Randomization Inside a Medicare Payment Model, It's Really Cool

Key Takeaways

• ACCESS Model randomizes 10% of enrollees to control for 12 months
• Fixed payments $90‑$420 per beneficiary, with 50% withhold
• Randomization addresses selection bias in digital health evaluations
• Successful design could inspire randomized trials in other federal programs

Pulse Analysis

The ACCESS Model represents a paradigm shift in Medicare’s approach to chronic disease management. By bundling fixed, outcome‑aligned payments ranging from $90 to $420 per beneficiary and tying them to four clinically relevant tracks—early cardio‑kidney‑metabolic, cardio‑kidney‑metabolic, musculoskeletal, and behavioral health—the model aligns financial incentives with measurable health improvements. The 50% withhold mechanism further ensures that providers are rewarded only when baseline outcomes are met, creating a clear link between payment and performance that could reshape value‑based care across the system.

What sets ACCESS apart is its embedded randomization. Ten percent of would‑be participants are automatically assigned to a control group, receiving standard Medicare services while the remaining 90% receive the digital intervention. This design directly tackles the chronic evaluation problem where engaged users are self‑selected and often healthier, skewing observational results. By creating comparable groups at enrollment, the model can isolate the true effect of technology‑enabled care on utilization, costs, and patient‑reported outcomes, offering policymakers robust evidence that has been missing from prior digital health pilots.

The implications extend beyond Medicare. If the randomized evaluation demonstrates cost‑effective quality gains, the Innovation Center can expand the model through rulemaking, setting a precedent for other federal programs. Agencies in labor, housing, and education could adopt similar randomization frameworks to test interventions rigorously without new legislation. While the approach may face political resistance—randomly denying eager participants a service—it provides the most credible causal evidence for scaling innovations, marking a significant step toward data‑driven, accountable governance.