Journal Archive > 2002 > February

Quality of care

Specialist or generalist? In treating diabetics, it may not matter

Chances are someone in your family has diabetes because at least 16 million Americans do. And chances are you assume that person would receive the best care from an endocrinologist. It's a natural assumption.

But it's not necessarily a valid assumption, says Dr. Sheldon Greenfield, professor of family medicine and community health and a pioneer in medical outcomes research.

Drs. Sherrie Kaplan and Sheldon Greenfield, study co-authors © Mark Morelli

True, there are studies that statistically indicate diabetics get better care from endocrinologists than from internists. However, Greenfield notes, those studies fail to take into account several important factors, including patient case mix and differences in practice styles among individual physicians—a subtle and elusive factor called physician-level clustering.

When Greenfield and his colleagues quantified and included these elements in their analysis of 1,750 adult diabetics receiving treatment at 29 sites in 13 states, they found, overall, no difference in the quality of care provided diabetics by specialists and generalists.

"We saw more variation in quality of care at the individual physician level in both groups—endocrinologists and internists—than between them," says Dr. Sherrie Kaplan, co-author of the study and a clinical research professor in family medicine and community health. "It appears that endocrinologists have done a very good job teaching generalists about how to treat diabetes."

The research, funded by the American Diabetes Association, was published in the January 15 issue of The Annals of Internal Medicine in an article titled "Profiling Care Provided by Different Groups of Physicians: Effects of Patient Case-Mix (Bias) and Physician-Level Clustering on Quality Assessment Results."

Accurate assessing is the goal
But diabetes is only a model for the researchers. Their real focus is to expand and refine methods for determining the quality of medical care. "A lot of quality-of-care assessment today is criticized as over-simplified. It's often a fair criticism," says Greenfield.

Fair or not, the demand is growing faster than the sophistication of methodology for measuring physicians' performance, care protocols and the quality of care provided by specific institutions. The demand is coming from patients, state and federal governments and health insurance companies. And it's not going away.

"Major HMOs are making decisions about a lot of things without statistically sound information. It's like telling people what their weight is without using a scale," Greenfield says.

To illustrate the impact of ignored factors, Greenfield likes to cite several studies that showed cardiologists treating heart attacks performed more consistently than did generalists. "But when month-out outcomes were factored in, the mortality rates were the same," says Greenfield.

He also tells of a recent meeting he attended for health care providers and administrators where one doctor's performance was being determined on the basis of satisfaction questionnaires filled out by four patients—a completely inadequate database, says Greenfield.

More alike than different
The bias of patient case mix—the fact that patients are not just randomly distributed among physicians but tend to choose doctors of certain characteristics based on their own particular characteristics—has long been recognized by researchers investigating quality of care, although that factor is not always integrated into quality assessment, notes Kaplan.

Kaplan and Greenfield found that diabetics who saw endocrinologists tended to be significantly younger, better educated and of generally overall better health, although they were longer-term diabetics (average 15 years) and more insulin-dependent than diabetics treated by generalists. This information was taken into account in their statistical analysis.

But the investigators also found another, ultimately more significant factor—"clustering" of physician performance at the level of the individual physician or group practice. Physician-level clustering refers to the consistency, across patients, of a single physician's performance in areas such as testing and treatment as well as patient outcomes and satisfaction.

The variations among patients have served as an argument against the viability of individual physician performance assessment. But despite patient variation, Greenfield and his colleagues found it is possible to measure individual physician style.

"We found that the cluster effect was strong enough that we could see a measurable 'thumbprint' for each physician," a clear pattern of practice characteristics across all patients, Greenfield says.

That 'thumbprint' was based on quality-of-care measures, including frequency of ordering a lipid profile, dilated eye examination, foot examination, blood pressure, urinalysis or test for urinary protein and self-monitoring of blood glucose.

Interpersonal behavior was assessed through a questionnaire filled out by patients. In fact, those questionnaires revealed the only quality measure that separated the specialists from the generalists, after the investigators had adjusted for patient mix and physician clustering: They found a significant number of diabetic patients felt endocrinologists offered better care than generalists.