Harold C. Sox is Emeritus Professor of Medicine and of the Dartmouth Institute at Geisel School of Medicine at Dartmouth, USA. Michael C. Higgins is Adjunct Professor at the Stanford Center for Biomedical Informatics Research, Stanford University, USA. Douglas K. Owens is a general internist and Professor and Chair of the Department of Health Policy, School of Medicine, and Director of Stanford Health Policy, Freeman-Spogli Institute for International Studies, Stanford University, USA. Gillian Sanders Schmidler is Professor of Population Health Sciences and Medicine at Duke University and Deputy Director of the Duke-Margolis Institute for Health Policy, Durham, USA.

• Foreword xiPreface xiii1 Introduction 11.1 How may I be thorough yet efficient when considering the possible causes of my patient’s problems? 11.2 How do I characterize the information I have gathered during the medical interview and physical examination? 11.3 How do I interpret new diagnostic information? 31.4 How do I select the appropriate diagnostic test? 41.5 How do I choose among several risky treatment alternatives? 42 Differential diagnosis 52.1 An introduction 52.2 How clinicians make a diagnosis 52.3 The principles of hypothesis- driven differential diagnosis 82.4 An extended example 14Bibliography 163 Probability: quantifying uncertainty 183.1 Uncertainty and probability in medicine 183.2 How to determine a probability 213.3 Sources of error in using personal experience to estimate the probability 233.4 The role of empirical evidence in quantifying uncertainty 303.5 Limitations of published studies of disease prevalence 353.6 Taking the special characteristics of the patient into account when determining probabilities 36Bibliography 374 Interpreting new information: Bayes’ theorem 384.1 Introduction 384.2 Conditional probability defined 404.3 Bayes’ theorem 414.4 The odds ratio form of Bayes’ theorem 454.5 Lessons to be learned from using Bayes’ theorem 504.6 The assumptions of Bayes’ theorem 524.7 Using Bayes’ theorem to interpret a sequence of tests 544.8 Using Bayes’ theorem when many diseases are under consideration 55Bibliography 575 Measuring the accuracy of clinical findings 585.1 A language for describing test results 585.2 The measurement of diagnostic test performance 625.3 How to measure diagnostic test performance: a hypothetical example 675.4 Pitfalls of predictive value 695.5 How to perform a high quality study of diagnostic test performance 705.6 Spectrum bias in the measurement of test performance 745.7 When to be concerned about inaccurate measures of test performance 795.8 Test results as a continuous variable: the ROC curve 815.9 Combining data from studies of test performance: the systematic review and meta- analysis 87A.5.1 Appendix: derivation of the method for using an ROC curve to choose the definition of an abnormal test result 89Bibliography 916 Decision trees – representing the structure of a decision problem 936.1 Introduction 936.2 Key concepts and terminology 936.3 Constructing the decision tree for a hypothetical decision problem 966.4 Constructing the decision tree for a medical decision problem 103Epilogue 112Bibliography 1127 Decision tree analysis 1137.1 Introduction 1137.2 Folding- back operation 1147.3 Sensitivity analysis 126Epilogue 133Bibliography 1338 Outcome utility – representing risk attitudes 1348.1 Introduction 1348.2 What are risk attitudes? 1358.3 Demonstration of risk attitudes in a medical context 1368.4 General observations about outcome utilities 1478.5 Determining outcome utilities – underlying concepts 151Epilogue 157Bibliography 1589 Outcome utilities – clinical applications 1599.1 Introduction 1599.2 A parametric model for outcome utilities 1609.3 Incorporating risk attitudes into clinical policies 1729.4 Helping patients communicate their preferences 181Epilogue 185A.9.1 Exponential utility model parameter nomogram 186Bibliography 18810 Outcome utilities – adjusting for the quality of life 18910.1 Introduction 18910.2 Example – why the quality of life matters 19010.3 Quality- lifetime tradeoff models 19310.4 Quality- survival tradeoff models 20310.5 What does it all mean? – an extended example 209Epilogue 217Bibliography 21711 Survival models: representing uncertainty about the length of life 21811.1 Introduction 21811.2 Survival model basics 21911.3 Medical example – survival after breast cancer recurrence 22611.4 Exponential survival model 22811.5 Actuarial survival models 23211.6 Two- part