the selected medical records. To conclude the discussion of data quality, the methodology presented in this study relies heavily on ICD9 N - and E-codes to identify injuries. Although we are not able to claim that the data collection system is error free, however we believe that as the procedures for the development and maintenance of surveillance systems mature, the quality, accuracy and consistency of data collection will undoubtedly improve. 7.3.2 Epidemiological Methodology There are many ways to measure the impact of injuries on the health of populations using health status indicators such as those shown in Figures 1.1 and 1.2. One of the innovative Chapter 7. Discussion 195 goals of this study was to integrate D A L Y estimates into public health surveillance as a new type of health status indicator that can be used at the community level. The resulting information can then be used to quantify the loss of healthy years due to premature death and disability in a single measure of disease burden. However, it is important to emphasize that commonly used health status indicators, such as mortality and morbidity rates, should not be blindly replaced by measures such as disability-adjusted life years (DALYs). Ultimately, the selection of a health status indicator should be based on the objectives of each study and the data resources available. On important lesson learned while implementing this new methodology was that conducting a burden of disease study is a complex task. Every step in this analysis involves judgments, assumptions and familiarity with epidemiological theory. Of particular relevance are several preference measures intrinsic in the D A L Y calculation involving which values to assign to the discounting rate, life expectancy, age weights, disability weights and durations of injury sequela (for the residual and non-residual categories). These assumptions have practical and theoretical implications on the estimated burden associated with each external cause of injury. The implications of using different preference measures on the estimation of disability-adjusted life years (DALYs) were assessed via three sensitivity analyses. The first sensitivity analysis explored the effect of using different sets of disability weights for the residual category of injury sequela on estimated D A L Y s . Conclusions regarding which external cause of injury imposed the greatest total burden on B C residents did not change despite higher weights being associated with larger D A L Y s ; i.e., the magnitude of change in these D A L Y s was not sufficient to change the ranking of D A L Y s for different external Chapter 7. Discussion 196 causes of injury. The second sensitivity analysis explored the effect of using different values for the discounting of future health rate on estimated D A L Y s . This analysis showed that the results for D A L Y based on different discounting rates are substantively different {e.g., for road traffic accident, the D A L Y s for males were 161,198 (A=0%), 97,980 (r=3%), 72,396 (r=6%), and 56,577 (r=10%)). Nonetheless, there is no convincing theoretical argument supporting any particular value of discounting, and as we explained in section 3.3.1.1 (page 61), in this study we based all our results on the conventional 3% discounting rate. However, further investigation of both theoretical and practical issues is recommended. The last sensitivity analysis explored the effect of using different values of age weighting on estimated D A L Y s . The effect of using (compared to not using) age weights was much smaller than the effect of changing the discounting rate on final estimates of D A L Y s . In the context of this study, uniform age weights were assumed due to concerns with the lack of solid theoretical grounds to make a better choice about which age weights to choose. HoweVer, it is important to mention that all three sensitivity analyses depend on the data being used, and we recommend that similar analyses be completed as a routine component of any burden of disease study. In addition, there are several important assumptions intrinsic to calculating disability weights that influence the estimation of D A L Y s . One assumption is associated with the method of health state valuations chosen to obtain the disability weights. In the original global burden of disease study, researchers employed the person trade-off (PTO) technique to obtain health state valuations to assist in decisions in the health care setting at the population level (Murray and Lopez 1996). PTO is a method that estimates the social value of different health states. When applying this technique, respondents are asked to make choices in the Chapter 7. Discussion 197 context of a decision involving individuals other than themselves (Green 2001; Pinto Prades 1997). This is an important distinction between PTO and other more frequently used methods such as the standard gamble (SG), visual analogue scale (VAS) and time trade-off (TTO), which ask respondents to make choices in terms of their own lives. PTO has not been widely used in the literature because more work is required to confirm the reliability, practicality, consistency and validity of the valuations. For example, studies have shown that the quality weights (or preferences) used in the calculation of quality-adjusted life years depend on the technique employed to measure health preference. Different techniques {e.g., PTO versus SG) tend to produce different quality weights, and potentially different disability weights when employed in studies involving the calculation of disability-adjusted life years (see for example Murray and Lopez 1996; Green 2000). Ideally, one should study how these disability weights would change when using methods other than PTO in health state valuations. More importantly, the conclusions drawn from the D A L Y analyses done in this study should be further investigated to see i f they hold given the application of different weights (e.g., weights obtained from SG). Two other assumptions intrinsic to the disability weights had an effect on the results of this study: the weights were assumed to be stationary over time, and the level of disability of different individuals with the same injury sequela(e) was assumed to be the same. However, as mentioned in section 3.3.1.4 (page 67), the data obtained for this study allowed for the identification of individuals with more than one injury sequela and adjustment of their disability weights, thus differentiating the severity of the same injury among different individuals. Nonetheless, this methodology requires further development to understand how the use of Chapter 7. Discussion 198 non-stationary weights might change various calculations and conclusions drawn from the final estimates of D A L Y s . The estimated D A L Y s might also be influenced by accounting for other comorbid prevalent conditions, in addition to controlling for coexisting comorbidities due to injury in the same individual (as done in this study). To improve the precision of health status estimates in future analyses, it is recommended that researchers develop methods of controlling for comorbid prevalent conditions. In terms of this study, it would be valuable to evaluate the extent to which such an adjustment would change our final results. At this time, we focus on how the handling of the residual categories (900-924, 930-939, 958-959 and 990-999)3 might have influenced the estimated D A L Y s through the calculation of disability weights and durations. As explained in the notes related to Table 3.1 (page 49), the WHO redistributed the ICD9 ,N-codes for the residual categories proportionately across all categories of injury sequela with disability weight or duration. Currently, no study has developed specific durations and disability weights for the residual categories of injury sequelae. However, we don't agree with the WHO approach, because depending on the distribution of the ICD9 N-codes, the level of disability associated with each external cause of injury can be overestimated or underestimated. It would therefore be useful to develop disability weights and durations for these residual categories. In particular, this new development would make possible the calculation of D A L Y s for those external 3 900-924 and 930-939 (injury to blood vessels; late effects of injuries, poisonings, toxic effects, and other external causes; superficial injury; contusion with intact skin surface; effects of foreign body entering through orifice), 958-959 (Certain traumatic complications and unspecified injuries) and 990-999 (Other and unspecified effects of external causes). Chapter 7. Discussion 199 causes of injury for which the distribution of ICD9 N-codes is highly concentrated in residual categories. One limitation of this study is the inability to apply the D A L Y methodology to injuries due to medical adverse events. We believe that the approach used to calculate the disability weights and durations for the other external causes of injury should not be applied to injuries due to medical adverse events because most of the injury sequelae associated with medical A E were classified as one of the ICD9 N-codes in the residual category of injury sequela. In addition, in the calculation of disability weights and durations for medical A E it is essential to take into consideration other causes of illness, since most AEs happened because patients were hospitalized because of some type of underlying condition other than injury (e.g., a cardiovascular disease), and during the course of treatment a medical A E occurred. Thus, we believe that the methodology used to obtain disability weights and durations for injuries due to medical adverse events should be carefully re-addressed, since the original methodology used to obtain these quantities is no longer applicable. Contrary to the WHO and several other researchers, we do not believe that it is appropriate to proceed with the D A L Y calculation for injuries due to medical AEs when the overall disability weight and duration were obtained in the same way as the other external causes of injury. We feel that it is important that the D A L Y methodology be appropriately modified to allow for separate D A L Y calculations for injuries due to medical AEs. The benefits of this analysis would be considerable, as we would gain a better understanding of both short-term and long-term health outcomes, such as impairments, functional limitations (disability) and potential handicaps associated with medical AEs (i.e., how many years in health states other than full health a patent would have to live). Chapter 7. Discussion 200 In summary, the D A L Y methodology can be improved by developing more appropriate disability weights and durations that address the issues discussed in this section {e.g., issues regarding the residual category of injury, comorbidity, the health states valuation method). If researchers continue to employ PTO methods to estimate the social value of different health states, it is important that health professionals from different health care settings be involved in this process so that more realistic weights are obtained. This would also facilitate the development of durations and weights that better reflect the reality of living in BC. When developing the new durations and weights, it is critical for researchers to fully understand the "natural history" of each of the injury sequela (injuries classified under the ICD9 N-Codes) in substantial detail, particularly the injury sequelae under the residual category. Thus, in spite of the potential for broad generalizability and applicability to other diseases and conditions, the D A L Y methodology presented in this study requires further refinement. Moreover, as we have demonstrated, the use of D A L Y s allows for an array of interesting results, however more research should be conducted to determine how researchers can realize the full potential of using D A L Y s in terms of policy making and health planning. Chapter 7. Discussion 201 7.3.3 Statistical Methodology Prior to applying the final statistical model used in this study, we experimented with a variety of spatio-temporal statistical models currently being used in Bayesian disease mapping. After examining measures of goodness-of-fit, residual plots, and diagnostic tools for full Bayesian analysis we decided to use model (3.1), as it best captured the health inequalities in BC, and offered a simple interpretation of the results. However, during our modelling process, we encountered the challenge of determining how big the number of cases in each HSDA x year combination should be in order to obtain adequate results. For example, in this study the spatio-temporal modeling of the number of hospitalization cases due to road traffic accidents for the younger age group (0-19 years) was accomplished without any problems, however we encountered difficulty when modeling the number of deaths for this same age group due to the large number of zero cells at the small-area level. To overcome this small-area issue we need to develop statistical modelling approaches that can accommodate the considerable number of zeros typically found in this type of data. For example, instead of modeling the rates via Poisson random effects specification (as was done in this study), a worthwhile future extension of this work might be to consider the use of a zero-inflated mixed effects Poisson model (Lambert 1992; Lee et al. 2002; Yau and Lee 2001; Van den Broek 1995; Bohning et al. 1999). Nevertheless, the small-area problem in Bayesian disease mapping does not impede the application of this methodology to external causes of injury, or diseases that have a satisfactory number of cases in each HSDA x year combination. We recommend that model (3.1) and its variants be tried before choosing a final model. This choice should be based on Chapter 7. Discussion 202 full Bayesian convergence diagnostics and measures of goodness-of-fit such as the examination of posterior density plots, Gelman-Rubin convergence statistic, deviance information criterion (DIC), and residual plots (Gelman et al. 1998; Carlin and Louis 2000; Spiegelhalter et al. 2002, 2003). In addition, for injuries or diseases in which the number of cases is sufficiently large in each HSDA x year combination, both the statistical and epidemiological analyses presented in this study can be repeated using a finer geographical classification of regions, such as at the local health area (LHA) level or even finer areas such as the micro-health area. These analyses would provide a deeper understanding of which small regions within each of the health services delivery areas are suffering more from the impact of injuries. Finally, the statistical model presented in this study can be extended to accommodate ecological analyses based on risk factors other than gender aiid age. These ecological analyses could be conducted by including predisposing (e.g., socio-demographic variables), enabling (e.g., social support, availability of health services, other neighbourhood variables) and need (e.g., comorbidities) risk factors associated with each injury that are theorized to explain a proportion the geographical variation in rates, ratios and burden (Andersen and Newman 1973). However, obtaining such risk estimates is challenging, since it would require extensive linkage to other databases (e.g., Canadian Community Health Survey). 