Explainable Transformer Models for Clinical Prediction Tasks on Structured Electronic Health Records

Researchers introduced BERT-LER, a BERT-style model for coded electronic health records (EHRs) that encodes laboratory test results as discrete tokens via percentile-based binning and integrates Integrated Gradients for token-level explanations. Trained on a de-identified dataset of 75 million patients, BERT-LER achieved predictive performance competitive with, and often exceeding, public benchmarks on the EHRShot suite and an asthma severity progression study, while providing attributions aligned with clinically known risk factors.

arXiv:2608.20315v1 Announce Type: new Abstract: Predictive models over structured electronic health records EHRs remain central to machine learning for healthcare, but few have jointly emphasized quantitative laboratory information and interpretability with respect to input medical events. We present BERT-LER, a BERT-style model for coded EHR timelines pretrained and fine-tuned from a de-identified EHR dataset of 75 million patients, that encodes laboratory test results as discrete tokens while retaining graded information through percentile-based binning, paired with Integrated Gradients for token-level attributions grounded in the input EHR sequence. We benchmark our approach on the public EHRShot benchmark suite and on an asthma severity progression study based on real-world data. This addresses a methodological gap in EHR foundation-style modeling by unifying laboratory value representation and explainability in a single framework, while assessing whether both predictive performance and explanations generalize beyond standard clinical prediction tasks. Across EHRShot and asthma tasks, BERT-LER achieves predictive performance that is competitive with, and on laboratory-related tasks often exceeds, publicly available benchmark models, and provides attributions that align with clinically known risk factors. Our architecture and explainability approach can be applied to many therapeutic areas and prediction tasks using language models trained on structured EHRs.