Estimating the occurrence of child maltreatment and risk-factor effects: benefits of a mixed-design strategy in epidemiologic research.