A Clinical Prediction Model for Self-Harm After Adolescent Traumatic Brain Injury

OBJECTIVE: Few prognostic models are available to assess risks of adverse outcomes among traumatic brain injury (TBI) patients beyond acute care. Self-harm is an important adverse outcome with modifiable risk factors. The authors developed and validated a risk model for self-harm after TBI in adolescence. METHODS: In this retrospective cohort study, the authors utilized a population-linked dataset based in Wales (the Secure Anonymised Information Linkage Databank). It included adolescents ages 10-24 years who sought health care services with a diagnosis of TBI between January 1, 2009, and October 16, 2023. Cox regression analysis was used to develop a prediction model for the association between risk factors and time to self-harm, with optimism-corrected bootstrapping for internal validation. Model performance was assessed as discrimination and calibration. RESULTS: In total, 34,092 adolescents were in the model development cohort, of whom 538 (1.6%) and 1,202 (3.5%) had recorded self-harm episodes within 1 year and 3 years of the index TBI, respectively. Twelve predictors were included in the risk model, including a history of self-harm, psychiatric conditions, and substance use as strong predictors. The optimism-adjusted area under the receiver operating characteristic curve was 0.74 at 1 year and 0.72 at 3 years post-TBI. The model demonstrated good calibration at both time horizons. CONCLUSIONS: A prognostic model based on data from electronic health records could accurately predict self-harm risk after adolescent TBI. Further research is needed to assess the model's clinical utility and transportability. Use of a prognostic model may assist clinical decision making about risk and resource allocation after TBI.