CMS Hold Harmless Period for Medicaid Work Requirements: Build Each Entity's Expansion-Adult Exposure From the Books, Phase the Loss by State Election, and Set the 2027 Payer-Mix Budget Before December 11

On November 13, 2026, CMS gave states an optional hold harmless period for Medicaid work requirements: from January 1 through June 30, 2027, an electing state assesses compliance but denies or disenrolls no one for failing it. A hospital group should now budget expansion-adult Medicaid revenue by entity and by state election, not by the national headline.
Why this matters now
CMS announced the hold harmless period in a press release on 2026-11-13 and in State Medicaid Director letter SMD 26-006 the same day. The statute, Public Law 119-21, requires states to implement the community engagement requirement by January 1, 2027, and that date did not move. What moved is enforcement. A state that elects the period must still assess whether each applicant and beneficiary meets the 80-hour requirement or qualifies for an exclusion, but it may not deny an application or disenroll an otherwise eligible person solely for failing it before July 1, 2027. States must tell CMS by December 11, 2026 whether they are in. Deputy Administrator Dan Brillman framed the purpose as data over paperwork: 'If the state can use data to show that someone is working or qualifies for an exclusion, that person should not have to track down paperwork to prove it.'
The requirement itself comes from the interim final rule CMS issued on 2026-06-01, summarized in its fact sheet: non-pregnant adults ages 19 to 64 in the expansion group across 43 states and D.C. must show 80 hours a month of work, work program participation, community service, or half-time education, or earn at least $580 a month in 2026. New applicants must meet it for at least one month before the month they apply; existing beneficiaries for one or more months between renewals. The rule's own impact analysis, published 2026-06-03, projects 2.3 million fewer enrollees in FY 2027 and 3.1 to 3.3 million fewer in each later year.
Coverage of the announcement has been fast and, for a finance team, incomplete. KFF on 2026-11-14, HFMA and the AHA on 2026-11-13, and Nextgov the same day all explain what a state may stop doing and when it must decide. None of them say what a hospital group with entities in two states should put in the Medicaid line of a 2027 budget that is due before December 11. The Commonwealth Fund estimated on 2025-09-18 that work requirements would cut hospital Medicaid revenue by up to 9.9 percent and add $7.0 to $8.0 billion of uncompensated care. That is a national average applied to a full year of enforcement. Your number is smaller in 2027 and lands in different months depending on which state each entity sits in.
What the hold harmless period changes in a provider's revenue, and what it does not
Start with the mechanics that drive timing, because timing is most of the 2027 budget question. In a state that enforces from January 1, the first losses come from two queues. New applicants who cannot show a compliant month before the application month are denied, so the self-pay share of new patients rises immediately. Existing enrollees are checked at renewal, and under the same law expansion adults now renew every six months instead of annually, so the renewal queue cycles through the whole population twice a year. A person found noncompliant gets a notice and 30 days to show compliance before coverage ends. Put those together and the first disenrollments in an enforcing state land in February or March, then ramp as renewals come due across the year.
In a state that elects the hold harmless period, both queues keep running but neither produces a denial or a disenrollment until July 1. The state still sends notices, still matches data, still records who failed. What happens to the accumulated noncompliance on July 1 is not spelled out in the letter, so treat it as a scenario rather than a fact: the conservative case is that action on it begins in July and reaches run-rate within a month or two. Either way, an electing state turns a first-quarter ramp into a third-quarter step.
Two more positions exist. States that planned to start before 2027 (KFF's April 30 survey listed Nebraska for May 1, 2026, Montana for July 1, 2026, and Iowa for December 1, 2026) already carry the effect in their run-rate, so their 2027 budget change is the annualization of what the books show today. And a state granted a good faith exemption defers implementation, at the latest to December 31, 2028; Georgetown's Center for Children and Families notes CMS expects to approve few of these.
