Part 2: The BEDPAN Award: Misinformed Research on SNF Policy

This is the second and final part of the Marc Zimmet series that premiered on Park Place last week titled: “The BEDPAN Award: Misinformed Research on SNF Policy.” The research in question came out in July in The New England Journal of Medicine -“Colliding Forces – The Aging of the Baby Boom Generation and Contracting Nursing-Home Supply.”

Where we left off…

All beds are not functionally equal. Roughly one-third of the nation’s SNFs fall outside the “Standard Set”:

Hospital-Based SNFs

The 215 figure is not “the answer” either (MedPAC counted 14,935 participating SNFs in 2011 and 14,720 in 2021, a net decline of 215 SNFs), It is a reminder to ask: 215 what?

MedPAC's long-run tables show hospital-based SNFs falling from 1,611 in 2002 to roughly 3 percent of 14,720 facilities in 2021 - about 440 units. “Freestanding” providers now represent 97 percent of the provider-class. It shows that a substantial portion of historical provider contraction occurred in a class that does not function like a community nursing home.

Hospital-based units proliferated after DRGs were introduced, for reasons that had less to do with building a long-term care safety net and more to do with managing the seams between Medicare payment systems. A short hospital-based SNF stay followed by transfer to a freestanding SNF is not simply “capacity.” It can reshape the payment curve, shift burden to the community provider, and distort shared-savings narratives. That makes a hospital-based TCU a different economic asset, not a community bed with better parking. Ignoring the difference distorts important SNF metrics.

Specifically, it lowers average length of stay and raises average Medicare rates (due to PDPM payment weighting). In the era of shared savings and payment recalibration, transfer economics should not be ignored.

A serious supply analysis should classify hospital-based SNFs, freestanding SNFs, swing beds, Distinct Parts, specialty units, and active-but-not-really-available beds separately. If the policy target is resident access, the unit of analysis must be the bed that can accept the resident without mucking up a dataset.

Licensed Beds: Value or Vestige?

The NEJM paper correctly notes that licensed beds can overstate supply, and it deserves credit for recommending staffed-bed surveillance. It still stops one layer too soon. I released the Post-Pandemic Capacity and Relative Occupancy framework in 2021 because the industry was overstating its recovery. A certified SNF bed has value; to decertify is to forfeit an asset. This is where national averages become policy platitudes and banks get nervous. Open facility, certified beds, mothballed unit, frustrated hospital discharge manager, scared and confused family. Anyone who has operated near the ground knows this. Anyone who has only counted rows in a file may not.

Functional Supply surveillance should track staffed beds, units offline, decertified beds, payer-specific acceptance, admission restrictions, dementia capacity, behavioral-health capacity, complex-care capacity, hospital discharge delays, and rural travel distance.

It is difficult because every major SNF dataset has a different purpose, timing convention, and failure mode. NHSN, Medicare Cost Reports, CMS Provider Info, Claims, PBJ, etc. are discrete domains with reporting inconsistencies. The answer is not selecting the file that agrees with the thesis. It is cross-contextualizing until the contradictions reveal the operating truth.

That is the backdrop almost nobody wants to say out loud: much of what the policy world thinks it knows about SNF data comes from sources that were never designed to answer the questions now being asked of them. A perfectly accurate dataset is like a witness; before we put it on the stand, we should ask what it actually observed.

Occupancy Is Not a Measure of Demand

Ready for the most meaningless statistic in policymaking? “Occupancy” as a term of comparative performance, should be retired. The article reports 1.24 million residents and an “average occupancy rate” of 84 percent for 2025, citing KFF. The issue brief reports 1.24 million residents and 14,742 facilities but does not identify an 84 percent national occupancy measure in the text. KFF's own State Health Facts series lists the national certified-facility occupancy rate at 79 percent for July 2025. That five-point difference is not rounding. It signals a different denominator or averaging method, neither of which the article explains.

An 84 percent result with 1.24 million residents implies a denominator of roughly 1.476 million beds. A 79 percent result implies roughly 1.570 million current certified beds. Using the 2019 comparison base in my working series - about 1.644 million beds - produces a more telling statistic: Relative Occupancy is roughly 75 percent. Each result can be mathematically valid. Each answers a different question. The NEJM article moves from one unexplained average to a national conclusion as though the choice does not matter.

One census. Four denominators or methods. Four different policy questions.

Relative Occupancy as the right measure for contraction. Aggregate current occupancy asks how full today's certified beds are. Practical Occupancy measures census within operable, clinically usable beds. Occupancy is not a measure of demand when trended - Relative Occupancy tells that story. A market can look comparably occupied against today's smaller denominator while serving far fewer residents than the system was built to serve.

Rate Construction makes the denominator financially operational. In states that impute days, apply minimum occupancy assumptions, or otherwise bake assumed utilization into cost-based rate methods, staffing-related unit closures can become payment penalties. The facility cannot open the beds because labor is unavailable; the state treats the beds as though they should have been filled; the rate gets worse; and the facility decertifies beds to stop the bleeding. The rate method then manufactures the decertification it later records as a market signal. Wonderful.

