Innovative Explainable AI Tool Predicts Diseases in Early Stages

Researchers from the University of Utah's Department of Psychiatry and the Huntsman Mental Health Institute have unveiled a groundbreaking open-source software called RiskPath. This innovative toolkit leverages explainable artificial intelligence (XAI) to forecast whether individuals are likely to develop chronic and progressive diseases years before any symptoms manifest. Published in the journal Patterns, the study highlights how RiskPath could revolutionize preventive healthcare by providing highly accurate predictions, ranging from 85% to 99%, based on extensive analysis of health data collected over several years.

Traditional disease prediction systems often fall short in handling long-term health data, typically identifying at-risk individuals correctly only about half to three-quarters of the time. In contrast, RiskPath employs advanced time-series AI algorithms that not only predict disease risk but also explain the contributing factors in an understandable manner for clinicians and patients alike. This transparency helps unravel how different risk factors evolve and influence disease development over time.

The significance of this technology lies in its ability to identify at-risk individuals early, allowing for targeted prevention strategies. Nina de Lacy, MD, MBA, and first author of the study, emphasized the potential impact:  by early detection and understanding which risk factors matter most at various life stages, more effective and personalized prevention measures can be implemented. This approach centers on proactive health management rather than reactive treatment.

Validated across three large-scale longitudinal cohorts with thousands of participants, RiskPath has successfully predicted eight different health conditions, including depression, anxiety, ADHD, hypertension, and metabolic syndrome. Notably, researchers discovered that even with just ten key factors—such as screen time and executive function—most conditions could be predicted with high accuracy, simplifying potential clinical adoption.

The system also provides intuitive visualizations that highlight critical periods in a person's life that contribute most to disease risks, thus identifying optimal windows for intervention. The research team is actively exploring integration with clinical decision support tools, preventive health programs, and investigating the neural mechanisms associated with mental illnesses. Future plans include expanding the application to additional diseases and more diverse populations.

This innovative approach signifies a major advancement in predictive medicine, promising more personalized, effective, and preventive healthcare in the near future.

For more information, see the original publication: Nina de Lacy et al, RiskPath: Explainable deep learning for multistep biomedical prediction in longitudinal data, Patterns (2025).