Cross-National Analysis of the Associations among Mental Disorders and Suicidal BehaviorFindings from the WHO World Mental Health Surveys

Suicide is one of those issues that hides in plain sight. The deaths are visible in the statistics, but the early steps—thinking about it, making a plan, and the first attempt—are where most of the story unfolds. That's where prevention lives. And here's the twist: we've known for years that mental disorders raise suicide risk, but most past work looked at one diagnosis at a time. Real people rarely have just one. Comorbidity is the rule, not the exception. So the real question is sharper: which disorders matter on their own, which combinations add up or don't, and at what stage do they matter most? As Nock, Hwang, Sampson, Kessler and colleagues showed using the World Mental Health Surveys, the answers change as you move from thoughts to actions. Picture the scope of their data. Twenty-one countries, from Europe and the Americas to Africa, Asia-Pacific, and the Middle East. A total of 108,664 people interviewed face to face, with an average response rate of around seventy-three percent. Everyone was assessed with the World Health Organization's Composite International Diagnostic Interview, Version 3.0—so standardized diagnoses of sixteen DSM-IV disorders, plus the ages those disorders began. The same interview captured suicidal ideation, plans, and attempts, also with ages of onset. The clever part is how they used time. They built a timeline for each person, one year at a time, and asked: by this age, which disorders had already started, and did suicidal ideation or an attempt happen this year? That's discrete-time survival analysis, but you can think of it as a year-by-year coin toss whose odds change when a disorder enters the picture. The models controlled for age, sex, and cohort, and they were run twice over: once in the full sample and again only among people who had already had suicidal thoughts. That second pass is crucial—it tells you which disorders signal who is likely to move from thinking to doing. Across the board, the initial story is simple and strong. In bivariate models—one disorder at a time—virtually every lifetime DSM-IV disorder predicted a later first suicide attempt. The odds ratios landed between about 2.9 and 8.9. Specific phobia sat on the low end; conduct disorder and post-traumatic stress disorder, or PTSD, were up near the top. That's the thirty-thousand-foot view: most diagnoses are red flags for a first attempt. But life isn't a one-variable model, and comorbidity blurs the lines. When the team adjusted for comorbidity, the signals faded but didn't vanish. In developed countries, most individual disorders settled into odds ratios roughly between one and three. In developing countries, several held on to stronger associations; post-traumatic stress disorder, conduct disorder, and drug abuse or dependence, for example, still stood out. Even after that adjustment, in both settings, about thirty of thirty-two disorder effects remained above one, and eighteen were statistically reliable. The core message holds: you can't ignore diagnoses just because they cluster together. You just have to be clear about what, exactly, they're predicting. Here's where the stage-by-stage view pays off. Most of the predictive power of mental disorders for suicide attempts comes from predicting who develops suicidal ideation in the first place, not from deciding who among ideators will go on to plan or try. Depression is a classic example. In developed countries, major depression strongly predicts ideation—think an odds ratio of around 3.3—but it's much weaker for the steps that follow. On transitions from ideation to planning or to attempts, depression hovers closer to 1.5, and for unplanned attempts, it's barely above one. Its grip is on thoughts. The disorders that tip ideators into action have a different flavor. They track with anxiety and impulse control. In developed countries, the standouts for unplanned attempts among ideators were post-traumatic stress disorder, bipolar disorder, and conduct disorder, with odds ratios just above two for each. In developing countries, post-traumatic stress disorder and conduct disorder again rise for planned attempts, roughly 2.8 and 3.2. For unplanned attempts among ideators there, oppositional defiant disorder and alcohol-related disorders were among the few significant predictors, around 3.1 and 1.9. Take a breath and connect the dots: arousal, reactivity, and poor control—those are the risk signatures when a person is already thinking about suicide. You can see the same theme when you zoom out to the population level. Using population-attributable risk proportions—the fraction of all attempts in a population that are associated with prior disorders—Nock and colleagues estimated that roughly sixty to seventy-five percent of attempts were tied to some DSM-IV disorder. The striking part is the decomposition. Between about sixty-one and seventy-six percent of that associated risk came from predicting ideation. The slices for later transitions were small: plans among ideators contributed up to five