Paid in Full, Underpaid Anyway

Executive Summary

In June 2026, MACPAC told Congress something remarkable in its plainness: after a decade in which managed care became the dominant Medicaid delivery model, "little is known about the accountability tools state Medicaid agencies use" to ensure managed care plans comply with their contracts and meet performance expectations. In the same reporting year, the federal government's own audit instrument, PERM, measured the improper payment rate in Medicaid managed care at exactly 0.00 percent.

Both statements are true. Together they describe a system that has stopped looking.

This white paper reports what we found when we did look, from inside the revenue cycle of two Kansas hospitals: a 25-bed rural emergency hospital and a physician-owned surgical hospital, 29,600 claim lines and $82.6 million in billed charges between them. It is written first for hospital leaders, because the dollars at stake are theirs. It is also written for the people who run state Medicaid agencies, because what we found is not primarily a story about bad actors. It is a story about a visibility gap that harms hospitals, misleads agencies, and hides in the one place nobody audits: claims that were paid.

The paper makes four arguments:

1. The most expensive claims problem in Medicaid managed care is not denial. It is the claim marked "paid in full." We document ER physician claims where the remittance shows payment at 100 percent of the allowed amount, while the allowed amount itself sits at roughly a quarter of the state fee schedule floor, identically, across two different managed care organizations. No denial report, no aging report, and no EOB review will ever surface this. Only independent repricing does.
2. Hospital leaders are managing to reports that structurally cannot show the loss. At one facility, Medicaid claims collected at 4.2 percent of charges over twelve months while Medicare fee-for-service collected 20.3 percent, on the same systems, with the same staff. Observed recoverable leakage across our two facilities runs 3 to 15 percent of annual collections, and most of it never appears in any denial queue.
3. The oversight system cannot see it either, and the agencies know it. PERM audits the state's capitation payment to the plan and stops there; the plan's adjudication behavior is invisible to it, which is how a 0.00 percent managed care error rate coexists with the patterns above. MACPAC has now said this to Congress. The gap is not a secret. What has been missing is an evidence stream.
4. The fix already exists, and it is cheaper than everyone thinks. The same claim-level dataset a hospital uses to recover its revenue is, read in the other direction, the accountability tooling MACPAC says states lack. In July 2026 we began operating exactly that loop with a state Medicaid agency: documented patterns, example claims their investigators can pull end to end, and a standing evidence rhythm. This paper describes the model so that any hospital, and any state, can run it.

The Claim That Said Paid in Full

Start with one claim, because the whole argument lives inside it.

A patient visits the emergency room of a 25-bed rural hospital in south central Kansas. The ER physician bills CPT 99282, a level 2 emergency visit, at $168.18. The KanCare managed care organization processes the claim and pays $28.84. The remittance advice shows the allowed amount: $28.84. Paid at 100 percent of allowed. The EOB reads, in every way that matters to a busy billing office, paid in full.

The hospital's billing staff reviewed this claim at our request and reported, reasonably: "According to the EOBs, the allowed amount on all of these physician claims is what we were paid. To me these have processed and paid in full."

Here is the problem. The Kansas Medicaid fee schedule prices that service at $105. KanCare managed care organizations are required to pay in-network providers at no less than 100 percent of the KMAP rate. The allowed amount is not evidence of correct payment. The allowed amount is the output of the payer's own rate table, and in this case the rate table is loaded at roughly 27 percent of the floor.

Now the detail that turns an error into a pattern: we found the same service, billed identically, paid identically at $28.84, on claim after claim, across two different KanCare MCOs. Identical variance across independent payers is not an adjudication judgment call. It is configuration.

The only way to see it is to do what almost no hospital does: independently reprice every encounter against the applicable benchmark, the fee schedule, the contract, the regulation, and compare that to the remittance. When we did that across two facilities, the paid-in-full problem stopped looking like an anecdote.

Three Patterns, One Mechanism

Underpayment against the contracted rate is one of three structural patterns we find everywhere we look. They share a mechanism: each one converts money the Medicaid program intended to reach a provider into a loss that is invisible at the point where anyone might act on it.

Pattern 1Retroactive Eligibility Gaps

A patient is determined Medicaid-eligible after the date of service. By the time the hospital learns of the determination, timely filing has run or nearly run. The claim dies on eligibility codes, the hospital writes the encounter off as uncompensated care, and the encounter never enters KanCare data as a claim at all. At our rural facility, an eligibility-weighted pool of roughly $606,000 in self-pay encounters is pending coverage discovery right now, with children and CHIP the strongest conversion lane in a non-expansion state. The program funded coverage the provider could not access, and the state's data shows a smaller program than the one it actually bought.

