Pay-for-Performance in Diabetes: Sorting the Good from the Not-so-Good

As P4P strategies gain popularity, the need for accurate assessment tools becomes more and more apparent-many physicians support PCP but believe that current measures are inaccurate. A common problem is that physicians appear to be penalized if they treat the elderly or patients who have reduced access to care or drug treatment.

Pay-for-performance (P4P) is a merit-based system that offers added incentive for healthcare providers and institutions to achieve optimal clinical outcomes in their patients. As P4P strategies gain popularity, the need for accurate assessment tools becomes more and more apparent—many physicians support PCP but believe that current measures are inaccurate. A common problem is that physicians appear to be penalized if they treat the elderly or patients who have reduced access to care or drug treatment.

A study conducted by Brown et al. explores the concept of evaluating primary care physician (PCP) performance in diabetes management relative to treatment expectations.

The researchers assessed a total of 14,033 patients treated by 133 PCPs. First, the researchers determined the likelihood that a patient would have uncontrolled diabetes (defined as HbA1C > 8%) based on patient-level characteristics that reflected both demographics and biomarkers (eg, BMI, serum albumin, LDL and HDL levels). All participating patients had seen their physician at least 3 times during the study period.

A second model predicted diabetes control from a physician-level perspective, accounting for characteristics like years since completion of residency, specialty, and proportion of patients taking insulin. The researchers then combined patient- and physician-level models to produce a third model that could explain HbA1C variance.

PCP performance was quantified as the difference between the expected and observed proportions of patients presenting with uncontrolled diabetes, and adjusted for relevant patient-level variables that are uninfluenced by physician behavior (eg, age, socioeconomic status). Based on this data, PCPs were successfully categorized as having performed either better or worse than expected.

Among the 133 PCPs, 116 performed as expected. Eight physicians performed worse than expected and 9 performed better than expected. PCPs who performed worse cared for an average of 164 diabetes patients compared to an average of 126 for PCPs lower than expected proportion of patients. The researchers indicate that patient volume was not significant, and it’s likely that PCPs who performed best made good patient management decisions.

Men, African Americans, and unmarried patients were at elevated risk for uncontrolled diabetes, a problem the researchers suggest may be resource-related.

The possible applications of this method of PCP evaluation could be extended to many other chronic conditions, and may provide new insights regarding the relative effectiveness of various treatment strategies. This study appears electronically ahead of print in the American Journal of Medical Quality.