Yes, Involuntary Commitment Really Is That Complicated

I do not even have a mustache to twirl

Tewksbury State Hospital Administrative Building, in Tewksbury, MA By John Phelan - Own work, CC BY 3.0

Those of you aware of Psychiatry Substack might be aware of a paper A Danger to Self and Others: Health and Criminal Consequences of Involuntary Hospitalization and the analysis of that paper Groundbreaking Analysis Upends Our Understanding of Psychiatric Holds posted by Awais Aftab over on his Substack At The Margins. Let me briefly summarize:

This paper examined the effects of psychiatric hospitalization on what we might call “judgement-call” patients who are being evaluated for the first time. Those patients for whom psychiatrists might reasonably disagree about the need for involuntary hospitalization. The results are quite striking:

“For being charged with a violent crime, our reduced form estimate finds that a 10 percentage point increase in a physician’s tendency to hospitalize is associated with a 2.6 percentage point increase in being charged with a violent crime within the following 3 months.” [Emanuel, et al. 2025]

“A 10 percentage point increase in a physician’s tendency to hospitalize yields an increase in the probability of death in the next three months of 0.98 percentage points.” [Emanuel, et al. 2025]

This paper has, to put it mildly, ruffled some feathers. Freddie deBoer — who has, in my estimation, generally done a very good job telling off people who think that involuntary commitment needs to be totally abolished — has written a piece called The New York Fed's New Involuntary Commitment Paper is Methodologically Shaky as Well as Irresponsible in which he… kinda misses the forest for the trees? Whether or not Freddie is right, I’ve become pretty perplexed by his view, and the view of many of the commenters on that post, that people like Awais and Emanuel et al. are trying to use this data as an excuse to not hospitalize people who are severely psychiatrically ill for various nefarious reasons and maybe even as a way to try and abolish involuntary psychiatric commitment all together. Indeed, Freddie has openly stated that he believes that Awais and the study authors simply treat the deaths that have resulted from (what Freddie believes to be) insufficient psychiatric hospitalization as “abstractions.”

My mistake-theorist approach makes me hope that such a view is simply due to a lack of a true understanding of the perspective of the clinician, the state of psychiatric healthcare, the diversity of patient presentation, and the trade-offs that would be demanded from the sort of approach that Freddie seems to advocate.

Fortunately, I am a practicing psychiatrist who works in an inpatient psychiatric hospital and has plenty of experience in emergency psychiatry settings, so have ample perspective to share.

I should probably do a little bit of table-setting before I go into this, so you know where I’m coming from. I am someone who feels very strongly about upholding civil liberties for various reasons: partially because I am suspicious of state power, partially because I am suspicious of the utility of involuntary commitment in marginal cases. I’d guess that if you made a bell curve of the rate of civil commitment for all psychiatrists in the US, I’m probably at least 1 standard deviation below the mean. That said, I also know that there are people who are so mentally ill that the only real place they can live lives that are even moderately decent is in the sort of highly structured environment that only exist in the nearly extinct institutions we refer to as asylums.

Those Marginal Admissions Are Not Just People With SMI

Based on how Freddie likes to trot out the parade of horribles — namely examples of psychotic patients who were not hospitalized and ended up doing horrific things — he seems to imagine that the vast majority of patients being considered for psychiatric admission are individuals with clear-cut and (relatively) severe mental illness who would obviously benefit from psychiatric hospitalization. This is the only way I can charitably make sense of how ferociously opposed he is to the idea that it might be reasonable for the individuals in the “complier” group (i.e. the cases that could go either way, what I’m going to call marginal cases) to be discharged, as well as his comments about deaths resulting from insufficient psychiatric hospitalization being treated as mere abstractions.

I sure wish Freddie was right.

In my experience, it was pretty unusual for a marginal case to be about whether or not a psychotic patient was sick enough to be admitted involuntarily. If I had to guess, those were probably <15% of the marginal cases I saw in the ED as a resident. Instead, marginal cases looked a lot like these:

This patient has a substance use problem, shitty-life syndrome, and are saying that they’re depressed and suicidal, but also have felt this way for 10 years and don’t want to go inpatient because they want to leave the hospital and keep using substances. Do we think that a 3 day hospitalization that will change roughly nothing about their social situation will do much?