Now the accounting, which does not change with timing. Under ASC 606 as applied to patient service revenue, a provider records revenue at the transaction price it expects to collect, net of explicit concessions (contractual adjustments) and implicit price concessions for patients who cannot pay. When a Medicaid encounter becomes a self-pay encounter because the patient lost coverage, the revenue for that encounter is lower from the start. It is a reduction of net patient service revenue, a flow item, not a bad debt expense. Bad debt expense is reserved for a later deterioration in a patient's ability to pay after revenue was recognized. Care delivered to the now-uninsured patient that the provider does not expect to collect goes into the uncompensated care disclosure at cost; it is never a revenue line. The point-in-time side is patient receivables: a lost-eligibility denial on an account already billed to Medicaid is a debit to the price concession (contra revenue) and a credit to receivables, and the self-pay receivable that replaces it carries a far higher concession estimate. The same discipline you apply when refreshing payer-mix assumptions with contract-level evidence applies here: the budget line moves only when the books can show which patients, in which entity, under which state rule.
One health system, three entities, two states: the walk-through
Call it Riverbend Health. Three entities, each a separate QuickBooks Online file: Riverbend Medical Center, a hospital in State A; Riverbend Physician Group in State B; and Riverbend Urgent Care, also in State A. State A has signaled it will elect the hold harmless period; State B has said it will enforce from January 1. Nothing is official until December 11. The numbers are illustrative; the sequence is the point.
Step one: pull Medicaid net revenue by entity
Trailing twelve months of net patient service revenue by payer class, by entity. The hospital shows $300.0 million total with $48.0 million Medicaid (16 percent). The physician group shows $60.0 million with $9.0 million Medicaid (15 percent). Urgent care shows $12.0 million with $2.4 million Medicaid (20 percent). Combined Medicaid revenue is $59.4 million.
Step two: isolate the expansion-adult share
'Expansion adult' is not a field in any ledger. Build it from what is: patient age at date of service between 19 and 64, no Medicare entitlement, aid category or managed care plan code that maps to the adult group, and no pregnancy or disability category. Riverbend's revenue cycle team runs that filter on the encounter detail behind each entity's Medicaid revenue and maps each plan code once, in writing. The hospital's expansion-adult share is 45 percent ($21.6 million); the physician group's is 55 percent ($4.95 million); urgent care, which sees a younger walk-in population, is 70 percent ($1.68 million). Total exposure is $28.23 million, not $59.4 million. Everything else in the Medicaid line, including children, pregnant patients, dual eligibles, and disability categories, is outside the requirement.
Step three: choose loss scenarios from published projections, not instinct
CMS's own rule projects 2.3 million fewer enrollees in FY 2027 against roughly 20 million expansion enrollees, about 11.5 percent, and 3.1 to 3.3 million in later years, about 16 percent. The Commonwealth Fund's 5.1 to 5.8 million uninsured implies a higher case near 28 percent. Riverbend carries three scenarios: 11.5, 16, and 28 percent of expansion-adult revenue at full-year run-rate. The walk-through below uses 16 percent.
Step four: phase the loss by each state's position
This is the step the national coverage skips. The annual loss rate is a run-rate; 2027 captures only part of it, and how much depends on the state. The table gives Riverbend's phasing factors, which are assumptions the CFO can defend from the rule's mechanics and will revise once a state's election is known.
| State position on January 1, 2027 | January to June factor | July to December factor | 2027 share of full-year run-rate | What drives it |
|---|---|---|---|---|
| Already enforcing before 2027 | 1.0 | 1.0 | 100 percent, and already in the 2026 base | Losses are in the trailing twelve months; budget the annualization |
| Enforcing from January 1, no hold harmless | Ramp from 0 in January to 1.0 by June | 1.0 | About 75 percent | 30-day notice and cure; renewals spread across the year |
| Hold harmless elected | 0 | 0.5 in July, then 1.0 | About 46 percent | No denials or disenrollments before July 1; accumulated noncompliance acted on after |
| Good faith exemption granted | 0 | 0 | 0 percent | Implementation deferred, at the latest to December 31, 2028 |
Apply it. The physician group in State B: $4.95 million times 16 percent is $792,000 at run-rate, times 75 percent is $594,000 of 2027 Medicaid revenue lost. The hospital in State A: $21.6 million times 16 percent is $3.456 million at run-rate, times 46 percent is $1.59 million. Urgent care in State A: $1.68 million times 16 percent is $269,000, times 46 percent is $124,000. Riverbend's 2027 Medicaid revenue reduction in the base case is about $2.3 million. The naive number, 16 percent of all Medicaid revenue from January 1, is $9.5 million, four times too high. And the naive number hides what matters more: the run-rate leaving 2027 is $4.5 million a year, and $3.7 million of it switches on in July in State A.
