![[Adobe Stock]](https://sp-ao.shortpixel.ai/client/to_webp,q_glossy,ret_img,w_750,h_375/https://www.psypost.org/wp-content/uploads/2024/07/sad-depressed-teen-girl-750x375.jpg)
An analysis of data from over 11,000 adolescent participants in the Adolescent Brain and Cognitive Development (ABCD) study found that sleep disturbances may be the most influential predictor of future psychopathology. Sleep issues were more predictive than adverse childhood experiences and family mental health history. Neuroimaging data did not improve the ability to forecast mental health risks. The findings were published in Nature Medicine.
As science and technology have rapidly advanced over the past century, medicine has found ways to cure or manage many once-deadly diseases. Conditions such as tuberculosis, pneumonia, typhoid fever, measles, and polio—which previously claimed millions of lives—are now treatable or nearly eradicated thanks to medical and pharmaceutical breakthroughs.
As a result, attention has increasingly shifted to medical conditions that remain difficult to treat. Mental health disorders are among the most prominent in this category. Despite some progress in treatment, many individuals continue to struggle with chronic mental health conditions that show limited response to current interventions.
Prevention has emerged as a promising strategy. Identifying individuals at risk before symptoms become severe could enable early support and reduce long-term impacts. This has led scientists to seek reliable methods for predicting who is most likely to develop psychiatric disorders.
In the new study, lead author Elliot D. Hill and his colleagues developed machine learning models to predict mental health risk based on psychosocial and neurobiological data.
The researchers used data from over 11,000 participants in the ABCD study, a large longitudinal project in the United States. Participants were between 9 and 15 years old at the time of assessment, and approximately 48% were female. They were followed for one to three years after enrollment.
Participants completed various psychosocial assessments and underwent magnetic resonance imaging (MRI). The researchers trained several machine learning models to predict future psychiatric risk based on the collected data.
A model trained on participants’ current symptoms was highly accurate in predicting which adolescents would transition into a high-risk category for psychiatric illness within the following year. Another model that relied solely on potential underlying causes—such as sleep problems, family dynamics, and adversity—performed slightly less well but still achieved respectable accuracy, even without symptom data.
Sleep disturbances emerged as the strongest predictor of increased psychiatric risk, surpassing other established factors such as adverse childhood experiences and family mental health history. Adding MRI data to the models did not improve predictive performance.
“These findings suggest that artificial intelligence models trained on readily available psychosocial questionnaires can effectively predict future psychiatric risk while highlighting potential targets for intervention. This is a promising step toward artificial intelligence-based mental health screening for clinical decision support systems,” the study authors concluded.
The study enhances scientific understanding of mental health risk factors and how early risk might be estimated. However, it is important to note that the findings are based on statistical associations. The study’s design does not permit conclusions about causality.
The paper, “Prediction of mental health risk in adolescents,” was authored by Elliot D. Hill, Pratik Kashyap, Elizabeth Raffanello, Yun Wang, Terrie E. Moffitt, Avshalom Caspi, Matthew Engelhard, and Jonathan Posner.
