Developing and Implementing a Quality Measure for Glycemic Control

Background: Many patients with diabetes are under sub-optimal glycemic control. Central to the clinician's task in improving glycemic control is the management of hypoglycemic medication therapy, including the use of drugs such as insulin and sulfonylureas. Clinical trials have demonstrated that more intensive hypoglycemic medication therapy results in improved glycemic control. Yet quality measures for this critical process of care have not been developed and we know little about how physicians actually manage hypoglycemic medications. Objectives: We propose to develop a quality measure that describes the intensity of physicians' hypoglycemic medication therapy. We will then provide feedback to VA physicians regarding their practices and access to experts in diabetes care to determine whether this intervention leads to improvements in glycemic control. Methods: The study was divided into two phases. During the first phase we used existing data to model the decision to increase hypoglycemic medications. At each medical visit, we determined whether an increase in medication therapy occurred. We then used recursive partitioning to develop a model that identified patient characteristics at the visit, such as recent laboratory results and diagnoses, associated with the decision to increase therapy. This model assigns a predicted probability of an increase in therapy to each visit. We used these predictions to define an intensity of hypoglycemic medication therapy for each physician that compared the actual to predicted number of increases over all patient-visits. The second phase was a randomized trial in which clinicians at experimental sites receive feedback on performance and access to expert opinion while usual care is provided at control sites. Feedback on performance was provided twice over 6 months. The change in intensity of treatment scores and glycosylated hemoglobin levels pre- and post-intervention at these sites were compared to performance of primary care physicians at control sites not receiving the intervention. Status: Completed.