7.3.4 General Issues Some researchers might suggest that our reliance solely on hospitalization data for the calculation of D A L Y s is a weakness since the analyses captures only those individuals who Chapter 7. Discussion 203 were seriously injured enough to require hospitalization. In particular, patients with minor injuries treated in health care settings other than hospitals (e.g., physician's offices) were not accounted for in this study. However, studies (assessing the burden from injuries) have shown that using hospitalization data is sufficient in terms of providing a valid estimate of the morbidity component of DALYs (Mathers et al. 2001). In a future work, data from the Medical Services Plan (MSP) payment information could also be used to identify those injury cases in which the patient did not heed in-hospital care. However, the usefulness of incorporating this new source of data into the D A L Y calculation depends on how reliable, valid and complete the injury diagnostic codes are in the out-of-hospital setting. In conclusion, the calculation of D A L Y s and associated health status indicators represents the first step in evaluating the impact of injuries in the communities of BC. In order to develop appropriate policies to address the problem of injuries, one should also conduct complementary studies that include assessments of the economic impact of injuries. In economic analyses, it would be important to assess the magnitude of direct health care costs such as the cost of medications, services involved in rehabilitation and treatment of injuries, as well as the magnitude of indirect costs such as the loss of productivity and income. In addition, the D A L Y analysis in this study can be extended to include projections of the total burden of injuries for future years. These projections could be used to inform decision makers of potential areas of high demand for hospital and rehabilitation services at the provincial and regional levels. Finally, in the future, D A L Y as a measure of cost/benefit per unit of injury burden avoided should be applied to rank health interventions in priority setting. Cost-effectiveness analysis of health interventions has gained popularity in the health policy and priority setting literature (World Bank 1993; Lyttkens 2003; Arnesen and Kapiriri Chapter 7. Discussion 204 2003; Paalman et al. 1998; Jamison and Mosley 1991; Mahapatra 2002; to name a few). The importance of setting priorities arises from a scarcity of resources in dealing with the health needs of the population. Health policy makers, in general, are faced with having to manage resources so that they maximize health outcomes, even i f it means allocating limited new resources or cutting back on the use of existing resources. However, needless to say this is a complex exercise, since policy makers have to base their decisions on: social, political and ethical issues; technical and practical feasibility; cost-effectiveness of interventions; and an array of other causes of disease with significant population burden. One of the main challenges in priority setting using cost-effectiveness analysis is to find a measure of disease burden capable of providing valid and reliable information. It is important to acknowledge that even though D A L Y has been extensively advocated to be used in such analysis, caution is needed in the interpretation of the results. As mentioned in previous sections, there are several intrinsic preference values in the D A L Y calculation, such as choices of discounting, age weighting, life expectancy and disability weights. Thus, it is expected that decisions will change based on different combination of such preference values. As I see it, the potential of using D A L Y in priority setting decisions is enormous, however until D A L Y is considerably improved, we should limit its use as a measure to provide descriptive information about the mortality and morbidity health impacts of specific illnesses, until there is great clarity regarding the reasonable estimation of preference values and demonstration of D A L Y ' s utility with those agreed upon values. Chapter 7. Discussion 205 7.4 Conclusion The disability-adjusted life years (DALY) methodology provides important baseline information about the past and present differential impact of injuries on subgroups of the population; it also shows how variable the distribution of burden and incidence of different causes of injuries are among regions in BC. This information is important to motivate different researchers to include injuries in future decisions regarding health care interventions. Moreover, unique to this study, we have monitored the health status of the BC population at the small-area level using the latest statistical models in disease mapping. We demonstrated how practical and appealing the models used in this study were in obtaining smoothed injury incidence rates to be later used in calculating the injury burden at regional and provincial levels. Inferences drawn from these statistical models in disease mapping enabled lis to assess whether the differences in health status indicators from one region to another were in fact statistically significant. In addition, the success in obtaining small-area estimates of rates, ratios and healthy years lost due to mortality and disability is a first step towards providing information to assist in health planning and priority setting at the regional and provincial levels. Furthermore, we demonstrated the tremendous potential of using administrative data and data linkage to ascertain hospital separations and deaths using the International Classification of Disease System (ICD) Version 9 coding system. For some injuries, such as medical adverse events, we showed that the use of E-codes is an easy and inexpensive method to unambiguously identify these injuries. Bibliography 206 Bibl iography Agency for Healthcare Research and Quality. Department of Health and Human Services. A H R Q Quality Indicators - Guide to Patient Safety Indicators. Rockville, M D : Agency for Healthcare Research and Quality, 2003. <ht1p://vv^vvw.qualityindicators.ahrq.gov/downloads/psi/psi_guide_rev3.pdf>. Anand S, Hanson K . "Disability-Adjusted Life Years: A Critical Review." Journal of Health Economics Vol . 16, No. 6 (1997): 685-702. Andersen R, and Newman JF. "Societal and Individual Determinants of Medical Care Utilization in the United States." Milbank Memorial Fund Quarterly Vol . 51 (1973): 95-124. Arnesen T, Kapiriri L. "Can the Value Choices in D A L Y s Influence Global Priority-Setting?" Health Policy Vol . 70, No. 2 (2004): 137-149. Baker GR, Norton PG, Flintoft V , Blais R, Brown A , Cox J, Etchells E, Ghali W A , Hebert P, Majumdar SR, O'Beirne M , Palacios-Derflingher L , Reid RJ, Sheps S, and Tamblyn R. "The Canadian Adverse Events Study: The Incidence of Adverse Events among Hospital Patients in Canada." Canadian Medical Association Journal Vol . 170, No. 11 (2004): 1670-1686. Bibliography 207 Barss P, Smith GS, Baker SP, and Mohan D. Injury Prevention: An International Perspective-Epidemiology, Surveillance, and Policy. New York: Oxford University Press, 1998. B C Population by Age and Gender. 1971-2004. Table. British Columbia: Ministry of Management Services. B C Stats, 2004. <http://www.bcstats.gov.bc.ca/data/pop/pop/BCPopage.htm>. Begg S, and Tomijima N . "Global Burden of Injury in the Year 2000: An Overview of Methods." World Health Organization Draft 30-01-05 (2005). <http://www3.who.int/whosis/menu.cfm?path=evidence,burden&language=english>. Bernardinelli L , Clayton D, Pascutto C, Montomoli C, Ghislandi M , and Songini M . "Bayesian Analysis of Space-Time Variation in Risk." Statistics in Medicine Vol . 14 No. 21-22 (1995): 2433-2443. Bernardinelli L , and Montomoli M . "Empirical Bayes versus Fully Bayesian Analysis of Geographical Variation in Disease Risk." Statistics in Medicine, Vol . 11, No. 8 (1992): 983-1007. Besag J, York J, and Mollie A . "Bayesian Image Restoration, with Two Applications in Spatial Statistics." Annals of the Institute of Statistical Mathematics 43 (1991): 1-21. Bibliography 208 Bohning D, Dietz E, Schlattmann P, Mendonca L, and Kirchner U . "The Zero-Inflated Poisson Model and the Decayed, Missing and Filled Teeth Index in Dental Epidemiology." Journal of the Royal Statistical Society, Series A (Statistics in Society) Vol.162, No. 2 (1999): 195-209. Bonneux L. "How to Measure the Burden of Mortality?" Journal of Epidemiology and Community Health Vol . 56. No. 2 (2002): 128-131. Bota GW, Therrien SA, and Rowe B H . " A Truncated E-Code System for Injury Surveillance in the Emergency Department: Description and Clinometric Testing." Academic Emergency Medicine Vol . 4. No. 4 (1997J: 291-296. Brennan TA, Leape L L , Laird N M , Heber L, Localio AR, Lawthers A G , Newhouse JP, Weiler PC, and Hiatt H H . "Incidence of Adverse Events and Negligence in Hospitalized Patients. Results of the Harvard Medical Practice Study I." The New England Journal of Medicine Vol . 324, No. 6 (1991): 370-377. BreslOw NE, and Clayton D G . "Approximate Inference in Generalized Linear Mixed Models." Journal of the American Statistical Association Vol . 88, No. 421 (1993): 9-25. Canada 2001 Census. Table. Canada: Statistics Canada, 2004. <http://wwwl2.statcan.ca/english/census01/home/index.cfin>. Bibliography 209 Canadian Institute for Health Information. Health Indicators 2004. 15 Dec. 2004a. <http://secure.cihi.ca/cihiweb/dispPage.jsp?cw_page=indicators_e>. Canadian Institute for Health Information. System Performance is the Real Problem. 15 Dec. 2004b. < http://secure.cihi.ca/cihiweb/en/downloads/error_e_HP2_lerrors.pdf>. Carlin BP, and Louis TA. Bayes and Empirical Bayes Methods for Data Analysis. 2nd ed. London: Chapman and Hall, 2000. Centers for Disease Control and Prevention. National Center for Health Statistics. Classification of Diseases and Functioning and Disability. International Classification of Diseases, Ninth Revision (ICD-9). 30 Aug. 2003. <http://www.cdc.gov/nchs/icd9.htm>. Christoffel T, and Gallagher SS. Injury Prevention and Public Health. Practical Knowledge, Skills, and Strategies. Gaithersburg, Maryland: Aspen Publication, 1999. Clayton DG, Bernardinelli L, and Montomoli C. "Spatial Correlation and Ecological Analysis." International Journal of Epidemiology Vol . 22 (1993): 1193-1201. Coale A J , and Guo G. "Revised Regional Model Life Tables at Very Low Levels of Mortality." Population Index Vol . 55, No. 4 (1989): 613-643. Bibliography 210 Davis P, Lay-Yee R, Briant R, A l i W, Scott A , and Schug S. "Adverse Events in New Zealand Public Hospitals I: Occurrence and Impact." The New Zealand Medical Journal Vol . 115, No. 1167 (2002): U271. —. "Adverse Events in New Zealand Public Hospitals II: Preventability and Clinical Context." The New Zealand Medical Journal Vol . 116, No. 1183 (2003a): U624. Davis P, Lay-Yee R, Scott A , Briant R, and Schug S. "Acknowledgement of "No Fault" Medical Injury: Review of Patients' Hospital Records in New Zealand." British Medical Journal Vol . 326 (2003b): 79-80. Department of Justice. Canadian Health Act. Chapter C-6. Laws. 15 Sep. 2004 < http://laws.justice.gc.ca/en/c-6/16839.html>. Diehr P, Cain K C , Kreuter W, and Rosenkranz S. "Can Small-Area Analysis Detect Variation in Surgery Rates? The Power of Small-Area Variation Analysis." Medical Care Vol . 30, No. 6 (1992): 484-502. Elliott P, Wakefield J, Nicola B, and Briggs D. Spatial Epidemiology. Methods and Applications. New York: Oxford University Press, 2001. Bibliography 211 Elliott P, and Wartenberg D. "Spatial Epidemiology: Current Approaches and Future Challenges." Environmental Health Perspectives Vol . 112, No. 9 (2004): 998-1006. 8 Aug. 2004. < http://ehp.niehs.nih.gov/members/2004/6735/6735.html>. Environmental Systems Research Institute Inc. Getting to Know Arcview GIS for Version 3.1. United States: The Editors of the Environmental Systems Research Institute Press, 1999. Erickson P. "Evaluation of a Population-Based Measure of Quality of Life: The Health and Activity Limitation Index (HALex)." Quality of Life Research Vol . 7, No. 2 (1998): 101-114. EuroQol Group. "EuroQol: A New Facility for the Measurement of Health-Related Quality of Life". Health Policy Vol . 16, No. 3 (1990): 199- 208. Ezzati M , Lopez A D , Rodgers A , Vander Hoorn S, and Murray CJ. "Selected Major Risk Factors and Global and Regional Burden of Disease." Lancet Vol . 360, No. 9343 (2002): 1347-1360. Field JM, and Gold M R , ed. Summarizing Population Health: Directions for the Development and Application of Population Metrics. Institute of Medicine. Washington, DC: National Academy Press, 1998. Bibliography 212 # Forster A J , Asmis TR, Clark HD, A l Saied G, Code CC, Caughey SC, Baker K , Watters J, Worthington J, and van Walraven C. "Ottawa Hospital Patient Safety Study: Incidence and Timing of Adverse Events in Patients Admitted to a Canadian Teaching Hospital." Canadian Medical Association Journal Vol . 