Why use a self-controlled design?
🔹deal with between-person confounding (exposed & unexposed may differ w/r/t characteristics that are related to the outcome)
🔹can examine risk of an ACUTE outcome following a point exposure with a transient effect
🔹includes only cases
🔹observation period duration determined a priori
🔹obtain info on exposures
🔹determine a risk interval following the exposure that makes sense
🔹remainder of observation period: control interval
@SattuiSEMD Caveats for SCCS?
🔹time dependent confounding (confounders that change over time)
🔹reverse causation
🔹misclassification of time periods (wrong length, wrong timing)