Step five: convert lost coverage into net revenue and uncompensated care
A lost Medicaid dollar does not vanish from the income statement in one line. Riverbend assumes 60 percent of the affected encounters still happen as self-pay and that self-pay collections run at 10 percent of the Medicaid rate, so 6 percent of lost Medicaid revenue returns as self-pay net revenue, about $140,000 in 2027. The other 40 percent of encounters are deferred care, which lowers variable cost slightly and shows up later as sicker presentations. The cost of care delivered to the 60 percent is tracked in the uncompensated care disclosure. The budget therefore carries three linked lines per entity: expansion-adult Medicaid revenue, self-pay conversion revenue, and uncompensated care cost, each with the same scenario and phasing switches.
Step six: carry two cases until December 11, then lock one
Until a state notifies CMS, each entity's Medicaid line holds both the January-enforcement and hold-harmless phasings, and the consolidated budget shows the range rather than a point. On December 11 the CFO locks one phasing per state and documents the source. From January, four monthly measures by entity tell Riverbend whether the phasing was right: the share of prior-Medicaid patients whose eligibility check fails at registration, self-pay conversions among patients with a Medicaid encounter in the prior twelve months, Medicaid-pending days on accounts awaiting a determination, and eligibility-related denial dollars. In State A those should stay near zero through June. If they do not, the state is doing something the letter did not describe, and the budget moves before the quarter ends.
Accounting controls for the Medicaid work requirements budget
Source data: encounter-level detail behind each entity's Medicaid net revenue for the trailing twelve months, with patient age at date of service, Medicare entitlement flag, aid category or plan code, state, and entity; the written plan-code mapping to the expansion adult group; CMS's published projections and the Commonwealth Fund estimates with their dates; each state's election status as notified to CMS by December 11, 2026; and, from January, the registration eligibility responses, self-pay conversion counts, Medicaid-pending aging, and denial codes by entity. Record the capture date of each item and the authority it came from, including the 2026-11-13 press release and letter.
Calculation: a fixed sequence anyone can rerun. Filter Medicaid revenue to expansion adults by the documented mapping. Multiply by the scenario loss rate to get the run-rate reduction. Multiply by the state's phasing factor to get the 2027 amount, month by month. Apply the self-pay conversion and collection assumptions to get self-pay revenue. Apply the encounter-retention assumption and cost-to-charge ratio to get uncompensated care cost. Sum across entities for the consolidated view. Polarity: the Medicaid reduction and the self-pay addition are both revenue flow items and net against each other on the income statement; uncompensated care is a disclosed cost, not revenue; receivables and their concession estimates are point-in-time balances that must agree to the entity's aging at each month-end.
Review: a second person, not the analyst who built the model, ties the expansion-adult revenue to the entity's Medicaid revenue in the ledger, checks that every plan code in the Medicaid line is mapped exactly once, confirms the phasing factor matches the state's notified election, recomputes one entity by hand, and confirms that no expense accrual was booked for expected future coverage loss. From January the reviewer compares the four monthly measures to the phasing curve and flags any entity where actuals leave the band.
Decision: the CFO chooses the scenario the budget carries and the phasing factor per state, and signs the assumptions on December 11. The board approves the range before then and the point after. The revenue cycle director owns the plan-code mapping and the monthly measures. The state decides its election; nobody at Riverbend decides that, which is exactly why both cases stay in the model until the notification is in.