Pattern 2Timely Filing and Processing Friction

Claims filed on time still die on timeliness: delayed in payer processing, returned without adjudication, or denied again after corrected resubmission. The denial code, CO-29, records provider error. The claim history records payer latency. At the surgical hospital, twelve months of remittances show $248,881 denied on CO-29 across 433 claim lines. At the rural hospital, 339 of 2,453 live denial lines, carrying $334,962, have been denied more than once: submitted, denied, corrected, denied again, each cycle consuming 30 to 90 days of a small hospital's cash flow. Fourteen percent of the live denial inventory is on that treadmill.

Pattern 3Underpayment Against the Contracted Rate

The paid-in-full problem of the previous section. In the rural hospital's current working inventory, 267 of 534 actionable claims, half of the worklist, are underpayments rather than denials. All three KanCare MCOs appear. This is the pattern nobody audits because its claims are, by every conventional definition, resolved.

The three patterns compound. A hospital fighting the repeat-denial treadmill has no staff hours left to reprice paid claims. A billing office that trusts the allowed amount writes off the variance automatically. A finance committee reviewing denial dashboards concludes, correctly by the numbers in front of it, that Medicaid is simply a poor payer, when a material share of the gap is recoverable and a further share is preventable upstream.

What the CFO's Reports Cannot Show

Hospital leaders are not missing this because they are careless. They are missing it because the standard reporting stack is built around events, denials, aging, write-offs, and the paid-in-full claim generates no event.

Consider the cleanest number we have. At the surgical hospital, across 27,182 claims in twelve months, collection rates by payer class were:

Same facility, same coders, same systems, same clinical services. A five-to-one collection gap between Medicare and Medicaid is not an operations problem. It is a structural signature, and it is exactly the kind of number that never appears on a standard revenue cycle dashboard, because no single report crosses payer class, charge, and remittance at the encounter level.

Scaled to each facility's own collections base, the recoverable leakage we have sized so far runs from roughly 3 percent of annual collections at the rural emergency hospital to roughly 15 percent at the surgical hospital. The range is wide because facility type drives the mix: an REH heavy in Medicare outpatient leaks differently than a surgical hospital heavy in commercial and Medicare Advantage. That variance is itself a finding. State-average assumptions about hospital revenue leakage will be wrong for nearly every actual hospital, in one direction or the other.

For a hospital leader, the operational conclusion is uncomfortable but freeing: you cannot delegate this to the denial queue, because the dollars are not in the denial queue. The dollars are in paid claims priced against the wrong table, in eligibility determinations that arrived after your filing window, and in resubmission loops that your staff experiences as workload rather than as evidence.

Why Oversight Cannot See It Either

If the hospital's reports cannot see the problem, surely the government's can. They cannot, and the reason is architectural.

PERM, the federal Payment Error Rate Measurement program, is the instrument of record for Medicaid improper payments. For reporting year 2025, PERM measured the national Medicaid improper payment rate at 6.12 percent, roughly $37 billion. The managed care component of that rate: 0.00 percent.

The zero is not evidence of clean adjudication. It is a measurement boundary. PERM reviews the payment the state makes to the managed care plan, the capitation. If the capitation was paid correctly for an eligible enrollee, the transaction is proper, and PERM's inquiry ends there. What the plan then does with claims, the allowed amounts it loads, the denials it issues, the filing clocks it runs out, sits below the audit's floor. Every pattern in this paper lives below that floor. KFF and the Bipartisan Policy Center have both flagged the managed care measurement gap as PERM's known blind spot; the 0.00 percent simply makes it vivid.

MACPAC has now put the institutional version of this on the congressional record. Its June 2026 Report to Congress devotes a chapter to managed care plan accountability and opens from the finding that little is known about the accountability tools states actually use to enforce MCO contract compliance. Its recommendations point at usable performance data and better tools for states to assess plan performance. Translated out of commission language: the entities responsible for overseeing the plans do not currently have an evidence stream that shows them what the plans are doing at the claim level.

Meanwhile the enforcement instruments that do exist run on the wrong fuel. Hospital complaints arrive as anecdote, single claims, phone calls, frustration, easy to dismiss as billing disputes. State claims-resolution logs are published as PDFs that no hospital system can ingest at scale. Contracted rate schedules are not machine-readable, so paid-versus-contract variance cannot be computed by anyone but the payer. The system is not corrupt. It is unobserved.

The View From the Agency

It would be easy to write this paper as an indictment of state agencies. It would also be wrong, and hospital leaders who carry that framing into meetings with their state will lose the room.

Here is what the view looks like from the agency side, stated as fairly as we can. A state Medicaid operations director sits atop a program serving hundreds of thousands of members through three or more MCOs, with federal reporting obligations, legislative oversight, eligibility redetermination workloads that are about to double in frequency, and a provider community whose complaints arrive without documentation. The director cannot fix what she cannot see, and what she can see is what the plans report about themselves, plus whatever auditors find years later. When a state inspector general or legislative auditor does look, the findings arrive as retrospective reports, and the recovery follow-through is minimal because the evidence is stale by the time anyone could act.