survival models 235Epilogue 247Bibliography 24712 Markov models 24812.1 Introduction 24812.2 Markov model basics 24912.3 Determining transition probabilities 25912.4 Markov model analysis – an overview 269Epilogue 277Bibliography 27713 Selection and interpretation of diagnostic tests 27813.1 Introduction 27813.2 Four principles of decision making 27913.3 The threshold probability for treatment 28113.4 Threshold probabilities for testing 28813.5 Clinical application of the threshold model of decision making 29313.6 Accounting for the non- diagnostic effects of undergoing a test 29613.7 Sensitivity analysis 29813.8 Decision curve analysis 300Bibliography 30214 Medical decision analysis in practice: advanced methods 30314.1 An overview of advanced modeling techniques 30314.2 Use of medical decision- making concepts to analyze a policy problem: the cost- effectiveness of screening for HIV 30514.3 Use of medical decision- making concepts to analyze a clinical diagnostic problem: strategies to diagnose tumors in the lung 31314.4 Calibration and validation of decision models 31714.5 Use of complex models for individual- patient decision making 319Bibliography 32115 Cost- effectiveness analysis 32315.1 The clinician’s conflicting roles: patient advocate member of society and entrepreneur 32315.2 Cost- effectiveness analysis: a method for comparing management strategies 32515.3 Cost–benefit analysis: a method for measuring the net benefit of medical services 33015.4 Methodological best practices for cost- effectiveness analysis 33215.5 Reference case for cost- effectiveness analysis 33315.6 Impact inventory for cataloguing consequences 33415.7 Measuring the health effects of medical care 33415.8 Measuring the costs of medical care 33515.9 Interpretation of cost- effectiveness analysis and use in decision making 33715.10 Limitations of cost- effectiveness analyses 337Bibliography 338Index 340

Harold C. Sox is Emeritus Professor of Medicine and of the Dartmouth Institute at Geisel School of Medicine at Dartmouth, USA. Michael C. Higgins is Adjunct Professor at the Stanford Center for Biomedical Informatics Research, Stanford University, USA. Douglas K. Owens is a general internist and Professor and Chair of the Department of Health Policy, School of Medicine, and Director of Stanford Health Policy, Freeman-Spogli Institute for International Studies, Stanford University, USA. Gillian Sanders Schmidler is Professor of Population Health Sciences and Medicine at Duke University and Deputy Director of the Duke-Margolis Institute for Health Policy, Durham, USA.

• Foreword xiPreface xiii1 Introduction 11.1 How may I be thorough yet efficient when considering the possible causes of my patient’s problems? 11.2 How do I characterize the information I have gathered during the medical interview and physical examination? 11.3 How do I interpret new diagnostic information? 31.4 How do I select the appropriate diagnostic test? 41.5 How do I choose among several risky treatment alternatives? 42 Differential diagnosis 52.1 An introduction 52.2 How clinicians make a diagnosis 52.3 The principles of hypothesis- driven differential diagnosis 82.4 An extended example 14Bibliography 163 Probability: quantifying uncertainty 183.1 Uncertainty and probability in medicine 183.2 How to determine a probability 213.3 Sources of error in using personal experience to estimate the probability 233.4 The role of empirical evidence in quantifying uncertainty 303.5 Limitations of published studies of disease prevalence 353.6 Taking the special characteristics of the patient into account when determining probabilities 36Bibliography 374 Interpreting new information: Bayes’ theorem 384.1 Introduction 384.2 Conditional probability defined 404.3 Bayes’ theorem 414.4 The odds ratio form of Bayes’ theorem 454.5 Lessons to be learned from using Bayes’ theorem 504.6 The assumptions of Bayes’ theorem 524.7 Using Bayes’ theorem to interpret a sequence of tests 544.8 Using Bayes’ theorem when many diseases are under consideration 55Bibliography 575 Measuring the accuracy of clinical findings 585.1 A language for describing test results 585.2 The measurement of diagnostic test performance 625.3 How to measure diagnostic test performance: a hypothetical example 675.4 Pitfalls of predictive value 695.5 How to perform a high quality study of diagnostic test performance 705.6 Spectrum bias in the measurement of test performance 745.7 When to be concerned about inaccurate measures of