Limits and common mistakes
The letter does not say how an electing state must treat noncompliance it records between January and June once July arrives. The 0.5 July factor is a scenario, and a state could move faster or slower.
Do not budget the headline against all Medicaid revenue. Children, pregnant patients, dual eligibles, and disability categories are outside the requirement, and in most hospitals they are more than half the Medicaid line.
Do not book an expense or a reserve for coverage you expect to lose. Revenue is recognized encounter by encounter at the expected transaction price. The only balance that changes before the encounter happens is the budget. The only balance that changes after is the concession estimate on accounts already billed.
Do not attribute every 2027 Medicaid loss to work requirements. Six-month renewals start in the same window and will produce procedural churn among people who do meet the requirement. Tag the reason at the account level or the budget-versus-actual explanation will be wrong in both directions.
State elections may not be public on December 11. Confirm with the state Medicaid agency rather than inferring from a press report. Medicaid supplemental payments, disproportionate share adjustments, and 340B interactions are outside this article; the 2027 OPPS budget rebuild covers the Medicare side of the same budget.
Riverbend Health and its figures are illustrative. The dates, the mechanics of the hold harmless period, the 80-hour and $580 tests, and the enrollment and revenue projections come from the CMS, Federal Register, KFF, HFMA, AHA, Commonwealth Fund, and Georgetown sources cited above.
Frequently Asked Questions
What did CMS announce on November 13, 2026 about Medicaid work requirements?
CMS announced an optional hold harmless period for states implementing the Medicaid community engagement requirement by January 1, 2027. From January 1 through June 30, 2027, an electing state must still assess whether each applicant or beneficiary meets the 80-hour requirement or qualifies for an exclusion, but it may not deny an application or disenroll an otherwise eligible person solely for failing to meet it. States must notify CMS by December 11, 2026 if they elect the period.
Does the hold harmless period delay the January 1, 2027 start date?
No. The implementation date is set in Public Law 119-21 and CMS said it cannot change it. The requirement, the assessments, the data matching, and the notices all begin January 1, 2027 in every state that has not received a good faith exemption. What the hold harmless period suspends in an electing state is the enforcement action: denials and disenrollments for noncompliance do not happen before July 1, 2027.
How does a hospital estimate its exposure to Medicaid work requirements from its own books?
Pull trailing twelve-month Medicaid net revenue by entity, then filter the encounter detail to adults ages 19 to 64 with no Medicare entitlement whose aid category or plan code maps to the expansion group, excluding pregnancy and disability categories. Multiply that expansion-adult revenue by a loss scenario drawn from published projections, such as CMS's 2.3 million enrollees in FY 2027 or 3.1 to 3.3 million in later years against about 20 million expansion enrollees. Then phase the result month by month according to each state's position: enforcing from January 1, electing the hold harmless period, already enforcing, or exempt.
Is lost Medicaid coverage recorded as a revenue reduction or as bad debt expense?
As a revenue reduction. Under ASC 606 a provider records patient service revenue at the amount it expects to collect, net of contractual adjustments and implicit price concessions. An encounter that becomes self-pay because the patient lost Medicaid is recorded at the lower expected self-pay amount from the start. Bad debt expense applies only when a patient's ability to pay deteriorates after revenue was recognized. Care the provider does not expect to collect for is disclosed as uncompensated care at cost, not run through revenue.
When will hospitals know whether their state is using the hold harmless period?
States must notify CMS of their election by December 11, 2026. Public announcements may lag that date, so finance teams should confirm directly with the state Medicaid agency. Until then, the budget should carry both the January-enforcement phasing and the hold-harmless phasing for each entity, and lock one per state once the election is confirmed.
Once each entity's expansion-adult Medicaid revenue sits in one model with the state's phasing and the monthly measures that test it, a hospital group can answer the board's question for any entity before the quarter closes rather than after. FinBoard reads every entity's QuickBooks Online file into one consolidated view, so a finance team can see Medicaid revenue by entity and payer class, compare budget to actual by month, and drill from a variance to the encounters behind it. See how FinBoard consolidates every entity's books on one page.