Our direct experience in Kansas is that the willingness to engage is real. In July 2026, following a working meeting on these patterns, Kansas Medicaid operations leadership asked us for detailed briefings on each pattern with example claims their investigation teams can look up end to end in state systems, provided their provider escalation guide so hospital issues route through the process the state actually runs, and opened a standing channel for documented evidence. The message, nearly verbatim: we want to solve these issues, and we cannot solve issues we do not know about.

That sentence deserves more weight than it will get. Agency leaders are not short on desire to hold plans accountable. They are short on exactly one input: claim-level, documented, verifiable evidence, delivered in a form their teams can act on, from a source that has done the verification work before sending it. The moment that input exists, the escalation infrastructure, contract compliance teams, and deputy directors who own MCO performance all have something to do with it.

There is also a timing argument for agencies to act now. Kansas just enacted HB 2731 over a gubernatorial veto, mandating cross-agency data matching for eligibility verification: monthly death-record checks, quarterly wage matches, incarceration and winnings data. Whatever one thinks of the bill's politics, its premise is now state law: eligibility integrity is a data-matching problem, and the state is expected to run the match. Payment integrity is the same discipline pointed at the other side of the transaction. A state that is statutorily required to verify that members are eligible has no principled reason not to verify that plans are paying at the contracted floor. The muscle is the same; only the direction of the query changes.

The Same Dataset, Read in Two Directions

The central claim of this paper is that hospital revenue recovery and Medicaid managed care accountability are not two problems. They are one dataset read in two directions.

When a hospital independently reprices its encounters, it produces, as a byproduct of chasing its own money: allowed amounts versus fee schedule floors by MCO, denial and overturn behavior by MCO, processing latency and repeat-denial rates by MCO, and retroactive eligibility volumes with their downstream write-offs. That is, item for item, the performance evidence MACPAC says states lack.

The working loop we now operate looks like this. Hospital-side analysis identifies a pattern and the specific claims that exhibit it. Hospital billing leadership verifies each claim as real, current, and unpaid or underpaid, so nothing sent forward is theoretical. The documented pattern, with payer ICNs the state can pull, goes to the agency through its own escalation routing. The agency's investigators look the claims up end to end in their systems, where they can see what neither the hospital nor the plan will show them. Findings feed both directions: the hospital recovers, and the agency gains contract-compliance evidence that no self-reported plan metric would ever surface. Then it repeats, weekly, in bite-size documented increments rather than annual data dumps.

Three design principles make the loop work, and they are worth stating because most attempts at provider-state collaboration violate all three:

• Send only what the receiving side can act on. We filter everything through one question: can the Medicaid agency actually touch this? Commercial payer disputes, Medicare issues, and internal hospital process failures do not go in the channel. Managed care and fee-for-service Medicaid claims do.
• Verify before sending, every time. One unverified claim that turns out to be patient responsibility or already reprocessed costs the channel its credibility. Hospital billing staff confirm every example. When our first verification pass came back with corrections, that was the system working. The corrections made the evidence airtight.
• Prefer simple encounters for complex arguments. A rate-configuration finding is best demonstrated on a one-line ER physician claim where the arithmetic fits in a sentence, not on a 40-line surgical stay. Investigators replicate simple examples in minutes; complexity buys nothing but delay.

What Hospital Leaders Should Do Monday Morning

None of this requires waiting for policy. A hospital leadership team can start with five moves:

First, treat "paid in full" as a question, not an answer. Direct your revenue cycle team, or your vendor, to independently reprice a sample of paid Medicaid managed care claims against the state fee schedule floor. Start with high-volume, low-dollar professional claims, ER visit levels, where rate configuration errors hide best. If the sample shows variance, size the full book.

Second, demand payer-class collection rates on one page. Charges, payments, and collection percentage by payer class, twelve months, one table. If your Medicaid collection rate is a fraction of your Medicare rate, you have a structural finding, not an operations complaint, and it deserves structural follow-through.

Third, inventory your retroactive eligibility exposure. Ask one question of your data: how many self-pay write-offs in the last year belonged to patients who were, or later became, Medicaid-eligible? In a non-expansion state, look hardest at children. Run the 270/271 eligibility batches before the filing windows close, not after.

Fourth, track your repeat denials as a named metric. Claims denied more than once are your best evidence of payer processing friction, and your staff's best-hidden workload. Count them, sum them, and age them.

