Variation at provider, team and facility level for profile measures related to diabetes care

Hofer TP, Krein S, Davis J, Hayward RA. Variation at provider, team and facility level for profile measures related to diabetes care. Paper presented at: Scientific Basis of Health Services International Annual Conference; 2000 Sep 1; Toronto, Canada.

PURPOSE: We used computerized hospital databases to define a set of indicators representing important processes of care related to the management of diabetes. We asked whether the variation in the indicator performance occurred more at the provider, primary care team or clinic site level within a group of 20 primary care clinical sites across eight hospitals comprising one Veterans Administration Service Network (VISN).METHODS: Data files were obtained from each facility containing patient demographic information, laboratory data, pharmacy data, and encounter data (dates and diagnoses). Separate databases contain patient assignments to primary care providers and teams.Inclusion criteria required that a patient have at least one outpatient visit during FY98, and a prescription for insulin or an oral hypoglycemic medication or home glucose monitoring supplies during FY 98. We included 18723 patients and 265 providers in 50 teams. The indicators were taken from the Veterans Health Administration (VHA) Diabetes Clinical Guidelines, and included receipt of eye care, and indicators related to obtaining a measurement, the mean level and proportion of patients above a threshold level of HgbA1c and LDL. Multilevel random effects models were used to estimate the variance explained by clinic site, team and provider for each of the indicators.RESULTS: Glycemic control at sites (HgbA1c 7.3%-8.8%) is similar or better than most patient populations reported in the literature although these reports predate the publication of recent trials showing the benefits of tighter control. However, some sites still have many patients with poor glycemic control (clinic rate of patients with a last HgbA1c > 9.5% ranges from 8%-31%). Monitoring and treatment of lipids is an area that can be improved at all facilities. Many patients are not being monitored and high LDLs are relatively common in those patients who have measures. The site explained the most variance in mean HgbA1c and LDL level at 8.7-9.1% while PCP explained from 0-2% across the facilities. The largest amount of variation at the provider level was found for obtaining a HgbA1c or LDL measurement in the past year and varied substantially across facilities (percent of variation due to provider ranging from 2-20%). Eye exam rates varied most at the clinic site level (9% of the variance in rates) and little or none at the physician level.CONCLUSION: Clinic site level profiles for important processes of care related to diabetes can be feasibly developed and will be reasonably reliable given the number of patients and the magnitude of facility differences. While physician level profiles will usually have too low reliability to be useful, for some indicators at some sites investigating provider level differences in performance may be possible.