Should structural interventions be evaluated using RCTs? The case of HIV prevention