The patient-specific surveys were based on chart information of 3 randomly chosen patients per physician on controller medication.