The Three Card Monte of CMS Stars: Mastering Variable Change

You picked your card in January. The board changed in September. And you are still pointing at the same spot. Here is why the math says that is the most expensive mistake in Medicare Advantage.

---

The Setup: A Fair Game You Still Lose

Picture a busy sidewalk. A folding table. Three cards face down: two black Jokers and one red Queen. The dealer moves them, stops, and looks at you. "Find the Queen, win the green."

In real life, Three Card Monte is rigged. The dealer palms the card. The "winners" in the crowd are plants. The whole thing is theater designed to separate you from your twenty dollars.

But forget the hustle for a moment. Imagine a perfectly fair version of this game. No sleight of hand. No confederates. Pure probability. Even in this fair game, most people still lose. And they lose for one specific reason: they do not understand what happens to probability when new information enters the system.

The Math That Doubles Your Odds

Three cards, face down. You point to the one on the left. Your odds of being right? One in three. A 33.3% chance. The remaining 66.7% probability sits across the other two cards.

The dealer, who knows exactly where the Queen is, flips one of the other two cards. Black Joker. Now he asks: "Do you want to stay, or do you want to switch?"

This is where the room always splits. I have done this exercise with quality professionals, data directors, CMOs. Smart, experienced people. The majority say they are staying. Because it feels like a 50/50 now. Two cards left. Coin flip. Why move?

If you switch, your odds of winning literally double.

When you made your initial pick, you locked in a 33.3% probability. That number froze the moment your finger touched the card. The "other group" of two cards held 66.7%. When the dealer reveals a Joker from that group, the 66.7% does not evaporate. It condenses entirely onto the one remaining card you did not choose.

Stay: one in three. Switch: two in three. Same game. New information. Completely different odds.

In statistics, this is called conditional probability, sometimes called Bayesian updating. The core principle: you must update your strategy the moment new information becomes available. A static strategy inside a dynamic environment will fail. Every single time.

★ Signal Over Noise Variable change is not wishy-washy strategy. It is not indecision. It is the mathematically disciplined act of updating your assumptions when the board changes. Which is exactly what every healthcare quality leader should be doing, and almost none of them are.

Welcome to the CMS Stars Table

Now leave the sidewalk and walk into the executive conference room of a Medicare Advantage health plan.

CMS grades Medicare Advantage plans on a 1 to 5 Star scale across dozens of measures: diabetes control, cancer screenings, patient experience, customer service. Reach 4 Stars or above, and CMS hands you a Quality Bonus Payment. We are talking tens of millions, sometimes hundreds of millions of dollars. Drop below 3.5? No bonus. Margins shrink. Members leave. The spiral is difficult to reverse.

Here is exactly where CMS Stars maps to our card table: CMS grades on a curve. They use cut points, performance thresholds you have to cross to earn a specific star level on each measure. But they do not tell you the cut points in advance. They evaluate national performance, then set the curve retroactively.

You are placing your bets without knowing what the winning hand looks like until the game is already over.

How This Plays Out on the Ground

It is January. Your quality team sits down for annual planning. Your CMO says: "This year, we are going all in on Hemoglobin A1c Control. If we invest three million dollars in outreach, free testing kits, and provider incentives, we can move from 70% to 75%. That should get us from 3 Stars to 4 Stars on that measure."

You lock in the bet. Contracts signed. Vendor on board.

September arrives. Benchmarking data starts flowing. Across the industry, A1c control improved substantially. A new class of diabetes medications hit the market. A national pharmacy adherence campaign moved the needle for everyone. The 4-Star cut point is no longer trending toward 72%. It is projected to land at 78% or higher.

Your initiative is working. You will hit 75%. And it will not be enough. Three million dollars spent. Still 3 Stars.

The dealer just flipped a card.

This is where nearly every health plan makes the same mistake. They say: "We have already invested the money. We just need to push harder." They call it "staying the course." The math calls it the sunk cost fallacy.

"What a quality leader who understands variable change would do is scan the remaining board and execute the switch. Not because the initial strategy was wrong. But because a higher probability of success now exists somewhere else."

You scan the other measures. Medication Adherence for Statins has stagnated nationally. The 4-Star cut point is holding steady around 83%. You are sitting at 82%. One percentage point from a full star improvement.

You redirect remaining Q4 budget and call center capacity to targeted pharmacy outreach. Members who have not refilled their statin prescriptions in 90 days get home delivery offers. You move from 82% to 83.5%. You earn 4 Stars on Statin Adherence.

That is variable change applied to healthcare quality. Not because it felt good. Because the math demanded it.

Bayesian Updating at the Patient Level

Conditional probability does not just apply to your 30,000-foot measure strategy. It applies to every individual patient interaction in your care management infrastructure.

A 72-year-old patient, Mr. Henderson, is discharged after a heart failure exacerbation. Your predictive model assigns a 15% readmission risk. Standard protocol: automated text on day three, phone call on day seven.

Day eight: Mr. Henderson was a no-show for his 7-day follow-up appointment. The dealer just flipped a card.

If your care management platform is static, nothing changes. The system logs the no-show. Mr. Henderson stays in the automated queue and gets another text message.

But run the conditional probability. A missed 7-day follow-up in heart failure patients spikes readmission risk to somewhere between 55% and 70%. That no-show is not a scheduling inconvenience. It is a loud, flashing clinical signal.

You execute the switch. Pull him out of the automated queue. Trigger a high-touch intervention. A community health worker discovers his car is in the shop and he never picked up his diuretic from the pharmacy. You arrange transportation. You get him the medication.

You just prevented a $25,000 readmission. You protected your Stars measure. You improved a real patient's quality of life. The entire chain of events was enabled by one thing: recognizing the variable change and being willing to act on it.

