Risk-adjusted payment and performance assessment for primary care.

Area of Science:

• Health economics
• Primary care analytics
• Health services research

Background:

• Primary care practices seek improved incentives via bundled payments and performance feedback.
• Effective risk adjustment for patient cost and outcome variations is currently underdeveloped.

Purpose of the Study:

• To develop and evaluate risk adjustment models for bundled payments in primary care.
• To assess the performance of these models in predicting patient outcomes and practice-level variation.

Main Methods:

• Utilized MarketScan claims data (17.4 million lives) to model bundled payments for primary care activity levels (PCAL) and 9 patient outcomes.
• Evaluated models on 457,000 individuals across 436 primary care panels and a distinct multipayer medical home cohort.
• Predicted outcomes using age, sex, and diagnoses; defined PCAL as a proxy for comprehensive primary care costs.

Main Results:

• The PCAL model explained 67% of outcome variation at the patient level and 72% at the practice level.
• Outcome-specific models explained 17%-86% of practice-level variation, significantly outperforming generic scores.
• Demonstrated superior performance in predicting specific outcomes like "prescriptions for antibiotics of concern" (R(2) of 47% vs. 5%).

Conclusions:

• Existing claims data are sufficient to develop robust risk-adjusted bundled payment calculations.
• These data and models can support performance assessments to drive primary care transformation.
• The developed models facilitate fairer and more accurate performance judgments for primary care physicians.