Predicting self-harm within six months after initial presentation to youth mental health services: A machine learning study

Abstract

Background A priority for health services is to reduce self-harm in young people. Predicting self-harm is challenging due to their rarity and complexity, however this does not preclude the utility of prediction models to improve decision-making regarding a service response in terms of more detailed assessments and/or intervention. The aim of this study was to predict self-harm within six-months after initial presentation. Method The study included 1962 young people (12–30 years) presenting to youth mental health services in Australia. Six machine learning algorithms were trained and tested with ten repeats of ten-fold cross-validation. The net benefit of these models were evaluated using decision curve analysis. Results Out of 1962 young people, 320 (16%) engaged in self-harm in the six months after first assessment and 1642 (84%) did not. The top 25% of young people as ranked by mean predicted probability accounted for 51.6% - 56.2% of all who engaged in self-harm. By the top 50%, this increased to 82.1%-84.4%. Models demonstrated fair overall prediction (AUROCs; 0.744–0.755) and calibration which indicates that predicted probabilities were close to the true probabilities (brier scores; 0.185–0.196). The net benefit of these models were positive and superior to the ‘treat everyone’ strategy. The strongest predictors were (in ranked order); a history of self-harm, age, social and occupational functioning, sex, bipolar disorder, psychosis-like experiences, treatment with antipsychotics, and a history of suicide ideation. Conclusion Prediction models for self-harm may have utility to identify a large sub population who would benefit from further assessment and targeted (low intensity) interventions. Such models could enhance health service approaches to identify and reduce self-harm, a considerable source of distress, morbidity, ongoing health care utilisation and mortality.

Assessment of machine learning algorithms in national data to classify the risk of self-harm among young adults in hospital: a retrospective study

“…The AUC-ROCs were 0.744–0.755 but the study was limited by sample size (n=1962 patients) and class imbalance (320 [16%] self-harm vs 1642 [84%] no self-harm). 24…”

“…The AUC-ROCs were 0.744-0.755 but the study was limited by sample size (n=1962 patients) and class imbalance (320 [16%] self-harm vs 1642 [84%] no self-harm). 24 In this study we focused on the use of commissioning level data, but there has also been interest in using clinical text from electronic health-care records in the USA. Analysis of these records has yielded high accuracy but the approach is only applicable to hospitals that universally use electronic records, which is rare in the UK.…”

Predicting Lifetime Suicide Attempts in a Community Sample of Adolescents Using Machine Learning Algorithms

“…Even at the early age of 9-10 years, 8.4% of 7,944 children interviewed in the US-based Adolescent Brain and Cognitive Development study reported having past or current suicide ideation and 1.3% confirmed attempted suicides (Janiri et al, 2020). Non-fatal self-harm and previous suicide attempts considerably increase the risk of subsequent suicide attempts (Beckmann et al, 2018;Geulayov et al, 2019;Iorfino et al, 2020), making it crucial to better understand the mechanisms leading to suicidal behaviors at an early age to prevent suicides. However, previous efforts to predict suicide were often unsatisfactory, mainly because effect sizes of individual factors that have been shown to correlate with suicidal behaviors were only small to moderate (Franklin et al, 2017).…”

Self-Harm in Aviation Medicine—A Complex Challenge During a Pandemic

“…The clinician should be aware of common factors predisposing to self-harm. In the recent Australian cohort study involving 1,962 individuals aged 12–30 years who visited a mental health clinic, researchers examined the predictors of within 6 months repeated self-harm ( 34 ). The three strongest predictors were a history of self-harm, younger age and social and occupational functioning measured by low value in the Social and Occupational Functioning Assessment Scale (SOFAS) ( 35 ).…”