Self-Controlled Designs in Epidemiologic Research

To draw causal inference using traditional epidemiologic designs (e.g., cohort, case-control), we principally ask: “Why me and not other individuals?” In self-controlled designs, we principally ask: “Why me now and not me other times?” Self-controlled designs can be thought as special cases (i.e., variants) of traditional designs in that they are either exposure-indexed (i.e., exposed only), or outcome-indexed (i.e., case only). The case-crossover design (including its main subvariant: case-time-control) follows the case-control logic. The self-controlled case series design (including its main subvariant: self-controlled risk interval) follows the cohort logic. Grab a cup of coffee (but preferably tea) and spend some quality time reading the papers in this collection by the proposed order. These papers, selected out of many other outstanding articles, include a few “classics” as well as some more recent articles that collectively provide a broad coverage of conceptual considerations and empirical evaluations of self-controlled designs in epidemiologic research.