TA-RNN-Medical-Hybrid: A Time-Aware and Interpretable Framework for Mortality Risk Prediction

Zahra Jafari; Azadeh Zamanifar; Amirfarhad Farhadi

cs.LGMar 9, 2026

TA-RNN-Medical-Hybrid: A Time-Aware and Interpretable Framework for Mortality Risk Prediction

Authors: Zahra Jafari, Azadeh Zamanifar, Amirfarhad Farhadi

Organizations: Department of Computer Engineering, SRC.,Islamic Azad university, Tehran, Iran · School of Computer Engineering, Iran university of Science and technology, Tehran, Iran

Abstract

Accurate and interpretable mortality risk prediction in intensive care units (ICUs) remains a critical challenge due to the irregular temporal structure of electronic health records (EHRs), the complexity of longitudinal disease trajectories, and the lack of clinically grounded explanations in many data-driven models. To address these challenges, we propose TA-RNN-Medical-Hybrid, a time-aware and knowledge-enriched deep learning framework that jointly models longitudinal clinical sequences and irregular temporal dynamics through explicit continuous-time encoding, along with standardized medical concept representations. The proposed framework extends time-aware recurrent modeling by integrating explicit continuous-time embeddings that operate independently of visit indexing, SNOMED-based disease representations, and a hierarchical dual-level attention mechanism that captures both visit-level temporal importance and feature/concept-level clinical relevance. This design enables accurate mortality risk estimation while providing transparent and clinically meaningful explanations aligned with established medical knowledge. We evaluate the proposed approach on the MIMIC-III critical care dataset and compare it against strong time-aware and sequential baselines. Experimental results demonstrate that TA-RNN-Medical-Hybrid consistently improves predictive performance in terms of AUC, accuracy, and recall-oriented F $_2$ -score. Moreover, qualitative analysis shows that the model effectively decomposes mortality risk across time and clinical concepts, yielding interpretable insights into disease severity, chronicity, and temporal progression. Overall, the proposed framework bridges the gap between predictive accuracy and clinical interpretability, offering a scalable and transparent solution for high-stakes ICU decision support systems.

Figures & tables

Method	Temporal Modeling	Irregular Time Handling	Attention Mechanism	External Medical Knowledge	Interpretability Level	Clinical Applicability
Traditional ML Models	Limited	No	No	No	Feature-level	Moderate
Deep Learning (MLP/CNN)	Limited	No	No	No	Low	Moderate
RNN / GRU	Sequential	Partial	No	No	Low	Moderate
Transformer-based Models	Long-range	Partial	Yes	No	Visit/Modality-level	Moderate
TA-RNN Al Olaimat and Bozdag (2024)	Time-aware sequential	Yes	Visit-level	No	Visit-level	High
GRU-D Che et al. (2018)	✓	✓	No	✗	Hidden-state

Table 1: Comparison of representative ICU mortality prediction models

Figure 1: Pipeline for constructing the ICD-to-SNOMED embedding matrix. ICD codes are normalized and mapped to SNOMED concepts, followed by semantic text embedding using BioClinicalBERT and structural embedding via SNOMED relational graphs.

Figure 2: Overall architecture of the proposed TA-RNN-Medical-Hybrid framework. Diagnosis codes are first mapped to SNOMED-based embeddings and aggregated at the visit level. Explicit temporal embeddings are incorporated to handle irregular visit intervals. The resulting sequence is processed by a BiGRU and multi-head self-attention module, followed by a dual-level attention mechanism (visit + disease-level) and demographic fusion… to generate the final mortality risk prediction.

Figure 3: Summary of model interpretation outputs. The figure illustrates mortality risk stratification, visit-level attention distribution, and ranked disease-level contributions for representative ICU patients.

Figure 4: Temporal disease progression patterns derived from disease-level attribution scores. Increasing, decreasing, and stable contribution trends illustrate how individual diseases influence mortality risk over time.

Figure 5: Example of the proposed clinical interpretation dashboard generated by TA-RNN-Medical-Hybrid. The dashboard integrates predicted mortality risk, visit-level temporal importance, SNOMED-based feature-level attribution, and longitudinal feature progression to support transparent clinical decision-making.