★ Precision Is Patient Care A data signal acted on in time is a patient protected. This is why real-time leading indicators are not a nice-to-have. They are a clinical imperative.

When the Game Itself Changes: Weights, the HEI, and the Final Rule

There is a version of variable change even more disorienting than a shift in cut points. It is when CMS changes the payout structure for the entire game.

Not all Stars measures carry the same weight. A measure weighted at 5 has five times the impact of a measure weighted at 1. For years, clinical HEDIS measures dominated the Stars framework. Plans built enormous infrastructures around closing clinical gaps.

Then CMS elevated the weight of CAHPS, the patient experience survey, so substantially that experience and administrative measures started accounting for more than half of a plan's overall rating. Plans that recognized this variable change and pivoted to member experience programs maintained their ratings. Plans that insisted they were "a clinical organization" and refused to redirect resources? Some dropped from 4.5 Stars to 3 Stars. That kind of revenue loss is catastrophic.

CMS Is Doing It Again. Right Now.

In 2023, CMS finalized the Health Equity Index (HEI) as a replacement for the legacy reward factor, beginning with 2027 Star Ratings. The message: average performance is no longer sufficient. CMS would evaluate whether plans are caring equitably for dual-eligible members, low-income subsidy recipients, and populations facing social risk factors.

Plans spent two years investing in that reality. They stratified data by social risk factors. They restructured outreach programs. In April 2025, CMS rebranded the HEI to "Excellent Health Outcomes for All" (EHO4All), signaling this was a priority.

Then in November 2025, the proposed rule for Contract Year 2027 said CMS would not implement the HEI. Not implement EHO4All. They proposed keeping the historical reward factor.

Two years of finalized policy. Billions of dollars in strategic investment across the industry. And a proposed rule that would pull it off the table.

As of this writing, the final rule has not been published. The comment period closed in January 2026. The final rule is expected any day. Which means every plan in the country is staring at a card mid-flip.

Some plans are going to freeze. They will say: "We do not know what CMS is going to do, so we are going to wait." That is the most natural reaction. And it is the wrong one.

If the final rule removes EHO4All, and you spent two years improving care for vulnerable populations, did you waste your money? No. Those members still exist in your denominators. Their improved outcomes still roll up into your overall measure scores. That investment is never wasted.

But if you froze, paused all health equity work, and the final rule keeps EHO4All? You are two years behind. That is a hole you cannot dig out of in a single measurement year.

A Bayesian thinker hedges. You continue the work that produces value regardless of which card gets flipped. And you build the data infrastructure to pivot your strategy within 30 days of the final rule dropping, because that is the window where competitive advantage is won or lost.

Three Principles for Building an Organization That Can Switch

Knowing the theory is one thing. Living it inside a healthcare organization is something else entirely. I have spent 22 years watching brilliant quality professionals lose millions of dollars, not because they lacked intelligence, but because they were operating inside organizations that were structurally incapable of executing a variable change.

Principle 1: Treat Your January Strategy as a Starting Position, Not a Commitment

Your initial bet was correct given the available information. What would be a failure of leadership is refusing to update that bet when new information proves the original hypothesis wrong. Build a culture where a VP of Quality can walk into a steering committee in July and recommend redirecting Q3 budget to a higher-probability measure, and be celebrated for it, not forced to defend herself.

Principle 2: Invest in Leading Indicators. You Cannot Update Probability Without New Data.

If you are only doing retrospective analysis at the end of the measurement year, you are watching a replay of a game that already happened. You need mid-year mock CAHPS surveys, pharmacy adherence feeds, monthly HEDIS-equivalent tracking, and predictive cut point modeling that shows where the national curve is trending before CMS publishes the final numbers. Right now, with the EHO4All final rule pending, you need scenario models for both outcomes.

Principle 3: Take the Emotion Out. Follow the Math.

Quality initiatives carry enormous political and emotional weight. Vendor relationships. Board presentations. People's careers. The math does not care about any of that. When the data shows that a higher probability of success exists somewhere else, the only rational response is to switch. Not because you were wrong. Because a better bet now exists, and your job is to be where the probability is highest.

The Card Is in the Air

Value-based care is the hardest game in American healthcare. The cut points move. The weights change. CMS introduces new frameworks, renames them, and proposes to remove them entirely, all within a 24-month window.

But it is not rigged. It is math. The cut points follow statistical distributions. The weights are published in advance. Every variable change CMS has made over the last decade has been telegraphed for anyone paying close enough attention to the data.

The problem is not a lack of information. The problem is the organizational refusal to act on information that contradicts the initial bet.

So the next time you are sitting in a steering committee and someone insists you just need to "push harder" on a failing initiative, or that you should "wait and see" before making any moves, think about our table on the sidewalk. Remember the dealer, the Queen, and the Joker.

Ask one question: Are we staying with this because the math still supports it, or because we are too proud, too political, or too comfortable to admit the board has changed?

The card is in the air right now. It will land. And when it does, the plans that move first will define the next five years of Medicare Advantage quality.

Be willing to change the variable. Be willing to make the switch.

★ Translate to Win When your quality team, analytics team, and clinical team all understand the same math; when they can all see the board changing in real time and speak the same language about what to do next; the pivot is no longer political. It becomes obvious. That is the organizational infrastructure worth building.

Topics Covered: Variable Change, Conditional Probability, CMS Cut Points, Mid-Year Pivot Strategy, Bayesian Updating, Patient-Level Care Management, Stars Measure Weighting, Health Equity Index, EHO4All, Dynamic Quality Strategy

ClearStars.AI gives your quality team the intelligence layer to track cut point trends, identify bubble measures, and execute strategic pivots before it is too late. Deterministic AI. No black boxes. Signal over noise.

© 2026 ClearStars.AI ★ Bert Rico, Founder & CEO ★ Houston, TX