Risk of unnatural death following self-harm in South Africa: development and validation of multivariable prognostic models - Report - MDSpire

Risk of unnatural death following self-harm in South Africa: development and validation of multivariable prognostic models

  • By

  • Veronika Whitesell Skrivankova

  • Roxanne Pelteret

  • Stephan Rabie

  • Mpho Tlali

  • Naomi Folb

  • Eliane Rohner

  • Chido Chinogurei

  • Yann Ruffieux

  • Soraya Seedat

  • Mary-Ann Davies

  • Gary Maartens

  • John Joska

  • Andreas D Haas

  • June 25, 2026

  • 0 min

Share

Clinical Report: Assessing the Risk of Unnatural Mortality After Self-Harm

Overview

This study developed and validated prognostic models to predict the risk of unnatural death following non-fatal self-harm in South Africa. The models demonstrated good discrimination and calibration.

Background

Self-harm is a significant public health issue, particularly in low- and middle-income countries where the burden of suicide is high. Accurate risk stratification is crucial for effective intervention and resource allocation, especially in settings with limited mental health care capacity. This study addresses the gap in prognostic models tailored for such contexts.

Data Highlights

ModelC-index2-year Risk of Unnatural Death
At Presentation0.74-
At Discharge0.75-0.760.15% (95% CI 0.06% to 0.32%)

Key Findings

  • The final models achieved optimism-corrected C-indices of 0.74 at presentation and 0.75-0.76 at discharge.
  • Models identified 86-88% of observed unnatural deaths within 2 years among the highest risk group.
  • Calibration intercepts ranged from -0.02 to -0.01, indicating little systematic miscalibration.
  • The observed 2-year risk of unnatural death was 0.15% (95% CI 0.06% to 0.32%) among the 60% of individuals with the lowest predicted risk, comparable to the risk among individuals without a self-harm history.
  • The models included variables such as age, sex, prior medication use, and mental disorder diagnoses.

Clinical Implications

The developed models can assist clinicians in identifying individuals at high risk for unnatural death after self-harm.

Conclusion

The prognostic models provide a valuable tool for risk stratification in individuals presenting with non-fatal self-harm.

Related Resources & Content

  1. Author(s)/Org, Source, Year -- Title
  2. DIGITAL HEALTH — Real-world implementation and predictive performance of suicide prediction models: A systematic review and meta-analysis
  3. Intensive Care Medicine — SAPS 3—Transitioning from Patient Assessment to Intensive Care Unit Evaluation: Part 2 - Creation of a Prognostic Model for In-Hospital Mortality at ICU Admission
  4. Frontiers in Psychiatry — Development and validation of a machine learning–based risk prediction model for non-suicidal self-injury in adolescents
  5. Intensive Care Medicine — A predictive model for assessing mortality risk among emergency department patients
  6. NICE Guidance on Self-Harm
  7. VA/DoD Clinical Practice Guideline for Suicide Risk
  8. WHO Suicide Fact Sheet
  9. Acute‐Phase Interventions After Self‐Harm for Preventing Suicide and Recurrence: A Systematic Review and Meta‐Analysis - Roncete - 2026 - Acta Psychiatrica Scandinavica - Wiley Online Library
  10. Prediction models for self-harm and suicide: a systematic review and critical appraisal | BMC Medicine | Full Text
  11. https://www.statssa.gov.za/publications/P03093/P030932023.pdf
  12. Incidence and prognostic factors of self-harm and subsequent unnatural death in South Africa: A cohort study | PLOS Medicine

Original Source(s)

Risk of unnatural death following self-harm in South Africa: development and validation of multivariable prognostic models

  • by Veronika Whitesell Skrivankova, Roxanne Pelteret, Stephan Rabie, Mpho Tlali, Naomi Folb, Eliane Rohner, Chido Chinogurei, Yann Ruffieux, Soraya Seedat, Mary-Ann Davies, Gary Maartens, John Joska, Andreas D Haas

  • June 25, 2026

Related Content

Unmasking microaggressions: Experiences of women in clinical and academic psychiatry and the need for systemic reform

Medical Oddities: Could GLP-1s Curb Violence?

Five studies, one question: are we overlooking hidden drivers of health?

The Promise of Artificial Intelligence–Powered Speech Biomarkers in Psychiatry