Points of view from the team running the programs.

Three things happened at once. CMS finalized Tukey outer-fence outlier removal for cut-point calculations, meaning a handful of top plans no longer drag the whole distribution up. The Health Equity Index reward, replacing the old Reward Factor starting with 2027 Star Ratings, pays plans for closing gaps in LIS, dual-eligible, and disabled cohorts specifically, not aggregate performance. And the 2025 ratings landed with the average MA-PD contract at 4.04 Stars after a sharp drop to 3.92 the year before, a distribution that's tighter and less forgiving than it was three years ago.

What that means operationally: a program tuned to hit 4 Stars on 2022's math is landing at 3.5 on 2025's, and the gap is almost always in the cohorts the HEI now weights. We see the same three moves across plans that are holding position, attributing reward dollars to the measures CMS weights heaviest, running distinct outreach cadences for LIS/DE/disabled members (not one blended list), and closing the last 8–12% of gaps that legacy vendors write off as unreachable.

None of it costs more. It costs a rewards engine that can see the measure, the cohort, and the touch history in the same query.

By the time the continuous-enrollment unwinding wrapped in mid-2024, roughly 25 million Medicaid and CHIP members had been disenrolled and about 69% of those disenrollments were procedural, paperwork, not eligibility. KFF's final tracker put the procedural share above 70% in more than a dozen states. What the unwinding exposed wasn't a one-time event; it was the baseline churn dynamic that always existed and no one had modeled honestly.

The 2026 and 2027 program years make that dynamic permanent. Section 71107 of the Budget Reconciliation Act of 2025 (OBBBA, P.L. 119-21) requires states to redetermine eligibility for the Medicaid expansion population every six months, with implementation required by December 31, 2026. The Urban Institute (March 2026) projects the six-month cadence alone could reduce expansion enrollment by 2.0 to 3.1 million by 2028, with procedural disenrollments the dominant driver. That is on top of annual renewals now in effect for the rest of the 74 million Medicaid and CHIP population (CMS March 2026 Highlights).

A Medicaid member who churns in month 9 costs the plan twice: the acquisition cost that never amortized against premium, and the state-level auto-assignment scorecard hit that follows. At procedural-disenrollment rates in the low double digits, redetermination is a bigger driver of Medicaid margin in any given year than any single HEDIS measure.

The mechanics that actually work are unglamorous. Multi-channel reminders on the state's timeline, not the plan's. Reward for form submission, not for eligibility outcomes the state controls. A fallback human touch for members who miss two digital nudges. Language matched to the household, not the member, the person opening the envelope is often not the person on the card.

The teams that got this right during the unwind held retention meaningfully above book average. It wasn't a new vendor. It was treating the renewal form like a HEDIS gap.

Across the HSA/FSA/HRA books we've modeled, active card utilization follows a stubborn distribution: roughly 30% of members swipe monthly, 30% swipe quarterly, 30% swipe once or twice a year, and 10% never swipe at all. The industry's own data is consistent with the shape. EBRI's HSA database shows that only about 20% of accountholders take a distribution above a low threshold in a given year, and average balances keep growing not because members are strategic savers but because most of them aren't spending.

The unspent balance is where the administrator's interchange revenue and the plan sponsor's satisfaction scores quietly leak. Every dollar sitting on a card at year-end is a dollar the sponsor paid to fund a benefit the member didn't feel.

The interventions that shift the bottom two segments aren't education campaigns. They're a top-of-wallet moment (first-swipe reward inside 30 days of card issuance), an eligibility nudge tied to a real receipt ("your recent pharmacy visit qualified"), and a quarterly balance reminder that names the actual expiring dollars. A five-point lift in the bottom-40% cohort is almost always worth more than a ten-point lift at the top, the top is already sticky.

There's a fashionable reading of the behavioral-economics literature that says rewards are mostly symbolic, that the presence of a reward matters more than the size. It's not wrong for low-friction actions (log in, complete a survey). It falls apart for the actions we actually need members to take.

Volpp and colleagues' landmark NEJM smoking-cessation trial paid participants up to $750 and tripled quit rates at nine and twelve months. Halpern's follow-up trial in NEJM 2015 tested reward versus deposit-contract structures at similar dollar amounts and again produced double-digit lift over a control. The dollar level in these studies wasn't cosmetic, it was calibrated to the friction of the behavior.

What we see on plan books rhymes with the literature. Below roughly $15, the reward reads as insulting and the intervention underperforms baseline outreach without a reward. Between $15 and $22 the response is noisy and depends on channel, cohort, and copy. From $25 up to about $50, response stabilizes. Above $50, ROI degrades, the marginal completion isn't worth the incremental spend, and you start attracting reward-shopping behavior that hurts the next campaign.