percent, planned attempts around six to nineteen percent, and unplanned attempts roughly ten to eighteen percent. In plain English: if you want to lower attempts across a population, you have to shrink the pool of people who ever reach sustained suicidal thoughts. Comorbidity adds another layer. In simple dose-response terms, risk climbs as diagnoses pile up. In developed countries, having exactly one prior disorder was associated with about four times the odds of an attempt; three disorders pushed that to around fourteen; six or more to roughly twenty-eight. In developing countries, the same staircase appeared—about three and a half, nine, and then thirty-two for those same rungs. But here's the nuance that matters for clinicians. When the models explicitly included both which disorders a person had and how many, the extra boost from each added diagnosis shrank dramatically. Effects for "two," "three," and beyond dropped toward one, even slipping below one at the very high counts. That's subadditivity. It doesn't mean comorbidity is protective; it means the first few disorders do most of the predictive work, and piling on more doesn't keep multiplying risk at the same rate. The cross-national contrasts are real, but they rhyme. Across both developed and developing countries, mood and anxiety disorders reliably predicted ideation. In developed settings, mood disorders—major depression and bipolar—were among the strongest predictors of attempts overall, while anxiety and impulse-control disorders did the steering from ideation to action. In developing settings, impulse-control disorders, substance use, and post-traumatic stress disorder carried more of the weight for attempts. And among ideation predictors in some developing countries, post-traumatic stress disorder, dysthymia, and intermittent explosive disorder could even outstrip depression, with odds ratios in the ballpark of three to four. Different contexts, similar motifs. One more way to ground this. Look at who actually had a diagnosable disorder before they thought about or attempted suicide. Among people who seriously contemplated suicide, just over half in developed countries—about fifty-two percent—and just under half in developing countries—about forty-three percent—reported a prior disorder. Among people who attempted, those figures rose to roughly sixty-six and fifty-five percent. Those are not edge cases or rare exceptions. They're a common backdrop. All of this rests on a careful but imperfect design, and the authors are upfront about that. The diagnoses come from structured interviews rather than full clinical workups, although subsamples in France, Italy, Spain, and the United States showed good concordance with blinded clinical assessments. The interviews were retrospective; people recalled the ages their symptoms and behaviors began—so recall bias is a real possibility. Some disorders weren't assessed at all, including certain personality disorders and nonaffective psychoses. And the diagnostic rules didn't use hierarchical exclusions, which probably inflated observed comorbidity. The team addressed representativeness with weighting and a two-stage sampling design—everyone answered core questions, and a probability subsample completed the full battery—but not every country sample perfectly mirrors its national population. These caveats don't erase the findings, but they do shape how confidently we read them. So where does this leave us? First, it suggests that risk assessment should be stage-specific. If someone hasn't had suicidal thoughts, the broad screen matters: mood and anxiety disorders dominate the landscape of who's likely to start thinking about suicide. Once someone is an ideator, the focus should shift. Symptoms and diagnoses tied to arousal and control—post-traumatic stress disorder, bipolar disorder, conduct problems, substance use, and oppositional defiance—are the sharper signals of who might cross from thought to plan to act, especially for unplanned attempts. Second, it tells us to respect comorbidity without overcounting it. The first few diagnoses carry the lion's share of predictive power; after that, the curve flattens. There's also a policy-level point. Those population-attributable risk proportions are significant. If sixty to seventy-five percent of attempts are associated with prior disorders, and most of that association flows through ideation, then investments that prevent, detect, and treat common mental disorders—especially depression and anxiety—are suicide prevention at scale. But when you're triaging scarce resources in emergency rooms or primary care, the anxiety-impulsivity profile is what helps you identify the ideators who are in the danger lane right now. Nock and colleagues didn't claim causality, and they didn't pretend to solve the whole puzzle. What they gave us is a map: how diagnoses line up with each step from thought to plan to attempt, how that map changes across countries, and how the number of diagnoses changes the elevation. It's not every contour, but it's the major ridgelines. And for a problem where timing and attention mean everything, that's immensely practical.