Transparent Early ICU Mortality Prediction with Clinical Transformer and Per-Case Modality Attribution
Not on the map. Why?
Abstract (arXiv)
Early identification of intensive care patients at risk of in-hospital mortality enables timely intervention and efficient resource allocation. Despite high predictive performance, existing machine learning approaches lack transparency and robustness, limiting clinical adoption. We present a lightweight, transparent multimodal ensemble that fuses physiological time-series measurements with unstructured clinical notes from the first 48 hours of an ICU stay. A logistic regression model combines predictions from two modality-specific models: a bidirectional LSTM for vitals and a finetuned ClinicalModernBERT transformer for notes. This traceable architecture allows for multilevel interpretability: feature attributions within each modality and direct per-case modality attributions quantifying how vitals and notes influence each decision. On the MIMIC-III benchmark, our late-fusion ensemble improves discrimination over the best single model (AUPRC 0.565 vs. 0.526; AUROC 0.891 vs. 0.876) while maintaining well-calibrated predictions. The system remains robust through a calibrated fallback when a modality is missing. These results demonstrate competitive performance with reliable, auditable risk estimates and transparent, predictable operation, which together are crucial for clinical use.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 22 of its 44 references that are in the map. First posted on arXiv 19 Nov 2025.
Builds on
22 of 44 references.
References · 22
- 2017Attention Is All You Needcited by 60,361
- 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
- 2017On Calibration of Modern Neural Networkscited by 4,960
- 2017Axiomatic Attribution for Deep Networkscited by 3,384
- 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
- 2022Large Language Models Encode Clinical Knowledgecited by 1,830
- 2018An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modelingcited by 1,476
- 2019Publicly Available Clinical BERT Embeddingscited by 1,002
- 2019Multimodal Transformer for Unaligned Multimodal Language Sequencescited by 645
- 2018Scalable and accurate deep learning for electronic health recordscited by 280
- 2023Multimodal Deep Learningcited by 258
- 2019GRU-ODE-Bayes: Continuous modeling of sporadically-observed time seriescited by 231
10 more
- 2019What Clinicians Want: Contextualizing Explainable Machine Learning for Clinical End Usecited by 135
- 2017Attend and Diagnose: Clinical Time Series Analysis using Attention Modelscited by 105
- 2015Learning to Diagnose with LSTM Recurrent Neural Networkscited by 99
- 2021Self-Supervised Transformer for Sparse and Irregularly Sampled Multivariate Clinical Time-Seriescited by 96
- 2022Clinical-Longformer and Clinical-BigBird: Transformers for long clinical sequencescited by 69
- 2022A Multimodal Transformer: Fusing Clinical Notes with Structured EHR Data for Interpretable In-Hospital Mortality Predictioncited by 48
- 2019Using Clinical Notes with Time Series Data for ICU Managementcited by 38
- 2024Advancing Anomaly Detection: Non-Semantic Financial Data Encoding with LLMscited by 3
- 2024Enhancing In-Hospital Mortality Prediction Using Multi-Representational Learning with LLM-Generated Expert Summariescited by 2
- 2025Before the Clinic: Transparent and Operable Design Principles for Healthcare AIcited by 1
Built on by
Citing papers per year: 2026: 1.
1 paper.