170, No. 8 (2004): 1235-1240. Forster A J , Clark HD, Menard A , Dupuis N , Chernish R, Chandok N , Khan A , and van Walraven C. "Adverse Events among Medical Patients after Discharge from Hospital." Canadian Medical Association Journal Vol . 170, No. 3 (2004): 345-349. German A , Carlin JB, Stern HS, and Rubin DB. Bayesian Data Analysis. London: Chapman and Hall, 1998. Gold M R , Stevenson D, and Fryback DG. " H A L Y S and Q A L Y S and D A L Y S , Oh My: Similarities and Differences in Summary Measures of Population Health." Annual Review of Public Health Vol . 23 (2002): 115-134. Government of British Columbia. Ministry of Health Services. Health Authorities. Roles and Responsibilities of Health Authorities. 30 Aug. 2004. <http://www.healthservices.gov.bc.ca/socsec/roles.html>. Bibliography 213 Government of New Zealand. Ministry of Health. "The Burden of Disease and Injury in New Zealand." Public Health Intelligence Occasional Bulletin No. 1. Jan. 2001. <http://www.moh.govt.nz>. Green C. "On the Societal Value of Health Care: What Do We Know about the Person Trade-Off Technique?" Health Economics Vol . 10, No. 3 (2001): 233-243. Green C, Brazier J, and Deverill M . "Valuing Health-Related Quality of Life. A Review of Health State Valuation Techniques." Pharmacoeconomics Vol . 17, No. 2 (2000): 151-165. Gridelli C. "Same Old Story? Do We Need to Modify Our Supportive Care Treatment of Elderly Cancer Patients? Focus on Antiemetics." Drugs & Aging Vol . 21, No. 13 (2004): 825-832. Hennekens C H , Buring JE. Epidemiology in Medicine. Boston: Little, Brown & Company, 1987. Hollinghurst S, Bevan G, and Bowie C. "Estimating the "Avoidable" Burden of Disease by Disability Adjusted Life Years (DALYs)." Health Care Management Science Vol . 3, No. 1 (2000): 9-21. Bibliography 214 Howe G M . "Historical Evolution of Disease Mapping in General and Specifically of Cancer Mapping." Cancer Mapping. Ed. Boyle P, Muir CS, Grundmann E. New York: Springer-Verlag, 1989. Hunter D, and Bains N . "Rates of Adverse Events among Hospital Admissions and Day Surgeries in Ontario from 1992 to 1997." Canadian Medical Association Journal Vol . 160, No. 11 (1999): 1585-1586. Iburg K M , and Kamper-Jorgensen F. "Summary Measures of Population Health. A n Overview." Danish Medical Bulletin Vol . 49, No. 3 (2002): 256-259. Iezzoni LI, Foley SM, Heeren T, Daley J, DunCan CC, Fisher ES, Hughes J. " A Method for Screening the Quality of Hospital Care using Administrative Data: Preliminary Validation Results." Quality Review Bulletin Vol . 18, No. 11 (1992): 361-71. Jelsma J, De Weerdt W, and De Cock P. "Disability Adjusted Life Years (DALYs) and Rehabilitation." Disability and Rehabilitation Vol . 24, No. 7 (2002): 378-382. Kaplan R M . " A General Health Policy Model: Update and Applications." Health Services Research Vol . 23 (1988): 203-235. Karson, AS, and Bates DW. "Screening for Adverse Events." Journal of Evaluation in Clinical Practice Vol . 5, No. 1 (1999): 23-32. Bibliography 215 Kleinschmidt I, Sharp B, Mueller I, and Vounatsou P. "Rise in Malaria Incidence Rates in South Africa: A Small-Area Spatial Analysis of Variation in Time Trends." American Journal of Epidemiology Vol . 155, No. 3 (2002): 257-264. Knorr-Held L , and Besag J. "Modelling Risk from a Disease in Time and Space." Statistics in Medicine Vol . 17, No. 18 (1998): 2045-2060. Kohn LT, Corrigan JM, and Donaldson MS, ed. To Err is Human. Building a Safer Health System. Institute of Medicine. Washington, DC: National Academy of Sciences, 2000. Lambert D. "Zero-Inflated Poisson Regression, with an Application to Defects in Manufacturing." Tecnometrics Vol . 34, No. 1 (1992): 1-14. Lahdis J.R., Koch G.G. "The Measurement of Observer Agreement for Categorical Data." Biometrics Vol . 33, No . l (1977): 159-174. Lawson A , Biggeri A , Bohning D, Lesaffre E, Viel J-F, and Bertollini R, ed. Disease Mapping and Risk Assessment for Public Health. New York: John Wiley & Sons, 1999. Bibliography 216 Leape L L , Brennan TA, Laird N , Lawthers A G , Localio AR, Barnes B A , Hebert L , Newhouse JP, Weiler PC, and Hiatt H . "The Nature of Adverse Events in Hospitalized Patients. Results of the Harvard Medical Practice Study II." New England Journal of Medicine Vol . 324, No. 6 (1991): 377-384. Lee A H , Stevenson MR, Wang K , and Yau K K . "Modeling Young Driver Motor Vehicle Crashes: Data with Extra Zeros." Accident; Analysis and Prevention Vol . 34, No. 4 (2002): 515-521. LeMier M , Cummings P, and West TA. "Accuracy of External Cause of Injury Codes Reported in Washington State Hospital Discharge Records." Injury Prevention Vol . 7, No. 4 (2001): 334-338. Lu TH, Lee M C , and Chou M C . Accuracy of Cause-of-Death Coding in Taiwan: Types of Miscoding and Effects on Mortality Statistics. International Journal of Epidemiology Vol . 29, No. 2 (2000): 336-343. Maclntyre CR, Ackland MJ , and Chandraraj EJ. "Accuracy of Injury Coding in Victorian Hospital Morbidity Data." Australian and New Zealand Journal of Public Health Vol . 21, No. 7 (1997): 779-783. MacKenzie EJ. "Epidemiology of Injuries: Current Trends and Future Challenges." Epidemiological Reviews Vol . 22, No. 1 (2000): 112-119. Bibliography 217 MacNab Y C . "Generalized Linear Mixed Models with Splines: Bayesian Disease Mapping and Spatio-temporal Surveillance." Proceedings of the Hawaii International Conference on Statistics (2002): 1-29. —. " A Bayesian Hierarchical Model for Accident and Injury Surveillance." Accident; Analysis and Prevention. Vol . 35, No. 1 (2003a): 91-102. —. "Hierarchical Bayesian modeling of spatially correlated health service outcome and utilization rates." Biometrics Vol . 52, No. 2 (2003b): 305-316. "Bayesian Spatial and Ecological Models for Small-Area Accident and Injury Analysis." Accident; Analysis and Prevention. Vol . 36, No. 6 (2004): 1019-1028. MacNab Y C , and Dean CB. "Parametric Bootstrap and Penalized Quasi-Likelihood Inference in Conditional Autoregressive Models." Statistics in Medicine Vol . 19, No. 15/16(2000): 2421-2435. —. "Autoregressive Spatial Smoothing and Temporal Spline Smoothing for Mapping Rates." Biometrics Vol . 57, No. 3 (2001): 949-956. —. "Spatio-Temporal Modelling of Rates for the Construction of Disease Maps." Statistics in Medicine Vol . 21, No. 3 (2002): 347-358. Bibliography 218 MacNab Y C , Farrell PJ, Gustafson P, and Wen S. "Estimation in Bayesian disease mapping." Biometrics Vol . 60, No. 4 (2004): 865-873. Manton K G , Woodbury M A , Stallard E, Riggan WB, Creason JP, and Pellom A C . "Empirical Bayes Procedures for Stabilizing Maps of US Cancer Mortality Rates." Journal of the American Statistical Association Vol . 84 (1989): 637-650. Manuel DG, Schultz SE, and Kopec JA. "Measuring the Health Burden of Chronic Disease and Injury using Health Adjusted Life Expectancy and the Health Utilities Index." Journal of Epidemiology and Community Health Vol . 56, No. 11 (2002): 843-850. Mathers CD, Vos ET, Lopez A D , Salomon J, and Ezzati M . National burden of Disease Studies: A Practical Guide. 2nd edition. Global Program on Evidence for Health Policy. Geneva: World Health Organization, 2001. Mathers CD, Vos ET, Stevenson CE. The Burden of Disease and Injury in Australia. Canberra: Australian Institute of Health and Welfare, 1999. 15 Jan. 2003 <http://vvww.aihw.gov.au>. Mathers CD, Vos ET, Stevenson CE, and Begg SJ. "The Australian Burden of Disease Study: Measuring the Loss of Health from Diseases, Injuries and Risk Factors." The Medical Journal of Australia Vol . 172, No. 12 (2000): 592-596. Bibliography 219 —. "The Burden of Disease and Injury in Australia." Bulletin of the World Health Organization Vol . 79, No. 11 (2001): 1076-1084. Melse JM, Essink-Bot M L , Kramers PG, and Hoeymans N . " A National Burden of Disease Calculation: Dutch Disability-Adjusted Life-Years. Dutch Burden of Disease Group." American Journal of Public Health Vol . 90, No. 8 (2000): 1241-1247. Michel P, Quenon JL, de Sarasqueta A M , and Scemama O. "Comparison of Three Methods for Estimating Rates of Adverse Events and Rates of Preventable Adverse Events in Acute Care Hospitals." British Medical Journal Vol . 328, No. 7433 (2004): 199-203. Miller M , Elixhauser A , Zhan C, Meyer G. "Patient Safety Indicators: Using Administrative Data to Identify Potential Patient Safety Concerns." Health Services Research Vol . 36, No. 6 Part II (2001): 110-132. Mollie A . "Bayesian Mapping of Disease." Markov Chain Monte Carlo in Practice. London: Chapman and Hail, 1996. Moyer L A , Boyle CA, and Pollock DA. "Validity of Death Certificates for Injury-Related Causes of Death." American Journal of Epidemiology Vol . 130, No. 5 (1989): 1024-1032. Bibliography 220 Murray CJ, and Acharya A K . "Understanding D A L Y s . " Journal of Health Economics Vol . 16, No. 6(1997): 703-730. Murray CJ, and Lopez A D . The Global Burden of Disease. Cambridge, Massachusetts: Harvard University Press, 1996. —. "Mortality by Cause for Eight Regions of the World: Global Burden of Disease Study." Lancet Vol . 349, No. 9061 (1997a): 1269-1276. —. "Regional Patterns of Disability-Free Life Expectancy and Disability-Adjusted Life Expectancy: Global Burden of Disease Study." Lancet Vol . 349, No. 9062 (1997b): 1347-1352. —. "Global Mortality, Disability, and the Contribution of Risk Factors: Global Burden of Disease Study." Lancet Vol . 349, No. 9063 (1997c): 1436-1442. —. "Alternative Projections of Mortality and Disability by Cause 1990-2020: Global Burden of Disease Study." Lancet Vol . 349, No. 9064 (1997d): 1498-1504. —. "Progress and Directions in Refining the Global Burden of Disease Approach: A Response to Williams." Health Economics Vol . 9, No. 1 (2000): 69-82. Bibliography 221 Murray CJ, Salomon JA, and Mathers C. " A Critical Examination of Summary Measures of Population Health." Bulletin of the World Health Organization Vol . 78, No. 8 (2000): 981-984. Murray CJ, Salomon JA, Mathers C, and Lopez A D (2002). Summary Measures of Population Health. Concepts, Ethics, Measurement and Applications. Geneva, Switzerland: Office of Publications, World Health Organization, 2002. Nelson DE, Sacks JJ, Parrish RG, Sosin D M , McFeeley P, Smith SM. "Sensitivity of Multiple-Cause Mortality Data for Surveillance of Deaths Associated With Head or Neck Injuries." Morbidity and Mortality Weekly Report. CDC Surveillance Summaries / Centers for Disease Control Vol . 42, No. 5 (1993): 29-35. Owen JL, Bolenbaucher R M , and Moore M L . "Trauma Registry Databases: A Comparison of Data Abstraction, Interpretation, and Entry at Two Level I Trauma Centers." The Journal of Trauma Vol . 46, No. 6 (1999): 1100-1104. Pinto Prades JL. "Is the Person Trade-Off a Valid Method for Allocating Health Care Resources?" Health Economics Vol . 6, No. 1 (1997): 71-81. Ribbeck B M , Runge JW, Thomason M H , Baker JW. "Injury Surveillance: a Method for Recording E Codes for Injured Emergency Department Patients." Annals of Emergency Medicine Vol . 21, No. 1 (1992): 37-40. Bibliography 222 Richardson S (1992). "Statistical Methods for Geographical Correlation Studies." Geographical and Environmental Epidemiology: Methods for Small-Area Studies. London: Oxford University Press, 1992. Rothman K J , and Greenland S. Modern Epidemiology. 2nd edition. Philadelphia: Lippincott-Raven Publishers, 1998. Samore M H , Evans RS, Lassen A , Gould P, Lloyd J, Gardner R M , Abouzelof R, Taylor C, Woodbury D A , Willy M , and Bright RA. "Surveillance of Medical Device-Related Hazards and Adverse Events in Hospitalized Patients." Journal of the American Medical Association Vol . 291, No. 3 (2004): 325-334. SAS Institute Inc. The SAS System for Windows, version 8.0. United States: Cary, North Caroline, 1999. <http://www.sas.com>. Sayers B M , Fliedner T M . "The Critique of D A L Y s : A Counter-Reply." Bulletin of the World Health Organization Vol . 75, No. 4 (1997): 383-385. Schneider, MJ. Introduction to Public Health. Gaithersburg, Maryland: Aspen Publishers, 2000. Bibliography 223 Schwartz RJ, Nightingale BS, Boisoneau D, and Jacobs L M . "Accuracy of E-Codes Assigned to Emergency Department Records." Academic Emergency Medicine Vol . 2, No. 7(1995): 615-620. Sniezek JE, Finklea JF, and Graitcer PL. "Injury Coding and Hospital Discharge Data." The Journal of the American Medical Association Vol . 262, No. 16 (1989): 2270-2272. Spiegelhalter DJ, Best N G , Carlih BP, and van der Linde A . "Bayesian Measures of Model Complexity and Fit." Journal of the Royal Statistical Society: Series B (Statistical methodology). Vol . 64, No. 4 (2002): 583-640. Spiegelhalter D, Thomas A , Best N , and Lunn D. WinBUGS User Manual Version 1.4. Jan. 2003. <http://www.mrc-bsu.cam.ac.uk/bugs>. Stouthard M E A , Essink-Bot M - L , and Bonsel GJ. "Disability Weights for Diseases. A Modified Protocol and Results for a Western European Region." European Journal of Public Health Vol . 10, No. 1 (2000): 24-30. Stouthard M E A , Essink-Bot M - L , Bonsel GJ, Barendregt JJ, Kramers PGN, P A van de Water H , Gunning-Schepers LJ , and van der Maas PJ. Disability Weights for Diseases in the Netherlands. Netherlands: Department of Public Health, Erasmus University Rotterdam, 1997. Bibliography 224 Sun D, Tsutakawa R K , Kim H , and He Z. "Spatio-Temporal Interaction with Disease Mapping." Statistics in Medicine Vol . 19, No. 15 (2000): 2015-2036. Taylor R. "Small Area Population Disease Burden." Australian and New Zealand Journal of Public Health Vol . 25, No. 4 (2001): 289-293. Thomas EJ, and Brennan TA. "Incidence and Types of Preventable Adverse Events in Elderly Patients: Population Based Review of Medical Records." British Medical Journal Vol . 320, No. 7237 (2000): 741-744. Thomas EJ, Studdert D M , Burstin HR, Orav EJ, Zeena T, Williams EJ, Howard K M , Weiler PC, and Brennan TA. "Incidence and Types of Adverse Events and Negligent Care in Utah and Colorado." Medical Care Vol . 38, No. 3 (2000): 261-271. Torrance GW. "Measurement Of Health State Utilities For Economic Appraisal." Journal Of Health Economics Vol . 5, No. 1 (1986): 1-30. Torrance GW. "Utility Approach to Measuring Health-Related Quality of Life." Journal of Chronic Diseases Vol . 40, No. 6 (1987): 593-603. Torrance GW, Boyle M H , and Horwood SP. "Application Of Multi-Attribute Utility Theory To Measure Social Preferences For Health States." Operations Research Vol . 