A National Program, Won Locally

The 35 million are people

They are our friends, parents, and grandparents. One is at a pharmacy counter right now doing the math on whether she fills all three prescriptions or two. One was in an ER bed last night and goes home tomorrow to an empty apartment. One is lonely. One is fine, and plays pickleball twice a week. Most carry three or four conditions at once and try to age with some dignity. That is who is inside the number.

Medicare Advantage now covers about 35.2 million people, 55 percent of everyone on Medicare, and the mix is the part people skip past. It skews older, poorer, and sicker than the Medicare you picture.

35.2M

members, 55% of Medicare

68%

of all dual-eligibles are in MA

7.6M

in Special Needs Plans, up ~10%/yr

Start with income and support. MA members are more likely than traditional Medicare to be dual-eligible, lower-income, and to have less than a college education. Sixty-eight percent of all dual-eligibles, the people who qualify for both Medicare and Medicaid, are now in MA. Many live alone, and a meaningful share face food or housing insecurity. These are not edge cases in the data. They are the median member in a lot of the organizations I work with.

Then the clinical picture. Most MA members live with two or more chronic conditions at once. Roughly seven in ten have hypertension, about one in three has diabetes, and large shares carry heart disease, COPD, chronic kidney disease, arthritis, or depression, usually in combination rather than alone. Special Needs Plans, built for the dual-eligible, the institutionalized, and the chronically ill, have up to 7.6 million enrollees and are still climbing about 10 percent a year. Their risk scores run roughly 10 percent above comparable fee-for-service, which is the data's way of saying these members are genuinely harder to care for.

And the faces are not the ones the old charts assumed. MA enrolls a higher share of Black and Hispanic members than traditional Medicare, concentrated in exactly the urban and Southern markets where the plan shelf is most crowded, and the stakes for getting care right are highest.

So a half-star is not a stock tick. It is a grandfather who missed his diabetic eye exam, or a widow who couldn't reach a human on the phone the week she needed one. Sometimes this is also a data problem; Stars has its own flaws in many cases, and half a star or one star may reflect a data gap rather than an actual quality-of-care gap. A member may have received an eye exam, but the data did not reach the health plan on time or correctly. That is another topic for a future piece.

The fast national read

The headline barely moved. The enrollment-weighted average went from 4.01 to 3.99. Four straight years now sit inside a 0.12-star band. That is noise wearing the costume of a verdict.

But the averages hide more than they show, so it is worth going one layer down.

Plan performance, up close

Three things stand out when you stop looking at the mean and start looking at the distribution.

First, the top of the market thinned. Five-star contracts fell from 22 to 15. About 37 percent of contracts landed at 4 stars or above, and because the big plans cluster there, roughly 71 percent of all enrollees are still in a 4-plus plan. Four contracts came in as low performers. The quality-bonus pool riding on all of this runs to roughly 13 billion dollars of taxpayer money, which is why a tenth of a star is a boardroom conversation, not a footnote.

Second, size does not decide stars, in either direction. Focused, regional operators do lead the table on average, Kaiser near 4.5, Alignment around 4.1, Devoted close to 3.9. But the data does not support a simple scale-versus-quality story. Humana, one of the two largest MA carriers, averages about 3.8 across thirty rated contracts and carries six at 4.5 stars. UnitedHealth holds three 5-star contracts. And some of the very best results this year come from regional and Puerto Rico plans: MCS Classicare and GuideWell's Triple-S Advantage both earned 5 stars, with Elevance's MMM close behind. The honest read is that the rating is earned at the contract level, not the enterprise level. Inside a single large parent, contracts can run from 5 stars down to 2.5. What travels is the operating model and the data discipline, and it shows up market by market, not on the logo.

Third, a lot of the movement this year was not care getting better or worse. It was the cut points moving. CMS raises the bar each cycle, trims statistical outliers, and tightens the thresholds, so a plan can deliver the exact same care as last year and still slip a half-star because the curve moved underneath it. That mechanic matters for how you read any single year, and it is a big part of why I argue the scoreboard and the actual care delivered are not the same thing.

