SALT LAKE CITY

For quite some time, the medical community has aimed to gain deeper insights into the long-term behavior of diseases, hoping it will lead to improved prevention and earlier treatment. Nina de Lacy, a psychiatry professor and part of the One-U Responsible AI Initiative’s executive committee, pointed out that chronic and progressive diseases account for roughly 90% of healthcare spending in the country, along with a significant portion of illness and death.

Researchers at the University of Utah have made noteworthy progress in this area by introducing an open-source software toolkit designed to predict the onset of progressive and chronic diseases years before symptoms actually manifest.

Called RiskPath, this technology analyzes health data trends accumulated over several years to pinpoint individuals at risk with an impressive accuracy rate ranging from 85% to 99%, according to research supported by the National Institute of Mental Health and shared last week by the university’s Department of Psychiatry and Huntsman Mental Health Institute.

The program employs explainable AI, which aims to clarify intricate decision-making processes in ways that are understandable to people. “Explainability means, can I explain enough about how AI accomplished this prediction such that it becomes understandable to humans?” de Lacy stated, highlighting what RiskPath aims to achieve.

She mentioned that a long-standing challenge in biomedicine is creating models and analyzing longitudinal data—data collected over multiple time points. “One major application of longitudinal data is understanding how individuals develop over time,” de Lacy noted. “RiskPath focuses on understanding progressive or chronic diseases, many of which significantly impact human health.”

The findings revealed that current medical systems for predicting risks using longitudinal data often do not perform well, accurately identifying at-risk individuals only about half to three-quarters of the time. In contrast to existing systems, RiskPath utilizes advanced time-series AI algorithms that provide critical insights into how risk factors interact and their significance evolves throughout the disease’s progression.

“By identifying high-risk people before symptoms show up or early in the disease’s development, we can focus on which risk factors are most relevant at different life stages. This allows us to create more targeted preventive strategies, which I believe is arguably the most vital aspect of healthcare today,” de Lacy explained.

De Lacy and her team validated RiskPath through assessments of three significant long-term patient groups with thousands of participants, successfully predicting eight different conditions, including depression, anxiety, ADHD, hypertension, and metabolic syndrome.

The technology presents several key benefits:

• Better understanding of disease progression: RiskPath can illustrate how various risk factors change over time, emphasizing critical moments for intervention. For instance, the study highlighted the growing importance of screen time and executive function as risk factors for ADHD as children near their teenage years.
• Simplified risk assessment: Although RiskPath can process numerous health variables, researchers found that they could predict most conditions accurately using just 10 essential factors, making it easier to implement in clinical settings.
• Visual risk representation: The system offers clear visualizations that indicate which life stages contribute most to disease risk, allowing researchers to identify optimal times for preventive measures.

While RiskPath serves primarily as a research tool aimed at improving risk stratification models, de Lacy hopes it will eventually find a place in healthcare settings to enhance disease management. “Some may employ it to create models for health care, and that’s our hope. A significant goal of my lab is to develop tools that enhance risk stratification. We’re really focused on prevention,” she remarked. “Ultimately, the aim of RiskPath and similar tools is to assist in developing better risk stratification and decision support systems.”

“These aim to help clinicians—and hopefully patients too—better understand their risk for chronic or progressive diseases sooner,” she concluded.