30, No. 6 (1982): 1043-1069. Bibliography 225 Torrance GW, Furlong W, Feeny D, and Boyle M . "Multi-Attribute Preference Functions. Health Utilities Index." Pharmacoeconomics Vol . 7, No. 6 (1995): 503-520. Trinka E. "Epilepsy: Comorbidity in the Elderly." Acta Neurologica Scandinavica. Supplementum Vol . 180 (2003): 33-36. Van den Broek J. " A Score Test for Zero Inflation in a Poisson Distribution." Biometrics Vol . 51, No. 2 (1995): 738-743. Venables WN, and Ripley BD. Modern Applied Statistics with S-PLUS. 2nd Edition. New York: Springer-Verlag, 1997. Victorian Department of Human Services. Public Health and Development Division. The Victorian Burden of Disease Study: Mortality. 1999a. 8 Jul. 2003. <http://www.health.vic.gov.au/healthstatus/bod/bod_vic.htm>. —. Public Health and Development Division. The Victorian Burden of Disease Study: Morbidity. 1999b. 1 Jan. 2004. <http://www.health.vic.gov.au/healthstatus/bod/bod_vic.htm>. Vincent C, Neale G, and Woloshynowych M . "Adverse Events in British Hospitals: Preliminary Retrospective Record Review." British Medical Journal Vol . 322, No. 7285 (2001):517-519. Bibliography 226 Vital Statistics Agency of British Columbia. Selected Vital Statistics and Health Status Indicators. One Hundred Thirtieth, Annual Report 2001. 19 Nov. 2002 <http://www.vs.gov.bc.ca/stats/pdf.html>. Volinn E, Diehr P, Ciol M A , Loeser JD. "Why Does Geographic Variation in Health Care Practices Matter? (And Seven Questions to Ask in Evaluating Studies on Geographic Variation)." Spine Vol . 19, Suppl.18 (1994): 2092S-2100S. Walshe K . "Adverse Events in Health Care: Issues in Measurement." Quality in Health Care Vol . 9, No. 1 (2000): 47-52. Wanzel K R , Jamieson CG, and Bohnen JMA. "Complications on a General Surgery Service: Incidence and Reporting." Canadian Journal of Surgery Vol . 43, No. 2 (2000): 113-117. Weingart SN, Wilson R M , Gibberd RW, and Harrison B. "Epidemiology of Medical Error." British Medical Journal Vol . 320, No. 7237 (2000): 774-777. Williams A. "Calculating the Global Burden of Disease: Time for a Strategic Reappraisal?" Health Economics Vol . 8, No. 1 (1999): 1-8. Bibliography 227 Wilson R M , Runciman WB, Gibberd RW, Harrison BT, Newby L , and Hamilton JD. "The Quality in Australian Health Care Study." The Medical Journal of Australia Vol . 163, No. 9 (1995): 458-476. Winn DG, Agran PF, Anderson CL. "Sensitivity of Hospitals' E-Coded Data in Identifying Causes of Children's Violence-Related Injuries." Public Health Reports Vol . 110, No. 3 (1995): 277-281. World Bank. World Development Report 1993: Investing in Health. New York: Oxford University Press, 1993. 21 May. 2004. <http://www-wds.worldbank.org/servlet/WDS_IB ank_Servlet?pcont=details&eid= 000009265_3970716142319>. World Health Organization. Department of Injuries and Violence Prevention. Injury: A Leading Cause of the Global Burden of Disease, 2000. Ed. Peden M , McGee K , and KrugE. 15 Sept. 2004a: < http://www.who.int/violence_injury_prevention/publications/surveillance/en/>. —. Department of Injuries and Violence Prevention. The Injury Chartbook: A Graphical Overview of the Global Burden of Injuries. Ed. Peden M , McGee K , and Sharma G. 15 Sept. 2004b. < http://www.who.int/violence_injury_prevention/publications/surveillance/en/>. Bibliography 228 —. World Health Organization Statistical Information System (WHOSIS). Global Burden of Disease Estimates 21 Oct. 2004c. <http://www3.who.int/whosis/menu.cfm?path=whosis,bod,burden,burden_estimates &language=english>. Yau K K , and Lee A H . "Zero-Inflated Poisson Regression with Random Effects to Evaluate an Occupational Injury Prevention Programme." Statistics in Medicine Vol . 20, No. 19 (2001): 2907-2920. Zhan C, and Miller MR. "Administrative data based patient safety research: a critical review." Quality & Safety in Health Care Vol . 12, Suppl. 2 (2003a): ii58-63. Zhan C, and Miller MR. "Excess Length of Stay, Charges and Mortality Attributable To Medical Injuries during Hospitalizations." Journal of the American Medical Association Vol . 290, No. 14 (2003b): 1868-1874. 229 Appendices A. Acronyms B. Glossary C. Important ICD9-CM E-Codes D. Example of Disability Weight Spreadsheet Assuming Zero Disability Weight for the Residual Category of Injury Sequela E. Example of Duration Spreadsheet Assuming Zero Duration for the Residual Category of Injury Sequela F. Sensitivity Analyses G. Chapters' General Tables and Figures H. Diagnostic Plots 230 Appendix A Acronyms AE Adverse event ASDALY Age-standardized per capita disability-adjusted life years per 1,000 population ASHR Age-standardized hospitalization rate per 1,000 population ASMR Age-standardized mortality rate per 10,000 population Age-standardized per capita healthy years lost due to disability per 1,000 ASYLD population ASYLL Age-standardized per capita years of life lost per 1,000 population BC British Columbia DALY Disability-adjusted life years DALY[0,0] Disability-adjusted life years for no age weighting and 0% discounting rate DALY[0,3] Disability-adjusted life years for no age weighting and 3% discounting rate D ^residual Disability weight for the residual category of injury DW Disability weight GBD Global Burden of Disease HA Health Authority HALY Health-adjusted life years HRQL Health related quality of life HSDA Health Services Delivery Area ICD9-CM International Classification of Disease version 9 with clinical modifications Appendix A. Acronyms ICD10 International Classification of Disease version IQR Merquantile range LOS Length of stay in a hospital PHA Public health approach QALY Quality-adjusted life years SMR Standardized mortality rate SHR Standardized hospitalization rate SMPH Summary measures of population health YLD Healthy years lost due to disability Y L L Years of life lost WHO World Health Organization 232 Appendix B Glossary AGE-SPECIFIC PER CAPITA DISABILITY-ADJUSTED LIRE Y E A R S PER 1,000 POPULATION - The disability-adjusted life years (DALY) per 1,000 population by age and gender. The population used was the mid-period (1995) population estimates obtained from British Columbia Vital Statistics Agency. AGE-SPECIFIC PER CAPITA H E A L T H Y Y E A R S LOST D U E TO DISABILITY PER 1,000 POPULATION - (See Age-specific per capita disability-adjusted life years per 1,000 population) AGE-SPECIFIC PER CAPITA Y E A R S OF LIFE LOST PER 1,000 POPULATION - (See Age-specific per capita disability-adjusted life years per 1,000 population) A G E STANDARDIZATION (Source: British Columbia Vital Statistics Agency) - Age standardization is a method of calculation that adjusts a statistical measure for differences in the age/gender structures between populations. With standardized measures, more meaningful comparisons can be made between genders, different time periods, or geographic areas, because the age standardized statistic is calculated as i f all populations had the same age/gender population distribution. Appendix B. Glossary 233 AGE-STANDARDIZED M O R T A L I T Y R A T E (ASMR) PER 10,000 POPULATION (Source: British Columbia Vital Statistics Agency) - A summary of age-adjusted mortality rates by age and gender, which have been standardized to a 'standard' population (2001 Canada Census) for the purpose of rate comparisons between genders, different time periods or different geographic locations. The A S M R is the theoretical number of deaths that would occur per 10,000 population, i f the specific population had the same age structure as the standard population. Next, we show an example of calculation of A S M R at H A level Note that the same method was used to calculate the A S M R at H A level is the same method to calculate the A S M R at H S D A and B C levels. Table B . l . Example of calculation of A S M R for road traffic accident for the Interior HA. Males, 1991-2000. H A Age Standard Estimated Death Observed Group Population Population Rate/10,000 Deaths i *t Pt mi d< 0-19 . 3,983,130 506,158 1.88 171 20-29 4,194,045 614,251 3.79 343 40-64 4,867,020 558,254 2.14 215 65-79 1,344,430 167,236 2.24 81 80+ 318,220 39,674 3.66 31 Total 14,706,845 1,885,573 841 Appendix B. Glossary 234 For the Health Authority (HA): , 1.88x3,989,130 + ... + 3.66x318,220 ASMR - — = 2.58 El 14,706,845 where: pt is the H A population in age group i 7ri is the standard population in age group i FI is the total standard population, ^ ni i dt is the number of deaths in the H A in age group i m- is the mortality rate per 10,000 H A population in age group i, — x 10,000 Pi AGE-STANDARDIZED PER CAPITA DISABILITY-ADJUSTED LIFE Y E A R S (ASDALY) PER 1,000 POPULATION (See A S M R for details on calculation) - A summary of age-specific per capita D A L Y s by age and gender, which have been standardized to a 'standard ' population (2001 Canada Census) for the purpose of comparisons between genders, different time periods or different geographic locations. The A S D A L Y is the theoretical number of D A L Y s that would occur per 1,000 population, i f the specific population had the same age structure as the standard population. AGE-STANDARDIZED HOSPITALIZATION R A T E (ASHR) PER 1,000 POPULATION -(See A S M R per 10,000 population) AGE-STANDARDIZED PER CAPITA H E A L T H Y Y E A R S LOST DUE TO DISABILITY (ASYLD) PER 1,000 POPULATION - (See A S D A L Y per 1,000 population) Appendix B. Glossary 235 AGE-STANDARDIZED PER CAPITA Y E A R S OF LIFE LOST (ASYLL) PER 1,000 POPULATION - (See A S D A L Y per 1,000 population) STANDARD POPULATION (Source: British Columbia Vital Statistics Agency) - A reference population of known age distribution used in the calculation of standardized indicators to adjust for variations in population age structures in different geographic areas or time periods. The 2001 Canadian Census is used as the standard population in the calculation of A S M R , ASHR, A S D A L Y , A S Y L D and A S Y L L . Please see Tables G.5 for the actual population numbers. Appendix B. Glossary 236 ICD9-CM E-CODES FOR E X T E R N A L CAUSES OF INJURY (Source: ICD-9 Volume 1, Modified for US Purposes) This document includes a modified version of Volume 1 of the International Classification of Diseases, Ninth Revision (ICD-9) (CDC 2003). The full version of this volume is no longer available in print. This version has been modified to more appropriately reflect US mortality use. This material was provided by NCHS, which is a subunit of CDC/ATSDR. It has been cleared for public distribution by CDC/ATSDR and will be authentic i f gotten from ftp://ftp.cdc.gov. CDC/ATSDR takes all effort to assure the authenticity of electronically distributed documents. Mortality Statistics Branch National Center for Health Statistics Centers for Disease Control and Prevention 6525 Belcrest Road Hyattsville, M D 20782 (301) 436-8884 Appendix B. Glossary 237 Road traffic accidents Note: Other road vehicle accidents are transport accidents involving road vehicles other than motor vehicles. Another rOad vehicle is any device, except a motor vehicle, in, on, or by which any person or property may be transported on a highway. A streetcar is a device designed and used primarily for transporting persons within a municipality, running on rails, usually subject to normal traffic control signals, and operated principally on a right-of-way that forms part of the traffic way. A trailer being towed by a streetcar is considered a part of the streetcar. A pedal cycle is any road transport vehicle operated solely by pedals. Includes: accidents involving other road vehicles being used in recreational or sporting activities. Excludes: collision of other road vehicle [any] with: aircraft (E840.0-E845.9), motor vehicle (E813.0-E813.9, E820.0-E822.9), railway train (E801.0-E801.9). K D9-CM E-Codcs Description E810 Motor vehicle traffic accident involving collision with train E811 Motor vehicle traffic accident involving re-entrant collision with another motor vehicle E812 Other motor vehicle traffic accident involving collision with motor vehicle E813 Motor vehicle traffic accident involving collision with other vehicle E814 Motor vehicle traffic accident involving collision with pedestrian E815 Other motor vehicle traffic accident involving collision on the highway E816 Motor vehicle traffic accident due to loss of control, without collision on the highway E817 Noncollision motor vehicle traffic accident while boarding or alighting E818 Other noncollision motor vehicle traffic accident E819 Motor vehicle traffic accident of unspecified nature E826 Pedal cycle accident E827 Animal-drawn vehicle accident E828 Accident involving animal being ridden E829 Other road vehicle accidents E929.0 Late effects of motor vehicle accident Appendix B. Glossary 238 Poisoning Includes: accidental overdose of drug, wrong drug given or taken in error, and drug taken inadvertently, accidents in the use of drugs and biologicals in medical and surgical procedures. Excludes: administration with suicidal or homicidal intent or intent to harm, or in circumstances classifiable to E980-E989 (E950.0-E950.5, E962.0, E980.0-E980.5), correct drug properly administered in therapeutic or prophylactic dosage, as the cause of adverse effect (E930.0-E949.9). ICD9-CM E-C odes Description E850 Accidental poisoning by analgesics, antipyretics, and antirheumatics E851 Accidental poisoning by barbiturates E852 Accidental poisoning by other sedatives and hypnotics E853 Accidental poisoning by tranquilizers E854 Accidental poisoning by other psychotropic agents E855 Accidental poisoning by other drugs acting on central and autonomic nervous system E856 Accidental poisoning by antibiotics E857 Accidental poisoning by other anti-infectives E858 Accidental poisoning by other drugs E860 Accidental poisoning by alcohol, not elsewhere classified E861 Accidental poisoning by cleansing and polishing agents, disinfectants, paints, and varnishes E862 Accidental poisoning by petroleum products, other solvents and their vapors, not elsewhere classified E863 Accidental poisoning by agricultural and horticultural chemical and pharmaceutical preparations other than plant foods and fertilizers E864 Accidental poisoning by corrosives and caustics, not elsewhere classified E865 Accidental poisoning from poisonous foodstuffs and poisonous plants E866 Accidental poisoning by other and unspecified solid and liquid