Early prediction of onset of sepsis in Clinical Setting

TL;DR

This work tackles early onset sepsis prediction in a clinical setting by leveraging deidentified EHR data from Montefiore Medical Center and training an XGBoost model on 107 engineered features. It demonstrates promising performance, achieving a normalized utility of $0.494$ and F1 of $80.8\%$ on retrospective test data, with a prospective utility of $0.378$ and F1 of $67.1\%$ at threshold $0.3$, highlighting the impact of data availability on time-window features. SHAP analysis reveals that a core set of features, including qSOFA and SOFA-derived metrics, drive predictions, supporting interpretability for clinical use. The study points to practical deployment potential and outlines future directions toward multimodal integration and NLP-enabled notes to further improve early sepsis detection.

Abstract

This study proposes the use of Machine Learning models to predict the early onset of sepsis using deidentified clinical data from Montefiore Medical Center in Bronx, NY, USA. A supervised learning approach was adopted, wherein an XGBoost model was trained utilizing 80\% of the train dataset, encompassing 107 features (including the original and derived features). Subsequently, the model was evaluated on the remaining 20\% of the test data. The model was validated on prospective data that was entirely unseen during the training phase. To assess the model's performance at the individual patient level and timeliness of the prediction, a normalized utility score was employed, a widely recognized scoring methodology for sepsis detection, as outlined in the PhysioNet Sepsis Challenge paper. Metrics such as F1 Score, Sensitivity, Specificity, and Flag Rate were also devised. The model achieved a normalized utility score of 0.494 on test data and 0.378 on prospective data at threshold 0.3. The F1 scores were 80.8\% and 67.1\% respectively for the test data and the prospective data for the same threshold, highlighting its potential to be integrated into clinical decision-making processes effectively. These results bear testament to the model's robust predictive capabilities and its potential to substantially impact clinical decision-making processes.

Figures (6)

• Figure 1: A high-level architecture diagram showing steps.
• Figure 2: Length of Stay (LOS) of patients in the Hospital
• Figure 3: Using dendrogram and ward distance to find representative features from each clusters.
• Figure 4: Training Loss curve.
• Figure 5: Prospective Data Analyses flow