A Novel Multi-Task Teacher-Student Architecture with Self-Supervised Pretraining for 48-Hour Vasoactive-Inotropic Trend Analysis in Sepsis Mortality Prediction

TL;DR

This work tackles sepsis mortality prediction by modeling 48-hour vasoactive–inotropic score (VIS) dynamics using a self-supervised Masked Autoencoder (MAE) pretraining stage, followed by a frozen Teacher and a learnable Student in a multitask, knowledge-distilled framework. The approach jointly optimizes mortality classification and severity regression, yielding an AUROC of approximately 0.82 on MIMIC-IV v3.0 and demonstrating robust performance against missing/irregular VIS data. SHAP analyses reveal that both organ dysfunction severity scores and sociodemographic factors meaningfully influence predictions, highlighting the importance of multifactorial ICU risk assessment. The combination of MAE, knowledge distillation, and multi-task learning improves discrimination, interpretability, and resilience, signaling practical potential for real-time ICU decision support in sepsis management.

Abstract

Sepsis is a major cause of ICU mortality, where early recognition and effective interventions are essential for improving patient outcomes. However, the vasoactive-inotropic score (VIS) varies dynamically with a patient's hemodynamic status, complicated by irregular medication patterns, missing data, and confounders, making sepsis prediction challenging. To address this, we propose a novel Teacher-Student multitask framework with self-supervised VIS pretraining via a Masked Autoencoder (MAE). The teacher model performs mortality classification and severity-score regression, while the student distills robust time-series representations, enhancing adaptation to heterogeneous VIS data. Compared to LSTM-based methods, our approach achieves an AUROC of 0.82 on MIMIC-IV 3.0 (9,476 patients), outperforming the baseline (0.74). SHAP analysis revealed that SOFA score (0.147) had the greatest impact on ICU mortality, followed by LODS (0.033), single marital status (0.031), and Medicaid insurance (0.023), highlighting the role of sociodemographic factors. SAPSII (0.020) also contributed significantly. These findings suggest that both clinical and social factors should be considered in ICU decision-making. Our novel multitask and distillation strategies enable earlier identification of high-risk patients, improving prediction accuracy and disease management, offering new tools for ICU decision support.

Figures (7)

• Figure 1: Overview of the data extraction and preprocessing workflow.
• Figure 2: MAE_Encoder architecture: shows token masking, [CLS] token prepending, linear projection to 64 dimensions, and a 2-layer Transformer (with $d_{\text{model}} = 64$, FFN = 256, 8 heads, 0.1 dropout) for VIS time series representation.
• Figure 3: MAE_Decoder schematic: reconstructs masked tokens from the latent embedding (excluding the [CLS] token) back to the original 7-dimensional VIS space.
• Figure 4: Multitask_Model schematic: depicts the shared encoder with classification and regression branches for ICU mortality and severity score predictions, along with the teacher–student knowledge distillation framework.
• Figure 5: ROC curves on the test set. baseline achieves the highest AUROC, followed by no_kd and no_mt, reflecting the synergy from knowledge distillation and multi-tasking.