Work

Predicting the unpredictable? Prior psychiatric hospitalization and self-harm as predictors of suicide and self-harm

Year: 2026

Type: article

Abstract: PURPOSE: Prior psychiatric hospitalization and previous self-harm are the two strongest risk factors for suicide and self-harm. Yet the prognostic performance of these predictors over the short-term risk windows remains unclear. This study examined the incidence rates and prognostic performances associated with using these to predict suicide and suicide attempts. METHODS: Using nationwide Danish registry-based cohort data (2010-2020), we followed 4.67 million individuals to estimate incidence rates (IR) and incidence rate ratios (IRR), sensitivity, specificity, positive predictive values (PPV), and positive likelihood ratios (LR+) of self-harm and suicide within predefined exposure windows following psychiatric hospitalization and self-harm. RESULTS: Within the 0-3 days exposure window of prior hospitalization, the IRs of suicide and self-harm were 8.03 and 137.35 per 100 person-years. Higher IRs were observed within 0-3 days after self-harm. IRs generally decreased with longer exposure windows. Prior hospitalization had low prognostic performance for both outcomes: sensitivity 4.5% and 6.7%, PPV 0.1% and 1.1%, and LR + 0.6 and 0.9. Prior self-harm was more informative for self-harm. Prognostic performance generally increased with longer exposure windows. Age and sex influenced risks but did not materially improve prognostic performance. CONCLUSION: Even the strongest risk factors for suicide and self-harm showed limited short-term prognostic performance at the individual level. These findings challenge approaches that rely on risk stratification to identify future cases and instead support systematic, needs-based follow-up and broad preventive interventions for vulnerable groups. Effective suicide prevention may depend less on precise prediction and more on timely intervention and population-level risk reduction.

Source: Social Psychiatry and Psychiatric Epidemiology

DOI: 10.1007/s00127-026-03245-7

Authors: Erik Christiansen, Matthew J. Spittal, Rory C. O´Connor, Katrine U. Madsen, Esben Agerbo, Mette Reilev