Disease mapping and risk assessment for public health

Editors’ Preface xiii List of Contributors xv PART I: IIISEASE MAPPING 1 Disease Mapping and Its Uses 3 1.1 Introduction 3 1.2 Simple statistical representation 4 1.3 Model-based approaches 8 1.4 Spatio-temporal modelling 11 1.5 Conclusions 13 2 Bayesian and Empirical Bayes Approaches to Disease Mapping 15 2.1 Introduction 15 2.2 Maximum likelihood estimation of relative risks of mortality 16 2.3 Hierarchical Bayesian model of relative risks 19 2.4 Empirical Bayes estimation of relative risks 24 2.5 Fully Bayesian estimation of the relative risks 25 2.6 Conclusion 29 3 Addressing Multiple Goals Evaluating Region-Specific Risk Using Bayesian Methods 31 3.1 Introduction 31 3.2 Models 32 3.3 Goals and inferences 35 3.4 Using Monte-Carlo output 38 3.5 Scottish lip cancer data analysis 38 3.6 Conclusion 45 4 Disease Mapping with Hidden Structures Using Mixture Models 49 4.1 Introduction 49 4.2 The empirical Bayes approach 51 4.3 The validity of the mixture model approach for map construction 55 4.4 Extensions of the mixture model approach 56 4.5 Discussion and conclusions 58 4.6 Appendix: Details about the program DismapWin 59 PART 11: CLUSTERING OF DISEASE 5 Inference for Extremes in Disease Mapping 63 5.1 Introduction 63 5.2 Spatial models for disease incidence or mortality 64 5.3 Bayesian inference via simulation 67 viii Contents 5.4 Bayes and constrained Bayes estimates 75 5.5 Loss functions for extreme values 78 5.6 Results for the Scotland lip-cancer data 80 6 Edge Effects in Disease Mapping 85 6.1 Introduction 85 6.2 Edge effect problems 86 6.3 Edge effect compensation methods 87 6.4 A hierarchical Bayesian model for disease mapping of tract count data 89 6.5 TheTuscany example 92 6.6 Conclusions 97 7 A Review of Cluster Detection Methods 99 7.1 Introduction 99 7.2 Reasons for studying disease clustering 102 7.3 Definition of clusters and clustering 104 7.4 Modelling issues 107 7.5 Hypothesis tests for clustering 108 8 Comparison of General Tests for Spatial Clustering 111 8.1 Introduction 111 8.2 Inappropriate tests 111 8.3 Available tests 113 8.4 Discussion 116 9 Markov Chain Monte Carlo Methods for Putative Sources of Hazard and General Clustering 119 9.1 Introduction 119 9.2 Definitions 120 9.3 The analysis of health risk related to pollution sources 121 9.4 The analysis of non-focused disease clustering 122 9.5 A general model formulation for specific clustering 123 9.6 Markov chain Monte Carlo methods 124 9.7 Putative hazard example 125 9.8 Non-focused clustering example 132 9.9 Conclusions 141 10 Statistical Evaluation of Disease Cluster Alarms 143 10.1 Introduction 143 10.2 Focused cluster tests applied at other similar locations 144 10.3 Post-alarm monitoring 145 10.4 The spatial scan statistic 146 10.5 A proactive approach 147 10.6 Discussion 148 11 Disease Clustering for Uncertain Locations 151 11.1 Introduction 151 11.2 Classical statistics and randomisation tests 152 11.3 Cluster statistics as randomisation tests 153 11.4 Sample-based randomisation tests are epidemiologically unreasonable 154 11.5 Statistical inference 155 11.6 Location models 157 11.7 Location model applications 161 11.8 Spatial randomisation 162 Contents ix 11.9 Statistical inference for uncertain locations 11.10 An application 11.11 Discussion 12 Empirical Studies of Cluster Detection-Different Cluster Tests in Application to German Cancer Maps 12.1 Introduction 12.2 Methods 12.3 Results of application to German cancer mortality data 12.4 Discussion PART 111: ECOLOGICAL ANALYSIS 13 Introduction to Spatial Models in Ecological Analysis 13.1 Introduction 13.2 Ecological fallacy in spatial data 13.3 Statistical models 13.4 Example 13.5 Conclusions 1 4 Bayesian Ecological Modelling 14.1 Introduction 14.2 Statistical issues 14.3 Data issues 14.4 Problems with the interpretation of ecological regression studies 14.5 Technical implementation 14.6 Conclusions 15 Spatial Regression Models in Epidemiological Studies 15.1 Introduction 15.2 Truncated auto-Poisson vs. random effects Poisson regression 15.3 Model fitting using Monte Carlo Newton-Raphson 15.4 Prostate cancer invalencia, 1975 -1980 15.5 