Attributing Sources of Variation in Patients' Experiences of Ambulatory Care

Background: Public reporting and pay-for-performance programs increasingly rely on patient experience data to evaluate individual physicians and guide quality improvement efforts. The extent to which performance variation is attributable to physicians versus other system-level units, however, remains unclear. Methods: Using ambulatory care experience survey data from 61,839 patients of 1729 primary care physicians in California (response rate = 39.1%), this study assesses the proportion of explainable performance variation attributable to various organizational units in composite measures of physician-patient interaction, organizational features of care, and global assessments of care. For each measure, multilevel regression models that controlled for respondent characteristics and used random effects to account for the clustering of patients within physicians, physicians within care sites, care sites within medical groups, and medical groups within primary care service areas, estimated the proportion of explainable performance variation attributable to each system-level unit. Results: System-level factors explained between 27.9% to 47.7% of variation, with the highest proportion explained for the access to care composite and the lowest explained for the quality of chronic care composite. Physicians accounted for the largest proportion of explainable variance for all measures (range: 35.1%-49.0%). Care sites and primary care service areas explained Substantial proportions of variance (>20% each) for the access to care and care coordination measures. Medical groups explained the largest proportions of variation (>20%) for global assessments of care. Conclusions: Individual physicians and their care sites are the most important foci for patient experience improvement efforts. Because markets contribute substantially to performance variation on organizational features of care, future research should clarify the extent to which associated performance deficits are modifiable.