← Decision Lab case study

Decision Lab — evidence base

The cited ground under the Michael Chen scenario. Every figure pulled from a primary source, dated, and sorted into what survives scrutiny, what needs qualifying, and what must not be claimed.

Version 2.0 Pulled & verified August 5, 202611 claims discarded Fictional client — Bellwether Health Plan

1The finding that carries the whole piece

One sentence from KFF does more work than any other line here. Read it before the rest.

Plans offer supplemental benefits and cannot see whether anyone uses them Verified

KFF, Medicare Advantage in 2026: Premiums, Out-of-Pocket Limits, Supplemental Benefits, and Prior Authorization. kff.org

“It is not known what share of enrollees have used these benefits because data are not yet available.”

And the trend underneath it: transportation benefit access for individual Medicare Advantage enrollees fell to 22% in 2026, down from 28% in 2025. Special needs plans run far higher at 73%. Bathroom safety devices dropped from 32% to 21%; over-the-counter benefits from 79% to 68%.

This is the argument for the entire piece, made by someone other than you. Plans are cutting supplemental benefits whose utilization they cannot measure. A benefit nobody can see being used is a benefit that looks like pure cost at renewal. That is a measurement problem producing a benefits decision — and it is exactly the shape of the case you are making about care coordination. If a plan cannot see whether its transportation benefit gets used, it certainly cannot see whether a coordinator noticed the member who needed it.

And it lands directly on the scenario

Michael Chen declined the transportation benefit eight months ago, when he did not yet need it. Nobody re-offered. That is not a contrivance — it is what a benefit with no utilization visibility looks like from the member's side.

2Star Ratings and where the money actually sits

Part D medication adherence measures carry weight 3 Verified

Centers for Medicare & Medicaid Services, 2026 Star Ratings Measures and Weights. cms.gov

Adherence for diabetes medications · hypertension (RAS antagonists) · cholesterol (statins) — all weight 3, classified as intermediate outcome measures

In the same cycle, CMS reduced the weight of Patients' Experience and Complaints from 4 to 2, beginning with the 2026 Star Ratings. Weights across the measure set range from 1 to 5.

This is why a late refill is not a soft signal. It is a triple-weighted measure tied to plan revenue, and CMS has just shifted emphasis away from how members feel and toward whether they actually took the medication. A coordinator who spots an access barrier behind an adherence gap is working directly on a weight-3 measure. Say it that way to a Stars director and the conversation changes.

3Transportation is an access problem, not a compliance problem

Adults without a vehicle and with poor transit forgo care at four times the rate Qualify

Urban Institute, More than One in Five Adults with Limited Public Transit Access Forgo Health Care Because of Transportation Barriers (April 2023). Health Reform Monitoring Survey, fielded June 2022. urban.org

GroupForwent needed care over transportation
All adults surveyed5%
No household vehicle access13%
No vehicle and fair/poor transit access21%
No vehicle and good/excellent transit access9%
Adults with a disability17% (vs 2% without)
At or below 138% FPL14% (vs 1% at 400%+ FPL)

⚠ Qualify this one hard, and do it in the copy. The survey covers nonelderly adults, ages 18–64. It is not a Medicare-population figure, and presenting it as one is precisely the sleight of hand a quality director would catch. Use it for the mechanism — losing a vehicle roughly triples the rate of forgone care, and poor transit doubles it again — and say the age range out loud. The disability and income gradients are the strongest part and they generalize more comfortably than the headline number does.

4Language access is a legal obligation, not a courtesy

Relying on a family member to interpret is prohibited Verified

National Health Law Program, What is Required Under Title VI and Section 1557 to Ensure Language Access (December 2025 update). healthlaw.org

Covered entities — including insurers, Medicare, and qualified health plans — must take reasonable steps to provide meaningful access for people with limited English proficiency, free of charge, accurately, and timely. A qualified interpreter must be proficient in both languages, interpret effectively and impartially without omission or addition, and follow accepted interpreter ethics.

