Ninety percent of all people who die by suicide have a diagnosable psychiatric disorder at the time of their death.
The statistic is used to imply that, first, mental illness, and not individual choice, is the cause of suicide, and is often cited to justify coercive suicide prevention policies; and, second, that money invested in mental health treatment will reduce suicides. I wish to question the reliability of this statistic, as well as the two implications that are often drawn from it.
What is the source of this alarming statistic? It may surprise advocates of evidence-based medicine to learn that many of the source studies hardly qualify as scientific studies at all, in that many of them are entirely uncontrolled. The studies rely on a technique known as a "psychological autopsy," which tries to diagnose mental disorders in a deceased person based on interviews with family members. The so-called first generation of studies simply chose a study group of known suicides, and tried to identify mental disorders within the study group, with no control at all. This 1996 study, for instance, has no control, but purports to find that 90.1% of suicides have a diagnosable Axis I mental disorder. This is the study that the National Institutes of Mental Health cite as their basis for the figure!
A new generation of studies "during the last decade" has attempted to apply basic scientific control procedures, however. In these studies, a group of known completed suicides was matched with a control group of living people with similar characteristics. Interviews, medical records, and "information from the coroner" are collected and evaluated by psychiatrists who are often supposedly "blind to outcome" - that is, they are not supposed to know who is a suicide and who is alive. If an evaluator knew someone was a suicide, he might be predisposed to look extra hard for information indicating a psychiatric disorder.
Keeping evaluating psychiatrists outcome-blind seems like a particularly difficult task, especially given that "information from the coroner" is included in the case reports. More importantly, those preparing case reports are necessarily not outcome-blind. The idea that their preparation would not be influenced by knowledge of outcome (suicide or living) is rather hard to swallow.
At any rate, one (dubiously) controlled study of young men found that 88% of the suicides, compared with 37.3% of the non-suicides, had a diagnosable mental disorder. To report this study as finding that "90% of suicides have a diagnosable mental disorder" is to ignore its more important implications: well over a third of this population of young males has a mental disorder! But 37.3% of young men do not commit suicide. Clearly, mental illness is not much of a "cause" of suicide. Some scientists characterize it as a necessary but not sufficient condition.
It is also important to point out what counts as a mental disorder in these studies. Depression counts, but also alcohol or drug dependence, and often any Axis I or even Axis II disorder (as in the study of young men). It is instructive (and suspicious) that the percentage of suicides found to have a "mental disorder" does not seem to vary depending on the investigator's definition of "mental disorder."
It is also important to think about the vague, unscientific definitions of mental disorders found in the DSM-IV and its earlier incarnations. Given the vague definition of depression, for instance, is it really any surprise that people who commit suicide would meet the criteria for depression? (Actually, studies vary extremely widely in how many suicides they find to have been depressed - all the way from 30% to 90%. Personality disorders vary even more widely - from 0% to 57%. This variance should make us very suspicious.) What person deciding to end his life wouldn't, for example, experience a loss of pleasure in ordinary activities, or changes in sleep or appetite, or feelings of hopelessness or guilt? As for drug and alcohol use, what person, faced with the desire to die, wouldn't try to assuage his pain by any means available - including alcohol and drugs? In my own case, as a suicide, I view alcohol and drugs as a temporary suicide prevention device. Recent research in nicotine use, for instance, has revealed that nicotine may help symptoms like anxiety and depression, and help people with ADHD to function:
An even more important reason for the link between depression and smoking may stem from the pleasure that smoking can bring. As Dr. Fowler's research suggests, smoking triggers higher dopamine levels in the brain; elevated levels of dopamine have been linked to feelings of well-being and pleasure and have been found in users of heroin and cocaine. Such emotions may be particularly welcome by individuals suffering from depression.
I would like to point out that I do not smoke. But it is easy to see how this logic would apply to alcohol and other drugs. We should expect suicidal people to be more willing to experiment with illicit ways of promoting happiness, compared to the general population. To say that people who commit suicide are likely to have used drugs or alcohol is not to say that alcohol or drug use caused suicide.
Does investing money in mental health care prevent suicide? The relationship is shaky at best. The Japanese government's recent efforts to reduce suicide, through both coercive and non-coercive means, including increased mental health spending, have failed miserably. A 2005 study published in JAMA found that "despite a dramatic increase in treatment, no significant decrease occurred in suicidal thoughts, plans, gestures, or attempts in the United States during the 1990s." While the frequency of treatment of individuals who engaged in suicidal behavior more than doubled, the suicide rate dropped only 6%.
The statistic that 90% of suicides have a diagnosable mental illness, so gleefully reported by those in the anti-suicide industry, is questionable. Even if it has some basis in fact, vagueness of diagnostic criteria and other special factors detract from any conclusions that can be drawn from it. What is most uncertain is whether investing in mental health treatment actually reduces suicide. (This is made even more uncertain by the failure of mental health treatment even to, well, treat mental illness.)
A more realistic and ethical route would be to accept suicide as a relatively rare but natural and acceptable way to end life, to provide means of suicide that are effective and not harmful to bystanders, to allow competent adults to opt out of coercive suicide "rescue," and to focus any government or private spending on alleviating suffering, rather than preventing suicide.
See also: What the DSM-II Got Right, my examination of changes in the diagnostic taxonomy for depression since the DSM-II and their implications for suicide rights.
Update: Jason Malloy points me to this 2004 meta-study, studying suicides in North America, Australia, Europe, and Asia. "Twenty-seven studies comprising 3275 suicides were included, of which, 87.3% (SD 10.0%) had been diagnosed with a mental disorder prior to their death," say the authors. This is far superior to the studies that attempt to backwards-infer mental illness. My main problem with this study is that it’s tracking any and all “mental disorders” and even crap like “alcohol use” is coded as a disorder, not to mention “intermittent depressive disorder” and “neurotic depression” (i.e., they’re not even using the piss-poor standards of the DSM-IV).
(See Malloy's admirably tolerant responses to my increasingly drunken arguments here.)