This patient came in drunk and suicidal. Now they’re sober and saying they’re not suicidal anymore, but still seem kinda sad and depressed. How much do we want to believe in vino veritas?

This patient obviously has Borderline Personality Disorder, just overdosed for the 20th time in the past 5 years after a breakup, and now says that they’re not suicidal and want to go home. Do we hospitalize them to let them cool down for a few days, or do we acknowledge that personality disorders are not something that short-term psychiatric hospitalization will fix?

The Law Is What It Is

The first thing that you need to understand is that there is a whole web of laws that psychiatrists must abide by when choosing to involuntarily hospitalize patients. Every state must abide by the US Supreme Court’s decision (henceforth SCOTUS) in O’Connor v. Donaldson, which makes it clear that it is unconstitutional for a state to involuntarily confine someone because of mental illness alone. From the opinion itself case syllabus:¹

Held:

(1) A State cannot constitutionally confine, without more, a nondangerous individual who is capable of surviving safely in freedom by himself or with the help of willing and responsible family members or friends…

A finding of “mental illness” alone cannot justify a State’s locking a person up against his will and keeping him indefinitely in simple custodial confinement. Assuming that that term can be given a reasonably precise content and that the “mentally ill” can be identified with reasonable accuracy, there is still no constitutional basis for confining such persons involuntarily if they are dangerous to no one and can live safely in freedom.

May the State confine the mentally ill merely to ensure them a living standard superior to that they enjoy in the private community? That the State has a proper interest in providing care and assistance to the unfortunate goes without saying. But the mere presence of mental illness does not disqualify a person from preferring his home to the comforts of an institution.

On top of that, psychiatrists must work within the bounds of their state’s civil commitment laws. For the most part, these laws are all a variation on the same theme. The person must have a psychiatric illness that renders them:

Dangerous to themselves: That is, they may attempt, maim, or mutilate themselves.

Dangerous to others: That is, they are likely to kill, maim, or mutilate someone else. In some states, such as MA, others may only need to be placed in reasonable fear of being harmed.

Unable to care for their basic needs in the community: This generally refers to the patient’s ability to do the minimum necessary for their survival in the community. The emphasis is on the word “minimum,” and so this usually excludes things like obtaining housing, taking medications independently, or attending routine medical appointments.

There is considerable nuance here that varies on a state-by-state basis. There are at least 11 states² that have language that requires the likelihood of harm to be “imminent,” likely to occur in the “near future,” or evidence of “present dangerousness.” Others simply require a “likelihood of serious harm” and/or evidence of recent behavior in support of that determination.

You might think that such definitionally broad terms make things easy for us, but that would ignore the existence of case law: legally controlling interpretations of those words by state appellate courts.

In some states, psychiatric commitment law is written and interpreted so narrowly that most patients with schizophrenia pretty clearly cannot be involuntarily hospitalized. To give you a sense of just how narrow, I’m going to crib from a previous essay I wrote on NC psychiatric commitment law:

We start with the [North Carolina Supreme Court] referencing a decision of the NC Court of Appeals, In re: Monroe (1980). In this case, the trial court found that Monroe (a schizophrenic) was a danger to himself and others because: he became “uncontrollable at times,” his sleep was irregular, he was standing in his front yard making “all kinds of loud noises [and] calling inappropriately to anyone passing by,” and he disregarded his nutritional needs by “fasting for some periods and then eating a whole chicken or a whole loaf of bread” and that he “[ate] about five pounds of sugar every two days.” The Court of Appeals held that:

…neither the facts recorded by the trial court nor the record supports a conclusion or ultimate finding of dangerousness to self…

on the basis that

even if indicative of some danger, the facts do not support the finding that there is a reasonable probability of serious physical debilitation to the [r]espondent within the near future.