The regulations explicitly prohibit relying on “an adult not qualified to interpret” or on minors, with narrow exceptions for emergencies or where the LEP individual specifically requests an accompanying adult and that is documented as appropriate.

This turns a texture detail into a finding. The scenario records that Michael Chen brings his daughter to appointments to translate, and that his cardiology consult was scheduled with no interpreter booked despite a preference on file. Under Title VI and Section 1557, that is not a nice-to-have the plan missed — it is a compliance exposure, and the coordinator who notices it is doing regulatory work as well as clinical work. It also tightens the plot: the daughter is simultaneously the transport, the interpreter, and the entire support system.

5Medication nonadherence — cost, and what actually causes it

Every figure below was opened and read in the source. Four citations offered for this section were discarded because the numbers attached to them were not in the papers.

Per-person annual cost of nonadherence, in 2024 dollars Verified

Achterbosch M, Aksoy N, Obeng GD, Ameyaw D, Ágh T, van Boven JFM. Clinical and economic consequences of medication nonadherence: a review of systematic reviews. Front Pharmacol. 2025 Jun 25;16:1570359. PMID 40635744. PMC12237677

43 systematic reviews · 430+ studies on clinical outcomes · 174 on economic outcomes

ConditionMean annual cost per person (SD), 2024 US$
Diabetes$8,327 ($2,335)
Respiratory disease$8,584 ($469)
Cardiovascular disease$12,146 ($5,320)
Mental health conditions$14,585 ($5,315)
Gastrointestinal disease$30,771 ($8,270)
Osteoporosis$43,372 ($14,266)

Use this instead of any national aggregate. A plan can multiply a per-person figure by its own non-adherent panel. It cannot do anything with a contested $100–300 billion number — and that number, as it happens, traces back through this review to Cambridge 2009 and Senst 2001. It is nobody's primary finding. ⚠ This article carries a correction notice (Front Pharmacol. 2026 Mar 12;17:1822926); check it before quoting a figure.

Two findings in that review that complicate the pitch — include them anyway

1. In older adults, nonadherence drives hospitalization but not ED visits. The aging-population review inside it (Walsh et al., 2019) found all-cause hospitalization OR 1.17 (95% CI 1.12–1.21), p<0.0001 — but ED visits at OR 1.05 (95% CI 0.90–1.22), p = 0.566, not significant. If your measurement plan leans on avoidable ED utilization, lean on hospitalization instead.

2. Total healthcare cost impact is genuinely mixed. The review reports it "mostly varying between increased costs and no significant change," because lower drug spend partly offsets higher medical spend. Saying so out loud is worth more than a clean number a quality director will already distrust.

Cost-related nonadherence in older adults, and what predicts it Verified

Briesacher BA, Gurwitz JH, Soumerai SB. Patients at-risk for cost-related medication nonadherence: a review of the literature. J Gen Intern Med. 2007;22(6):864–871. PMID 17410403. PMC2219866 · 19 independent studies.

FindingFigure
Medicare beneficiaries reporting any cost-related nonadherence26% · 37% without drug coverage (n = 17,685)
By coverage source, chronically ill 50+ (n = 4,055)VA 12% · private 15% · Medicaid 25%
Out-of-pocket >$100/month vs <$50OR 5.5 (95% CI 4.0–7.7)
Income <$20,000 vs ≥$50,000OR 4.2 (95% CI 2.2–7.8)
Symptoms of depressionOR 2.4 (95% CI 1.3–4.5)
Lowest-quality physician relationshipOR 2.1 (95% CI 1.4–3.2)
Older ageDecreases risk — 11 of 13 studies
PolypharmacyNo relationship — 5 of 5 studies

The physician-relationship row is the one that matters to you. A trusted relationship roughly halves cost-related nonadherence risk, independent of income and drug costs. That is the closest thing in this literature to evidence that the human in the loop changes the outcome — which is the entire premise of training a care coordinator. The two counterintuitive rows (older age protects; polypharmacy does not predict) are worth citing precisely because they show you read the paper rather than the abstract.