The practical rule we give plan actuaries: pay $25 for the actions that require an appointment, $10 for the actions that don't, and never $5 for anything you actually want done. And structure the payout, deposit contracts, streaks, tiered payouts on completion, because the shape of the reward matters at least as much as the number on it.

In the 2021 Final Rule, CMS quadrupled the weight of patient-experience, complaints, and access measures. The 2023 Final Rule pulled that back to 2x for 2026 Star Ratings, but even at 2x these are the highest-weighted category on the sheet and the hardest to move quickly. A plan can close every open colonoscopy gap in Q4 and still lose a half-Star because members can't reach a nurse line inside three minutes.

The uncomfortable truth is that most of what shows up in CAHPS is upstream of the rewards program. It's IVR wait times. It's whether the appeals letter reads like a lawyer wrote it. It's whether the pharmacy call-back happens the same day. You cannot reward your way past a call center that treats members like tickets.

What rewards can do is fix the parts of member experience that get measured through action, not survey, the coordination touches around a new diagnosis, the follow-up after a hospital discharge, the reminder that a prior authorization is approved and the member should stop worrying. Those touches don't move a CAHPS question directly, but they change whether the member is annoyed on the day the survey arrives.

Medicare Advantage's expansion of non-medical supplemental benefits: SSBCI, VBID flex allowances, food-and-produce cards, transportation, OTC, is the single largest bet the industry has made on member engagement in a decade. It's also the least measured. The HHS Office of Inspector General reported in 2023 that MA plans lacked the data to demonstrate whether these benefits were actually being used, and CMS has since tightened marketing rules specifically because of the gap between what members were promised and what they redeemed.

In the books we've modeled, non-medical benefit utilization looks a lot like CDH card utilization: a small stickiness at the top, a broad middle of members who used the card once and forgot, and a long tail of zero-swipe members. Plans are funding a benefit whose value proposition, reduced medical spend downstream, depends on members actually redeeming it, and then not measuring whether they did.

The mechanics that lift utilization are the same three we use on CDH cards, tuned for the population. First-swipe reward inside the first 30 days. Merchant-specific nudges ("your card works at this Kroger") rather than category-level education. A quarterly reminder that names the dollars, in the language the household actually speaks. It isn't complicated; it's just unglamorous and requires infrastructure most plans don't own.

The National Assessment of Adult Literacy, still the most cited federal measurement, found that 36% of U.S. adults have basic or below-basic health literacy. The AHRQ has spent two decades documenting the downstream cost: worse adherence, more ED visits, worse outcomes on exactly the measures Stars and Medicaid quality programs weight.

Health-plan outreach is written by people who read at a college level and reviewed by lawyers who read higher than that. The result is copy that averages a 9th- to 11th-grade reading level, targeted at populations that skew several grades lower. We've seen SMS response rates jump 40–60% from a rewrite that changed nothing about the offer, just moved the reading level from 10th grade to 5th, shortened sentences, replaced clinical nouns with everyday nouns, and put the action verb first.

It is the single highest-ROI change most rewards programs never make. It also gets harder, not easier, once your language stack expands past English and Spanish, which is why the plans that translate first from English into plain English, and only then into other languages, get materially better response than plans that translate the legal version.

The Medicare Shared Savings Program covered about 11 million assigned beneficiaries in 2024 across 480 ACOs, a real number, but a small fraction of the roughly 34 million Medicare Advantage lives and the fee-for-service Medicare book the ACOs sit inside. For any given health plan, some meaningful share of the network is not attributed to a value-based arrangement, and the gaps in that share are exactly where Stars performance decays.

The temptation is to wait for the network to consolidate into risk. It won't, at least not on the timeline Stars operates on. What actually closes gaps at non-integrated FFS practices is a worklist that fits the front-desk staff's day, a per-closure incentive that clears in weeks not quarters, and a data return that arrives before the next visit, not a portal login the practice has to remember.

The plans getting real lift here have stopped treating provider engagement as a separate program from member engagement. They're the same gap, closed from two directions, on the same infrastructure.

Every rewards program in the industry reports engagement. Open rates. Click rates. Portal logins. Cards activated. It is, almost without exception, the wrong denominator. A campaign can double its open rate and not move a single HEDIS gap. A member can log in every week for a year and never fill their statin.

The measurement that matters is attributed closure: the specific gap, closed inside the specific member, inside the window the measure counts, traceable back to a specific touch. Everything else is proxy work, sometimes correlated, often not.

