Confounding, causality, and confusion: The role of intermediate variables in interpreting observational studies in obstetrics