⚠ Do not cite Briesacher for "32%." That figure appears in its opening line citing three other surveys, and it is not this review's finding.

Current prevalence, and the subsidy effect Verified

Dusetzina SB et al. Cost-Related Medication Nonadherence and Desire for Medication Cost Information Among Adults Aged 65 Years and Older in the US in 2022. JAMA Netw Open. May 18, 2023. · Loh FE, Stuart BC, Hunt RJ, Negari M, McRae J. How lack of subsidized prescription drug coverage affects medication use by low-income Medicare beneficiaries. Health Aff Sch. 2025;4(5):qxaf186. PMID 42093841.

≈1 in 5 (20%) of adults 65+ reported cost-related nonadherence in 2022 · national panel, n = 2,000

Loh et al. (2025) add the coverage mechanism: low-income-subsidy recipients averaged 49 prescriptions annually versus 25 for subsidy-eligible non-enrollees, and were 10% less likely to report cost-related nonadherence.

Together these give you a current prevalence figure and a demonstrated lever. ⚠ Confirm Dusetzina's article number against the paper — the number circulating for it does not reconcile with its DOI.

6Transportation — what the primary sources actually say

How older adults actually get to appointments Qualify

Suntai Z, Kubanga K, Adanu E, Lidbe A. Modes of Transportation to Medical and Primary Care Among Older Adults. Innov Aging. 2021;5(Suppl 1):128. doi:10.1093/geroni/igab046.490. PMC8682120 · NHATS 2018, adults 65+.

70.0% drive themselves · 34.8% rely on family, friend or paid person · 2.4% home visit · 2.1% public transit · 1.5% walk · 1.1% taxi

⚠ Two qualifications, both mandatory. These sum to 111.9%, so respondents reported more than one mode — they are not exclusive shares and must never be written as though they partition the population. And this is a conference abstract in a supplement, not a peer-reviewed article. Cite it as such or not at all. It is still the best older-adult modal breakdown available, and the substantive point survives: roughly a third of people over 65 depend on another human being to reach care.

Who actually cancels a medical transport trip Verified

MacLeod KE, Ragland DR, Prohaska TR, Smith ML, Irmiter C, Satariano WA. Missed or Delayed Medical Care Appointments by Older Users of Nonemergency Medical Transportation. The Gerontologist. 2014;55(6):1026–1037. PMID 24558264. PMC4668763

125,913 trips · 2,913 older adult Medicaid clients · Delaware

Over half of canceled trips were attributed to client reasons — no-show, refusal — rather than to the transport provider. Regularly scheduled trips were significantly less likely to be canceled.

This one is more useful than the version that was circulating, and it points the opposite way. A widely repeated claim reverses this — provider no-shows causing missed care. The paper says the client side dominates, which reframes the problem from vendor performance to engagement, scheduling friction, and whether anyone knows the ride is needed. And "regularly scheduled trips are less likely to be canceled" is a concrete operational finding a plan can act on. ⚠ This paper is frequently miscited as Silver, Blustein & Weitzman (2012), The Gerontologist 52(5):663–673. Those authors, that year, that volume, and that PMID all belong to something else.

The review that frames it Verified

Syed ST, Gerber BS, Sharp LK. Traveling Towards Disease: Transportation Barriers to Health Care Access. J Community Health. 2013;38(5):976–993. PMID 23543372. PMC4265215 · 61 US studies.

Transportation barriers lead to rescheduled and missed appointments, delayed care, and delayed medication use, disproportionately among lower-income and under/uninsured populations. The frequently quoted 3.6 million Americans figure is Wallace et al., cited inside this review — attribute it that way or cite Wallace directly.

⚠ The "3.0% to 7.7% of community-dwelling older adults" figure attached to this paper in circulation does not appear anywhere in it. Dropped.

