We present GenHAI, a hybrid generative modeling framework designed to predict and reason about hospital-acquired MRSA infections during a patient’s continuous hospital stay. The model supports time-evolving risk estimation, forecasting of future test outcomes, and counterfactual what-if analysis to aid clinical decision-making and infection control.

KAMELEON is a knowledge-augmented multimodal framework that integrates structured EHR data, clinical notes, and external medical knowledge using Large Language Models (LLMs) to improve hospital risk prediction. The framework emphasizes interpretability and robustness, enabling human-readable rationales alongside accurate predictive modeling.

This work introduces a hybrid spatio-temporal framework, VaxHesSTL, that combines a Graph Neural Network (GNN) to model spatial relationships between ZIP codes with a Recurrent Neural Network (RNN) to capture temporal dynamics, enabling ZIP-code–level prediction of vaccine hesitancy.

The model leverages an aggregated contact network derived from population mobility rather than geographic proximity alone, and uses an active learning strategy to accurately forecast hesitancy even when ground-truth data is available for only a subset of regions.