Key result
Feature-Token Clinical Transformer outperforms XGBoost for predicting in-hospital mortality, improving AUROC by ~0.03.
Why the study?
Conventional scores and flat-vector ML models may not preserve individual variable identity or context in heterogeneous EHR data during the first 24 h after ICU admission in critically ill heart failure patients.
Does a Feature-Token Clinical Transformer (FTCT) improve prediction of in-hospital all-cause death in critically ill adults with heart failure compared to conventional machine learning and clinical scores?
Cohort (n=20,993)
Yes
Does a Feature-Token Clinical Transformer (FTCT) improve prediction of in-hospital all-cause death in critically ill adults with heart failure compared to conventional machine learning and clinical scores?
Effect estimate: AUROC difference 0.032 (95% CI 0.014-0.050)
Absolute Event Rate: 0.804% vs 0.772%
A feature-token clinical transformer model utilizing first 24-hour ICU structured EHR data provides accurate, externally validated stratification of in-hospital mortality risk in critically ill heart failure patients, outperforming standard flat-vector machine learning models.
May support transformer models for HF ICU mortality prediction; leaves open clinical utility pending prospective validation.
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.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) conducted a cohort in Critically ill patients with heart failure (n=20,993). Feature-Token Clinical Transformer (FTCT) model vs. XGBoost and other machine learning/clinical scores was evaluated on In-hospital all-cause death (AUROC difference 0.032, 95% CI 0.014-0.050). The Feature-Token Clinical Transformer achieved an external AUROC of 0.804 for predicting in-hospital mortality, demonstrating better discrimination than XGBoost (AUROC difference 0.032).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: