SHM Converge Scientific Abstract ePoster Gallery-Explainable Ai-Empowered Analysis of Diverse Infection Outcomes Among People With Type 1, Type 2, and Prediabetes

Pdf Summary

This study used explainable artificial intelligence, specifically Bayesian networks, to examine how infection risks differ among people with type 1 diabetes (T1D), type 2 diabetes (T2D), and prediabetes (PRED) using University of Utah Health electronic health record data.<br /><br />The researchers aimed to improve understanding of infection risk, which is difficult to quantify because infectious diseases are highly diverse and complex. They compared the diabetes groups with nondiabetic and general populations, and analyzed both broad infection categories and 1,401 individual outcomes.<br /><br />Key findings showed that T1D, T2D, and PRED were each associated with a manyfold increase in lifetime risk for most infectious diseases. The diabetes groups had higher risk across all major organ system infection categories, all pathogen types, and severe infections. Risk patterns were broadly similar across the three diabetes subgroups: T1D and T2D shared increased risk for 57% of individual outcomes, T1D and PRED for 55%, and T2D and PRED for 63%.<br /><br />The study also found that age mattered: comparative risk for severe infections and for each pathogen type decreased with age. In addition, significant differences in infection risk were observed across sociodemographic subgroups of patients with T1D and/or T2D, suggesting important disparities.<br /><br />Overall, the study demonstrates that explainable AI can identify both known and previously unrecognized infection risks in diabetes populations, and may help guide future research and clinical risk assessment.

Asset Subtitle

Boomer Olsen

Meta Tag

Author List Boomer Olsen, Javier Hernandez, Karen Eilbeck, Mark Yandell, Martin Tristani-Firouzi

Category Research

Concept Bayesian Network

Concept Infection Outcome

Concept Infection Risk

Concept Type 1 Diabetes

Concept Type 2 Diabetes

Distinguished Non-Finalist

Presenter Organization University of Utah

Presenting Author Boomer Olsen

Track Outcomes Research

Keywords

explainable artificial intelligence

Bayesian networks

diabetes infection risk

type 1 diabetes

type 2 diabetes

prediabetes

electronic health records

infectious diseases

risk disparities

clinical risk assessment

Bayesian Network

Infection Outcome

Infection Risk

Type 1 Diabetes

Type 2 Diabetes