Speakers: Prerna Dua, PhD

Prerna Dua, PhD

Professor
Department of Biomedical and Health Informatics
School of Medicine
University of Missouri-Kansas City

 

Dr. Prerna Dua is a Professor in the Department of Biomedical and Health Informatics at the University of Missouri–Kansas City and holds fellowship in American Medical Informatics Association (AMIA). Her research focuses on developing and applying computational and machine learning methods to analyze large-scale healthcare data for predictive modeling, clinical decision support, and population health improvement. She leads interdisciplinary projects in maternal health, chronic disease outcomes, and health disparities, with particular emphasis on explainable AI, fairness, and translational impact. A consistent theme throughout her work is the integration of machine learning with large-scale health datasets while addressing issues of imbalanced representation. She has been actively funded by NIH, HRSA, Louisiana Board of Regents and Louisiana Department of Health. She has served as a reviewer for NSF grants and for multiple journals.

Dr. Dua serves as a peer reviewer for Health Informatics program and is a member of Health Informatics Accreditation Council (HIAC) at the Commission on Accreditation for Health Informatics and Information Management Education (CAHIIM). She’s an active member of AMIA.

 

From Claims to Care: Predicting Healthcare Utilization for Medicaid Through Machine Learning

Tobacco use exacerbates COVID-19 severity by impairing lung function and increasing systemic inflammation, disproportionately affecting Medicaid enrollees who already face barriers to preventive care and greater environmental exposure. Despite this dual burden, few studies have examined how tobacco use and COVID-19 jointly influence healthcare utilization in this population using predictive analytics. Using Louisiana Medicaid claims data (January 2020–February 2023; 6.3 million encounters from 270,000 enrollees), we developed seven machine learning models to predict clinical service utilization. The most influential predictors were hospital length of stay, disability status, history of stroke, age, and geographic region, aligning with prior Medicaid research linking disability and previous utilization to future healthcare demand. Our findings demonstrate that Medicaid administrative data can effectively support predictive modeling for population-level decision-making while providing actionable insights to identify high-risk subgroups and optimize resource allocation.