A fourth point is worth naming for anyone who runs a plan. The gap between the enrollment-weighted average, near 3.99, and the simple contract average, closer to 3.83, shows quality is concentrated. The large, higher-rated contracts carry most of the members, while a long tail of smaller contracts sits a full star or more below them. That tail is where most of the real improvement opportunity lives, and it is disproportionately the plans serving duals, SNP members, and harder markets. Add the pressure on the drug side, where adherence and medication-safety measures carry heavy weight, and you have a map of where a focused operator can gain ground while the market stands still.

2026 to 2027: what moved

Lay the two landscape files side by side, and the calm national number hides a strong current. Contracts fell from 679 to 632, a net loss of 47, with 151 contract-state exits against just 58 entries. The average state-level star still ticked up slightly, from 3.80 to 3.83.

The state map is where it gets real. New Hampshire jumped the most, up 0.44 of a star, with Wyoming, Montana, Washington, Idaho, and Vermont close behind. On the other side, Rhode Island fell the hardest, with Delaware, Florida, Massachusetts, and Nevada all down about a quarter star. Florida losing nearly a quarter star matters more than New Hampshire gaining almost half of one, because far more members live there. The shelf shifted too: Mississippi, Alabama, and Virginia gained plan choices, while Ohio shed 1,238 plan-county offerings and Texas 1,015.

The forced-to-shop map

32%

of 2026 plans disrupted

1,006

terminated or non-renewed

19,004

county offerings lost to termination

Of the 8,048 plans offered in 2026, 2,603, about one in three, were terminated, consolidated, or pulled out of counties. One in eight, 1,006 plans, were terminated or non-renewed outright. At the county level, that is 19,004 offerings lost to termination, and another 5,100 dropped when plans reduced their service areas. Every one of those is a member who gets a letter in October saying the plan they had is changing.

And it is concentrated. The share of 2026 offerings disrupted runs highest in Nebraska at 60 percent, New Mexico at 55 percent, Minnesota and Colorado near 50 percent, and then Ohio and Texas, where the absolute numbers are largest. A flat national star average sits on top of all of that. The number barely moved. The ground did.

Where Stars has to go?

Stars has done real good. It gave the market one honest-looking signal in a world where most plans are free, and members cannot see much else. But let me be direct about where it falls short: many of the measures are not the right measures. Too many of them reward documentation, process, and administrative completeness rather than whether a member's life actually got better. A plan can score well on paper while a member still cannot get an appointment, still bounces between specialists with no one coordinating, still goes home from the hospital to no follow-up. The score and the lived experience of care drift apart, and when they do, the program measures the wrong thing with great precision.

CMS knows this, and to its credit it has started to evolve. It has removed and retired measures, many of them administrative or process measures that stopped telling us anything new, and it has reweighted toward outcomes. That is the right direction. But the instruments we rely on to capture members' own experiences are still blunt. A long-running, legitimate debate asks whether survey tools like HOS, the Health Outcomes Survey, and CAHPS, the member experience survey, measure true outcomes or capture memory, mood, and noise as much as care. When a plan's rating can swing on a survey with modest response rates and a long recall window, we should be honest that we are estimating quality, not measuring it.

So we have a lot of evolution still to make. My real hope, and my optimism, is that we get to genuinely good measurement in about five years, as NCQA and the broader system move to digital means of consuming clinical data. Digital quality measurement, pulling real clinical results through FHIR and the CMS data APIs rather than chasing charts eleven months later, is the path from did the plan document it to did the member actually get better. When the measure can see blood pressure, A1c, functional status, and follow-up in near real time, the score and the care finally start to describe the same thing.

There is a cost side to this too, and it is not small. The current program asks plans and providers to carry an enormous measurement and abstraction burden, much of it retrospective, much of it duplicative, and much of it invisible to the member. Every hour a nurse spends chasing a chart to close a gap on paper is an hour not spent with a patient. A better-designed program would measure more from existing data and ask people to document less, benefiting the member, the clinician, and the plan at the same time.

