Accounting for Social Risk in Medicaid: Insights from North Carolina
Tina Cheng, Kun Li, Kelley A Jones, Katie Huber, Alida Austin, Rushina Cholera, Robert Saunders, Rebecca WhitakerAbstract
Background/Objective
Risk adjustment can incentivize high-quality care for individuals with complex needs. Incorporating social drivers of health into risk adjustment models, known as social risk adjustment (SRA), may improve fairness and payment accuracy.
Methods
Using North Carolina Medicaid claims data (2021–2023) for 687,856 adults enrolled ≥6 months per year, we compared spending models incorporating clinical risk alone versus models that additionally included area-level social risk factors from the social-deprivation index (SDI), demographic characteristics, and interactions. We assessed how alternative specifications distributed expected spending by calculating observed-to-expected (O:E) spending ratios for clinically and social defined subgroups.
Results
Adding social risk and demographic factors produced little improvement in overall model performance, and clinical risk remained strongly associated with spending across models. However, these factors changed the distribution of expected spending across subgroups, improving alignment between observed and expected spending for several racial and ethnic populations. Interactions further suggested that the relationship between clinical risk and spending varied across SDI dimensions.
Conclusion
Although SRA may yield limited gains in overall model performance, it can improve subgroup level payment distribution and inform debates about fairness, resource allocation, and incentive alignment in Medicaid payment models.