Prisons Are Built With Bricks Of Law And Brothels With Bricks Of Religion, But That Doesn’t Prove A Causal Relationship

Research Suggests Psychiatric Interventions Like Admission To A Mental Hospital Could Increase Suicide Risk says an Alternet article about a study that specifically mentions that it should not be used to conclude that psychiatric interventions like admission to a mental hospital could increase suicide risk.

But I wouldn’t be so worried if it wasn’t based on a very similar editorial written by field experts and published in the Journal of Social Psychiatry and Psychiatric Epidemiology.

The study involved is Rygaard-Hjorthøj, Madsen, Agerbo, and Nordentoft (2013), hereafter just “Hjorthøj” because I like saying that word. Hjorthøj finds that people who receive psychiatric treatment are much more likely to commit suicide than people who don’t. For example, someone who gets psychiatric medication is six times more likely to commit suicide than someone who doesn’t; someone who gets admitted to a psychiatric hospital is a whopping 44 times more likely to commit suicide than someone who doesn’t. The authors observe a “dose-response relationship”, which means that the more psychiatric treatment you get, the more likely you are to kill yourself.

Now, you’re probably asking yourself at this point “Wait, were they just using perfectly healthy people with no psychiatric problems as a control group?” and the answer is yes. Yes they were. So this study is basically finding that people who get committed to psychiatric hospitals are more likely to be the sort of people who are going to commit suicide than people who do not get committed to psychiatric hospitals. I for one find this result rather reassuring.

The authors of the study are absolutely on board with this, saying that “observational studies such as the present one cannot establish causality, but merely associations”, and their conclusion is that “not only people with a history of of psychiatric hospitalization, but also those receiving only psychiatric medication, outpatient treatment, or emergency room treatment should be monitored more closely”. Sure. If you absolutely must have a snappier conclusion than “psych patients often mentally ill, more at eleven,” I guess that fits the bill.

But according to an editorial published in the same journal by two people who are not the original authors, it says something much more sinister:

The results of a study in this issue of the Journal…raise the disturbing possibility that psychiatric care might, at least in part, cause suicide.

A…bold hypothesis. Why should we privilege this hypothesis over the alternative possibility that suicidal people are more likely to seek (or get forced into) psychiatric treatment?

The authors understandably caution that ‘the association is likely one of selection rather than causation, in that people with increasing levels of psychiatric contract are also more severely at risk of dying from suicide.’ This is undoubtedly part of the reason for the association, but it is not possible to be sure that an element of causation may not also be contributing. Associations that are strong, demonstrate a dose-effect relationship, and have a plausible mechanism are more likely to indicate a causal relationship than associations that lack these characteristics.

And then the Alternet article picks this up and adds a different argument:

The Danish researchers argued that we were seeing the results of something like a cancer treatment study. Sicker people were appropriately getting into more intensive treatments, but unfortunately the sicker they were the more likely it was that they would still die, despite even the best of medicines. They also suggested that we may have therefore discovered the most accurate predictor of suicide we’ve ever found: The more someone seeks or is forced into psychiatric care, the closer they probably are on the trajectory towards suicide.

The only problem with this line of reasoning is that there’s no evidence to support it. Suicide is not a progressive illness like cancer; that is, there’s no evidence that people with suicidal feelings travel on a trajectory of ever-intensifying, ever-more-constant suicidal feelings while getting into ever more intensive psychiatric care until they die at steadily increasing rates along the way. If suicidality was in fact progressive in that way, we’d be much better at identifying where people are along that path and intervening at the right time to prevent suicides. Instead, completed suicides tend to be impulsive, related to a myriad of cascading, confounding, unpredictable factors, not much more common overall in people diagnosed with mental disorders than in the general population, and most often surprising to even those closest to the victims.

Okay, let’s stop talking about psychiatric disease and shift to murder.

Probably the best risk factor for murder that you will ever find, better than being abused as a child or doing drugs or having the MAOA warrior gene or whatever, is “previous contact with the police”.

Murder is not “progressive” (shut up, neoreactionaries). Much like suicide, there’s no evidence that murderers “travel on a trajectory of ever-intensifying, ever-more-constant murderous feelings while getting into more intensive police custody until they kill at steadily increasing rates along the way.” Instead it seems to be “impulsive, related to a myriad of cascading, confounding, unpredictable factors, and surprising even to those closest to the perpetrators.”

The link between murder and previous contact with the police will be strong. For example, previous murderers released from prison have a 1.2% chance of getting arrested for another murder within three years, compared to about a 0.0001% murder rate per three years among the general population. That’s a relative risk of 10,000x, which blows Hjorthøj’s relative risk of 44x out of the water.

The link will be dose-dependent. People who have previously only gotten warnings from the police will be less likely to murder than people who have gotten small fines, who are less likely to murder than people who have gotten probation, who are less likely to murder than people who have gotten short jail sentences, who are less likely to murder than people who have gotten long jail sentences.

The link even has a plausible causal mechanism. Contact with the police can seriously disrupt people’s lives, making them stressed and anxious and angry and hopeless, all of which are the sort of emotions that predispose someone towards violence.

Therefore, the police cause murder?

Here are some other links that are non-progressive, strong, dose-dependent, and have plausible causal mechanisms.

The link between getting detention and dropping out of school. Therefore, detentions cause students to become demoralized and drop out from school.

The link between ice cream sales in a city and heatstroke cases in that city. Therefore, ice cream contains toxic chemicals that cause heatstroke.

The link between having lots of bruises and being in an abusive relationship. Therefore, abusers only abuse their victims because they’re angry about how many bruises they have.

The editorial authors seem to have gotten the “strong, dose-dependent, plausible” criteria from an article on epidemiology (God only knows where the journalist got the non-progressive criterion from). I would bet that the epidemiology article either did not intend for it to be used in this way, or that it meant that these criteria provide only the most tenuous of possible links.

This is why the saying is “correlation doesn’t imply causation” and not “correlation does not imply causation, unless it’s really strong correlation, in which case knock yourself out.”

And this is why the article finds that even going to a psychiatric emergency room and being turned down for treatment increases your risk of suicide almost twenty times. I mean, in my ER patients only even see a psychiatrist for like half an hour. You’re saying a half an hour with a psychiatrist leads to a vigintupling of suicide rates months down the road? We might be bad. But we’re not that bad.

The sad thing is, I think there might be a point buried underneath all this.

You can’t conclude from an increased murder rate among people with criminal histories that the police cause murder. But the justice system does contribute to murder in its way by sticking hardened criminals together, traumatizing them, and failing to give them enough resources to rebuild their lives. The contribution of the criminal justice system to crime isn’t exactly a secret, it’s just not accessible with that methodology.

Likewise, I don’t disagree that contact with the psychiatric system can sometimes be harmful. Forced commitment can sometimes make people lose their jobs, or cause them stigma, or stick them in an unpleasant psychiatric hospital where they don’t want to be. While there are no doubt potential benefits as well, the weighing of the costs and benefits is something that hasn’t been investigated nearly as much as it deserves. I think forced commitment is an overused tool and would be glad to get some evidence backing me up.

But this paper contributes nothing to the discussion. All we know is there’s an association between psychiatric care and suicide, which was entirely obvious already. We don’t know how much of that association is causal, how much of it is selection, and how much of it is “it would be even worse without psychiatric care but psychiatric care can’t do everything.

The exact effect of psychiatric care on suicide is a topic worthy of further high-quality research and discussion. But this isn’t it.