ABSTRACT
This chapter continues the overview of epidemiologic analyses that are most applicable to EHR-based research. The material is framed via the analytic epidemiology lens, where a testable hypothesis has been specified in the context of cross-sectional, cohort, or case–control study designs. To begin the chapter, theoretical and statistical confounder selection is used to justify the variables included in an adjusted model. Then, multivariate regression procedures are presented specific to cross-sectional, cohort (with and without clustered or repeated data), and case–control study designs. Such regression models include linear, logistic, Poisson, Cox proportional hazards, and mixed effects. Post-analysis, the discussion shifts to statistical and causal inference of the study findings. Although causal inference merits its own book, a high-level overview is included to provide readers with a general understanding of how to evaluate causality in the context of observational EHR research.
