Results: We used data passively transferred from EMR databases to assess patterns of diabetes care and quality in a large number of primary care clinics, to determine the feasibility, accuracy, and possible uses of such profiling activities.
We identified 66 clinics in two large care systems that used Epic Electronic Medical Records (EMR), and collaborated with programmers at Epic (Madison, WI) to develop programs to extract clinical data that can be used to assess patterns of diabetes care and quality in primary care settings. Data extracted and analyzed included confirmation of a diabetes diagnosis, dates and values of blood pressure (BP), glycated hemoglobin (A1c) and low-density lipoprotein (LDL) tests, selected other laboratory tests, gender, age in years, selected medications, primary care physician code, and clinic code.
Checks of data accuracy at both medical groups detected some programming errors, most of which were related to identification of medications, that were iteratively corrected before finalizing the analytic data set. There was statistically and clinically significant variation in patterns of diabetes care and in measures of diabetes quality of care across primary care clinics. The proportion of patients who simultaneously met evidence-based goals of A1c < 7%, SBP < 130 mm Hg, and LDL < 100 mg/dl ranged from a low of 3.6% to a high of 32.5% across clinics, with a median clinic value of 13.4% in one medical group and 17.9% in the other medical group. Patterns of use of insulin, metformin, thiazolidenediones, ACE/ARBs, statins, and other recommended medications also varied substantially.
There is wide variation in diabetes care quality across clinics and physicians, and efforts to understand the dynamics of this variation are likely to lead to identification and implementation of clinical care systems that may broadly improve care. EMR-derived data may be used to passively monitor patterns diabetes care at the level of medical groups, clinics, and physicians. These electronic data are demonstrably accurate, permit adjustment for patient factors, and are inexpensive to obtain and analyze. This source of information has the potential to accelerate efforts to improve care delivered for diabetes, hypertension, heart disease, and other conditions in clinics that use EMRs.