The NCSC points out that even though these findings demonstrated a deficiency in self-care, the statute “mandates a specific finding of a probability of serious physical debilitation resulting from the more general finding of lack of self-caring ability.”

Most states also require “least restrictive means” testing. That is, we must try and discharge the patient to the lowest level of care that can safely handle the patient. We cannot hold patients forever simply because there is some amount of risk present if they were to leave the hospital.

Some have argued that they think it is likely that the current Supreme Court would overrule/narrow O’Connor if given the chance. Maybe, but (1) psychiatrists are still bound to follow the law as it stands today, and (2) even if O’Connor was overturned…

The Current System Can’t Hold These People Forever Without Everything Breaking

Let’s say that O’Connor gets totally wiped out tomorrow. In a shocking 9-0 opinion, Justice Gorsuch suddenly decides that his dalliance with civil liberties is over and that paternalistic state and federal government is actually a great idea.

This solves approximately zero problems because there is nowhere for all of these people to go.

Depending on where you look, somewhere between 0.25% and 0.64% of the US population has schizophrenia. Let’s use 0.5% to keep things simple. For the US population of ~340 million, this means that there are about 1.7 million people in the United States with schizophrenia. Let’s also be very conservative and say that we’re only really interested in institutionalizing the sickest 10% of schizophrenics. The ones that are so sick that they probably need to just live in a state hospital for the rest of their lives. A mere 170,000 people.

Seems like a bit of a problem, then, that there are only 36,150 beds in state psychiatric hospitals, that 50% of those beds are occupied by forensic patients (i.e. individuals who are legally required to be there by the courts for various reasons), and that the median occupancy rate for state-run hospitals is 90%. For the laypersons reading this, state psychiatric hospitals are pretty much the only facilities in modern times that are actually designed and run with the idea of keeping patients for years at a time, outside of some very niche private facilities.

So, even if we assume that 100% of those beds are being used to hold the sickest 10% of schizophrenics (I can promise you that they are absolutely not), we’re just about at capacity already, and have only hospitalized about 20% of the sickest 10% of schizophrenics.

Well, what if we put all inpatient psychiatric beds on the table? Beds on psychiatric units that are designed only for short-term stabilization. Where most patients³ will: Share their room with 1-3 other very mentally ill individuals, with minimal privacy, being checked on every 15 minutes (great for light sleepers, let me tell you). Not have access to outdoor spaces, save for perhaps a 15-20 minute walk once a day (but only if you’re not sick enough that the facility isn’t worried that you’ll elope). Have no means to exercise beyond walking around the unit. Have no daytime activities beyond attending some one-size-fits-all groups. Have no prospects of developing any sense of purpose or participating in anything that even might give their lives any iota of meaning.

So, yeah, what if we consider putting them in somewhere like that for an indefinite period of time? I guess that’s maybe better than letting them be homeless?

Lindenfield at al. looked at CMS data from 2023 and found that there were 28.4 inpatient psychiatric beds per 100,000 people in the US, which is roughly 95,200 beds. I’m not sure if this includes state hospital beds, but let’s say it doesn’t. That brings us to a grand total of 131,350 beds.

So, even if we filled up every state hospital to maximum capacity and were willing to keep about 70% of those patients for years in acute psychiatric units that were never designed to hold them for more than a handful of weeks (in conditions that might actually be worse than being homeless)… We’re still 38,650 beds short of housing just the sickest 10% of schizophrenics.

Let that sink in.

We would still need to more than double our current number of state-hospital beds just to cover the remaining 10% of schizophrenics if we used every available psychiatric bed in the United States.

Of course, this would be a total disaster, because we would have no beds for anything else. This is a problem, because this report from the AHRQ⁴ tells us that there were approximately 83,725 hospitalizations per month for psychiatric illnesses that were not schizophrenia spectrum disorders. This cashes out to approximately 2,790 hospitalizations a day,⁵ and given that the average length of stay is about 5-6 days, that means that we’d need at least an additional 17,000 beds. That number almost seems quaint at this point, doesn’t it?

