Developing a multi-domain EHR foundation model for predicting Hepatitis B liver disease: a clinical perspective
Weis, C. V.; Grazioli, F.; Visona, G.; Kania, A.; Burg, M. F.; Horn, M.; Golob, J. L.; Schwab, P.
Show abstract
Foundation models trained on patient electronic health records (EHRs) hold promise for transforming clinical care by enabling effective decision support and personalized healthcare delivery, but have been limited by a focus on intensive care objectives. Here we present a multi-domain transformer-based EHR foundation model designed to predict two liver disease outcomes in patients with Chronic Hepatitis B, an infection characterized by diverse and uncertain medical trajectories. Through case studies employing attention maps, we demonstrate that the transformer model identifies patterns similar to one-liners employed by clinical staff and depends on distinct clinical events to estimate disease progression. Our findings underscore both the utility and challenges of EHR foundation models in clinical care and the necessity to evaluate EHR-models on less-regimented diseases.
Matching journals
The top 7 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
Similar papers in this journal
- Modular Clinical Decision Support Networks (MoDN)—Updatable, Interpretable, and Portable Predictions for Evolving Clinical Environments 96%
- Explainable deep learning for disease activity prediction in chronic inflammatory joint diseases 94%
- Uncovering the effects of model initialization on deep model generalization: A study with adult and pediatric chest X-ray images 93%
Similar papers in this journal
- EHR Foundation Models Improve Robustness in the Presence of Temporal Distribution Shift 97%
- Developing Machine Learning Models for Predicting Intensive Care Unit Resource Use During the COVID-19 Pandemic 94%
- Mitigating Machine Learning Bias Between High Income and Low-Middle Income Countries for Enhanced Model Fairness and Generalizability 94%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.