Interpreting Epidemiologic Evidence

David A. Savitz, PhD, is Professor of Epidemiology and Obstetrics and Gynecology at Brown University. He has held leadership positions at the University of North Carolina (Chair of the Department of Epidemiology) and at Brown University (Vice President for Research), as well as in a number of professional organizations. His research is focused primarily on reproductive and environmental epidemiology. He was elected to the National Academy of Medicine in 2007.Gregory A. Wellenius, ScD, is Associate Professor of Epidemiology at Brown University. His research focuses on the effects of our environment on cardiovascular health, with an emphasis on the adverse health effects of air pollution and the built environment. He has taught advanced graduate courses in epidemiologic methods, provided invited expert testimony before the US House of Representatives and US Senate, and mentored a number of undergraduate, medical, masters, and doctoral students.

• Interpreting Epidemiologic Evidence: Connecting Research to Applications1. IntroductionSynopsisLearning ObjectivesPerspectiveApproach to the Evaluation of EvidenceOrganization of Book 2. The Nature of Epidemiologic Evidence SynopsisLearning Objectives Goals of Epidemiologic Research Measurement of Causal Relations Between Exposure and Disease Applications of Epidemiologic Research Framework for Examining Epidemiologic Evidence Relationship of Epidemiology to Health Policy Exercise: Critical Assessment of Study Methods, Results, and Applications 3. Causal Diagrams for Epidemiologic InferenceSynopsisLearning ObjectivesIntroductionCausal Diagrams in Epidemiology Purpose and TerminologyDAGs Encode Our AssumptionsStatistical AssociationsConnection to Data AnalysesDepicting Passage of TimeDirect vs. Indirect EffectsConcluding ThoughtsRecommended Additional ReadingsExercise: Application of Causal Diagrams for Epidemiologic Inference4. Strategy for Drawing Inferences from Epidemiologic Evidence SynopsisLearning Objectives Conceptual Framework for the Evaluation of Error Estimation of Measures of Association Systematic Evaluation of Sources of ErrorObjective Evaluation of Sources of ErrorIdentifying the Most Important Sources of ErrorSpecifying Bias ScenariosExercise: Specifying Scenarios of Bias5. Confounding I: Theoretical ConsiderationsSynopsisLearning ObjectivesDefinition Identifying Potential ConfoundersTraditional Approach to Assessing ConfoundingModern Approach to Assessing ConfoundingInappropriate AdjustmentsAssessing the Direction and Magnitude of Potential ConfoundingMethods of Controlling ConfoundingRandomizationSelection of Study Setting Free of ConfoundingRestrict Study Groups to Enhance Comparability Statistical Adjustment for ConfoundingRecommended Additional ReadingsExercise: Conceptual Basis of Confounding6. Confounding II: Practical ConsiderationsSynopsisLearning ObjectivesEvaluating the Presence and Impact of ConfoundingSpecifying Scenarios of ConfoundingAssessing Whether Confounding is PresentConsider Potential for Complete ConfoundingAssess Consequences of Inaccurate Confounder MeasurementApplying Knowledge of Confounding Based on Other StudiesAssessing Confounding When Risk Factors are UnknownDose-Response Gradients and Potential for ConfoundingIntegrated Assessment of Potential Confounding Exercise: Connecting Conceptual and Statistical Assessment of Confounding7. Selection Bias and Confounding Resulting from Selection in Cohort Studies SynopsisLearning ObjectivesStudy DesignsDefinition and Examples of Selection BiasSelection Bias Versus Confounding Evaluation of Bias in Cohort StudiesCompare Those Included to Those Not Included Compare Disease Rates Among Unexposed to External Populations Assess Whether Expected Patterns of Disease are PresentAssess Pattern of Results in Relation to Markers of Susceptibility to Bias Due to Participant SelectionAssess Rates for Diseases Known Not to Be Affected by the ExposureIntegrated Assessment of Potential for Bias in Cohort Studies Exercise: Assessment of Bias Due to Selection in Cohort Studies 8. Selection Bias in Case-Control Studies SynopsisLearning ObjectivesControl SelectionParticipant Selection in Case-Control and Cohort StudiesSelection