Predicting future organ dysfunction in Intensive Care Unit (ICU) patients is critical for early clinical intervention, yet existing machine learning approaches have largely treated the Sequential Organ Failure Assessment (SOFA) score as an input to binary mortality prediction rather than as a continuous clinical outcome in its own right. We investigate the extent to which a Temporal Convolutional Net work (TCN) can predict next-day SOFA scores from multivariate ICU time-series data extracted from MIMIC-IV, characterise the relative contribution of each organ system to total SOFA variance and deterioration, and identify distinct trajectory patterns across ICU stays. A residual TCN trained on three-day sliding windows achieved a five-fold cross-validation R2 of 0.740 +- 0.013 and MAE of 1.431 +- 0.022, outperforming a naive persistence baseline on RMSE and R2. SHAP interpretability analysis revealed that the model functions primarily as a severity-anchoring mechanism rather than a true sequence model, with predictions dominated almost entirely by the most recent observation day. Cardiovascular dysfunction emerged as the strongest discriminator of both cross-sectional severity and acute deterioration, and unsupervised trajectory clustering identified two clinically meaningful phenotypes, an improving group (58.9%) and a persistently severe group (41.1%), differentiated by cardiovascular, hepatic, coagulation, and renal involvement. We conclude that TCNs can extract meaningful predictive signal from ICU physiological data, but that short input windows and complete-case selection bias currently limit their clinical utility, motivating future work on longer input horizons, alternative missing-data strategies, and external validation.
Figures & tables
MAE
RMSE
R2
CV mean ± SD
1.431±0.022
1.882±0.027
0.740±0.013
Holdout
1.462
1.906
0.715
Persistence baseline
1.401
1.933
0.707
Table 1: TCN performance across five-fold ICU stay-level cross-validation, single holdout evaluation, and the naïve persistence baseline.
Figure 1: MAE, RMSE, and R2 across five cross-validation folds, with mean performance shown as a dashed line.
Figure 2: Predicted vs actual SOFA Evaluation (left), error distribution (centre), and training history (right).
Figure 3: SHAP feature importance (left) and temporal SHAP heatmap by feature and time step (right).
Figure 4: SOFA subscore variance contribution (left) and Pearson correlation with total SOFA score (right).
Table 4: TCN regression performance stratified by SOFA severity band (supports Section 4.2 ).
Figure 8: Predicted vs actual SOFA trajectories for six sampled ICU admissions, illustrating the model’s smoothing bias under rapid deterioration or recovery.
Figure 9: Silhouette score by number of clusters (left) and elbow plot of inertia (right), supporting Section 4.6 .
Figure 10: Average organ system contribution to SOFA score by severity band, supporting Section 4.5 .
Figure 11: Mean subscore difference between deteriorating and stable days by organ system, supporting Section 4.5 .
Figure 12: Box plot of SOFA subscore distributions by organ system, supporting Section 4.5 .
Organ system
Variables (source tables)
Cardiovascular
MAP, vasopressor rates (chartevents, inputevents)
Respiratory
FiO 2 , PaO 2 (chartevents, labevents)
Neurological
GCS components (chartevents)
Renal
Serum creatinine (labevents)
Hepatic
Total bilirubin (labevents)
Coagulation
Platelet count (labevents)
Appendix
Table 5: SOFA organ system variables and their MIMIC-IV source tables, supporting Section 3.2 .
Early identification of patients at risk for clinical deterioration in the intensive care unit (ICU) remains a critical challenge. Delayed recognition of impending adverse events, including mortality, vasopressor initiation, and mechanical ventilation, contributes to preventable morbidity and mortality. We present a multimodal deep learning approach that combines structured time-series data (vital signs and laboratory values) with unstructured clinical notes to predict patient deterioration within 24 hours. Using the MIMIC-IV database, we constructed a cohort of 74,822 ICU stays and generated 5.7 million hourly prediction samples. Our architecture employs a bidirectional LSTM encoder for temporal patterns in physiologic data and ClinicalBERT embeddings for clinical notes, fused through a cross-modal attention mechanism. We also present a systematic review of existing approaches to ICU deterioration prediction, identifying 31 studies published between 2015 and 2024. Most existing models rely solely on structured data and achieve area under the curve (AUC) values between 0.70 and 0.85. Studies incorporating clinical notes remain rare but show promise for capturing information not present in structured fields. Our multimodal model achieves a test AUROC of 0.7857 and AUPRC of 0.1908 on 823,641 held-out samples, with a validation-to-test gap of only 0.6 percentage points. Ablation analysis validates the multimodal approach: clinical notes improve AUROC by 2.5 percentage points and AUPRC by 39.2% relative to a structured-only baseline, while deep learning models consistently outperform classical baselines (XGBoost AUROC: 0.7486, logistic regression: 0.7171). This work contributes both a thorough review of the field and a reproducible multimodal framework for clinical deterioration prediction.
Forecasting evolving clinical risks relies on intrinsic pathological dependencies rather than mere chronological proximity, yet current methods struggle with coarse binary supervision and physical timestamps. To align predictive modeling with clinical logic, we propose the Medical-semantics Aware Time-ALiBi Transformer (MATA-Former), utilizing event semantics to dynamically parameterize attention weights to prioritize causal validity over time lags. Furthermore, we introduce Plateau-Gaussian Soft Labeling (PSL), reformulating binary classification into continuous multi-horizon regression for full-trajectory risk modeling. Evaluated on SIICU -- a newly constructed dataset featuring over 506k events with rigorous expert-verified, fine-grained annotations -- and the MIMIC-IV dataset, our framework demonstrates superior efficacy and robust generalization in capturing risks from text-intensive, irregular clinical time series.
Zhichong Zheng, Xiaohang Nie, Xueqi Wang +3
Tongji University · Shanghai Innovation Institute · Harbin Institute of Technology +1
Early prediction of severe clinical deterioration and remaining length of stay can enable timely intervention and better resource allocation in high-acuity settings such as the ICU. This has driven the development of machine learning models that leverage continuous streams of vital signs and other physiological signals for real-time risk prediction. Despite their promise, existing methods have important limitations. Contrastive pretraining treats all patients as equally strong negatives, failing to capture clinically meaningful similarity between patients with related diagnoses. Meanwhile, downstream fine-tuning typically ignores complementary modalities such as clinical notes, which provide rich contextual information unavailable in physiological signals alone. To address these challenges, we propose OC-Distill, a two-stage framework that leverages multimodal supervision during training while requiring only vital signs at inference. In the first stage, we introduce an ontology-aware contrastive objective that exploits the ICD hierarchy to quantify patient similarity and learn clinically grounded representations. In the second stage, we fine-tune the pretrained encoder via cross-modal knowledge distillation, transferring complementary information from clinical notes into the model. Across multiple ICU prediction tasks on MIMIC, OC-Distill demonstrates improved label efficiency and achieves state-of-the-art performance among methods that use only vital signs at inference.
Zhongyuan Liang, Junhyung Jo, Hyang-Jung Lee +2
Computational Precision Health UC Berkeley, UCSF · Samsung Advanced Institute of Technology (SAIT) · Computational Precision Health UC Berkeley, UCSF[cs.LG]