DT-Transformer: A Foundation Model for Disease Trajectory Prediction on a Real-world Health System

2026-05-13

AI Summary

This study addresses the limited generalizability of existing disease trajectory prediction models, which typically rely on single-center data and struggle to perform in real-world, multi-institutional clinical settings. To overcome this, the authors propose DT-Transformer, a foundational disease trajectory model pretrained on a large-scale multicenter dataset encompassing 170 million structured electronic health records from 1.7 million patients across 11 healthcare institutions. Built upon the Transformer architecture, DT-Transformer enables next-event prediction across diverse disease categories. Evaluated on 896 disease categories stratified by age and sex, the model achieves a median AUC of 0.871, with all categories significantly outperforming random chance. Robust performance in both retrospective and prospective validations demonstrates markedly improved generalizability and clinical applicability.

Abstract

Accurate disease trajectory prediction is critical for early intervention, resource allocation, and improving long-term outcomes. While electronic health records (EHRs) provide a rich longitudinal view of patient health in clinical environments, models trained on curated research cohorts may not reflect routine deployment settings, and those trained on single-hospital datasets capture only fragments of each patient's trajectory. This highlights the importance of leveraging large, multi-hospital health systems for training and validation to better reflect real-world clinical complexity. In this work, we develop DT-Transformer, a foundation model trained on 57.1M structured EHR entries over 1.7M patients from Mass General Brigham (MGB), spanning 11 hospitals and a broad network of outpatient clinics. DT-Transformer achieves strong discrimination in both held-out and prospective validation settings. Next-event prediction achieves a median age- and sex-stratified AUC of 0.871 across 896 disease categories, with all categories exceeding AUC 0.5. These results support health system-scale training as a path toward foundation models suited to real-world clinical forecasting.

Problem

• disease trajectory prediction
• electronic health records
• real-world clinical complexity
• multi-hospital health systems
• foundation model

Innovation

• DT-Transformer
• disease trajectory prediction
• foundation model
• electronic health records
• multi-hospital EHR