Why Functional Supply Contracts

The article's Medicaid-law direction is fair but undisciplined. Work requirements principally affect ACA expansion adults and certain waiver populations, not the aged, blind, and disabled pathway that dominates long-stay nursing-home eligibility. The access risk is broader: reduced retroactive eligibility, shorter renewal cycles, limits on state financing tools, verification burdens, and general administrative inertia can make coverage slower and payment less reliable. The resident with dementia may not personally face a work test; the facility may still finance months of care while eligibility catches up. That lag is painful in some buildings and immaterial in others and that variance is the point.

The authors say Medicaid pays less than cost and Medicare post-acute payment is higher, creating an incentive to prioritize Medicare residents. Ya think? But a nursing home is not choosing between two cartoon residents named Medicare Good and Medicaid Bad. Irrespective of payer-mix, the SNF cost center obligations don’t change much – there is an irreducible minimum cost to compliantly operate. Academics can posit as all they want, hospitals have trouble discharging specific patients for a reason, and it’s not because no one wants another Medicaid resident. This is SNFonomics. No shirt. No shoes. No dice.

Medicaid is the room-and-board backbone for long-stay care, but it is not the only reimbursement signal attached to a Medicaid resident. Part B and Part D change the resident-level economics. State case-mix methods, nursing CMI, acuity add-ons, provider taxes, and rebasing timing all matter. That is why the same "Medicaid shortfall" can mean different things in different states and buildings.

Medicare Advantage is the missing middle that policy treated as background noise for two decades. The old story assumes Medicare fee-for-service margins cross-subsidize Medicaid shortfalls. That story weakens as MA penetration rises. On a net basis, MA admissions are becoming less attractive than a Medicaid dual in more markets after administrative friction and revenue per admission compression are factored in. That’s especially true for admissions with high ancillary costs (e.g., medications) because the SNF is responsible under MA, while Medicare Part B and D are charged directly for Medicaid duals.

As Medicaid rates rise and MA rates do not, providers may be more interested in filling beds with Medicaid duals than churning burdensome MA admissions. States are paying for federal MA largesse. Why are they not suing the federal government over it?

So yes, Medicaid stabilization matters. But a useful recommendation has to model the full revenue-side ecosystem: Medicaid base rate, case-mix, Part B, MA, FFS Medicare, I-SNP where relevant, provider tax, supplemental payments, Medicare bad debt, occupancy assumptions, and direct-care spending. Rate setting must be dynamic and elastic. Otherwise, it is not SNF policy; it’s writing Mazel Tov on a Christmas card. Rinse, repeat, on to the next crises that could have been avoided.

Targeted Payment Requires Target Practice

The article recommends targeted Medicaid stabilization payments. Good. The idea is not new; the only interesting question is whether the targeting is intelligent. An old proposal does not become innovative because it comes home from college wearing a blazer and asks to be called Dr. Targeted Payment. The point was never that targeted payments are clever.

A useful stabilization framework targets Functional Access risk, not public sympathy. It identifies markets with limited alternative supply, high Medicaid dependence, high dementia or behavioral-health need, documented staffing shortages, hospital discharge delays, and Rate Construction mechanics that accelerate decertification. It protects facilities that preserve hard-to-place capacity. It does not sprinkle money over every provider as it falls to the bottom line for some but does not save others. It does not reward facilities with more revenue for higher staffing levels. A useful framework should not require stabilization.

Then comes the awkward question: who is going to build and administer this policy? Many states once had robust Medicaid rate-setting departments with institutional knowledge, Medicare Cost Report fluency, desk-review discipline, and people who understood how one assumption could move an entire market. In too many places, those departments are now ghost towns. The people who knew the machinery retired, left, or were replaced by budget-neutrality memos, and consultants who can model a formula without knowing what the formula is touching. That is not a staffing problem. It is a governance problem.

Why Is the Paper So Long?

Because I have issues - and because the article got me going because it’s not just an article, it’s another in a decades long misunderstanding. It moves the nursing-home supply problem into a public frame, but public policy built on vague categories is not harmless. It tells legislators the wrong story. It lets advocates choose the scariest denominator. It lets agencies regulate certified beds as though they are staffed beds. It lets hospitals discuss throughput without acknowledging transfer economics. It lets Medicaid policy ignore Part B and MA. It lets staffing policy ignore immigration. It lets everyone feel informed while missing the machinery.

That is the part that concerns me. The alarm is being sounded by people who are still mixing up the component parts. The country does not need another elegant warning that nursing homes are under pressure. It needs a precise explanation of which assets are disappearing, which are changing function, which remain unavailable despite being counted, and which policy choices are accelerating the decline.

If we keep writing the same paper with the same vague terms, we will continue to receive ill-conceived policy. Five years from now, someone will publish another prestigious warning announcing that the country is short of nursing-home capacity. The rest of us, calculators still in hand, will be asked to act surprised.

Marc Zimmet is the CEO of Zimmet Healthcare Services Group.

Comments on this article? Contact Patrick Connole at pconnole@parkplacelive.com.