Model	Time-aware Modeling	Irregular Sampling	Attention Mechanism	External Medical Knowledge	Interpretability Level
Logistic Regression	✗	✗	✗	✗	Feature-level
Random Forest / XGBoost	✗	✗	✗	✗	Feature-level
RNN / GRU	✓	✗	✗	✗	Hidden-state
BiGRU	✓	✗	✗	✗	Hidden-state
TA-RNN	✓	✓	Visit-level	✗	Visit-level
Transformer-based ICU Models	✓	Partial	Self-attention	✗	Token-level

Table 2: Methodological comparison of TA-RNN-Medical-Hybrid with representative ICU mortality prediction models.

Model	Accuracy	AUC	F2-score
GRU-D	0.84	0.71	0.88
BEHRT	0.88	0.78	0.92
Med-BERT	0.89	0.80	0.93
TA-RNN	0.86	0.73	0.90
TA-RNN-Medical-Hybrid	0.91	0.82	0.95

Table 3: Performance comparison on ICU mortality prediction.

Model Variant	Accuracy	AUC	F2-score
TA-RNN (Baseline)	0.86	0.73	0.90
TA-RNN-Medical-Hybrid (Full Model)	0.91	0.82	0.95
w/o structured feature embeddings	0.88	0.75	0.91
w/o temporal attention	0.89	0.79	0.91
w/o contextual feature attention	0.89	0.80	0.92
w/o continuous-time encoding	0.89	0.80	0.92

Table 4: Component-wise ablation analysis of TA-RNN-Medical-Hybrid (feature-based).

Model	Accuracy	AUC	F2-score
TA-RNN-Medical-Hybrid	0.91 ± 0.01	0.82 ± 0.02	0.95 ± 0.01

Table 5: Performance of TA-RNN-Medical-Hybrid across 5 random seeds (Mean ± Std).

Explore similar work

Apr 2, 2026

MATA-Former & SIICU: Semantic Aware Temporal Alignment for High-Fidelity ICU Risk 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

Clinical Risk Prediction Longitudinal EHR Modeling Irregular Time-Series Modeling Temporal Attention

Ask PDF

Aug 30, 2026

INTERVenE: Temporal-Abstraction-Interval Based Transformers for Short-Horizon Medical Event Prediction

Electronic Health Record (EHR) prediction models in the intensive care unit must learn from sparse and irregular measurements while preserving the clinical meaning of time and supporting transparent decision-making. We present INTERVenE, a family of Transformer architectures whose input is an interval-based, knowledge-based temporal abstraction (KBTA), a token stream of named clinical concepts (states, trends, events, contexts) drawn from a curated medical ontology, rather than an unnamed bin index or a raw measurement triplet. This naming layer is what we ask KBTA to do: it makes the model's per-token attributions resolve to clinical concepts by construction. INTERVenE offers two complementary variants: an auto-regressive decoder that generates future abstraction trajectories with a per-step risk readout (localizing when and after which events risk rises), and a bidirectional encoder for single-pass joint risk and time-to-event prediction. Evaluated on 57,078 MIMIC-IV admissions against GRU-D, STraTS, and KarmaLego, INTERVenE-Enc reaches a support-weighted AUPRC

_w

of 0.672, improving by 0.041 over the strongest neural baseline with non-overlapping 95% bootstrap CIs, while also taking the best AUROC

_w

(0.901) and length-of-stay MAE (44.4,h). INTERVenE-Ar (AUROC

_w

0.854

, AUPRC

_w

0.587

under the same evaluation contract - a strictly harder generative readout) provides a complementary token-level risk trajectory. An input-representation ablation confirms the lift transfers across structured discretizations, positioning KBTA-based intervals as the interpretable substrate that makes per-token attributions resolve to meaningful clinical concepts within the deployed model.

Shahar Oded, Yuval Shahar

The Stein Faculty of Computer and Information Science, Ben-Gurion University of the Negev, Beer-Sheva, Israel

Event Time Prediction Interpretable ML Clinical Outcome Prediction Electronic Health Records Time Series Transformers

Ask PDF

Mar 16, 2026

Multimodal Deep Learning for Early Prediction of Patient Deterioration in the ICU: Integrating Time-Series EHR Data with Clinical Notes

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.

Binesh Sadanandan

Clinical Risk Prediction Longitudinal EHR Modeling Multimodal Learning

Ask PDF