Predictive Model Finds Social Conflict Strongest Predictor of Teenage Mental Health Disorders
Mental health disorders among U.S. pre-teens and teenagers are on the rise. Determining who is most at risk remains challenging due to the complex interplay of genetic and lifestyle factors, even with advanced predictive models. Researchers from WashU Medicine analyzed an enormous data set and found that social conflicts were the strongest predictors of short- and long-term mental health issues.
The team — which included co-senior author Aristeidis Sotiras, PhD, assistant professor of radiology at WashU Mallinckrodt Institute of Radiology (MIR) and the Institute for Informatics, Data Science & Biostatistics — utilized data including neuroimaging scans, psychological tests, and personal and family mental health histories from participants in the Adolescent Brain Cognitive Development study. With this data, they created a machine learning model that predicts current and future mental health symptoms as well as changes in symptoms over time. The results, published in Nature Mental Health, show that social conflicts — particularly family fighting and reputational damage or bullying from peers — were the strongest predictors. Additionally, the research highlighted sex differences in how boys and girls experience stress from peer conflict, emphasizing the need for nuanced assessments of social stressors in teens.
Sotiras noted that machine learning enables researchers to analyze highly dimensional data and identify patterns that predict outcomes. “It’s a powerful way to move beyond single-factor explanations toward a more comprehensive, data-driven understanding of risk,” he said in a press release. “As we’re developing better computational models, it’s important to make sure our predictions are grounded in biologically meaningful information.”
Read more from WashU Medicine.
Image: Sara Moser, WashU Medicine
