Viewpoint | Detecting selection bias in observational studies—when interventions work too fast

Detecting Selection Bias in Observational Studies—When Interventions Work Too Fast – JAMA Internal Medicine (free for a limited period)

Commentary from the author on Twitter

Our paper in @JAMAInternalMed helps recognize confounding/biases in obs studies.https://t.co/qhptGVAIev

Look at image on left-a randomized study in which K better than V (slow separation of curves, plausible).

On right-retrospective obs study of VRD vs KRd.Early separation!! pic.twitter.com/mma9vGpueQ

— Manni Mohyuddin (@ManniMD1) June 12, 2023