In: Coleman K, Reiter KL, Fulwiler D. " The impact of pay-for-performance in a large network of community healthy centers." J Health Care Poor Underserved. 2007 Nov; 18(4):966–83.
Errors were discovered in the data provided that led to an overstatement of patients and encounters in 2004. The revised Table 1 below presents descriptive statistics for Access Community Health Network.
Removing erroneous encounters from the data resulted in an increase in the total number of patient-doctor pairs eligible for inclusion in the final sample (n=1,288). The reanalysis shows no changes in the direction or significance of the associations between the pay-for-performance program and the process and outcome measures; however, the nature and strength of some other relationships changed. In Table 3, the reanalysis indicates that the average annual number of encounters per diabetic patient remained about the same from 2003 to 2004, at 6.13 and 5.71, respectively (t-value 2.55). In Table 5, among people with at least one HbA1c test in 2003 and 2004, HbA1c scores are, on average, 0.55 points lower for those using an internal medicine physician for their primary care than for those using a family practice physician, holding all else constant (z-value = -1.97, p=.0490). Neither insurance status nor location of physician training are significantly related to HbA1c score.
One interesting finding occurred in the reanalysis of the relationship between physicians' baseline performance in 2003 and their performance after implementation of the pay-for-performance program in 2004. In Table 6, baseline performance on diabetes outcome, measured as percent of patients with HbA1c,7, was predictive of performance after the intervention (t-value=2.97, p-value=.0051), although the model explained only 16% of the variation.
Overall, the revised results are consistent with the authors' expectations and with the original conclusions regarding the impact of the pay-for-performance program on diabetes care at Access Community Health Network.
Profile of Access Community Health Network
|Total unique patients||83,695||138,572||159,881|
|Average # encounters/patient||2.87||2.92||3.21|
|Total diabetic patients||3,697||5,423||8,141|
|Percent diabetic patients||4.4%||3.9%||5.1%|