Emergency departments board psychiatric patients for days — and no national measure counts the wait

What is happening?

People who arrive at a United States emergency department in psychiatric crisis are held in that department until an inpatient bed opens somewhere, and for a large share of them the wait is measured in days rather than hours. Among Medicaid-enrolled youths aged 5 to 17 across 44 states, 30,359 of 255,139 emergency department visits for mental health conditions — 11.9 percent — became a boarding stay of three to seven days, with a mean of 4.5 days. That is 2022 claims data, published in August 2025.

The supply side is the reason the wait exists. 43 of 48 responding state mental health agencies, or 90 percent, reported a shortage of inpatient psychiatric beds in 2025, against 50 percent of states when the same question was first asked in 2002. The country has 28.4 inpatient psychiatric beds per 100,000 people, against the 60 per 100,000 that researchers are reported to cite.

The wait itself has never been counted nationally. The federal measure that would produce such a number, Emergency Care Access and Timeliness, was adopted only in the Calendar Year 2026 Medicare Hospital Outpatient Prospective Payment System final rule announced in November 2025.

Whose problem is this?

The party that decides to close an inpatient psychiatric unit is not the party whose emergency department absorbs the boarding that follows, and in many cases it is not even the same hospital.

Where does this problem end?

The boundary matters here because the wait is not the same thing as the shortage that produces it, and the two are counted by different bodies at different times.

What is the state now, and what should it be?

Now

Needs a new measurementThe target state: no source opened here states an official national target for how long a psychiatric patient may board. The 60 beds per 100,000 figure is reported as a level researchers frequently refer to, not as a government standard. The nearest thing to a federal threshold is the 4 hour mark that the new Emergency Care Access and Timeliness measure treats as poor performance, and that measure was adopted in November 2025 for boarding generally, not for psychiatric boarding specifically.

How big is it?

The counted population is 30,359 — the number of emergency department visits by Medicaid-enrolled youths aged 5 to 17, across 44 states in 2022, that turned into a boarding stay of three to seven days.

That number counts one payer, one age band, one duration band, and one year. It is a floor and not an estimate of the whole. The study behind it excluded boarding events longer than six midnights to limit coding error, so the longest stays fall outside it, and it publishes no band shorter than three days, so a patient who waited 40 hours is invisible in it. Adults, commercially insured patients and the uninsured are absent entirely.

No source opened here gives a national all-ages count of psychiatric boarding episodes. Scaling this figure up by the Medicaid share of children, or by the national pediatric visit total, would be proportional allocation rather than measurement, so it was not done.

Under what conditions does it arise?

Four conditions hold this arrangement in place, and none of them is about clinical judgment in the emergency department itself.

1. The door and the bed are controlled by different parties. Who arrives at an emergency department is set by crisis, police practice and the state of community services. How many inpatient beds exist is set by hospital boards, payer rates and state capital budgets. Neither side adjusts to the other in the same cycle. 2. The emergency department cannot refuse the overflow. When inpatient capacity is full, the shortfall does not appear as a closed door or a canceled service. It appears as a patient still sitting in a treatment space, which is a form that no capacity statistic records as a failure. 3. The units that close are the ordinary ones. In the 2012 to 2022 hospital analysis, the hospitals with the largest reductions in psychiatric beds were more likely than the comparison group to be general medical hospitals and non-profit hospitals, so the losses come out of the general acute system that emergency departments feed into. 4. The wait has had no national number. Until the November 2025 rule there was no federal reporting measure for boarding duration at all, which means the quantity being argued about could not be compared across hospitals or tracked over time.

What has been tried?

Two of these run in opposite directions on purpose. Adding beds and diverting people away from beds are both live state strategies, and 35 states reported the second while 12 reported the first.

What was found?

Why is it still unsolved?

Supply shortage — the number of inpatient psychiatric beds is smaller than the number of people the emergency system decides to admit, and the two quantities are set by parties who never meet.

Nobody involved is behaving irrationally, and that is the difficulty. A hospital board closing a marginal psychiatric unit is responding to what that unit costs on its own books. A state mental health agency choosing community services over beds is following four decades of policy consensus and is often right about the individual patient. An emergency physician who admits a person in crisis is doing the only safe thing. The result of all three correct decisions is a person on a gurney for four days.

The shortage also has an unusual property: it does not present as a shortage. When a service runs out of capacity it normally turns people away, and being turned away is a countable event. Psychiatric bed scarcity instead converts into waiting inside a department that has already accepted the patient, so the deficit is stored as time rather than as refusals. Time stored this way shows up on no official ledger until somebody decides to measure it, and the decision to measure it nationally was made in November 2025.

Finally, the geography has been stable for a decade. The share of the population living in a region with an inpatient psychiatric bed shortage moved from 60.7 percent to 64.9 percent between 2012 and 2022, which is a system holding still rather than one deteriorating or recovering. A condition that neither worsens dramatically nor improves does not generate the kind of event that forces a response.

What observation would mean it is solved?

Candidates — (a) a national median boarding time for psychiatric patients that falls below the 4 hour threshold the new measure treats as poor performance (b) a fall in the number of states reporting an inpatient psychiatric bed shortage from the current 43 of 48 (c) a fall in the share of the population living in a region classified as short of beds, currently 64.9 percent.