substances E867 Accidental poisoning by gas distributed by pipeline E868 Accidental poisoning by other utility gas and other carbon monoxide E869 Accidental poisoning by other gases and vapors E929.2 Late effects of accidental poisoning Appendix B. Glossary 239 Falls Excludes: falls (in or from): burning building (E890.8, E891.8),into fire (E890.0-E899), into water (with submersion or drowning) (E910.0-E910.9),machinery (in operation) (E919.0-E919.9),on edged, pointed, or sharp object (E920.0-E920.9),transport vehicle (E800.0-E845.9),vehicle not elsewhere classifiable (E846-E848). ICD9-CM E C odes Description E880 Fall on or from stairs or steps E88I Fall on or from ladders or scaffolding E882 Fall from or out of building or other structure E883 Fall into hole or other opening in surface E884 Other fall from one level to another E885 Fall on same level from slipping, tripping, or stumbling Fall on same level from collision, pushing, or shoving, by or with E886 other person E887 Fracture, cause unspecified E888 Other and unspecified fall E929.3 Late effects of accidental fall Appendix B. Glossary 240 Medical adverse events Misadventures to patients during surgical and medical care Excludes: accidental overdose of drug and wrong drug given in error (E850.0-E858.9) surgical and medical procedures as the cause of abnormal reaction by the patient, without mention of misadventure at the time of procedure (E878.0-E879.9). ICD9-C M E-C odes Description Accidental cut, puncture, perforation, or hemorrhage during E870 medical care E871 Foreign object left in body during procedure E872 Failure of sterile precautions during procedure E873 Failure in dosage E874 Mechanical failure of instrument or apparatus during procedure Contaminated or infected blood, other fluid, drug, or biological E875 substance E876 Other and unspecified misadventures during medical care Surgical and medical procedures as the cause of abnormal reaction of patient or later complication, without mention of misadventure at the time of procedure Includes: procedures as the cause of abnormal reaction, such as: displacement or malfunction of prosthetic device, hepatorenal failure, postoperative, malfunction of external stoma, postoperative intestinal obstruction, rejection of transplanted organ. Excludes: anesthetic management properly carried out as the cause of adverse effect (E937.0-E938.9), infusion and transfusion, without mention of misadventure in the technique of procedure (E930.0-E949.9). rC D9-CM E-Codes Description E878 E879 Surgical operation and other surgical procedures as the cause of abnormal reaction of patient, or of later complication, without mention of misadventure at the time of operation Other procedures, without mention of misadventure at the time of procedure, as the cause of abnormal reaction of patient, or of later complication Appendix B. Glossary 241 Drugs, medicinal and biological substances causing adverse effects in therapeutic use Includes: correct drug properly administered in therapeutic or prophylactic dosage, as the cause of any adverse effect including allergic or hypersensitivity reactions. Excludes: accidental overdose of drug and wrong drug given or taken in error (E850.0-E858.9), accidents in the technique of administration of drug or biological substance such as accidental puncture during injection, or contamination of drug (E870.0-E876.9) administration with suicidal or homicidal intent or intent to harm, or in circumstances classifiable to E980-E989 (E950.0-E950.5, E962.0, E980.0-E980.5). ICD9-CM E C odes Description E930 Antibiotics E931 Other anti-infectives E932 Hormones and synthetic substitutes E933 Primarily systemic agents E934 Agents primarily affecting blood constituents E935 Analgesics, antipyretics, and antirheumatics E936 Anticonvulsants and anti-Parkinsonism drugs E937 Sedatives and hypnotics E938 Other central nervous system depressants and anesthetics E939 Psychotropic agents E940 Central nervous system stimulants E941 Drugs primarily affecting the autonomic nervous system E942 Agents primarily affecting the cardiovascular system E943 Agents primarily affecting gastrointestinal system E944 . Water, mineral, and uric acid metabolism drugs Agents primarily acting on the smooth and skeletal muscles and E945 respiratory system Agents primarily affecting skin and mucous membrane, E946 ophthalmological, otorhinolaryngological, and dental drugs E947 Other and unspecified drugs and medicinal substances E948 Bacterial vaccines E949 Other vaccines and biological substances Appendix B. Glossary 242 Other unintentional injuries (excluding road traffic accidents, poisoning, falls, and medical adverse events) Railway accidents Excludes: accidents involving railway train and aircraft (E840.0-E845.9), motor vehicle (E810.0-E825.9), watercraft (E830.0-E838.9). ICD9-C M E-C odes Description E800 Railway accident involving collision with rolling stock E801 Railway accident involving collision with other object Railway accident involving derailment without antecedent E802 collision E803 Railway accident involving explosion, fire, or burning E804 Fall in, on, or from railway train E805 Hit by rolling stock E806 Other specified railway accident E807 Railway accident of unspecified nature Motor vehicle non-traffic accidents Includes: accidents involving motor vehicles being used in recreational or sporting activities off the highway collision and non-collision motor vehicle accidents occurring entirely off the highway. Excludes: accidents involving motor vehicle and aircraft (E840.0-E845.9), watercraft (E830.0-E838.9), accidents, not on the public highway, involving agricultural and construction machinery but not involving another motor vehicle (E919.0, E919.2, E919.7). 1( D9-C M E-Codcs Description L820 E821 E822 E823 E824 E825 Non-traffic accident involving motor-driven snow vehicle Non-traffic accident involving other off-road motor vehicle Other motor vehicle non-traffic accident involving collision with moving object Other motor vehicle non-traffic accident involving collision with stationary object Other motor vehicle non-traffic accident while boarding and alighting Other motor vehicle non-traffic accident of other and unspecified nature Appendix B. Glossary 243 Water transport accidents Includes: watercraft accidents in the course of recreational activities. Excludes: accidents involving both aircraft, including objects set in motion by aircraft, and watercraft (E840.0-E845.9). ICD9-CM K-C'odcs Description E830 Accident to watercraft causing submersion E831 Accident to watercraft causing other injury Other accidental submersion or drowning in water transport E832 accident E833 Fall on stairs or ladders in water transport E834 Other fall from one level to another in water transport E835 Other and unspecified fall in water transport E836 Machinery accident in water transport E837 Explosion, fire, or burning in watercraft E838 Other and unspecified water transport accident Air and space transport accidents ICD9-CM E-Codes Description E840 Accident to powered aircraft at takeoff or landing E841 Accident to powered aircraft, other and unspecified E842 Accident to unpowered aircraft E843 Fall in, on, or from aircraft E844 Other specified air transport accidents E845 Accident involving spacecraft Vehicle accidents not elsewhere classifiable ICD9-CM E-Coilvs Description Accidents involving powered vehicles used solely within the E846 buildings and premises of industrial or commercial establishment E847 Accidents involving cable cars not running on rails E848 Accidents involving other vehicles, not elsewhere classifiable Appendix B. Glossary 244 Accidents caused by burns/fires/scalds Includes: asphyxia or poisoning due to conflagration or ignition burning by fire, secondary fires resulting from explosion. Excludes: arson (E968.0) fire in or on: machinery (in operation) (E919.0-E919.9), transport vehicle other than stationary vehicle (E800.0-E845.9), vehicle not elsewhere classifiable (E846-E848). ICD9-CM E-Codes Description E890 Conflagration in private dwelling E891 Conflagration in other and unspecified building or structure E892 Conflagration not in building or structure E893 Accident caused by ignition of clothing E894 Ignition of highly inflammable material E895 Accident caused by controlled fire in private dwelling Accident caused by controlled fire in other and unspecified E896 building or structure E897 Accident caused by controlled fire not in building or structure E898 Accident caused by other specified fire and flames E899 Accident caused by unspecified fire Natural and environmental factors IIP'CM'E'-Codes Description E900 Excessive heat E901 Excessive cold E902 High and low air pressure and changes in air pressure E903 Travel and motion E904 Hunger, thirst, exposure, and neglect Venomous animals and plants as the cause of poisoning and toxic E905 reactions E906 Other injury caused by animals E907 Lightning E908 Cataclysmic storms, and floods resulting from storms E909 Cataclysmic earth surface movements and eruptions Appendix B. Glossary 245 Accidents caused by drowning, suffocation, and foreign bodies ICD9-CM K-Codcs Description E910 Accidental drowning and submersion Inhalation and ingestion of food causing obstruction of respiratory E911 tract or suffocation Inhalation and ingestion of other object causing obstruction of E912 respiratory tract or suffocation E913 Accidental mechanical suffocation E914 Foreign body accidentally entering eye and adnexa E915 Foreign body accidentally entering other orifice Other accidents ICD9-CM K-Codes Description E916 Struck accidentally by falling object E917 Striking against or struck accidentally by objects or persons E918 Caught accidentally in or between objects E919 Accidents caused by machinery Accidents caused by cutting and piercing instruments or objects (accidental injury by object or fall on object: edged, pointed, E920 sharp) E921 Accident caused by explosion of pressure vessel E922 Accident caused by firearm missile E923 Accident caused by explosive material Accident caused by hot substance or object, caustic or corrosive E924 material, and steam E925 Accident caused by electric current E926 Exposure to radiation E927 Overexertion and strenuous movements E928 Other and unspecified environmental and accidental causes E929.1.E929.4, E929.5, E929.8, E929.9 Late effects of accidental injury Appendix B. Glossary 246 Suicide and self-inflicted injuries Includes: injuries in suicide and attempted suicide, self-inflicted injuries specified as intentional. ICIW-CM E-Codcs Description E95.0 Suicide and self-inflicted poisoning by solid or liquid substances E951 Suicide and self-inflicted poisoning by gases in domestic use E952 Suicide and self-inflicted poisoning by other gases and vapors Suicide and self-inflicted injury by hanging, strangulation, and E953 suffocation E954 Suicide and self-inflicted injury by submersion [drowning] E955 Suicide and self-inflicted injury by firearms and explosives E956 Suicide and self-inflicted injury by cutting and piercing instrument E957 Suicide and self-inflicted injuries by jumping from high place E958 Suicide and self-inflicted injury by other and unspecified means E959 Late effects of self-inflicted injury Other intentional injury Homicide and injury purposely inflicted by other persons Includes: injuries inflicted by another person with intent to injure or kil l , by any means. Excludes: legal intervention (E970-E978), operations of war (E990-E999). 1CD9-CM E-Codes Description E960 Fight, brawl, rape E961 Assault by corrosive or caustic substance, except poisoning E962 Assault by poisoning E963 Assault by hanging and strangulation E964 Assault by submersion [drowning] E965 Assault by firearms and explosives E966 Assault by cutting and piercing instrument E967 Child battering and other maltreatment E968 Assault by other and unspecified means E969 Late effects of injury purposely inflicted by other person Appendix B. Glossary 247 Legal intervention Excludes: injuries caused by civil insurrections (E990.0-E999). IC D9-CM E-C odes Description E970 Injury due to legal intervention by firearms E971 Injury due to legal intervention by explosives E972 Injury due to legal intervention by gas E973 Injury due to legal intervention by blunt object E974 Injury due to legal intervention by cutting and piercing instrument E975 Injury due to legal intervention by other specified means E976 Injury due to legal intervention by unspecified means E977 Late effects of injuries due to legal intervention E978 Legal execution Injury resulting from operations of war Excludes: accidents during training of military personnel, manufacture of war material and transport, unless attributable to enemy action. 