The reason most programs report engagement instead is that attributed closure is harder to instrument. It requires reward-side data (who got what touch, when) joined to claims-side data (what closed, when) on a cadence claims don't naturally support. The plans that solved it built the join once, on infrastructure they owned, and stopped arguing about vendor attribution in every QBR.

Behavioral economists Hengchen Dai, Katy Milkman, and Jason Riis found that people are measurably more likely to start a goal-oriented behavior right after a temporal landmark: a Monday, a birthday, the first of a month, a new year. They named it the fresh start effect, and the pattern held across gym attendance, diet logging, and goal-related searches. The mechanism is mental accounting. A landmark lets people wall off a past, less-than-perfect self from a present, aspirational one, and start clean.

We see the same shape in member outreach. A screening reminder sent the Monday after a member's birthday outperforms the same reminder sent on a random Wednesday, by a margin that surprises plan teams who have never tested it. The same nudge sent the first week of the plan year outperforms the same nudge sent mid-Q2. It costs nothing to align the calendar. What it costs is a rewards engine that can schedule a touch to a date, not just a cadence.

The practical move is to anchor outreach to the member's landmarks, not the plan's. Birthday, plan-year anniversary, the month after a qualifying life event, the first Monday of the quarter. The fresh start window is short. If the touch lands outside it, the lift is gone.

Decades of behavioral research show that forming an intention, wanting to quit smoking, get a screening, fill a prescription, explains only a fraction of whether someone actually does it. Paschal Sheeran's review of the intention-behavior literature put the correlation at a medium effect; intentions alone leave most of the variance in behavior unexplained. Peter Gollwitzer's work on implementation intentions showed that converting a vague goal into a specific when-and-where plan roughly doubles follow-through, with a meta-analytic effect size in the medium-to-large range.

Most health-plan outreach stops at the intention. 'It is time for your annual wellness visit.' 'You are due for a colorectal cancer screening.' That copy creates an intention and leaves the action to chance. The lift from adding the when-and-where is one of the largest we see per dollar spent, because it is free and it attacks the exact gap the literature says is widest.

The operational version is not 'book your screening.' It is 'Tap here to book at your nearest in-network site on the next available Tuesday.' The reward engine should be able to resolve the member's nearest site, the next available slot, and the channel they actually answer, then write that specificity into the nudge.

Standard economics says a member offered a free, fully covered preventive visit with a clear health benefit should take it. They mostly don't. The Medicare Annual Wellness Visit is fully covered with zero cost-sharing, and still a large share of eligible members never schedule one. Behavioral economics calls this present bias. The cost of acting now, making a call, taking a morning off, is concrete and immediate, while the benefit is abstract and deferred. Ted O'Donoghue and Matthew Rabin formalized this as hyperbolic discounting, and it is the best single explanation for why 'covered' and 'used' are different words.

The fix that works is not more education about why prevention matters. It is removing the action cost and making the healthy choice the default. Richard Thaler and Cass Sunstein's nudge framework, and the work of the Penn Medicine Nudge Unit under Mitesh Patel and Kevin Volpp, shows that defaults move behavior more than any amount of persuasive copy. Opt-out rather than opt-in. Pre-scheduled rather than 'call to book.' Auto-assigned rather than 'select your preference.' When a clinician order set defaults to the evidence-based option, adherence jumps, and the same physics applies to member outreach.

For a health plan, the implication is structural. Stop asking members to opt into their own prevention. Pre-schedule the wellness visit and let them cancel. Pre-select the closest in-network pharmacy for a 90-day fill and let them change it. Default reward enrollment to 'on' with a one-tap exit. Every opt-in you convert to an opt-out is a gap that closes itself.

The behavior-change literature converged on a finding most rewards programs ignore. What predicts whether a member keeps doing the healthy thing six months later is not the reward they got. It is whether the behavior crossed into habit. Wendy Wood and colleagues' experience-sampling work put a large share of daily behavior in the habitual range, performed repeatedly in the same context with little conscious deliberation. The implication for a health plan is that the goal of a single intervention is not the completion. It is the repetition that makes the next completion automatic.

The mechanism is context-cueing. A behavior becomes habitual when it is performed consistently in the same situation, after the same trigger, in the same place, at the same time. A rewards program that pays for a one-time colonoscopy moves one gap. A program that pairs a quarterly refill reminder with the same channel, the same day, the same reward, builds a cue that survives the campaign. The third swipe of a benefit card, the third quarterly refill, the third walking-reward redemption, is the inflection point where completion stops costing willpower.

The operational lesson: design for the repeat, not the event. Streak mechanics, schedule-anchored touches, and a reward structure that pays the second and third completion more than the first are how you convert a campaign into a habit. It is slower, and it is the only thing that holds once the reward budget tightens.