test performance 795.8 Test results as a continuous variable: the ROC curve 815.9 Combining data from studies of test performance: the systematic review and meta- analysis 87A.5.1 Appendix: derivation of the method for using an ROC curve to choose the definition of an abnormal test result 89Bibliography 916 Decision trees – representing the structure of a decision problem 936.1 Introduction 936.2 Key concepts and terminology 936.3 Constructing the decision tree for a hypothetical decision problem 966.4 Constructing the decision tree for a medical decision problem 103Epilogue 112Bibliography 1127 Decision tree analysis 1137.1 Introduction 1137.2 Folding- back operation 1147.3 Sensitivity analysis 126Epilogue 133Bibliography 1338 Outcome utility – representing risk attitudes 1348.1 Introduction 1348.2 What are risk attitudes? 1358.3 Demonstration of risk attitudes in a medical context 1368.4 General observations about outcome utilities 1478.5 Determining outcome utilities – underlying concepts 151Epilogue 157Bibliography 1589 Outcome utilities – clinical applications 1599.1 Introduction 1599.2 A parametric model for outcome utilities 1609.3 Incorporating risk attitudes into clinical policies 1729.4 Helping patients communicate their preferences 181Epilogue 185A.9.1 Exponential utility model parameter nomogram 186Bibliography 18810 Outcome utilities – adjusting for the quality of life 18910.1 Introduction 18910.2 Example – why the quality of life matters 19010.3 Quality- lifetime tradeoff models 19310.4 Quality- survival tradeoff models 20310.5 What does it all mean? – an extended example 209Epilogue 217Bibliography 21711 Survival models: representing uncertainty about the length of life 21811.1 Introduction 21811.2 Survival model basics 21911.3 Medical example – survival after breast cancer recurrence 22611.4 Exponential survival model 22811.5 Actuarial survival models 23211.6 Two- part survival models 235Epilogue 247Bibliography 24712 Markov models 24812.1 Introduction 24812.2 Markov model basics 24912.3 Determining transition probabilities 25912.4 Markov model analysis – an overview 269Epilogue 277Bibliography 27713 Selection and interpretation of diagnostic tests 27813.1 Introduction 27813.2 Four principles of decision making 27913.3 The threshold probability for treatment 28113.4 Threshold probabilities for testing 28813.5 Clinical application of the threshold model of decision making 29313.6 Accounting for the non- diagnostic effects of undergoing a test 29613.7 Sensitivity analysis 29813.8 Decision curve analysis 300Bibliography 30214 Medical decision analysis in practice: advanced methods 30314.1 An overview of advanced modeling techniques 30314.2 Use of medical decision- making concepts to analyze a policy problem: the cost- effectiveness of screening for HIV 30514.3 Use of medical decision- making concepts to analyze a clinical diagnostic problem: strategies to diagnose tumors in the lung 31314.4 Calibration and validation of decision models 31714.5 Use of complex models for individual- patient decision making 319Bibliography 32115 Cost- effectiveness analysis 32315.1 The clinician’s conflicting roles: patient advocate member of society and entrepreneur 32315.2 Cost- effectiveness analysis: a method for comparing management strategies 32515.3 Cost–benefit analysis: a method for measuring the net benefit of medical services 33015.4 Methodological best practices for cost- effectiveness analysis 33215.5 Reference case for cost- effectiveness analysis 33315.6 Impact inventory for cataloguing consequences 33415.7 Measuring the health effects of medical care 33415.8 Measuring the costs of medical care 33515.9 Interpretation of cost- effectiveness analysis and use in decision making 33715.10 Limitations of cost- effectiveness analyses 337Bibliography 338Index 340

Harold C. Sox is Emeritus Professor of Medicine and of the Dartmouth Institute at Geisel School of Medicine at Dartmouth, USA. Michael C. Higgins is Adjunct Professor at the Stanford Center for Biomedical Informatics Research, Stanford University, USA. Douglas K. Owens is a general internist and Professor and Chair of the Department of Health Policy, School of Medicine, and Director of Stanford Health Policy, Freeman-Spogli Institute for International Studies, Stanford University, USA. Gillian Sanders Schmidler is Professor of Population Health Sciences and Medicine at Duke University and Deputy Director of the Duke-Margolis Institute for Health Policy, Durham, USA.