1CD9-CM E-C odes Description E990 Injury due to war operations by fires and conflagrations E991 Injury due to war operations by bullets and fragments E992 Injury due to war operations by explosion of marine weapons E993 Injury due to war operations by other explosion E994 Injury due to war operations by destruction of aircraft Injury due to war operations by other and unspecified forms of E995 conventional warfare E996 Injury due to war operations by nuclear weapons Injury due to war operations by other forms of unconventional E997 warfare Injury due to war operations but occurring after cessation of E998 hostilities E999 Late effect of injury due to war operations 248 Appendix C Important ICD9-CM E-Codes Road traffic accidents For road traffic accidents, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: •(E812.0, E812.1) Agent: Motor vehicle traffic accident involving collision with another motor vehicle parked, stopped, stalled, disabled, or abandoned on the highway. Host: The driver of the motor vehicle. • (E814.7) Agent: Collision between motor vehicle, of any kind, and pedestrian. The pedestrian was dragged, hit, or run over by the motor vehicle. Host: Pedestrian. •(E816.0,E816.1,E816.9) Agent: Failure to make a curve; going out of control due to blow out; burst tire; overturning; driver falling asleep; driver inattention; excessive speed; and failure of mechanical part and as a consequence colliding with an object off the highway or overturning or stopping abruptly off the highway. Host: The driver or a passenger or unspecified person of a motor vehicle other than a motorcycle. Appendix C. Important ICD9-CM E-Codes 249 •(E819.0, E819.1.E819.9) Agent: Motor vehicle traffic accident of unspecified nature. Host: A driver or passenger or an unspecified person of a motor vehicle other than a motorcycle. P o i s o n i n g For poisoning, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: •(E850.0) Agent: Accidental poisoning by analgesics, antipyretics, and antirheumatics -specifically heroin /diacetylmorphine. •(E854.1,E854.8,E855.2) Agent: Accidental poisoning by other drugs acting on central and autonomic nervous system - specifically local anesthetics such as cocaine, lidocaine [lignocaine], procaine and tetracaine. • (E858.8, E858.9) Agent: Accidental poisoning by other drugs: central appetite depressants and other unspecified drugs. •(E860.1,E860.2,E860.9) Agent: Accidental poisoning by alcohol (other and unspecified ethyl alcohol and its products; methyl alcohol; unspecified alcohol). Appendix C. Important ICD9-CME-Codes 250 • (E868.2) Agent: Accidental poisoning by motor vehicle exhaust gas. Fall For falls, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: • (E887) Agent: Fracture of cause unspecified. • (E888) Agent: Other and unspecified fall. • (E880.9) Agent: Fall on or from stairs or steps other than escalator and sidewalk curb. • (E884.9) Agent: Other fall from one level to another such as: embankment, haystack, stationary vehicle and tree. Medical adverse events For medical A E , all ages combined, the Y L D contributed more to the D A L Y than the Y L L . The important ICD9-CM E-codes contributing to Y L D are: • (E878, E879) Agent: surgical and medical procedures as the cause of abnormal reaction of patient or later complication, without mention of misadventure at the time of procedure. Appendix C. Important ICD9-CME-Codes 251 Other transport injuries For other transport injuries, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: • (E805.2) Agent: Hit by rolling stock Host: Pedestrian • (E822.7) Agent: Other motor vehicle non-traffic accident involving collision with moving object Host: Pedestrian • (E830.0, E830.9) Agent: Accident to watercraft causing submersion Host: The occupant of a small boat (unpowered) or an unspecified person • (E832.0) Agent: Other accidental submersion or drowning in a water transport accident Host: The occupant of a small boat (unpowered) •(E841.3.E841.9) Agent: Accident to powered aircraft (unspecified) Host: Passenger on powered aircraft (not specified anywhere else) or other person Appendix C. Important ICD9-CM E-Codes 252 Burns/fires/scalds For burns/fires/scalds, all ages combined, the Y L D contributed more to the D A L Y than the Y L L . The important ICD9-CM E-codes contributing to Y L D are: • (E890.3) Agent: Burning caused by conflagration • (E894.0) Agent: Ignition of highly inflammable material • (E924.0) Agent: Hot liquids and vapours including steam. Drowning For drowning, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: • (E910.1, E910.2, E910.8, E910.9) Agent: Accidental drowning and submersion while: water-skiing; engaged in other sport Or recreational activity without diving equipment; in a quenching tank or swimming pool. Or an accidental drowning or submersion of an unspecified nature. Appendix C. Important ICD9-CM E-Codes 253 Sports injuries For sports injuries, all ages combined, the Y L D contributed more to the D A L Y than the Y L L . The important ICD9-CM E-codes contributing to Y L D are: • (E886.0) Agent: Fall oh same level from colliding, or pushing, or shoving, by or with other person in sports • (E917.0) Agent: Striking against or struck accidentally by objects or persons in sports {e.g., kicked or stepped on during game (football) (rugby), knocked down while boxing, struck by hit or thrown ball, or struck by hockey stick or puck) • (E927.0) Agent: Overexertion and strenuous movements from lifting, pulling, pushing, recreational activities and other activities. Natural and environmental factors For natural and environmental factors, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: •(E901.0, E901.9) Agent: Excessive cold due to weather conditions or of unspecified origin Appendix C. Important ICD9-CME-Codes 254 • (E904.3) Agent: Exposure to weather conditions (not elsewhere classifiable) combined with hunger, thirst, exposure, and neglect. • (E906.8) Agent: other specified injury caused by animal (e.g., butted by an animal, fallen on by horse or other animal (not being ridden), gored by animal, run over or stepped on by an animal not being ridden) • (E909) Agent: Cataclysmic earth surface movements and eruptions Machinery injuries For machinery injuries, all ages combined, the Y L D contributed more to the D A L Y than the Y L L . The important ICD9-CM E-codes contributing to Y L D are: •(E919.2,E919.4,E919.8) Agent: Machinery, mainly: lifting machines and appliances; woodworking and forming machines; and other machines for manufacture of clothing, foodstuffs and beverages, paper; printing machines; recreational machinery and spinning, weaving and textile machines. •(E920.1.E920.4) Agent: Powered lawn mower or other hand tools and implements Appendix C. Important ICD9-CM E-Codes 255 Suffocation and foreign bodies For suffocation and foreign bodies, all ages combined, the Y L D contributed more to the D A L Y than the Y L L . The important ICD9-CM E-codes contributing to Y L D are: •(E911.0) Agent: Injury caused by inhalation and ingestion of food causing obstruction of respiratory tract or suffocation. • (E912.0) Agent: Inhalation and ingestion of an object causing obstruction of respiratory tract or suffocation • (E914.0) Agent: Foreign body accidentally entering eye and adnexa. •(E915.0) Agent: Foreign body accidentally entering other orifice than respiratory tract, eye or adnexa. Appendix C. Important ICD9-CM E-Codes 256 Other unintentional injuries excluding medical AE For other unintentional injuries, all ages combined, the Y L D contributed more to the D A L Y than the Y L L . The important ICD9-CM E-codes contributing to Y L D are: • (E916.0) Agent: Struck accidentally by falling object such as: rock, snowslide (nowhere else specified), stone, tree, object falling from machine not in operation or from stationary vehicle. • (E917.9) Agent: Struck accidentally by an air rifle. •(E918.0) Agent: Caught accidentally in or between objects. • (E920.3, E920.8) Agent: Cutting and piercing instruments or objects mainly: knives, swords and daggers, and other cutting and piercing objects (e.g., broken glass, nail, tin can lid, splinter, etc.) • (E929.8) Agent: Late effect of other accidents (excluding all the external causes mentioned above) Appendix C. Important ICD9-CME-Codes 257 Suicide and self-inflicted injury For suicide and self-inflicted injury, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are: • (E950.0, E950.3, E950.4, E950.5) Agent: Suicide and self-inflicted poisoning by solid or liquid substances such as: analgesics, antipyretics, antirheumatics, tranquilizers, other psychotropic agents, and other specified or unspecified drugs and medicinal substances. •(E952.0, E952.1) Agent: Suicide and self-inflicted poisoning by other gases and vapours such as motor vehicle exhaust gas and other carbon monoxide. • (E953.0, E953.1, E953.8, E953.9) Agent: Suicide and self-inflicted injury by hanging, strangulation, suffocation and other unspecified means. • (E955.4) Agent: Suicide and self-inflicted injury by firearms other than handgun, shotgun, hunting rifle and military firearms. •(E957.0, E957.1) Agent: Suicide and self-inflicted injury by jumping from residential premises and other man-made structures. Appendix C. Important ICD9-CM E-Codes 258 Other intentional injury For other intentional injury, all ages combined, the Y L L contributed more to the D A L Y than the Y L D . The important ICD9-CM E-codes contributing to Y L L are mainly due to homicide and violence and they are: • (E960.0) Agent: Homicide and injury inflicted by unarmed fight or brawl. • (E963) Agent: Assault or homicidal attempt by hanging and strangulation such as garrotting or ligature, hanging, strangulation and suffocation. • (E965.4) Agent: Assault by firearms other than handgun, shotgun, hunting rifle and military firearms. It also includes some cases of unspecified firearm. • (E966) Agent: Assault or assassination attempt and homicidal attempt by cutting, puncturing and stabbing or by a piercing instrument classifiable under E920 (please see Glossary). • (E968.8, E968.9) Agent: Assault or assassination attempt and homicidal attempt by other specified or unspecified means 259 Appendix D Example of Disability Weight Spreadsheet Assuming Zero Disability Weight for the Residual Category of Injury Sequela In this section we demonstrate the methodology used to calculate the overall disability weight by gender and age group for each of the external causes of injury in this study. As an example we decided to show the calculations for the disability weights for males that were hospitalized due to a road traffic accident. Note that the tables corresponding to each of the steps that we describe next are named at the top of each page as Table from Step .... In practice all the steps described in this example were implemented as a macro in SAS version 8 and obtained automatically. Step 1. We selected all males that suffered a road traffic accident and were hospitalized during 1991-2000, using the ICD9-CM E-codes E810-E819, E826-E829, E929.0; Step 2. We classified these cases of hospitalization according to the nature of injuries described in Table 3.1 (page 49). The natures of injury described in page 50 (notes (1) to (6)) were proportionately redistributed according to the recommendations in this same page; Step 3. In this step we redistributed the ICD9-N-codes 958 - 959, 990 - 999 (coded in table as N40-N41) proportionately across all categories that have disability weights (coded in table as N1-N33) as described in the recommendation in page 50. These ICD9-Ncodes represent natures of injury that have no disability weight assigned to. So at this time we calculated the Appendix D. Example of Disability Weight Spreadsheet Assuming Zero Disability 260 Weight for the Residual Category of Injury Sequela percentage of cases each of the 33 categories N1-N33 would receive from the categories N40-N41; Step 4. We calculated how many cases each of the 33 categories N1-N33 would receive from the categories N40-N41 based on Step 3. In SAS we randomly selected these cases that would be assigned to each of these 33 categories of nature of injury; Step 5. Now we added the original number of cases of hospitalization to the number of cases obtained in Step 4 to obtain the new cases of hospitalization for the categories N1-N33. ; In SAS, we made a table with all the disability weights obtained from the global burden of disease (GBD) study specific to each age group (Murray and Lopez 1996). We also have these weights displayed in Table 3.2 (page 68). Note that some injuries can have long-term and short-term disability weights. We assigned the disability weights according to what is described in column "GBD assumptions" in Table 3.2: (a) We adjusted the weight for individuals with more than one nature of injury (or sequela) using the formula DWU...K=H^-DWI)*(1-DW2)*... *(l-DWK), where the subscript K represents the number of injury sequela for each individual; (b) Then, for each age group, we calculated the overall DW (Table 3.3) as a weighted average of the adjusted (to other injury sequelae) disability weights for each individual: Individuals^ *DWadJi +... + IndividualsN *DWadj N Individuals^ +... + Individuals N where DWadj,N represents the corresponding DW adjusted for multiple sequelae in the same individual, Individuals^ represents the number of individuals with DW equal to F>Wadj, N and N is the total number of individuals with that particular DWadj,N. Injury Nature ICD9-CM N-Codes Code Original number of hospitalizations 0-19 20-39 40-64 65-79 80+ Fractured skull 800 - 801 NI- ' 505 624 293 87- 31 Fractured face bones 802 N2 660 1161 512 117 • 28 Fractured vertebral column 805 - N3 535 1284- 669 164 45 Injured spinal cord 806 and 952 - N 4 . . 121 370 193 ' 64 • 17.-Fractured rib or sternum 807 • N5 • ' 277 1218 • 1292 ' • 535 • 222 Fractured pelvis • 808 N6 r 336-, 712: 466 • . - 178 64 . Fractured clavicle, scapula or humerus - • 810-812 - N7- ; : 780: 1151 749 • 192 68. Fractured radius or ulna • •• 813 • N8- 1-343 . 1105 574- ' - 1-32 29 Fractured hand bones 814-817 •• N 9 - ' 322 , 839 389. . 52 14. Fractured femur 820-821 N10 • 638. 854 . 452 - 175 I l l Fractured patella, tibia or fibula 822-823 •- N i l " . - ' 691 . 1291 861 274 ' 82-Fractured ankle 824 ' N12- , 363' 835 466 - 98 " 30 Fractured foot bones 825 - 826 -• N13- 148 485 258 41- 8 Other dislocation 830,833 - 834, 836 -839 N14 122 448 244 35 5 Dislocated shoulder, elbow or hip 831,832, 835 N15 - 115 404 213 37 9 Sprains 840 - 848 N16 399 1221 610 125 41 Intracranial injury 850 - 854 N17 : 2617 . 2864 1265 430 116 Internal injuries 860 - 869 • N18 1177 1892 1125 391 117 Open wound 870, 872 - 884, 890 - 894 N19 2170 3676 1575 454 128 Injury to eyes 871,950 N20 28 47 32 15 5 Amputated thumb 885 N21 ; 0 3 5- 0 0 Amputated finger 886 N-22 8 13 12 0 0 Amputated arm 887 N23 0 6 5 0 0 Amputated toe 895 N24 0 6 1 0 0 Amputated foot 896, 897.0-897.1 N25 6 4 7 1 0 Amputated leg 897.2 - 897.3 N26 ' 0 2 0 0 0 Crushing 925 - 929 N27' 25 54 30 9 0 • Bums<20% 940-947,948.0-948.1 N28 17 48 28 5 0 Bums > 20% and < 60% 948.2 - 948.5 . N29 0 1 1 0 0 Bums > 60% 948.6-948.9 N30 ~ 0 1 0 0 0 Injured nerves 951,953 -957 N31 135 302 103 20 5 Poisoning 960-979,980. -989 ' - N32 ' 4 15 - 6 3 0 Residual 900 - 924, 930 - 939 N33" •. 2424 4488 2395 684 247 Injuries in Note (7) (A.7) 958-959, 990-999 • N40 581 1330 805 255 90 Injuries in Note (8) (A.8) 829,849,855-859,888-889,898-899,897.8-897.7 "'' '•• N41 ' 0 1 1 ' 1 0 Total (A.7) + (A.8) Total without (A.7) and (A.8) 16,547 28,755 15,637 • 4,574 1,512 581 1,331 806 256 . 90 15,966. 27,424 14,831 4,318 1,422 a" o 3 on a> to TO^ 5^ TO 3 >*. TO .to Co-¥ Ox; O Si-TO Co SS s TO TO Co s» TO TO Co Co S S • TO • >!