Feature-token clinical transformer for 24-h landmark mortality risk stratification in critically ill patients with heart failure: development and multicenter external validation
- 1Department of Cardiology, The First Affiliated Hospital of Soochow University
Abstract
Background: The first 24 h after ICU admission in critically ill patients with heart failure yield heterogeneous EHR evidence across vital signs, laboratory tests, treatments, organ support, and nurse-charted monitoring. Conventional scores and flat-vector ML models may not preserve individual variable identity or context.
Methods: We developed FTCT, encoding each predictor recorded during the first 24 h after ICU admission as a clinically tagged token (value, missingness, identity, group). MIMIC-IV (2008–2019) served for development, validation, and internal testing; eICU-CRD (2014–2015) served for external validation. Included were adults with heart failure, a complete 24-h ICU observation window, and known discharge status. The endpoint was in-hospital all-cause death.
Results: MIMIC-IV included 11,156 stays and 1,484 deaths, partitioned into training ( n = 7,809), validation ( n = 1,673), and internal test ( n = 1,674) sets. eICU-CRD included 9,837 stays and 1,319 deaths. FTCT achieved an AUROC of 0.849 (95% CI, 0.835–0.863) internally and 0.804 (0.788–0.820) externally. Patient-paired AUROC differences vs. XGBoost were 0.027 (0.011–0.043) internally and 0.032 (0.014–0.050) externally. A matched FT-Transformer-style baseline achieved an external AUROC of 0.781 (0.760–0.801). In eICU-CRD, FTCT achieved an AUPRC of 0.296 (0.278–0.314), Brier score of 0.091 (0.085–0.097), and expected calibration error of 0.036 (0.030–0.042). In a post hoc cohort-specific descriptive operating analysis targeting approximately 85% specificity, external sensitivity was 0.662 (0.636–0.687), positive predictive value 0.406 (0.385–0.427), negative predictive value 0.942 (0.937–0.947), and F1 score 0.503. Removing feature-token encoding produced the largest ablation loss in external AUROC (−0.031).
Conclusions: FTCT combines established feature-token Transformer modeling with clinical-group embeddings, explicit missingness representation, and group-aware pooling for 24-h landmark mortality risk stratification from structured EHR data recorded during the first 24 h after ICU admission. Its external validation and ablation results support prospective investigation, but do not yet demonstrate clinical utility, workflow integration, admission-time applicability, or autonomous use.
Related
- Development and validation of a simple clinical prediction model for post-stroke seizures in critically ill patients: a multicenter retrospective cohort study
- Development and validation of a risk stratification model for mild cognitive impairment among older adults with possible sarcopenia integrating traditional Chinese medicine constitution: a multicenter cross-sectional study
- Development and validation of an interpretable machine learning model for risk stratification among individuals with latent tuberculosis infection: a multicenter case-control study
- Identification and validation of early vital sign trajectory subphenotypes for risk stratification in sepsis-associated liver injury: a multicenter retrospective cohort study
- Development and Internal Validation of a Preoperative Risk Stratification Model for Difficult Laparoscopic Cholecystectomy in Acute Cholecystitis
BibTeX
@article{li2026featuretoken,
title = {{Feature-token clinical transformer for 24-h landmark mortality risk stratification in critically ill patients with heart failure: development and multicenter external validation}},
author = {Mianxian Li and Yuefei Zhou and Chun Li},
journal = {Frontiers in Cardiovascular Medicine},
year = {2026},
month = oct,
doi = {10.3389/fcvm.2026.1945947},
url = {https://www.frontiersin.org/articles/10.3389/fcvm.2026.1945947/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Feature-token clinical transformer for 24-h landmark mortality risk stratification in critically ill patients with heart failure: development and multicenter external validation
AU - Mianxian Li
AU - Yuefei Zhou
AU - Chun Li
PY - 2026
DA - 2026/10/15
JO - Frontiers in Cardiovascular Medicine
DO - 10.3389/fcvm.2026.1945947
UR - https://www.frontiersin.org/articles/10.3389/fcvm.2026.1945947/full
AB - Background: The first 24 h after ICU admission in critically ill patients with heart failure yield heterogeneous EHR evidence across vital signs, laboratory tests, treatments, organ support, and nurse-charted monitoring. Conventional scores and flat-vector ML models may not preserve individual variable identity or context. Methods: We developed FTCT, encoding each predictor recorded during the first 24 h after ICU admission as a clinically tagged token (value, missingness, identity, group). MIMIC-IV (2008–2019) served for development, validation, and internal testing; eICU-CRD (2014–2015) served for external validation. Included were adults with heart failure, a complete 24-h ICU observation window, and known discharge status. The endpoint was in-hospital all-cause death. Results: MIMIC-IV included 11,156 stays and 1,484 deaths, partitioned into training ( n = 7,809), validation ( n = 1,673), and internal test ( n = 1,674) sets. eICU-CRD included 9,837 stays and 1,319 deaths. FTCT achieved an AUROC of 0.849 (95% CI, 0.835–0.863) internally and 0.804 (0.788–0.820) externally. Patient-paired AUROC differences vs. XGBoost were 0.027 (0.011–0.043) internally and 0.032 (0.014–0.050) externally. A matched FT-Transformer-style baseline achieved an external AUROC of 0.781 (0.760–0.801). In eICU-CRD, FTCT achieved an AUPRC of 0.296 (0.278–0.314), Brier score of 0.091 (0.085–0.097), and expected calibration error of 0.036 (0.030–0.042). In a post hoc cohort-specific descriptive operating analysis targeting approximately 85% specificity, external sensitivity was 0.662 (0.636–0.687), positive predictive value 0.406 (0.385–0.427), negative predictive value 0.942 (0.937–0.947), and F1 score 0.503. Removing feature-token encoding produced the largest ablation loss in external AUROC (−0.031). Conclusions: FTCT combines established feature-token Transformer modeling with clinical-group embeddings, explicit missingness representation, and group-aware pooling for 24-h landmark mortality risk stratification from structured EHR data recorded during the first 24 h after ICU admission. Its external validation and ablation results support prospective investigation, but do not yet demonstrate clinical utility, workflow integration, admission-time applicability, or autonomous use.
SN - 2297-055X
ER -