We Could Do More, But Not Overnight

Listen, I’m not saying that we couldn’t fix this problem.

It probably seems unfathomable to most psychiatrists my age, but we used to have about 500,000 beds in state psychiatric hospitals in the 1950s. These places were not exactly bastions of ethical and humane psychiatric care, but in the 1950s we were only spending about $8,400 in 2024 dollars per patient per year.

So, we have to spend more money. How much more money? Virginia paid about $1,221 per patient per day in 2024; that’s about $445,000 per patient per year. Texas paid between $17,828 and $19,771 per ~30 days. About $594-$659 per day, or $213,936 - $237,252 per year.

Let’s make some assumptions. First, assume every state, on average, is willing to be as cheap as Texas and spend about $240,000 per patient per year, and is willing to build and maintain enough state beds to hospitalize a proportion of the sickest 10% of schizophrenics (~170k people) proportional to their population size.

I had ChatGPT help me extract state budget numbers minus federal funding from the NASBO State Expenditure Report for FY 2024 to get a sense of how much maintaining these hospital beds would cost relative to state budgetary numbers (excluding federal funds). You can see my data for yourself here on this spreadsheet, feel free to let me know if these numbers seem off.

Obviously, the top five most populous states will incur the largest absolute costs, but the impact on a state’s budget varies by a couple of percentage points in either direction. The average state would have to spend about 2.10% of its yearly budget. though some wouldn’t spend quite so much, like Hawai’i (+0.9%) and Alaska (+0.8%), while others, like New Hampshire (+5.0%!) and Nevada (3.30%), will have to spend considerably more. These numbers seem small, but in practical terms they represent considerable increases in state expenditures that will be permanent and thus will require considerable political wrangling to figure out where all of that extra money will need to come from.

This is, of course, back-of-the-spreadsheet math that does not capture other obvious and important complexities and cost considerations. States will have to build many new psychiatric hospitals, though these costs could be fairly small relative to the yearly costs per patient and be built fairly quickly; North Carolina’s 432-bed Central Regional Hospital only took a mere 2 years (2006-2008) to build and open and cost a modest $183 million in 2025 dollars.

A target of 170,000 state hospital beds also seems likely to represent a bare-minimum ideal. There are more than just schizophrenics out there that will need long-term psychiatric care. State psychiatric hospitals are also generally obligated to hold forensic cases for court-ordered evaluations or attempts to restore individuals to competency. In states like Massachusetts, this population has probably accounted for more than 50% of state psychiatric beds in recent years⁶ (so about 330 beds) though it’s not clear to me if it’s reasonable to expect that number to significantly increase with the number of beds available.

Some Of These People Are Never Getting Better

Sometimes, I get the sense from Freddie and others who make similar arguments that hospitalization will just be A Temporary Thing. We’ll get these patients on the right medications, they’ll become functional enough to return to the community, and we’ll send ‘em out the door with an outpatient appointment for their next long-acting injectable in a month.

That is, unfortunately, wishful thinking, particularly for patients with primary psychosis. Schizophrenia, unlike bipolar and major depressive illness, is not an episodic condition. Once you have it, you’ve pretty much got it for good, and the outcomes are pretty poor.

The SOHO⁷ study followed 6,642 individuals with schizophrenia to see their real-world outcomes with treatment. For the 1,009 patients who had never been treated prior to study entry, 70% achieved an illness severity (CGI-S) score of ≤3 — meaning mild symptoms or better — after 2 years of treatment. Sounds pretty good, right? Not so fast.

When the authors did the 3-year analysis on the full group of 6,642 participants, they found that (operational definitions for each in the footnotes) “33% achieved long-lasting symptomatic remission,⁸ 13% long-lasting functional remission,⁹ 27% long-lasting adequate quality of life,¹⁰ and 4% achieved recovery.¹¹”

As if that wasn’t dismal enough, patient characteristics that predicted success were basically just indicators of milder illness. That is to say, the sickest patients (i.e. the ones most likely to need long term hospitalization) are the least likely to get better.