• o Co SS to 0\ Injury Nature ICD9-CM N-Codes Code Percentage without categories N40 and N41. % 0 -19 % 20 - 39 % 40 - 64 % 65 - 79 % 80+ Fractured skull 800-801 N l 3.163 2.275 1.976 2.015 2.180 Fractured face bones • 802 N2 4.134 • 4.234 • 3.452 2.710 1.969 Fractured vertebral column 805 N3 . 3.351 4.682 4.511 3.798 3.165 Injured spinal cord 806 and 952 ' N4 • 0.758 1.349 1.301 1.482 1.195-Fractured rib or sternum 807 • • N5 - • 1.735- 4.441 ' 8.711- • 12.390 15.612 Fractured'pelvis 808 N6 2.104 2.596 3.142 4.122 . ' 4.501 Fractured clavicle, scapula or humerus 810-812 • .N7 4.885 4.197 : -5.050 - 4.447 4.782 Fractured radius or ulna 813 N8 8.412 4.029 - 3.870 3.057 • 2.039 Fractured hand bones 814-817 N9 . 2.017 3.059 2.623 1.204 0.985 Fractured femur 820-821 • N10 3.996 3.114 3.048 - 4.053 . 7.806 Fractured patella, tibia or fibula 822 - 823 N i l 4.328 4.708 5.805 6.346 5.767 Fractured ankle 824 N12 2.274' 3.045 3.142 . 2.270 2.110 Fractured foot bones 825 - 826 N13 . 0.927 1.769 1.740 0.950 0.563 Other dislocation 830, 833 - 834, 836 - 839 N14 • 0.764 1.634 1.645 0.811 0.352 Dislocated shoulder, elbow or hip 831,832,835 N15 0.720 1.473 1.436 0.857 0.633 Sprains 840 - 848 N16 2.499 4.452 4.113 2.895 2.883 Intracranial injury 850 - 854. N17 16.391 10.443 8.529 9.958 8.158 Internal injuries 860 - 869 N18 7.372 6.899 7.585 9.055 8.228 Open wound 870, 872 - 884, 890 - 894 N19 13.591 13.404 10.620 10.514 9.001 Injury to eyes 871,950 N20 _ 0.175 0.171 0.216 0.347 0.352 Amputated thumb 885 N21 0.000 0.011 0.034 0.000 0.000 Amputated finger 886 N22 0.050 0.047 0.081 0.000 0.000 Amputated arm 887 N23 0.000 0.022 0.034 0.000 0.000 Amputated toe 895 N24 0.000 0.022 0.007 0.000 0.000 Amputated foot 896, 897.0-897.1 N25 0.038 0.015 0.047 0.023 0.000 Amputated leg 897.2 -897.3 N26 0.000 0.007 0.000 ' 0.000 0.000 Crushing 925 - 929 N27 0.157 0.197 0.202 0.208 0.000 Bums < 20% 940-947,948.0-948.1 N28 0.106 0.175 0.189 0.116 0.000 Bums > 20% and < 60% 948.2 - 948.5 N29 0.000 0.004 0.007 0.000 0.000 Bums > 60% 948.6 - 948.9 N30 0.000 0.004 0.000 0.000 0.000 Injured nerves " • 951,953 -957 N31 0.846 1.101 0.694 0.463 0.352 Poisoning 960-979 ,980-989 N32 0.025 0.055 0.040 0.069 0.000 Residual 900 - 924, 930 - 939 N33 - 15.182 • 16.365 16.149 - 15.841 17.370 Injuries in Note (7) (A.7) 958-959,990-999 N40 - . - - - -Injuries in Note (8) (A.8) 829,849,855-859,888-889,898-899,897.8-897.7 N41 - - - - - -H. i : O' jo, . § : ' . r. 3',-Total 100 100 100 100 100 to ov to Injury Nature ICD9-CM N-Codes Code > Number cases of hospitalizations borrowed from N40 and N41 0-19 20-39 40 - 64 65-79 80+ Fractured skull' 800 - 801 NI 18.4 30:3 15.9 5.2 2.0 Fractured face bones 802 N2 24.0- 56.3 27.8 6.9 1.8 Fractured vertebral column 805 . • N3 19.5 62.3 . 36.4 9.7 2.8-Injured spinal cord 806 and 952 - • N4 • 4.4 18.0- 10.5 3.8 1.1 Fractured rib or sternum- 807 N5 10.1- • 59.1 70.2 31.7 - 14.1 -Fractured pelvis 808 N6 12.2 : 34.6 25.3 10.6 4.1 Fractured clavicle, scapula or humerus 810-812 • N7 28.4- - 55.9 40.7 11.4 .. 4.3 Fractured radius or ulna 813 • N8 48.9 53.6 31.2 .7.8- 1.8 Fractured hand bones 814-817 N9 11.7 40.7 • ' 21.1 3.1 • 0.9 Fractured femur 820-821 N10 23.2 41.4- • 24.6 10.4 7.0 Fractured patella, tibia or fibula 822 - 823 N i l 25.1 62.7 46.8 16.2 5.2 Fractured ankle' . 8 2 4 N12- . 13.2 • • 40.5 • 25.3 5.8 ' 1.9 Fractured foot bones 825 - 826 N13 5.4 23.5- 14.0 2.4 0.5 Other dislocation 830, 833 - 834, 836 - 839 N14 4.4 21.7 13.3 2.1 0.3 Dislocated shoulder, elbow or hip 831, 832, 835 N15 4.2 19.6 11.6 2.2 0.6 Sprains 840 - 848 N16 14.5 59.3 33.2 7.4 2.6 Intracranial injury 850 - 854 N17 95.2 139.0 68.7 25.5 7.3 Internal injuries 860 - 869 N18 42.8 91.8 61.1 23.2 7.4 Open wound 870,872-884, 890-894 N19 79.0 178.4 85.6 26.9 8.1 Injury to eyes 871,950 N20 1.0 2.3 1.7 0.9 0.3 Amputated thumb 885 N21 0.0 0.1 0.3 0.0 0.0 Amputated finger 886 N22 0.3 0.6 0.7 0.0 0.0 Amputated arm 887 N23 0.0 0.3 0.3 0.0 0.0 Amputated toe 895 N24 0.0 0.3 0.1 0.0 0.0 Amputated foot 896,897.0-897.1 N25 0.2 0.2 0.4 0.1 0.0 Amputated leg 897.2 - 897.3 N26 0.0 0.1 0.0 0.0 0.0 Crushing 925 - 929 N27 0.9 2.6 1.6 0.5 0.0 Bums < 20% 940-947,948.0-948.1 N28 0.6 2.3 1.5 0.3 0.0 Bums > 20% and < 60% 948.2 - 948.5 N29 0.0 0.0 0.1 0.0 0.0 Bums > 60% 948.6 - 948.9 N30 0.0 0.0 0.0 0.0 0.0 Injured nerves • 951,953 -957 N31 4.9 14.7 5.6 1.2 0.3 Poisoning 960 - 979, 980 - 989 N32 0.1 0.7 0.3 0.2 0.0 Residual 900 - 924, 930 - 939 N33 - 88.2 ' 217.8 130.2 40.6 15.6 Injuries in Note (7) (A.7) 958-959, 990-999 N40 - • - - -Injuries in Note (8) (A.8) 829,849,855-859,888-889,898-899,897.8-897.7 N41 - - -O Lo H-o OT3 i—K T 3 -Oq. TO. s- s - S3 TO. S? » TO- q. al ^ «" TO* Co OX) TO^ s TO TO a Co 3-TO TO Co Co £ 3 Ox? -- TO O Co a Total 581 1,331 806 256 90 to Injury, Nature Code New hospitalization total 0-19 20-39 40 - 64 65-79 80+ Fractured skull - short and long terms (already adjusted); - N l - • 523 654 309 • 92 33 • Fractured face bones - N2 684 1217 540 124 30 Fractured vertebral column - short term- N3 , 554 1346 705 • 174 48 Injured spinal cord - life long ' N4 125 388 203 ''• 68 . 18 Fractured rib or sternum - short term . N5 287 1277 1362 - 567 236 Fractured pelvis - short term N6 , 348 747 • 49r- 189 68 . Fractured clavicle, scapula or humerus - short term N7 ' 808 1207 790 . 203 72 Fractured radius or ulna - short term N8 1392 . 1159- 605 140 31-Fractured hand bones - N9 . 334- 880- 410" 55 15 Fractured femur - short and long terms (already adjusted)- • N10' -661 895 477 185 118. Fractured patella, tibia or fibula - short term • N i l - ••- 716 1354 908 290 87 Fractured ankle - short term, --• ... N12 376 876 491' 104" 32 . Fractured foot bones N13 153 509' - 272' 43 9 Other dislocation N14 126 470 257 37 - • 5 Dislocated shoulder, elbow or hip - short term N15~ 119 424 225 39 10 Sprains N16 414 1280 643 132 44 Intracranial injury - short and long terms (already adjusted) N17- - 2712 3003 1334 455 123 Internal injuries - short term N18 1220 1984 1186 414 124 Open wound N19 2249 3854 1661 481 136 Injury to eyes - life long N20 29 49 34 16 5 Amputated thumb - life long N21 0 3 5 0 0 Amputated finger - life long N22 8 14 13 0 0 Amputated arm - life long N23 0 6 5 0 0 Amputated toe - life long N24 0 6 1 0 0 Amputated foot - life long N25 6 4 7 1 0 Amputated leg - life long N26 0 2 0 0 0 Crushing - short term N27 26 57 32 10 0 Bums < 20% - l ifelong N28 18 50 30 5 0 Bums > 20% and < 60% - life long N29 0 1 1 0 0-Bums > 60% - life long N30 0 1 0 0 0 Injured nerves - life long N31 140 317 109 21 5 Poisoning - short term • N32 4 16 6 3 0 Residual N33 , 2512 4706 2525 725 263 ciT O, ' H" •o. OO" ' r-t-; 3^- " ^ " ^ Z b-a-TO Co ~ 5* TO" 5. .a 3 Co S3" ^ If' 3^ TO se' <§'" TO £^3 K TO >! TO S3 Co a-TO TO Co Co S a OQ " .TO Total Total without residual 16,547 28,755 15,637 • 4,574 1,512 14,035 24,049 13,112 3,849 "1,249 Co ON 4^ 265 Appendix E Example of Duration Spreadsheet Assuming Zero Duration for the Residual Category of Injury Sequela In this section we demonstrate the methodology used to calculate the overall durations by gender and age group for each of the external causes of injury in this study. As an example we decided to show the calculations for the durations for males that were hospitalized due to a Road Traffic Accident. In practice all the steps described in this example were implemented as a macro in SAS version 8 and obtained automatically. Steps 1 to 5. Same as described in Appendix D (Example of disability weight spreadsheet assuming zero disability weight for the residual category of injury sequela); In SAS, we made a table with all the durations obtained from the global burden of disease (GBD) study specific to each age group (Murray and Lopez 1996). We also have these durations displayed iii Table 3.5 (page 76). Note that some injuries can have long-term and others can have short-term durations. We distributed the cases of injury according to what is described in column "GBD assumptions" in Table 3.5. Note that the long-term durations are based on the B C life expectancy as described in section 3.3.1.5 (page78). (a) We adjusted the duration of individuals with more than one nature of injury (or sequela) by calculating an average between different durations; Appendix E. Example of Duration Spreadsheet Assuming Zero Duration for the 266 Residual Category of Injury Sequela (b) Then, for each age group, we calculated the overall D (Table 3.7, page 79) as a weighted average of the adjusted (to other injury sequelae) duration of disability of each individual: Individuals^ * Dadj, +... + IndividualsN * DadjN . Individuals^ +... + Individuals N where D^N represents the corresponding D adjusted for multiple sequelae in the same individual, Individuals^ represents the number of individuals with D equal to Dadj,N and is the total number of individuals with that particular Dadj,N.. 267 Appendix F Sensitivity Analyses In this appendix, we examine the robustness of D A L Y estimates for the following six categories of external cause of injury: road traffic accidents; poisoning; falls; other unintentional injuries (other transport injuries, burns/fires/scalds, drowning, sports injuries, natural and environmental factors, machinery injuries, suffocation and foreign bodies); suicide and self-inflicted injuries; and other intentional injuries (Figure 4.1, page 90). The robustness of estimated disability-adjusted life years (DALY) was examined using two sensitivity analyses. The first sensitivity analysis was conducted to investigate the sensitivity of D A L Y estimates to different values of the discounting future health rate. The relevance of performing this sensitivity analysis was to emphasize the importance of understanding the impact of different discount rates on the calculation of D A L Y for injuries in BC. The discount rates examined in this sensitivity analysis were 0%, 3%, 6% and 10%. The literature recommends the use of a 3% discounting rate, however it is important to stress that there is no convincing theoretical argument for this choice. Based on the results of the sensitivity analysis conducted in this study, we decided to use a 3% discount rate in all D A L Y calculations presented this study. The second sensitivity analysis was conducted to investigate the robustness of D A L Y to the impact of different disability weights used for the residual category of injury sequela in the D A L Y calculation. The importance of performing this sensitivity analysis resides in the fact that there are no such weights developed for BC, and future policy actions based on the results of this thesis could accommodate different scenarios (using different weights), Appendix F. Sensitivity Analyses 268 depending on what the demonstrated effects of weights are. We demonstrated that using different weights for the residual category of sequela did not affect the conclusions drawn from this study. 1. Sensitivity Analysis for the Discounting of Future Health Rate To study the robustness of D A L Y to different discounting rates (0%, 3%, 6% and 10%), we compared: (1) the ranking of D A L Y s (for all external causes of injury combined) for each age group; (2) the ranking of D A L Y s (for all age groups combined) for each external cause of injury (road traffic accidents, poisoning, falls, other unintentional injuries (excluding medical AE) , suicide and self-inflicted injuries and other intentional injuries); and (3) the logarithm of the Y L D : Y L L ratio for each external cause of injury and age group. Note that all sensitivity analyses were performed separately for males and females. Throughout this section we assumed fixed zero weight and duration for the residual category of injury. The ten-year D A L Y (for all external causes of injury) for each age group and the ratio between D A L Y s for different values of discounting are presented in Figure F. 1 and Table F . l . From this table and figure, we can see that the ten-year D A L Y by each age group, for both males and females, are very different depending on the magnitude of the discount rate used. Appendix F. Sensitivity Analyses Table F . l . Ten-Year D A L Y by Discounting Rate, Gender and Age Group. Description D A L Y (1991-2000) A l l Injuries llll 11 I 1111111 20-39 4-i -o4 65 - 79 S I B H P I I 80+ Discount Rate (0%) 206,483 407,803 164,103 23,058 8,109 Discount Rate (3%) 126,119 255,416 125,069 20,262 7,622 Discount Rate (6%) 91,948 186,230 99,948 17,988 7,179 Discount Rate (10%) 68,669 140,441 78,660 15,579 6,649 3% vs 0% (ratio) 0.611 0.626 0.762 0.879 0.940 6% vs 3% (ratio) 0.729 0.729 0.799 0.888 0.942 10% vs 6% (ratio) 0.747 0.754 0.787 0.866 0.926 D A L Y (1991 -2000)~A11 Injuries SillE ! ! a : * f i i i | ^ Females Description' p-.'. :.:.lL:|::.:i::!H!::|!|!!=:!lHH!|!|!|!|!|!in!