Fifth, use your state's escalation guide, and document as you go. Most states publish provider escalation paths; Kansas's current guide routes MCO claims issues through the plan, then to the agency, then to a named deputy director. Escalations that follow the state's own process, with claim-level documentation attached, are the ones that build the relationship this paper describes. Escalations that arrive as frustration build nothing.

What Agency Leaders Should Ask

For the state Medicaid director or operations chief reading this, the equivalent list is shorter, because the leverage is higher:

Ask your MCOs for overturn rates, not just denial rates. A denial rate measures member and provider burden. An overturn-on-appeal rate measures whether the initial denials were right. Plans that overturn a large share of appealed denials are telling you their first-pass adjudication is wrong at scale, and providers who cannot afford to appeal are absorbing that error silently.

Commission an allowed-amount audit against the fee schedule floor. Take the top 25 professional codes by volume, pull each MCO's loaded allowed amounts, and compare to the state floor. It is days of work, not months, and it will either clear the plans or find money that belongs to your provider network. Either result is worth having.

Make your claims-resolution artifacts machine-readable. Publishing resolution logs as structured data instead of PDFs costs little and lets every hospital in the state act on them at scale. The providers who most need the information run the smallest billing shops.

Count retroactive eligibility determinations and their claim outcomes. Your eligibility systems know how many determinations were retroactive. Your encounter data knows how many of those windows produced a paid claim. The gap between the two numbers is program benefit that was funded and never delivered, and it is currently nobody's metric.

Accept verified evidence from providers, and route it. The cheapest accountability tool available to any state is a structured intake for documented, claim-level provider evidence, with disposition reporting back. It converts complaint volume into audit leads and costs approximately one inbox.

The Policy Horizon

Four developments make this the right eighteen months to build the loop.

MACPAC's June 2026 recommendations put managed care accountability tooling on the congressional agenda, with usability of plan performance data as the explicit gap. States that build claim-level evidence streams now will be ahead of whatever reporting requirements follow.

PERM's eligibility penalty gets teeth in 2029. Under P.L. 119-21, beginning October 2029, states with PERM eligibility error rates above 3 percent face FMAP reductions, and nearly a quarter of states currently exceed that line. Eligibility data integrity is about to carry a direct fiscal price, and the same data infrastructure serves both eligibility and payment integrity.

Six-month redeterminations arrive January 2027 for expansion adults and comparable populations, doubling eligibility churn and, with it, retroactive eligibility exposure for providers. Pattern 1 in this paper gets mechanically worse on that date unless the notification and filing-clock fixes land first.

The CMS state directed payment proposed rule, with comments closing July 21, 2026, will reshape hospital-directed payment flows through managed care. Every dollar that moves through an MCO under an SDP arrangement is a dollar whose delivery depends on the adjudication integrity this paper documents. The comment docket is the right place to say so.

Methodology & Sources

Every figure in this paper derives from primary data: 44 raw X12 835 remittance files (7,721 remittance envelopes, 27,182 claims) spanning twelve months at the surgical hospital; a 2,453-line live denial inventory reconciled against 835s and the income statement cash register at the rural hospital; CARC classification against a 1,485-code reference library; independent encounter repricing layered national (OPPS, MPFS, NCCI) then state (KMAP fee schedule, refreshed monthly) then facility (contract exhibits); a Medicare fee-for-service pricing control that must pass before Medicare or MA underpayment figures publish; and eligibility-weighted, non-expansion-calibrated estimates for coverage-discovery pools. Contractual write-offs (CO-45) are excluded from all recovery figures. No dollar is reported recoverable if the hospital's cash register shows it collected. Hospitals are described by type rather than name; claim examples cited to the state carry payer ICNs, not patient identifiers.

1. MACPAC, June 2026 Report to Congress on Medicaid and CHIP, Chapter 3, Managed Care Plan Accountability. macpac.gov
2. CMS, Payment Error Rate Measurement (PERM) Program, Medicaid Improper Payment Rates, RY2025. cms.gov
3. KFF, A Look at the Medicaid Payment Error Rate Measurement (PERM) Program and Upcoming Changes and Impacts. kff.org
4. Kansas Legislature, Substitute for HB 2731 (2025-26 session); Kansas Health Institute analysis. khi.org
5. Federal Register, Medicaid Managed Care State Directed Payments (proposed rule, May 22, 2026; comments close July 21, 2026). federalregister.gov
6. CMS, SMDL #26-001, Implementation of Eligibility Redeterminations under P.L. 119-21 (April 2026). medicaid.gov
7. KDHE, Provider Escalation Guide v1.5 (June 2026).
8. HRN Group primary analyses: surgical hospital 835 reassessment (May 2026); rural hospital recovery reassessment (May 2026) and uncompensated-care bucket analysis (June 2026); active worklist refresh (June 2026). Client data cited in aggregate.