Unattributed Medicare beneficiaries have greater social disadvantage than those linked to primary care physicians.

INTRODUCTION: Attributing patients to physicians using claims data is a key design feature of most payment and delivery models. However, claims-based attribution makes model participants less representative by excluding beneficiaries who use less care. METHODS: We used the attribution methodology from the largest completed primary care model, Comprehensive Primary Care Plus (CPC+), to describe differences in attributable and unattributable Medicare beneficiaries living in CPC + regions from 2016 to 2021. RESULTS: Unattributed beneficiaries were more male (56% vs 43%), eligible for the low-income drug subsidy (15% vs 12%), dual eligible (13% vs 10%), and non-White (22% vs 15%). CONCLUSION: Tests of models that use claims-based attribution may fail to generate evidence relevant to underserved groups. Policymakers should consider attribution designs that could improve representativeness of model test populations.