What is the clinical evidence from this study?

Study design: Other. Population: Venous thromboembolism (VTE) risk (n=14808). Intervention: EMAX (attention-based AI model using LLMs). Primary outcome: Expert-adjudicated Caprini RAM (2010) risk prediction (AUC 0.9513).

What question did this study set out to answer?

This research aims to develop an interpretable AI model to predict VTE risk using electronic health records.

July 4, 2026Computer Methods in Biomechanics & Biomedical Engineering

Attention-based AI model EMAX accurately predicts expert-adjudicated Caprini VTE risk with ~0.95 AUC.

View Full Paper

Ask AI

Bookmark

Why the study?

Does an AI paradigm using LLMs and an attention-based model accurately predict expert-adjudicated Caprini RAM risk for VTE from EHR narratives?

Population

14,808 eligible electronic health record encounter records with documented expert Caprini scores, curated…

Design

Other

Key result

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.

Authors

AGAn Gong SWS S Wu SWShujing Wang

Discussion

Member takes

Overview

May enable auditable VTE risk scoring in EHRs; leaves open prospective outcome validation.

Key Points

This research aims to develop an interpretable AI model to predict VTE risk using electronic health records.
Utilized 113,836 raw encounter records and selected 14,808 based on expert Caprini scores for supervised learning.
Employed locally deployed LLMs to structure EHR narratives and an attention-based model named EMAX for risk prediction.
Evaluated model performance using the AUC metric on a designated test set.
EMAX achieved an AUC of 0.9513 on the test set, indicating high predictive performance.
The model provides a clinically aligned approach, enhancing current VTE risk assessment methods.
Demonstrated an end-to-end pathway for accurate and adoptable VTE risk stratification.

Structured PICO

Does an AI paradigm using LLMs and an attention-based model accurately predict expert-adjudicated Caprini RAM risk for VTE from EHR narratives?

Population

14,808 eligible encounter records with documented expert Caprini scores used to train and test a VTE risk prediction model.

Exposure

An interpretable AI paradigm using locally deployed Large Language Models (LLMs) to structure EHR narratives and an attention-based model (EMAX)

Outcome

Prediction of expert-adjudicated Caprini RAM (2010) risk

Main Result

Effect estimate: AUC 0.9513

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.

Cite This Study

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.

synapsesocial.com/papers/6a48a4c689561a0c2d78de56 https://doi.org/10.1080/10255842.2026.2698065

View Full Paper

Ask AI

Bookmark

July 4, 2026Computer Methods in Biomechanics & Biomedical Engineering

Attention-based AI model EMAX accurately predicts expert-adjudicated Caprini VTE risk with ~0.95 AUC.

View Full Paper

Ask AI

Bookmark

Why the study?

Does an AI paradigm using LLMs and an attention-based model accurately predict expert-adjudicated Caprini RAM risk for VTE from EHR narratives?

Population

14,808 eligible electronic health record encounter records with documented expert Caprini scores, curated…

Design

Other

Key result

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.

Authors

AGAn Gong SWS S Wu SWShujing Wang

Discussion

Member takes

Overview

May enable auditable VTE risk scoring in EHRs; leaves open prospective outcome validation.

Key Points

This research aims to develop an interpretable AI model to predict VTE risk using electronic health records.
Utilized 113,836 raw encounter records and selected 14,808 based on expert Caprini scores for supervised learning.
Employed locally deployed LLMs to structure EHR narratives and an attention-based model named EMAX for risk prediction.
Evaluated model performance using the AUC metric on a designated test set.
EMAX achieved an AUC of 0.9513 on the test set, indicating high predictive performance.
The model provides a clinically aligned approach, enhancing current VTE risk assessment methods.
Demonstrated an end-to-end pathway for accurate and adoptable VTE risk stratification.

Structured PICO

Does an AI paradigm using LLMs and an attention-based model accurately predict expert-adjudicated Caprini RAM risk for VTE from EHR narratives?

Population

14,808 eligible encounter records with documented expert Caprini scores used to train and test a VTE risk prediction model.

Exposure

An interpretable AI paradigm using locally deployed Large Language Models (LLMs) to structure EHR narratives and an attention-based model (EMAX)

Outcome

Prediction of expert-adjudicated Caprini RAM (2010) risk

Main Result

Effect estimate: AUC 0.9513

Cite This Study

synapsesocial.com/papers/6a48a4c689561a0c2d78de56 https://doi.org/10.1080/10255842.2026.2698065

View Full Paper

Ask AI

Bookmark