Email Record: Applying propensity scores estimated in a full cohort to adjust for confounding in subgroup analyses