AI predicts suicide risk
Approximately every forty seconds a person dies from suicide and nearly 800,000 people die each year according to Our World in Data. A new meta-analysis study published in the Journal of Psychiatric Research shows that artificial intelligence (AI) machine learning outperforms traditional suicide risk prediction models, potentially improving suicide risk detection.
Artificial intelligence is used as a tool to help clinicians and healthcare providers diagnose and predict diseases and disorders at an early stage in hopes of early intervention for better outcomes. In 2022, the global artificial intelligence in healthcare market size value is an estimated USD 15.4 billion and is expected to increase at a compound annual growth rate (CAGR) of 38.4 percent during 2022 to 2030 to reach USD 208.2 billion by 2030 according to Grand View Research.
AI is also being used as an assistive tool for mental health clinicians and providers. The researchers affiliated with the Black Dog Institute and the Centre for Big Data Research in Health at the University of New South Wales sought to conduct a systematic review and meta-analysis of AI machine learning models in predicting longitudinal outcomes of not only suicide ideation, but also suicide attempts and death by suicide.
The researchers analyzed the potential covariates of AI model performance, including AI algorithm type, data, and outcomes of scientific studies from PsycINFO, PubMed, Embase, and Web of Science. The team customized a prediction model called the Risk of Bias Assessment Tool to help evaluate the risk of bias for each study.
The scientists evaluated 54 AI machine learning models’ ability to predict suicidal ideation, attempts, and behaviors. Overall, the research demonstrated that AI machine learning outperformed traditional clinical, theoretical and statistical risk prediction models. The AI machine models achieved a pooled AUC (area under the curve) of 0.86 and a specificity of 0.87.
“Findings suggest that machine learning has the potential to improve suicide risk detection, with pooled estimates of machine learning performance comparing favorably to performance of traditional suicide prediction models,” the researchers wrote.