(a) is not yet observable and may never be observable in this form. The measure exists on paper as of November 2025, and the announcement opened here does not say that its results will be broken out for mental health patients. If they are not, a falling overall boarding median can coexist with a rising psychiatric one, because psychiatric boarders are a small share of all admissions and a long tail rarely moves an aggregate.

(b) is self-reported by the parties that request the budget. The same agencies that answer the shortage question also submit the capital requests that a reported shortage supports, and the survey asks about perception of shortage rather than a measured ratio.

(c) depends on a threshold researchers chose. What counts as a shortage region is a definition, and a definition can be met by redrawing regions or by the population moving.

There is a fourth trap that cuts across all three. Boarding duration can fall because patients stop being admitted rather than because beds opened. In Massachusetts, nearly 26 percent of patients who boarded more than 12 hours were ultimately discharged home, which means a shorter wait and a discharge without admission look identical in the boarding statistic.

What is it connected to?

Fills with researchplausible links to Medicaid financing of inpatient psychiatric care, to the workforce shortage that leaves licensed beds unstaffed, to jail and police custody as the alternative destination for people in crisis, and to rural hospital service closures generally. Relation type and evidence grade were not researched in this round.

What these sources do not say

• A national median or mean boarding time for psychiatric patients, at any age. Every duration figure gathered here is either one state, one payer, one age band or one hospital. The number that the phrase psychiatric boarding most obviously refers to does not exist as a published national statistic, which is itself the finding.
• Adult boarding duration nationally. Both duration studies opened are pediatric or youth. Adults are the larger group and the least documented one.
• What happens to the patient during the wait. No source opened states whether psychiatric treatment is delivered while boarding, or whether the days are custodial. The clinical content of the interval is unrecorded.
• How many inpatient psychiatric units closed in 2025 and 2026, and how many beds that removed. Closures surface one at a time in local and trade coverage. No opened source gives a running national tally for those two years, so the closure axis of this problem is asserted here only through the 2012 to 2022 hospital analysis.
• Whether the new federal measure will report mental health separately. The rule announcement opened here names the measure and what it times, and says nothing about stratification. The reporting years, voluntary and mandatory, were also not confirmed from any opened source.
• Why Massachusetts improved. Waiting times there fell between 2022 and 2024 while beds were added and while a monthly public census ran. Which of those mattered, or whether neither did, is not established by anything read here, and no source tests whether the improvement generalizes.

See the evidence

Ten source documents were opened directly for this entry. Two of them are PDFs — the NRI state profile and the Massachusetts hospital association census — which were retrieved over HTTP and then read as extracted text from those same retrieved bytes, so the figures above come from the documents themselves rather than from summaries of them.

Four cautions belong with the numbers. First, the NRI document is stamped as a draft for state review, and its own footers are inconsistent, labeling it as both the 2025 and the 2024-2025 state profiles; its headline infographic gives 36,542 patients in state hospitals while its body text gives 36,510 for the same fiscal year 2023 reporting system, and that difference is not explained in the document. Second, the two bed rates are not the same measurement and must not be read against each other: 28.4 per 100,000 counts inpatient psychiatric beds of all types, while 11 per 100,000 counts residents of state psychiatric hospitals only, and the second is fiscal year 2023 data published in July 2025. Third, the KFF Health News article gives both 28.4 beds per 100,000 people and a total of about 61,000 beds, and those two do not reconcile against any United States population base — 61,000 beds at 28.4 per 100,000 implies a population of roughly 215 million — so the article does not appear to be describing one bed universe with both numbers, and they are carried here as separate statements rather than merged. Fourth, no figure here was confirmed in two independent sources — each row rests on one document, and the Massachusetts duration figures reach this entry through press coverage of a Health Policy Commission analysis that was not itself opened.

People affected

Estimated range 30,359–30,359 As of 2022

Sensitivity The interval has zero width because this is not an estimate with a confidence interval but one measured count from one claims analysis. The figure is a floor on four axes at once: one payer (Medicaid), one age band (5 to 17), one duration band (3 to 7 days) and one year (2022). The study excluded boarding events longer than six midnights to limit coding error, so the longest stays are cut off at the top, and it publishes no band below three days, so a youth who waited 40 hours does not appear at all. Adults, commercially insured patients and the uninsured are outside the sample entirely, and adults are the larger group. Six states are also outside the 44-state sample. No source opened for this entry gives a national all-ages count of psychiatric boarding episodes, and scaling this number up by the Medicaid share of children or by national pediatric visit totals would be proportional allocation rather than measurement, so it was not attempted. The true all-ages figure is larger than this by an unknown factor.

Regional breakdown The study publishes the boarding share by state, from 2.7 percent in Arkansas to 27.3 percent in Iowa, but not the per-state visit denominators, so the per-state counts cannot be recovered from the published values. Six states are missing from the sample. Splitting the national count by state population would be proportional allocation, and it would be especially unfounded here because the state shares differ by roughly a factor of ten, which is exactly the variation that such a split would erase.