So, the sickest 10% of schizophrenics I mentioned earlier? They’re not going to get much better, they’re not going to live independently, and right now the system we have cannot possibly hold all of them. In the absence of available long-term state hospital beds, most psychotic patients are going to get admitted and discharged inside of 2 weeks.

Do You See It Now?

Hopefully, you’re starting to understand the quandary that we face when deciding whether or not to involuntarily hospitalize and why we might be genuinely interested in asking whether or not those patients on the margins are likely to benefit from psychiatric hospitalization.

It is not because we are cartoon villains who, for some inexplicable reason, are totally fine with psychotic patients dying of exposure or murdering innocent people. Instead we have to contend with the reality of what it means to send a marginal case inpatient.

For non-psychotic, non-manic patients, there are genuine questions about whether or not a brief inpatient stay actually has the potential to do anything positive for them. The borderline patient might benefit from hospitalization to give them a day or two to chill out, but the attention that they get from the medical system and their friends/family also might reinforce their parasuicidal behaviors. The suicidal-but-only-when-drunk patient with alcohol problems might get better connected with substance use resources, but they also might lose their job when they don’t show up for 4 days. These are real, serious clinical considerations that have real consequences for our patients, and good, thoughtful psychiatrists reasonably disagree about them all the time!

Even for our marginal, chronically psychotic patients the “best” case scenario might look like being held on a locked acute care unit that is not designed for long-term hospitalizations for >12 months waiting for a state hospital bed. Is that obviously better than letting them return to their group home, homeless shelter, or the streets? For some patients, I’m really not sure.

In the real world, where we all know that there is not enough room to hold all of these patients while we wait for the states to build an extra 140,000 beds worth of capacity, it means less than 2 weeks of psychiatric treatment (assuming the patient takes medications on day 1) before they are discharged right back to the community, with the same chronic condition that they always had.

Maybe, in the future, we’ll have those 140,000 beds and acute inpatient units will just be a stepping-stone on the way to inpatient hospitalization. Until then, though, I welcome any data that helps us weigh the costs of involuntary commitment to our patients and their communities.

1. Thanks to HW for catching this mistake

2. This was by my counting with the help of ChatGPT, but it is probably more since I know it missed imminence language in the NC statute)

3. There are some very nice units like the one I work on at Duke where there is an exercise room, a large outdoor space that is basically escape-proof, and patients have single rooms. This is very, very much the exception and not the rule; Duke maybe has the nicest inpatient unit I have ever seen.

4. Agency for Healthcare Research and Quality - a department within Health and Human Services

5. Assuming a 30 day month

6. Schizophrenia Outpatients Health Outcomes

7. “Achieving a level of severity that was mild or less (i.e., a score of < 4 on the scale from 1 to 7) in the CGI-SCH positive, negative, cognitive and overall severity scores, plus no inpatient admission, for a minimum period of 24 months and maintained until the 36-month visit.”

8. “Fulfilling the following three criteria for a minimum period of 24 months and maintained until the 36-month visit: (1) a positive occupational/vocational status (i.e., paid or unpaid full- or part-time employment, being an active student, or a housewife); (2) living independently; and (3) having active social interactions (i.e., having more than one social contact during the last 4 weeks or having a spouse or partner).”

9. “This was defined as achieving an [EuroQol 5 Dimensions Visual Analogue Scale] (EQ-5D VAS) score of ≥ 70 [out of 100] for a minimum period of 24 months and maintained until the 36-month visit.”

10. “Recovery was defined as simultaneously achieving long-lasting symptomatic and functional remission and an adequate quality of life for a minimum period of 24 months and maintained until the 36-month visit.”

11. This news article has quotes from union representatives that claimed 2/3rd’s, so take that number with a shaker of salt, but this is generally in-line with the literally years-long state hospital waitlist that I saw when I was a resident between 2021-2024, and this 2008 editorial which more concretely cited 45% of MA’s state hospital beds occupied by forensic cases.