New Explainable AI Tool Enhances Mortality Risk Prediction for Diabetes

AI Model Offers Personalized Mortality Risk Assessment for Diabetic Patients

Moscow: A team of scientists at the Institute of Biology and Biomedicine at Lobachevsky University in the Russian Federation has developed an advanced artificial intelligence (AI) model capable of predicting the risk of all-cause mortality among patients with diabetes, while also providing clear explanations of the factors behind its predictions for medical specialists.

According to a statement issued by the university, the AI model was trained using a comprehensive dataset that included physiological indicators of more than 550 diabetic patients, monitored over a 17-year period.

Mikhail Ivanenko, lead researcher and Director of the Institute of Biology and Biomedicine, explained that the innovation of the study lies in the development of a highly accurate and interpretable predictive tool. He noted that the system enables the identification of complex interrelationships among dozens of clinical and biological parameters.

He added that the analysis demonstrated that age, duration of diabetes, and the number of disease-related complications represent the most influential risk factors for mortality. The model can also generate personalized risk profiles, explaining how specific indicators contribute to an individual’s risk. For example, it can show that a 68% increased mortality risk may be primarily driven by elevated creatinine levels, advanced age, and the presence of multiple complications.

The research also highlighted the importance of several lesser-known biomarkers, including NT-proBNP, which reflects myocardial stress; creatinine, an indicator of kidney function; and certain N-glycan structures found in blood serum, which are associated with immune regulation and biological aging.

The researchers emphasized that this development represents a significant step in transforming explainable artificial intelligence from a theoretical concept into a practical clinical tool, capable of supporting physicians in diagnostic decision-making. The innovation holds promising potential to improve disease management and enhance the quality of life for millions of people worldwide living with diabetes.