To develop and validate prognostic models to stratify individuals by risk of unnatural death after healthcare presentation for non-fatal self-harm, in the context of recurrent self-harm and its associated risks.
Approach:
Validation: Utilized bootstrapping for internal validation and estimated optimism-corrected concordance indices (C-index), calibration intercepts and slopes, and proportions of unnatural deaths captured across predicted-risk groups.
Key Findings:
Final models achieved optimism-corrected C-indices of 0.74 (95% CI not provided) at presentation and 0.75–0.76 (95% CI not provided) at discharge.
Identified 86–88% of observed unnatural deaths within 2 years among the 40% of individuals with the highest predicted risk.
Observed 2-year risk of unnatural death was 0.15% among the lowest predicted risk group, comparable to individuals without a self-harm history.
Interpretation:
The prediction models rank individuals accessing private sector care in South Africa by their probability of unnatural death following healthcare presentation for non-fatal self-harm.
Limitations:
Models showed some overfitting with optimism-corrected calibration slopes ranging from 0.84 to 0.92, which may affect their applicability to broader populations.
The study is based on a specific cohort of medically insured individuals, which may limit generalizability.
Conclusion:
The models provide a framework for prioritizing individuals at high risk for targeted interventions.
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