of Controls from the Source PopulationCoherence of Cases and ControlsEvaluation of Selection Bias in Case-Control StudiesTemporal Coherence of Cases and ControlsDiscretionary Health Care of Cases and ControlsCompare Exposure Prevalence in Controls to an External PopulationDetermine Whether Exposure Prevalence Varies as Expected Among Controls Examine Markers of Potential Selection Bias in Relation to Measures of Association Adjust Measures of Association for Known Sources of Non- ComparabilityDetermine Whether Established Associations Can Be ConfirmedIntegrated Assessment of Potential for Selection Bias in Case-Control StudiesExercise: Assessing Selection Bias in Case-Control Studies 9. Bias Due to Loss of Study Participants SynopsisLearning Objectives Conceptual Framework for Examining Bias Due to Loss of Study Participants Evaluation of Bias Due to Loss of Study Participants Characterize NonparticipantsConsider Gradient of Difficulty in RecruitmentStratify Study Base by Markers of ParticipationImpute Information for Nonparticipants Integrated Assessment of Potential for Bias Due to Loss of Study ParticipantsExercise: Examining Implications of Non-Participation 10. Measurement and Classification of Exposure SynopsisLearning ObjectivesIntroductionIdeal Versus Operational Measures of Exposure Biologically Relevant ExposureTemporally Relevant ExposureOptimal Level of Exposure Aggregation Comparison of Optimal to Operational Measures of ExposureDoes Exposure Misclassification Differ by Disease Status?DefinitionsMechanisms of Differential Exposure MisclassificationEvaluation of Exposure MisclassificationCompare Routine Measure to Superior MeasuresExamine Multiple Indicators of ExposureExamine Subsets of the Population with Differing Exposure Data QualityEvaluate Known Predictors of ExposureEvaluate Known Consequences of ExposureExamine Dose-Response GradientsEvaluate Whether Exposure Misclassification Differs by Disease StatusIdentification of Subgroups with Nondifferential Exposure MisclassificationIntegrated Assessment of Bias Due to Exposure Misclassification Exercise: Assessing the Presence and Impact of ExposureMisclassification11. Measurement and Classification of Disease SynopsisLearning ObjectivesFramework for Evaluating Disease Misclassification Sources of Disease Misclassification Impact of Differential and Nondifferential Disease Misclassification Evaluation of Disease Misclassification Verify Diagnostic Accuracy for Subset of Study ParticipantsExamine Results Across Levels of Diagnostic CertaintyEvaluate Alternate Methods of Disease GroupingDetermine Whether Misclassification is Differential by Exposure StatusCreate Subgroups with Accurate Ascertainment or Non-Differential UnderascertainmentRestrict Inference to Disease Outcome That Can Be Ascertained AccuratelyIntegrated Assessment of Potential for Bias Due to Disease Misclassification Exercise: Assessing the Presence and Impact of Disease Misclassification 12. Random Error SynopsisLearning ObjectivesNature of Random VariationSequential Approach to Considering Random and Systematic Error Special Considerations in Evaluating Random Error in Observational Studies Statistical Significance Testing Interpretation of Confidence Intervals Multiple Comparisons and Related Issues Integrated Assessment of Random ErrorExercise: Assessing Random Error 13. Integration of Evidence Across Studies SynopsisLearning ObjectivesIntroduction Systematic Evidence Reviews Data Pooling and Comparative Analyses Meta-Analysis Interpreting Consistency and Inconsistency Among StudiesInconsistent FindingsConsistent FindingsEvolution of Epidemiologic ResearchIntegrated Assessment from Combining Evidence Across Studies Exercise: Interpreting Evidence from a Collection of Studies 14. Characterization and Communication of Conclusions SynopsisLearning ObjectivesPresenting Clear, Objective, and Informed ConclusionsApplications of EpidemiologyIntegration of Epidemiologic Evidence with Other Information Identification of Key Concerns Controversy over InterpretationThe Case Against Algorithms for Interpreting Epidemiologic Evidence Exercise: Communicating Summary Assessment of Epidemiologic Evidence

Highly relevant to any practicing epidemiologist.