UnitedHealth's Optum is developing an AI-powered risk scoring system for Medicare patients that does not rely solely on physician-submitted diagnosis codes
Bob Herman / STAT :
Context & Ripple Effects
Optum’s project extends insurer use of predictive models into Medicare risk assessment, following reporting that insurers have used algorithms to identify points at which treatment payments can be curtailed. The distinction between scoring risk and making coverage decisions will matter because CMS has said Medicare Advantage insurers cannot use AI to determine care or deny coverage.
The move also lands as public programs explore AI in clinical and payment settings, including CMS’s outcome-based model for AI-driven care. It places a major Medicare-linked operator closer to the data and model layer of care financing.
First-order effects
- Optum can evaluate Medicare patient risk with a signal set that is not limited to physician-submitted diagnosis codes, potentially changing how it identifies and prioritizes members for internal risk-related workflows.
- The system will require governance that keeps risk scoring separate from prohibited automated coverage determinations, given the existing CMS policy boundary around AI use in Medicare Advantage.
Second-order effects
- Providers may face greater pressure to understand how insurer-side risk assessments interact with diagnosis documentation, care-management outreach, and payment-related processes.
- Other Medicare insurers and health-services vendors may need to clarify the inputs, validation, and human oversight behind their models as scrutiny grows from earlier reporting on insurer predictive algorithms and newer AI pilots in Medicare claims evaluation.
Third-order effects
- If insurers increasingly build risk models around data beyond submitted codes, competitive advantage may shift toward organizations that combine insurance operations, care delivery, and analytics infrastructure.
- The durable policy question will be whether rules designed for AI coverage denials adequately govern upstream scoring systems whose outputs can still shape care-management and financial decisions.
The trend: This is one data point in the expansion of AI from discrete clinical tools into the insurer and public-program infrastructure that organizes medical risk, payment, and care operations.