7Do not claim

Hard limits
  • Do not present the Urban Institute figures as Medicare-population data. Ages 18–64. If you need an older-adult figure, it has to be pulled separately and cited separately.
  • No ROI number. Star Ratings and avoidable ED utilization give you real denominators to model against. Label every input as modeled, exactly as the cohort dashboard does.
  • Do not claim this training reduces ED visits, readmissions, or improves a Star measure. It has not been deployed. Kirkpatrick L3 and L4 are a measurement plan, not a result.
  • Do not imply Bellwether reflects any real plan's practices. It is fictional, and the fiction boundary needs to be visible on the piece itself, not buried in a footer.
  • No statistic reproduced from a vendor blog or a news summary. Everything above came from CMS, KFF, Urban Institute, or NHeLP directly. Hold the line.
Discarded in verification — do not reinstate

Every item below was offered as a citation, opened, and found not to say what was claimed. They are listed so they do not come back.

  • "$100–290 billion annually (Cutler 2018)" — appears in Cutler's introduction citing reference 6. Not Cutler's finding.
  • "10% of hospitalizations in older adults / $2,000 per patient per year (Cutler 2018)" — Cutler's references 10 and 11.
  • "32% of older adults (Briesacher 2007)" — opening line, citing three other surveys.
  • "Briesacher found an 11–29% range" — no such range exists in the paper.
  • "3.0% to 7.7% of community-dwelling older adults (Syed 2013)" — not in the paper.
  • "Silver, Blustein & Weitzman (2012), The Gerontologist 52(5):663–673" — wrong authors, year, volume, pages and PMID. The paper is MacLeod et al. (2014).
  • "125,913 trips, New York City" — Delaware.
  • "12.6% of NEMT users missed appointments" — not in the paper.
  • "Over half of cancellations were provider no-shows" — inverted. Over half were client reasons.
  • "Silver, H.J. & Williams, I.C. (2021), Journal of Applied Gerontology 40(12):1735–1744" — the data is real; that citation is not. It is Suntai et al., Innov Aging.
  • "Hung et al. (2026), JAMA Internal Medicine" — no such paper. PubMed returns zero results.

8Gap statement

The problem statement the piece is built on. Every claim traces to §1–§6.

Gap statement v1.0 — cited

Medicare Advantage plans are measured on whether members take their medication: the three Part D adherence measures each carry weight 3 in the 2026 Star Ratings, while CMS cut the weight of patient experience from 4 to 2 in the same cycle (CMS, 2026). Adherence gaps are therefore revenue events, not administrative ones.

But an adherence gap is a symptom with several possible causes, and the systems that detect it are built to see only one. A late refill is coded as non-adherence — a member choosing not to take a drug — when it may be a member who can no longer get to the pharmacy. Losing vehicle access roughly triples the rate at which adults forgo needed care, and poor transit roughly doubles it again (Urban Institute, 2023; ages 18–64).

The plan's own remedies are going the wrong way. Transportation benefit access for individual MA enrollees fell from 28% to 22% between 2025 and 2026 — and KFF notes plainly that “it is not known what share of enrollees have used these benefits because data are not yet available” (KFF, 2026). Plans are cutting benefits whose utilization they cannot see. Meanwhile, language access is a standing legal obligation under Title VI and Section 1557 — qualified interpreters, free of charge, with family members explicitly prohibited from interpreting (NHeLP, 2025) — and it lives in yet another system again.

The frontline care coordinator sits at the one point where these records converge, and is the least equipped person in the organization to read them together: non-clinical, high-volume, high-turnover, and trained by annual click-through.

The gap is not that coordinators do not care. It is that the signal is distributed across four systems that each see one symptom, no workflow joins them, and nothing the coordinator does or fails to do is currently measured.

Design implication

The intervention has to do two things at once: rehearse the join under realistic conditions, and emit data about whether it happened. The second half is not decoration — it is the direct answer to the sentence KFF wrote about benefit utilization.

9Sources

Before publishing

Re-check the KFF supplemental benefits page and the CMS measure list each plan year — both move annually, and the whole lead argument depends on a year-over-year comparison that will need updating. Section 1557 has been subject to repeated rulemaking; verify the explainer is still current at publication.