Key result
Attention-based AI model EMAX accurately predicts expert-adjudicated Caprini VTE risk with ~0.95 AUC.
Why the study?
Manual VTE risk assessment is often inconsistent, while many high-performing AI models remain difficult to audit clinically.
Does an AI paradigm using LLMs and an attention-based model accurately predict expert-adjudicated Caprini RAM risk for VTE from EHR narratives?
Does an AI paradigm using LLMs and an attention-based model accurately predict expert-adjudicated Caprini RAM risk for VTE from EHR narratives?
Answers are grounded in this paper and cite their sources.
An AI framework combining LLMs and an attention-based model can accurately predict expert-adjudicated VTE risk scores from EHR narratives, offering a scalable and interpretable tool for clinical risk stratification.
Manual VTE risk assessment is often inconsistent, while many high-performing AI models remain difficult to audit clinically. We propose an interpretable and clinically aligned paradigm that uses locally deployed LLMs to structure EHR narratives and an attention-based model (EMAX) to predict expert-adjudicated Caprini RAM (2010) risk. From 113,836 raw encounter records, we curated 14,808 eligible records with documented expert Caprini scores for supervised learning. On the test set, EMAX achieved an AUC of 0.9513. The framework provides an end-to-end pathway for accurate and adoptable VTE risk stratification.
No takes yet. Share an insight, caveat, or question.
Gong et al. (2026) studied Venous thromboembolism (VTE) risk (n=14,808). EMAX (attention-based AI model using LLMs) was evaluated on Expert-adjudicated Caprini RAM (2010) risk prediction (AUC 0.9513). An attention-based AI model (EMAX) using locally deployed LLMs accurately predicted expert-adjudicated Caprini VTE risk, achieving an AUC of 0.9513 on the test set.
Synapse has enriched 10 related papers. Consider them for comparative context: