Medicare Advantage 2026: How Forced Disenrollment Reshapes Risk Pools and Locks In 2027 Bid Exposure

Approximately 2.9 million Medicare Advantage enrollees are switching plans in 2026 after carrier exits reached 10% of the MA population, a tenfold jump from the 1% annual average between 2018 and 2024. Plans that absorbed those members filed 2027 bids in June 2026, before the shift was fully visible in their own data. CMS's 2.48% effective rate against 8 to 10% medical cost trend sets the frame the coding gap then lands on.

Key Takeaways

• Forced disenrollment hit 10% of MA enrollees for 2026, up from 6.9% in 2025 and just over 1% annually across 2018 to 2024, across 192 million enrollee-years of Plan Finder data.
• Displaced and retained members show no statistically significant difference in fee-for-service HCC risk scores, so this is a payment correction rather than demographic self-selection.
• Rural beneficiaries faced a 28% forced disenrollment rate against 15% urban, and smaller insurers accounted for 48.8% of forced disenrollments against 27.1% of retained enrollees.
• A receiving plan does not inherit the exiting plan's Annual Wellness Visit encounter data, so year-one V28 risk scores understate the chronic condition burden the plan is actually carrying.
• CMS estimated the 2027 risk model and normalization changes reduce payments by 3.32%, running in the same direction as the rate gap rather than offsetting it.

The Displacement and the Rate Set Against It

The Johns Hopkins research letter published February 18, 2026 in JAMA analyzed 192 million enrollee-years of CMS Plan Finder data. Annual forced disenrollment averaged just over 1% between 2018 and 2024, climbed to 6.9% in 2025, and reached 10% for 2026. Smaller insurers accounted for 48.8% of forced disenrollments against 27.1% of retained enrollees, and rural beneficiaries faced a 28% rate against 15% for urban. PPO enrollees were displaced more often than HMO members.

The risk score finding is the one that reframes the problem. The authors found no statistically significant difference in fee-for-service HCC scores between members forced to disenroll and members who kept their plans. "We are probably overpaying for Medicare Advantage, and so efforts to reduce how much we pay are going to mean that some plans can no longer be profitable," Mark Meiselbach told AJMC. Receiving plans did not inherit a sicker cohort in aggregate.

The rate environment they filed into is tight independently of that. CMS finalized the 2027 effective rate at 2.48% on April 6, 2026, up from the 0.09% proposed in the Advance Notice, representing over $13 billion of additional payments against 2026. Oliver Wyman projected medical cost trend of 8 to 10% against that floor and described the result plainly: revenue growth will not keep pace with expense growth.

Rate geography and exit geography are not independent. Effective rate growth varies by county benchmark, Star Rating, and risk score trend, and lower-benchmark rural counties see below-average growth even in a year when the national average improves. Those are the counties where exits ran highest.

The Coding Gap the Bid Cannot Reach

The exposure is mechanical rather than compositional. The V28 HCC model reached full weight for payment year 2026 after phasing in at one-third in 2024 and two-thirds in 2025, using 115 HCC categories and 7,770 diagnostic codes against V24's 86 categories and 9,797 codes. The narrower structure means each surviving code carries more weight, so a missed code costs more under V28 than it did under V24.

Annual Wellness Visits are where chronic conditions get coded. A member who completed an AWV at their prior plan carries that coding into that plan's risk scores. The receiving plan does not inherit the encounter data. It has to conduct its own AWV under its own NPI and submit those records to generate risk scores for the following payment year.

V28 Coding Gap Mechanics for Displaced MA Members

Stage	Mechanism	Actuarial Effect
Year 0 (exiting plan, 2025)	AWV and chronic condition codes submitted by exiting plan	Risk scores accrue to the plan that exits, not the receiving plan
Year 1 (receiving plan, 2026)	No inherited encounter history; AWV completion delayed by transition friction	V28 risk scores suppressed below member's true acuity; bid filed June 2026 cannot reflect this
Year 2 (receiving plan, 2027)	AWV and coding catch-up as member stabilizes; chronic conditions re-established in records	Risk score increases above year-1 baseline; looks like medical trend if not modeled separately

Displaced members are exactly the population least likely to complete one promptly. Establishing a new primary care relationship, working out a different network, and resolving prescription continuity all compete for attention in a first enrollment year, and the transition friction suppresses AWV completion. The plan's year-one risk score reflects whatever partial coding it happens to hold.

The actuarial damage arrives a year later, as an accounting problem rather than a cost problem. When those conditions are finally documented, the catch-up presents in 2027 experience as apparent medical cost trend. A plan modelling trend against its stable historical base reads that increase as organic, and a plan filing 2028 bids on 2027 trend data without separating the displaced cohort embeds the misattribution in forward pricing. The bid cycle offers no correction: 2027 bids were filed in June 2026, before H1 utilization on those members existed.

Recalibration and the Shape of the Pool

Two further effects run in the same direction rather than offsetting.

For 2027, CMS recalibrated V28 using 2023 diagnosis data and 2024 expenditure data, replacing a calibration anchored to 2018 diagnoses and 2019 expenditures, and estimated that the model and normalization changes together reduce payments by 3.32%. CMS kept the V28 clinical classification system, deferring the larger revisions floated in the Advance Notice, so this is a coefficient change rather than a structural one. For a receiving plan it is still a reclassification: inherited members carry prior-year profiles generated under another plan's documentation practice and scored on the old calibration, and their conditions may weight differently under the new one in either direction.

The geographic effect is the subtler one, and it does not show up in individual scores at all. Where carrier exits left a rural county without a viable MA alternative, displaced members revert to traditional Medicare rather than re-enrolling. Vermont saw 92.2% of MA enrollees facing forced disenrollment, and at least 40% did in six further states, so reversion is the default outcome in those markets rather than a tail case.

That changes the composition of the surviving pool without any member becoming sicker. Members who actively re-enroll in another MA plan tend to be those most engaged with the product, often the chronically ill who rely on care coordination and supplemental benefits; those who revert to traditional Medicare tend to be healthier. The same differential sorting appears when ACA carrier exits shrink the pool of remaining issuers. Risk adjustment normalizes at county and plan level, so a book that becomes more urban because its rural members left the program entirely carries a geographic mix shift in its risk revenue that individual HCC tracking will not surface.