One Question About Suicidality, 9.5 Years of Mortality: What a Korean National Cohort Tells Us
- Nationwide linkage of KNHANES (2007–2018) with cause-of-death records (2007–2022): 62,631 adults, 597,807.5 person-years, mean follow-up 9.5 years.
- 10.8% (n=6,733) reported a history of suicidality. After full covariate adjustment, that history predicted suicide mortality HR=2.8 (95% CI 2.0–4.1), external-cause mortality HR=1.8 (1.3–2.6), and intrinsic-cause mortality HR=1.3 (1.2–1.4).
- Plans or attempts carried roughly twice the risk of ideation alone — suicide HR=4.3 (2.1–8.6), external-cause HR=3.0 (1.4–6.1).
- After Korea revised its national survey item to "seriously considered suicide" in 2013, the suicide-mortality risk ratio rose from 2.1 (1.3–3.2) to 4.6 (1.7–12.0) — a sharper screen, not a different population.
A self-report endorsed in a community survey, a decade later: dead by suicide nearly three times as often, dead from any external cause almost twice as often, dead from "natural" disease 30% more often. The Korean team linked KNHANES with national death records and watched 62,631 people for almost ten years. The signal does not decay. Suicidality is not a transient state captured by a screener; it is a marker of a life-course trajectory that includes accidents, overdoses, untreated chronic disease.
What the data shows
Three things matter clinically. First, the gradient inside the suicidality variable is real: ideation only is bad (HR 2.8 for suicide), plans or attempts are worse (HR 4.3). The screening dichotomy hides a dose. Second, the "improvement" of the survey wording in 2013 — from "thought about suicide" to "seriously considered suicide" — more than doubled the predictive HR (2.1 → 4.6). The earlier item was contaminated by passive ideation that does not predict death. Wording is not cosmetic. Third, the elevated intrinsic mortality HR=1.3 means that even after deaths from suicide and accidents are excluded, people with a suicidality history die earlier from cancer, cardiovascular, metabolic disease. Whatever this trait is, it travels with somatic neglect.
For your practice
If a patient told you, even five years ago, that they had seriously considered suicide — that disclosure still matters in 2026. Long-term tracking of these patients is not over-clinical; it is calibrated to actual mortality risk. Two practical shifts:
- Treat past plans/attempts as a separate risk stratum from ideation. The Kwon data show roughly double the hazard. Your safety planning, lethal-means counselling, and contact frequency should mirror that asymmetry.
- Co-ordinate with primary care. The intrinsic-mortality finding (HR 1.3 for cancer, cardiovascular, metabolic deaths) means your suicidality-history patients are also somatically high-risk. A standing communication channel with their GP — annual check-ups confirmed, not assumed — is part of suicide prevention as much as a Columbia scale.
A self-reported "yes" to seriously considering suicide is not an event — it is a trait that predicts how a person dies for at least a decade.
Self-reported exposure with the usual under-reporting bias; KNHANES did not capture treatment received between waves, so the residual hazard might be lower in actively treated subgroups. The 2013 wording change makes pre/post comparisons indirect.