Εffect measure modification and confounding in applied medical research