Discussion 16 Multilevel Modelling of Area-Based Health Data 16.1 Introduction 16.2 Developing a Poisson spatial multilevel model 16.3 Incidence of prostate cancer in Scottish local authority districts 16.4 Discussion PART IV: RISK ASSESSMENT FOR PUTATIVE SOURCES OF HAZARD 17 A Review of Modelling Approaches in Health Risk Assessment around Putative Sources 17.1 Introduction 17.2 Problems of inference 17.3 Exploratory techniques 17.4 Models for point data 17.5 Models for count data 17.6 Modelling vs. hypothesis testing 17.7 Conclusions 164 164 165 169 169 170 172 176 181 181 182 185 191 191 193 193 194 196 199 200 200 203 203 205 208 210 214 217 217 2 19 22 3 227 231 231 233 2 34 2 35 242 244 24 5 x Contents 18 Disease Mapping Using the Relative Risk Function Estimated from Area1 Data 247 18.1 Introduction 247 18.2 Definition of the relative risk function 248 18.3 Estimation of the relative risk function 249 18.4 AppIication to childhood cancer data 249 18.5 General heterogeneity of risk 252 18.6 Selecting the degree of smoothing 254 18.7 Discussion 254 19 The Power of Focused Score Tests Under Misspecified Cluster Models 257 19.1 Introduction 257 19.2 Models of clustering and score tests 258 19.3 Homogeneous population results 261 19.4 Post hoc power analysis: NewYork leukaemia data 266 19.5 Discussion 268 20 Case-Control Analysis of Risk around Putative Sources 271 20.1 Introduction 271 20.2 General definitions 2 72 20.3 Estimation 2 74 20.4 Crude trend tests on distance from source 278 20.5 Stratified analysis 282 20.6 Logistic regression analysis 282 20.7 Conclusions 285 21 Lung Cancer Near Point Emission Sources 287 21.1 Introduction 287 21.2 Industrial areas 288 21.3 Urban areas 289 21.4 Methodological implications 290 21.5 Conclusions 292 PART V: PUBLIC HEALTH APPLICATIONS AND CASE STUDIES 22 Environmental Epidemiology, Public Health Advocacy and Policy 295 22.1 Introduction 295 22.2 ‘Washing Whiter’: A pitfall to avoid 296 22.3 A societal context to take into account 298 22.4 Conclusion 299 23 The Character and the Public Health Implications of Ecological Analyses 301 23.1 Introduction 301 23.2 Some remarks on the assessment of ecological exposures 302 23.3 Comment on study designs 303 23.4 Credibility of ecological analyses 304 23.5 Scientific aspects 305 23.6 Societal attitudes 307 23.7 Conclusions 308 24 Computer Geographic Analysis: A Commentary on Its Use and Misuse in Public Health 311 24.1 Introduction 311 24.2 Hypothesis generation 312 Contents xi 24.3 Ecological studies 316 24.4 Descriptive/administrative uses of GIS and geographic analysis 317 24.5 Hypothesis testing 318 24.6 Conclusion 319 25 Estimating the Presence and the Degree of Heterogeneity of Disease Rates 321 25.1 Introduction 321 25.2 The Poisson case 322 25.3 Binomial case 324 25.4 Discussion 326 26 Ecological Regression with Errors in Covariates: An Application 329 26.1 Introduction 329 26.2 Background and data 331 26.3 The statistical model 333 26.4 The data 340 26.5 Estimation 340 26.6 Results 340 26.7 Discussion 346 27 Case Studies in Bayesian Disease Mapping for Health and Health Service Research in Ireland 349 27.1 Introduction 349 27.2 Background 350 27.3 Low birth weight and area deprivation 352 27.4 Avoidable mortality for asthma 358 27.5 Conclusion 363 28 An Analysis of Determinants of Regional Variation in Cancer Incidence: Ontario, Canada 365 28.1 Introduction 365 28.2 Background and objectives 366 28.3 Methods 367 28.4 Results 369 28.5 Conclusions 380 29 Congenital Anomalies Near Hazardous Waste Landfill Sites in Europe 383 29.1 Introduction 383 29.2 Methods 384 29.3 Results 388 29.4 Discussion 391 30 An Analysis of the Geographical Distribution of Leukaemia Incidence in the Vicinity of a Suspected Point Source: A Case Study 395 30.1 Introduction 395 30.2 Materials and methods 396 30.3 Results 401 30.4 Discussion 406 30.5 Conclusion 408 xii Contents 31 Lung Cancer Mortality in Women in Germany 1995:A Case Study in Disease Mapping 411 31.1 Introduction 411 31.2 Thedata 412 31.3 The methods 413 Appendix: Disease Mapping and Risk Assessment for Public Health Decision Making: Report on a WHOIBiomed2 International Workshop 453 Index 469