This project utilizes data-driven optimization and machine learning methods to improve the quality of inpatient fall risk assessment and decision-making. Our team has developed novel optimization models to increase the sensitivity and specificity of identifying patients at high risk for falls, while maintaining interpretable model structure using the Johns Hopkins Fall Risk Assessment Tool (JHFRAT) and additional electronic health record (EHR) data. The new models account for rarity of falls and uncertainty in true underlying risk due to risk-obscuring interventions. We are continuing to work on data-driven methods for understanding the relationships between EHR measures of risk, fall prevention interventions, and fall incidence in order to provide automated clinical decision support.

Publications
An interpretable data-driven approach to optimizing clinical fall risk assessment
Fardin Ganjkhanloo, Emmett Springer, Erik H Hoyer, Daniel L Young, Holley Farley, Kimia Ghobadi
PLOS Digital Health
Joint Score-Threshold Optimization for Interpretable Risk Assessment Under Partial Supervision
Fardin Ganjkhanloo, Emmett Springer, Erik H Hoyer, Daniel L Young, Kimia Ghobadi
arXiv:2510.21934
Falls in Hospitals: Challenging Traditional Risk Assessments With New Insights Into Patient Mobility
Erik H Hoyer, Daniel L Young, Chi Zhang, Elizabeth Colantuoni, Kimia Ghobadi
Journal of Advanced Nursing
Association of Longitudinal Mobility Levels in the Hospital and Injurious Inpatient Falls
Erik H Hoyer, Daniel L Young, Victor Ke, Jia Yi Zhang, Elizabeth Colantuoni, Holley Farley, Anton Dahbura, Kimia Ghobadi
American Journal of Physical Medicine & Rehabilitation