Learning from RCTs in Public Health and Medicine
summary
Argues observational data can be judged against randomized benchmarks by target trial emulation where each eligible person is nested into every trial they could join, comparing concrete regimes like one-year versus never and estimating effects with blip functions. Proposes structural approaches—structural nested models, marginal structural models, proximal inference and ML—as tools, stresses coordinated RCT–observational pairs, logs of all analyses, and warns unmeasured confounding often makes faithful emulation impossible.
AI-written summary · about