Integrating Physiological Time Series and Clinical Notes with Transformer for Early Prediction of Sepsis
This work addresses the critical problem of early sepsis detection for ICU patients, but it is incremental as it builds on existing multimodal and Transformer approaches.
The paper tackled early sepsis prediction in ICU patients by proposing a multimodal Transformer model that integrates physiological time series and clinical notes within 36 hours of admission, and it outperformed six baselines on metrics across two large datasets (MIMIC-III and eICU-CRD).
Sepsis is a leading cause of death in the Intensive Care Units (ICU). Early detection of sepsis is critical for patient survival. In this paper, we propose a multimodal Transformer model for early sepsis prediction, using the physiological time series data and clinical notes for each patient within $36$ hours of ICU admission. Specifically, we aim to predict sepsis using only the first 12, 18, 24, 30 and 36 hours of laboratory measurements, vital signs, patient demographics, and clinical notes. We evaluate our model on two large critical care datasets: MIMIC-III and eICU-CRD. The proposed method is compared with six baselines. In addition, ablation analysis and case studies are conducted to study the influence of each individual component of the model and the contribution of each data modality for early sepsis prediction. Experimental results demonstrate the effectiveness of our method, which outperforms competitive baselines on all metrics.