M-BEHRT demonstrates improved prediction of disease-free survival in breast cancer patients, suggesting EHR utilization is critical.
Key Points
M-BEHRT achieves an AUC-ROC of 0.77 for predicting disease-free survival in breast cancer patients, outperforming traditional methods.
The approach utilizes electronic health records, modeling patient trajectories over time, thus capturing critical clinical data for analysis.
Retrospective analysis included 15,000 breast cancer patients, with M-BEHRT effectively identifying subsets of patients with particularly affected outcomes.
M-BEHRT, based on transformer architecture, showcases the potential of deep learning in extracting valuable patterns from EHRs for better patient care.