|!Hl^HU:|iyK^  I s ^ ^ i i i i i h i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i p i i i i n ^ l P 0-19 i 20 - 39 40 - 64 : " :r :" '•' ' i ' l ^ ^ p l l l | | 65 - 79 ilillllpi iii^^^^S 80 i Discount Rate (0%) 99.232 117,988 58,993 18,470 13.65S Discount Rate (3%) 59,310 75,620 44,241 16,060 12,802 Discount Rate (6%) 43,324 57,018 35,278 14,163 12,031 Discount Rate (10%) 32,880 44,968 28,054 12,218 11,119 3% vs 0% (ratio) 0.598 0.641 0.750 0.870 0.937 6% vs 3% (ratio) 0.730 0.754 0.797 0.882 0.940 10% vs 6% (ratio) 0.759 0.789 0.795 0.863 0.924 For example, the ratio using a discount rate equal to 3% vs 0% was calculated using the formula DALY[0,3] DALY[0,0] Appendix F. Sensitivity Analyses 270 DALY (1991-2000) - All injuries Males 0 -19 20-39 40-64 65-79 80+ Discount Rate (0%) Discount Rate (3%) Discount Rate (6%) Discount Rate (10%) DALY (1991-2000) - All injuries Females 20,000 Discount Rate (0%) Discount Rate (3%) Discount Rate (6%) Discount Rate (10%) 0 -19 20 - 39 40 - 64 65 - 79 80+ Figure F . l . Ten-Year D A L Y By Discounting Rate, Gender and Age Group. Appendix F. Sensitivity Analyses 271 f DALY[0,rt] DALY[0,r2] Note that i f the ratio between D A L Y s , , is approximately one, it means that the D A L Y s for the two different discount rates are very similar in magnitude; the further this ratio is from one, the more different the D A L Y s are. Depending on the age group, the ratio between D A L Y s for two different discount ratios ranged from 0.61 (ratio between DALY[0,3] and DALY[0,0] for males aged 0-19) to 0.94 (ratio between DALY[0,6] and DALY[0,3] for females aged 80+). In Table F . l , for both males and females, we observe that for younger age groups (those individuals whose age is less than 40 years, with the age group 0-19 having the largest difference) there is a large impact on D A L Y when the discount rate is increased from 0% to 3% (smaller ratios). For the later age groups (age > 40 years), increasing the discount rate does not have a large impact on the value of D A L Y . However, as age increases within this older group, there is a small impact on the value of D A L Y when the discount rate is increased from 6% to 10%. For a quick visual summary (by age group) of the D A L Y ratios for successive discount rates see Figure F.2. Appendix F. Sensitivity Analyses 272 — • - 3 % vs 0% - Males - » - 6 % vs 3% - Males ~sr-10% vs 6% - Males - x — 3 % vs 0% - Females —A—6% vs 3% - Females - e - 1 0 % vs 6% - Females 0 - 19 2 0 - 3 9 4 0 - 6 4 6 5 - 7 9 80+ Figure F.2. Percentage of Change in Ten-Year D A L Y for Different Discounting Rates by Gender and Age Group. In summary, we observed that use of a low discount rate increases the impact of the ten-year burden from injury in the younger age groups and as the discount rate increases, the impact of the ten-year burden from injury shifts to the older age groups. Table F.2 shows that the ranking of D A L Y for each external cause of injury shifts depending on the magnitude of the discount rate. Appendix F. Sensitivity Analyses 273 Table F.2. Ranking of Ten-Year D A L Y by Discount Rate, Gender and External Cause of Injury. (Note: Changing ranks highlighted in colour) D A L Y (19c)l-2(»iit)i Description Males Discount Rate (0%) Ranking Discount Rate (3%) Ranking Discount Rate (6%) ||||^ »^|||||f Rankinu Disc.mill Rate (10%) Ranking Road traffic accidents 161,198 2 97,980 2 72,396 2 56,577 2 Poisoning 110,292 4 63,611 4 42,387 4 28,679 4 Falls 42,640 f. 33,923 " 5 29,552 5 26,200 5 Other unintentional injuries1'' 301,168 1 219,946 1 173,306 1 134,575 1 Suicide and self-inflicted injuries 146,784 3 88,221 3 62,007 3 44,917 3 Other intentional injuries 47,475 • V;5 30,806 6 23,646 6 19,048 6 Description — D A L Y (1991-2000) iJlilBlB^ Discount Rate (0%) Ranking Discount Rate (3%) Ranking Discount Rate «h"..i Ranking Discount Rate (10%) Ranking Road traffic accidents 72,637 2 45,280 2 34,639 2 28,032 2 Poisoning 34,746 4 19,668 5 13,104 5 8,980 Falls 26,926 5 23,371 4 21,246 4 19,328 4 Other unintentional injuries* 97,520 1 70,360 1 54,983 1 42,437 1 Suicide and self-inflicted injuries 51,369 3 33,852 3 26,135 3 21,099 3 Other intentional injuries 25,145 6 15,503 6 11,708 f l i l l 9,363 5 Excludes medical adverse events Appendix F. Sensitivity Analyses 274 There was variation in the ranking of D A L Y s when we increased the discount rate from 0% to 10%, especially for females. For males, the external causes of injury that were robust {i.e., did not change in ranking) to changing values of discounting were road traffic accidents, poisoning, other unintentional injuries (excluding medical AE) and suicide and self-inflicted injuries which were the external causes of injury with the largest D A L Y . For females, the robust external causes of injury were the same as males with the exception of poisoning. The ten-year D A L Y (for all ages combined) for each external cause of injury and the ratio between D A L Y s for different values of discounting are presented in Figure F.3 and Table F.3. This table and figure indicate that the ten-year D A L Y by each age group, for both males and females, is very different depending on the magnitude of the discount rate. Appendix F. Sensitivity Analyses 275 Table F.3. Ten-Year D A L Y by Discount Rate, Gender and External Cause of Injury. Description1 DA I.Y (1991-2000) A l l Injuries Males Road Ira flic accidents Poisoning .V Falls Other unintentional injuries* Suicide and self-inflicted injuries Other intentional injuries Discount Rate (0%) 161,198 110,292 42,640 301,168 146,784 47,475 Discount Rate (3%) 97,980 63,611 33,923 219,946 88,221 30,806 Discount Rate (6%) 72,396 42,387 29,552 173,306 62,007 23,646 Discount Rate (10%) 56,577 28,679 26,200 134,575 44,917 19,048 3% vs 0% (ratio) 0.608 0.577 0.796 0.730 0.601 0.649 6% vs 3% (ratio) 0.739 0.666 0.871 0.788 0.703 0.768 10% vs 6% (ratio) 0.782 0.677 0.887 0.777 0.724 0.806 Description' - DAI.Y (1991-2000) A l l Injuries Road ^ traffic ' Poisoning accidents 1 alU Other unintentional injuries* Suicide and self-inflicted injuries (Hhci intentioiul injuries Discount Rate (0%) 72,637 34,746 26,926 97,520 51,369 25,145 Discount Rate (3%) 45,280 19,668 23,371 70,360 33,852 15,503 Discount Rate (6%) 34,639 13,104 21,246 54,983 26,135 11,708 Discount Rate (10%) 28,032 8,980 19,328 42,437 21,099 9,363 3% vs 0% (ratio) 0.623 0.566 0.868 0.721 0.659 0.617 6%vs3% (ratio) 0.765 0.666 0.909 0.781 0.772 0.755 10% vs 6% (ratio) 0.809 0.685 0.910 0.772 0.807 0.800 Excludes medical adverse events Appendix F. Sensitivity Analyses 276 Depending on the external cause of injury, the ratio between D A L Y s for two different discount ratios ranged from 0.57 (ratio between DALY[0,3] and DALY[0,0] for males -poisoning) to 0.91 (ratio between DALY[0,10] and DALY[0,6] for females - falls). For both males and females, the largest impact on the value of D A L Y s was associated with increasing the discount rate from 0% to 3%. For both males and females, the smallest change in the value of D A L Y associated with increasing the value of discounting, was for falls (i.e., the ratios were closer to one). As a visual summary of the D A L Y ratios for each external cause of injury by increasing amount of discounting can be found in Figure F.3. 0.95 -0.90 -0.85 -•o 0.80 -E 0.75 -> _i 0.70 -< Q 0.65 -0.60 -0.55 -0.50 -Road Traffic Poisoning Falls Other Suicide and Other Accidents unintentional Self-inflicted Intentional injuries^ Injuries Injuries 3% vs 0% - Males x 3% vs 0% - Females -6% vs 3% - Males - * — 6 % vs 3% - Females 10% vs 6% - Males • 10% vs 6% - Females ' 'Excludes medical adverse events Figure F.3. Ten-Year D A L Y Ratio for Different Discount Rates by Gender and External Cause of Injury. Appendix F. Sensitivity Analyses 277 We also examined the influence of different discount rates on the natural logarithm of ( Y L D ^ the ratio ( Y L D : Y L L ) [i.e., In ]. This ratio is similar to the approach that the WHO v Y L L J Global Burden of Disease study adopted (Murray and Lopez 1996). In this study we applied the logarithm of this ratio to make the comparisons between Y L D and Y L L graphically clear. If this ratio is approximately zero, it means that Y L D and Y L L are very similar in magnitude; i f this ratio is greater than zero it means that Y L D > Y L L ; conversely, i f the ratio is negative, Y L D < Y L L . For each external cause of injury (all age groups combined), we observed that as the discounting rate increased, the proportion of the burden of injuries attributed to Y L D also increased (Figure F.4). 2.00 >-ti o E ro D) O Males Road Traffic Poisoning Accidents "Excludes medical adverse events Falls Other Suicide and Other unintentional Self-inflicted Intentional injuries*. Injuries Injuries Figure F.4. Logarithm of the Ratio Y L D : Y L L by Discount Rate, Gender and External Cause of Injury. Appendix F. Sensitivity Analyses 278 Females 2.00 >-ti _ i >-o E ro O) o -2.00 Road Traffic Poisoning Accidents Falls Other Suicide and Other unintentionalSelf-lnflicted Intentional injuries^ Injuries Injuries * Excludes medical adverse events ->-0% -m-3% __™6%" Figure F.4. (continued) Logarithm of the Ratio Y L D : Y L L by Discount Rate, Gender and External Cause of Injury. Next, we calculated the Y L D : Y L L ratio for all external causes of injury combined for each age group separately. We observed that as the discount rate increased, the proportion of the burden of injuries due to Y L D also increased. For the younger age groups, the ratio varied across different values of discounting. However, as age increased, the ratios became more similar for the different values of discount rate used (Figure F.5). Appendix F. Sensitivity Analyses 279 0.80 0.60 _l _l 0.40 >-d _i 0.20 >-M— o 0.00 E < -0.20 nj o -0.40 -0.60 -0.80 Males 0 -19 20-39 40-64 65-79 80+ -•—0% -a-3% A 6% -K-10% >-Q _i >-. O E co o 0.80 0.60 0.40 0.20 0.00 -0.20 -0.40 -0.60 -0.80 -1.00 Females 0-19 20-39 40-64 65-79 80+ -0% -3% •is 6% -x-10% Figure F.5. Logarithm of the Ratio Y L D : Y L L by Discount Rate, Gender and Age Group. Last, we observed that the values of the ratio In ^ Y L D ^ V Y L L y (for all external causes of injury and age combined) were constantly parallel between males and females, regardless of the value of discounting (Figure F.6). Appendix F. Sensitivity Analyses 280 Appendix F. Sensitivity Analyses 281 2. Sensitivity Analysis for Disability Weights for the Residual Category of Injury Sequela The set of disability weights for the residual category of injury that we examined in this sensitivity analysis were {0, 0.25, 0.50}. Throughout the analysis we assumed a 3% discounting rate and for simplicity we refer to the disability weight for the residual category of injury as DWresidual. In Table F.4 we show the overall disability weight for each external cause of injury by age group and gender. Table F.4. Influence of Different DWresidual on the Disability Weights for each External Cause of Injury, by Gender and Age Group. Zero weight for residuals Description Disability weight by age group - Males 0-19 20 - 39 40 - 64 65 - 79 80-Road traffic accidents 0.262 0.247 0.244 0.260 0 264 Poisoning 0.625 0.642 0.636 0.620 0.624 Falls 0.224 0.203 0.223 0.270 0.300 Other unintentional injuries* 0.187 0.165 0.183 0.257 0.311 Suicide and self-inflicted injuries 0.546 0.556 0.598 0.559 0.492 Other intentional injuries 0.249 0.258 0.303 0.334 0.387 1 »escription Disability weight by age firoup - Females 0 - 19 20 - 39 40 - 64 65 - 79 80+ Road traffic accidents 0.262 0.242 0.241 0.247 0.252 Poisoning 0.631 0.652 0.644 0.625 0.615 Falls 0.220 0.197 0.220 0.272 0.308 Other unintentional injuries* 0.191 0.192 0.225 0.304 0.336 Suicide and self-inflicted injuries 0.624 0.624 0.648 0.620 0.554 Other intentional injuries 0.401 0.365 0.414 0.419 0.420 ^Excludes medical adverse events Appendix F. Sensitivity Analyses 282 Table F.4. (continued) Influence of Different DWresidual on the Disability Weights for each External Cause of Injury, by Gender and Age Group. 0.25 weight for residuals Description Disability .weight by age group - Males 0 - ]9 20 - 39 40 - 64 65 - 79 80-Road traffic accidents 0.318 0.305 0.301 0.316 0.327 Poisoning 0.629 0.647 0.638 0.623 0.625 Falls 0.237 0.228 0.251 0.305 0.342 Other unintentional injuries* 0.250 0.210 0.241 0.338 0.396 Suicide and self-inflicted injuries 0.557 0.569 0.610 0.574 0.499 Other intentional injuries 0.289 0.301 0.357 0.413 0.458 Description Disability weight by-age group - Females 0 - 19 20 - 39 40 - 64 65 - 79 80-Road traffic accidents 0.324 0.314 0.308 0.312 0.317 Poisoning 0.631 0.653 0.647 0.626 0.617 Falls 0.236 0.224 0.244 0.302 0.344 Other unintentional injuries* 0.293 0.247 0.287 0.369 0.399 Suicide and self-inflicted injuries 0.630 0.630 0.652 0.626 0.563 Other intentional injuries 0.475 0.452 0.497 0.492 0.479 % Excludes medical adverse events 0.50 weight for residuals Description Disability weight b\ age group - Males 0 - 19 20 - 39 40 - 64 65-79 80+ Road traffic accidents 0.365 0.356 0.351 0.365 0.383 Poisoning 0.632 0.652 0.641 0.626 0.626 Falls 0.250 0.251 0.277 0.338 0.381 Other unintentional injuries* 0.310 0.253 0.297 0.417 0.479 Suicide and self-inflicted injuries 0.567 0.581 0.621 0.589 0.505 Other intentional injuries 0.325 0.338 0.405 0.485 0.523 Description Disability weight by age group - Females II - Vi 20 - 39 4"i-n4 65 - 79 80+ Road traffic accidents 0.376 0.376 0.366 0.367 0.372 Poisoning 0.632 0.654 0.651 0.628 0.618 Falls 0.250 0.250 0.266 0.329 0.376 Other unintentional injuries* 0.393 0.301 0.348 0.433 0.460 Suicide and self-inflicted injuries 0.635 0.636 0.656 0.631 0.571 Other intentional injuries 0.536 0.525 0.565 0.552 0.531 Excludes medical adverse events Appendix F. Sensitivity Analyses 283 Tables F.5-F.7 document the variability of the disability weights for each external cause of injury, by gender and age group by using different values of DWresidual. In summary, it appears that the greatest amount of variation associated with increased DWresidual is observed when we compare the final disability weights across external causes of injury. Comparatively less variation in the final disability weights was observed across age groups. Table F.5. Variation in Disability Weight by Age Group, by Gender and External Cause of Injury. Description Variation in disability weight by age group - Males 0- 19 20 - 39 40 - 64 65 - 79 Xii Road traffic accidents 0.052 0.055 0.054 0.053 0.059 Poisoning 0.003 0.005 0.003 0.003 0.001 Falls 0.013 0.024 0.027 0.034 0.041 Other unintentional injuries* 0.061 0.044 0.057 0.080 0.084 Suicide and self-inflicted injuries 0.011 0.012 0.011 0.015 0.007 Other intentional injuries 0.038 0.040 0.051 0.076 0.068 Description Variation in disability weight by age group - 1 emales 0 - 19 20 - 39 40 - 64 65 - 79 80-Road traffic accidents 0.057 0.067 0.062 0.060 0.060 Poisoning 0.000 0.001 0.003 0.002 .0.001 Falls 0.015 0.027 0.023 -0.028 0.034 Other unintentional injuries* 0.101 0.055 0.061 0.064 0.062 Suicide and self-inflicted injuries 0.006 0.006 0.004 0.005 0.009 Other intentional injuries 0.068 0.080 0.076 0.067 0.056 Excludes medical adverse events Appendix F. Sensitivity Analyses 284 Table F.6. Variation in Disability Weight among Age Groups, by Gender and External Cause of Injury. Description \ .1 -iation in disability weight among age groups - Males Zero weight 0.25 weight 0.50 weight Road traffic accidents 0.009 0.010 0.012 Poisoning 0.009 0.010 0.011 Falls 0.040 0.049 0.058 Other unintentional injuries''' 0.061 0.077 0.094 Suicide and self-inflicted injuries 0.038 0.041 0.043 Other intentional injuries 0.057 0.072 0.088 Description Variation in disability weight among age groups - Females Zero weight 0.25 weight 0.50 weight Road traffic accidents 0.009 0.006 U.UU5 Poisoning 0.015 0.015 0.015 Falls 0.046 0.051 0.056 Other unintentional injuries1 0.067 0.063 0.064 Sui