The Economics of Patient-Centered Care

The patient-centered medical home (PCMH) is a widely-implemented model for improving primary care, emphasizing care coordination, information technology, and process improvements. The PCMH has been extensively promoted by organizations representing primary care practitioners and substantially supported by funding from public and private payers. However, despite numerous pilot projects and extensive efforts at evaluating the PCMH, the evidence regarding the model’s impacts on patient experience, utilization and expenditures remains mixed.

One reason for these null findings may be the tendency of health services researchers to treat the PMCH model as an undifferentiated intervention, which obscures meaningful variation in implementation. Practices are able to choose which subset of over 100 possible practice improvements to adopt, and resulting differences in emphasis may lead to practice styles with differing impacts on patient outcomes. This heterogeneity leads to contracting inefficiencies between insurers and practices and may account for mixed evidence on its success.

Using a novel dataset including over 150,000 patients in 104 primary care practices, we use hierarchical clustering to identify groups of practices with meaningful variation in implementation, and then link these clusters with detailed patient claims data. This process points to three distinct approaches to PCMH implementation: one lower-performing cluster that focuses on meeting minimum thresholds for PCMH recognition and associated reimbursement increases; and two higher-performing clusters, with comparable overall performance scores but different emphasis. The first of these implemented more “patient-facing” features, such as electronic care management support, while the second focused on “physician-facing” improvements. We find implementation choice affects performance, suggesting that generally-unobserved features of primary care reorganization influence patient outcomes. Reporting these features may be valuable to insurers and their members.

An extension of this work seeks to address important limitations related to the limited sample of 104 practices included by extending the clustering to nearly 10,000 practices operating nationwide, repeating the clustering algorithm on a sufficiently large group to confirm the stability of the method used, assess which features differentiate clusters in the national sample, and repeat the patient-level analyses on a larger and more geographically diverse population.