• Foreword xiPreface xiii1 Introduction 11.1 How may I be thorough yet efficient when considering the possible causes of my patient’s problems? 11.2 How do I characterize the information I have gathered during the medical interview and physical examination? 11.3 How do I interpret new diagnostic information? 31.4 How do I select the appropriate diagnostic test? 41.5 How do I choose among several risky treatment alternatives? 42 Differential diagnosis 52.1 An introduction 52.2 How clinicians make a diagnosis 52.3 The principles of hypothesis- driven differential diagnosis 82.4 An extended example 14Bibliography 163 Probability: quantifying uncertainty 183.1 Uncertainty and probability in medicine 183.2 How to determine a probability 213.3 Sources of error in using personal experience to estimate the probability 233.4 The role of empirical evidence in quantifying uncertainty 303.5 Limitations of published studies of disease prevalence 353.6 Taking the special characteristics of the patient into account when determining probabilities 36Bibliography 374 Interpreting new information: Bayes’ theorem 384.1 Introduction 384.2 Conditional probability defined 404.3 Bayes’ theorem 414.4 The odds ratio form of Bayes’ theorem 454.5 Lessons to be learned from using Bayes’ theorem 504.6 The assumptions of Bayes’ theorem 524.7 Using Bayes’ theorem to interpret a sequence of tests 544.8 Using Bayes’ theorem when many diseases are under consideration 55Bibliography 575 Measuring the accuracy of clinical findings 585.1 A language for describing test results 585.2 The measurement of diagnostic test performance 625.3 How to measure diagnostic test performance: a hypothetical example 675.4 Pitfalls of predictive value 695.5 How to perform a high quality study of diagnostic test performance 705.6 Spectrum bias in the measurement of test performance 745.7 When to be concerned about inaccurate measures of test performance 795.8 Test results as a continuous variable: the ROC curve 815.9 Combining data from studies of test performance: the systematic review and meta- analysis 87A.5.1 Appendix: derivation of the method for using an ROC curve to choose the definition of an abnormal test result 89Bibliography 916 Decision trees – representing the structure of a decision problem 936.1 Introduction 936.2 Key concepts and terminology 936.3 Constructing the decision tree for a hypothetical decision problem 966.4 Constructing the decision tree for a medical decision problem 103Epilogue 112Bibliography 1127 Decision tree analysis 1137.1 Introduction 1137.2 Folding- back operation 1147.3 Sensitivity analysis 126Epilogue 133Bibliography 1338 Outcome utility – representing risk attitudes 1348.1 Introduction 1348.2 What are risk attitudes? 1358.3 Demonstration of risk attitudes in a medical context 1368.4 General observations about outcome utilities 1478.5 Determining outcome utilities – underlying concepts 151Epilogue 157Bibliography 1589 Outcome utilities – clinical applications 1599.1 Introduction 1599.2 A parametric model for outcome utilities 1609.3 Incorporating risk attitudes into clinical policies 1729.4 Helping patients communicate their preferences 181Epilogue 185A.9.1 Exponential utility model parameter nomogram 186Bibliography 18810 Outcome utilities – adjusting for the quality of life 18910.1 Introduction 18910.2 Example – why the quality of life matters 19010.3 Quality- lifetime tradeoff models 19310.4 Quality- survival tradeoff models 20310.5 What does it all mean? – an extended example 209Epilogue 217Bibliography 21711 Survival models: representing uncertainty about the length of life 21811.1 Introduction 21811.2 Survival model basics 21911.3 Medical example – survival after breast cancer recurrence 22611.4 Exponential survival model 22811.5 Actuarial survival models 23211.6 Two- part survival models 235Epilogue 247Bibliography 24712 Markov models 24812.1 Introduction 24812.2 Markov model basics 24912.3 Determining transition probabilities 25912.4 Markov model analysis – an overview 269Epilogue 277Bibliography 27713 Selection and interpretation of diagnostic tests 27813.1 Introduction 27813.2 Four principles of decision making 27913.3 The threshold probability for treatment 28113.4 Threshold probabilities for testing 28813.5 Clinical application of the threshold model of decision making 29313.6 Accounting for the non- diagnostic effects of undergoing a test 29613.7 Sensitivity analysis 29813.8 Decision curve analysis 300Bibliography 30214 Medical decision analysis in practice: advanced methods 30314.1 An overview of advanced modeling techniques 30314.2 Use of medical decision- making concepts to analyze a policy problem: the cost- effectiveness of screening for HIV 30514.3 Use of medical decision- making concepts to analyze a clinical diagnostic problem: strategies to diagnose tumors in the lung 31314.4 Calibration and validation of decision models 31714.5 Use of complex models for individual- patient decision making 319Bibliography 32115 Cost- effectiveness analysis 32315.1 The clinician’s conflicting roles: patient advocate member of society and entrepreneur 32315.2 Cost- effectiveness analysis: a method for comparing management strategies 32515.3 Cost–benefit analysis: a method for measuring the net benefit of medical services 33015.4 Methodological best practices for cost- effectiveness analysis 33215.5 Reference case for cost- effectiveness analysis 33315.6 Impact inventory for cataloguing consequences 33415.7 Measuring the health effects of medical care 33415.8 Measuring the costs of medical care 33515.9 Interpretation of cost- effectiveness analysis and use in decision making 33715.10 Limitations of cost- effectiveness analyses 337Bibliography 338Index 340