And that is the point of all of it. We are not doing this for a scoreboard. We are doing it for better quality of life, and for healthy aging and longevity for the people in these plans. If in five years a member's rating reflects whether she is actually living better, not whether a form was filed, then Stars will have grown into what the demonstration always hoped it could be. I have spent fifteen years believing it can. The next five are where it gets proven.

What actually moves the number

Once you know who the member is, the work is obvious, and almost none of it is analysis. For the plans and provider groups I work with, this is the heart of the program: access, meaning can the member get the appointment, the ride, the refill, the call back; clinical data, meaning wrangling supplemental and chart feeds to close the gap at the point of care rather than eleven months later; interoperability, meaning the FHIR and CMS API plumbing that turns a retrospective chart chase into a near-real-time member view; and the hard-to-reach member, disproportionately the dual-eligible and the SNP member with four conditions and no ride. No model closes that gap. People do. The analysis just tells you where to point.

What each of us does on Monday

• Health plan Presidents. Treat Stars as a P&L line by market, and say out loud which two or three markets have a half-star in reach this cycle and which do not. Invest where the membership is going: SNP and dual capabilities, member experience, and the data infrastructure under both. Show your team where you want to go in the next three to five years, while also focusing on financial discipline, doubling down on the mission, and, last but not least, empathizing with and recognizing the people doing the work.
• Quality VPs and Directors. Run a blameless retrospective on any measure that slipped. Sunset two or three low-ROI interventions and redeploy that capacity to the heavier clinical and experience measures. Map performance by market, not just by contract, and put your best field resources where the local cut is winnable. Clear is nice. I learned this the hard way. You need partners in every area within the plan, with providers and with vendors; partnership is key. Being clear with your partners pays dividends over a long time. Relentless, consistent execution, without burning out your team, can drive real progress.
• Consultants and solution providers. Bring a current-year, county-level gap view, not a post-mortem on a frozen year. Turn the CY2027 changes into a 90-day reprioritization, not a 200-slide deck. Put plan incentives into shared-savings language a CFO or Plan President acts on, and earn trust by naming what will not move the rating this year.
• Health systems and small practices. You are where Stars are won or lost. Document the eye exam, the depression screen, and the med reconciliation at the visit. Engage with the patient after the ED visit or discharge, before the plan's vendor does; it is the right move for quality and cost. Ask your plans for your own gap list and work it within your clinician's workflow. Invest in smoothing your clinician's workflow over the next year or two; clinical measure weighting in Stars is higher moving forward, and demand will increase to meet higher performance thresholds. Be ready.
• Stars and HEDIS analysts. Rebuild the dashboard on the 2027 weights, not last year's. Report gap-closure velocity by market, weekly, so leaders see momentum rather than a year-end rate. Put a responsible AI on the abstraction and reconciliation grunt work, and spend the time you save on root cause. Get ready to move your career to the next level: with AI-powered systems, analysis is going to get faster, and it already is, so grow into the judgment work the agents cannot do.
• Measure owners in pharmacy, operations, and clinical. Every measure needs one named owner who can recite its top three drivers from memory. Pharmacy owns adherence. Operations owns access and the appeals clock. Clinical owns point-of-care closure and getting documentation to the plan ahead of the claims lag.
• Investors and private equity. Underwrite the care model and the data infrastructure, not the rating headline. Give portfolio plans patient capital for root-cause fixes that take more than one measurement year to show up. The fastest way to destroy value here is to buy a rating and cut the operations that earned it.

The close, and what is coming next

Today's number reflects care delivered in 2025. It is done. The only question that matters is what we each do next, in our own markets, for the members behind it. Watch the scoreboard, but play the game more. The 35 million-plus people in this program deserve better care.

This is the first in a short series. Coming next from Vivega Health: a measure-by-measure look at what changed in the 2027 methodology and which measures are most winnable in 2027, and a deeper piece on digital quality measurement and what the move to FHIR-based clinical data will really take. If you want those as they land, the newsletter is the place.

Until then, with Clarity, Subbu Ramalingam.