Physician Quality Can't Be Boiled Down to a Few Measures

PITTSBURGH — Preliminary results from the American Board of Internal Medicine's Comprehensive Care Project confirm what many internists already know intuitively: Overall physician “quality” can't accurately be described using a single or limited number of conditions and/or measures.

“General internists play a central role in caring for patients with multiple medical conditions and comorbidities. However, current physician performance measurement typically focuses on quality measures for a single or a limited number of conditions,” Dr. Eric S. Holmboe said at the annual meeting of the Society of General Internal Medicine.

Results of the ABIM's medical chart audit for 22,526 patients seen by 236 general internists showed wide variation both within and between physicians across a range of medical conditions, suggesting that “assessment of quality using limited numbers of existing performance measures appears not to be an accurate measure of comprehensive physician practice quality,” according to Dr. Holmboe, senior vice president for quality research and academic affairs for the ABIM.

Invitations to participate in the project were mailed to 6,709 general internists with time-limited certification residing in 13 states that were stratified by 2005 state rankings of quality by the Agency for Healthcare Research and Quality (AHRQ). Participants in the performance project received an honorarium and points toward completion of maintenance of certification. Of the 254 physicians who began the project, a total of 236 completed on-site medical record audits.

Medical records were audited for six chronic conditions (diabetes, hypertension, coronary artery disease, heart failure, atrial fibrillation, and osteoarthritis); four acute conditions (upper respiratory infection, urinary tract infection, low back pain, and depression); and six preventive processes of care (smoking cessation counseling, influenza and pneumococcal immunizations, and screening for breast cancer, colon cancer, and osteoporosis). In all, 56 performance measures were abstracted for each physician's practice, said Dr. Holmboe, also of Yale University, New Haven, Conn.

The mean age of the 236 internists was 42 years, and 36% were female. Of the 190 who completed a survey component on practice systems, 36% were in solo practice, 30% in single-specialty practices, 25% in multispecialty practices, and 6% in academic faculty practice.

The medical record audit showed that characteristics of the 22,526 patients varied widely among the physician practices. The mean patient age per physician sample was 60 years, with a per-practice range of 44–77 years. The percentage of women patients averaged 60%, ranging from 10% to 75% per practice. Ethnicity was 37% white, 9% black, 8% Hispanic, and 46% undetermined.

An average of 95 charts was abstracted per physician. Most of the physicians were able to meet the request for at least 20 performance measures each for patients with hypertension (235) and diabetes (215), but fewer could provide the requested 20 for combined upper respiratory infection/urinary tract infection (187) and the three cardiovascular conditions (120), or the 10 charts for low back pain (208). Only about a third of the physicians were able to meet the targets for all six conditions, although half were able to meet the requested targets for at least five, Dr. Holmboe said.

Performance on process and outcome measures for each of the six conditions varied considerably. For example, the two measures with the least successful overall results were foot exams for diabetic patients (11%, range 0%-100%) and appropriate use of nasal decongestants for upper respiratory infections (5%, range 0%-100%). The two measures with the most successful results were weight documentation for heart failure patients (86%, range 0%-100%) and not prescribing drugs for low back pain (86%, range 36%-100%).

For physicians, although there was modest correlation between performance on chronic condition measures and on prevention measures, performance on chronic condition measures correlated poorly with performance on acute condition measures.

Correlation was even lower between performance measures of acute care and prevention.

These preliminary findings suggest that it is feasible to measure quality performance in a general internal medicine practice for some conditions but it may not be possible for all conditions in general internal medicine practices, Dr. Homboe concluded.