A colleague once told me that there are two types of psychiatrist: those who have cared for a patient who died by suicide, and those who will. As much as we hate the fact, suicides happen.

In their May 2017 paper in BJPsych Open, Large et al note that the risk of suicide is greatest in psychiatric inpatients (i.e. those in hospital, on leave or having absconded). The exact risk varies wildly between studies, from 9 to 10,490 suicides per 100,000 years of admission time.

The high risk isn’t surprising, they say, as people are admitted to psychiatric hospitals for good reason; often because they might hurt themselves. And those who are most likely to hurt themselves stay on the wards longest, skewing the statistics.

Large et al aren’t content with that simple interpretation though. They posit the notion that some inpatient suicides might be caused by being on a psychiatric ward, rather like some infections might be caused by being in a general hospital. Stigma, loss of social role and abuse are suggested as mechanisms of this possible effect.

In the absence of conclusive research data to back up the theory, the paper uses Bradford Hill’s criteria to investigate whether psychiatric admission could be a cause of suicide, not just correlated with it. Here, I outline Large et al’s comments, and add my own.

The criteria

Strength (of association)

• Suicide rates are 40-120 times higher in psychiatric inpatients than in the general population, and significantly higher than in non-admitted psychiatric patient control groups.
• So this criterion is met, but it does no damage to the sensible theory that inpatients are simply sicker than outpatients.

Consistency

• Yes, the link between suicide and admission has been noted consistently by researchers all over the world, but the inconsistency in suicide rates is striking, just like we know the variation in standard of care is. We’ll come back to that.

Specificity

• Is suicide the only form of harm that being an inpatient might increase your risk of? If it were, that might increase the chances of it being caused by admission, says Bradford Hill.
• Large et al decide this criterion is hard to apply to suicide, but I disagree. Do these patients self-harm more? Do they get depressed more, or suffer functional decline? Possibly, but Large et al don’t say, and I’m not sure anyone knows. In any case, whether this could really cast judgement on a causative link is dubious.

Temporality

• Suicide risk is at its highest immediately after discharge from hospital, and many patients who die by suicide on wards are admitted without suicidal intent, so this criterion is more convincing for a causative link.

Biological gradient

• Does a bigger dose of the risk factor (being on a psychiatric ward) lead to a higher risk of suicide?
• Well yes, but as we said, we don’t know if that’s because these patients are sicker to start with. This is a long way from “more cigarettes increase your risk of lung cancer”, the archetypal causative link, because wards are a treatment for those already at risk, and cigarettes aren’t.

Plausibility and coherence

• Suicide due to stressors and loss of role that might occur due to an inpatient admission would be plausible, and wouldn’t contradict what we already know about suicide.
• Fair comment, but again, it depends on the ward.

Experiment

• Data that does exist, namely from lower-hierarchy types of research like observational studies, is conflicting. Some papers say suicide rates go up when beds are closed down, others say they go down, or don’t change at all.

Analogy

• Large et al suggest that suicide due to psychiatric admission is more likely because there is an analogy, namely hospital-acquired infections occurring in general hospitals.
• This, for me, is purely fanciful in logic and adds nothing to the argument.

Discussion

Large et al state that five of Bradford Hill’s criteria for causation are met (strength, consistency, plausibility, coherence and analogy) and another three are ‘partially’ met (temporality, biological gradient and experiment). But as you have read, I am more stringent in my interpretation.

The strength of association is indisputable. Temporality and biological gradient are clear but again, not at all distinctive of causation. The call to analogy is flaky, and the experimental evidence is more ‘inconsistent’ than ‘partial’.

Plausibility and coherence (i.e. does a causative link make sense?) are, to my mind, dependent on other factors for their validity. Factors like the quality of ward care, pressure on the ward from the rest of the system, treatment standards and social circumstances outside of hospital. This would easily explain the inconsistency in the actual suicide rates of those with inpatient status.

To be clear, I think the question ‘does psychiatric inpatient admission cause suicide?’ is naïve and not meaningfully answerable. It assumes all wards, all patients, and all wider social circumstances, are the same. Could a badly run, cruel ward cause someone to kill themselves? Of course. And could a well-run ward with caring staff, good community links and excellent clinical results cause someone to kill themselves? It’s hard to see how. Large et al themselves note that one study showed an inverse relationship between length of admission and risk of suicide; clearly those wards are doing something right before discharge.

Investigating whether wards cause people to die by suicide can only lead in one direction; we suggest improvements to bad wards, if we find them. No study, however strongly it implicates suicide and inpatient admission (which I doubt it would, certainly without being huge), is going to justify the closure of psychiatric wards, as their role is too pivotal. They aren’t going anywhere.

So the real question we should be examining, to my mind, is ‘how do we make psychiatric wards as conducive to recovery as possible?’ I would be glad to see that answered.

Links

Primary paper

Large MM, Chung DT, Davidson M, Weiser M, Ryan CJ. (2017) In-patient suicide: selection of people at risk, failure of protection and the possibility of causation. British Journal of Psychiatry Open May 2017, 3 (3) 102-105; DOI: 10.1192/bjpo.bp.116.004309

Photo credits

• Elliott Brown CC BY 2.0
• Photo by Goh Rhy Yan on Unsplash