To estimate the expected magnitude of error produced by uncontrolled confounding from health behaviours in observational medical record-based studies evaluating ...