Uses of routine data sets in the evaluation of health promotion interventions: opportunities and limitations