Intelligent Prediction Platform for Sepsis Risk Based on Real-Time Dynamic Temporal Features: Design Study.

Abstract

Background: The development of sepsis in the intensive care unit (ICU) is rapid, the golden rescue time is short, and the effective way to reduce mortality is rapid diagnosis and early warning. Therefore, real-time prediction models play a key role in the clinical diagnosis and management of sepsis. However, the existing sepsis prediction models based on artificial intelligence still have limitations, such as poor real-time performance and insufficient interpretation.

Objective: Our objective is to develop a real-time sepsis prediction model that integrates high timeliness and clinical interpretability. The model is designed to dynamically predict the risk of sepsis in ICU patients and establish a practical, tailored sepsis prediction platform.

Methods: Within a retrospective analysis framework, the model comprises a real-time prediction module and an interpretability module. The real-time prediction module leverages 3-hour dynamic temporal features derived from 8 noninvasive, real-time physiological indicators: heart rate, respiratory rate, blood oxygen saturation, mean arterial pressure, systolic blood pressure, diastolic blood pressure, body temperature, and blood glucose. Three linear parameters (mean, SD, and endpoint value) were calculated to construct the prediction model using multiple ML algorithms. The interpretability module uses the TreeSHAP (Tree-Based Shapley Additive Explanations) method to enhance model transparency through both individual prediction and global explanations. Further, it added a link between the output interpretation of the explainable artificial intelligence method and its potential physiological or pathophysiological significance, including the relationship among the output interpretation and the patient's hemodynamics, thermoregulatory response, and the balance between oxygen delivery and oxygen consumption. Finally, a web-based platform was developed to integrate prediction and interpretability functions.

Results: The sepsis prediction model demonstrated robust performance in the test cohort (224 patients), achieving an accuracy of 0.7 (95% CI 0.68-0.71), precision of 0.69 (95% CI 0.68-0.71), F₁-score of 0.69 (95% CI 0.67-0.70), and area under the receiver operating characteristic curve of 0.76 (95% CI 0.74-0.77). The TreeSHAP method effectively visualized feature contributions, enabling clinicians to interpret the model's prediction logic and identify anomalies. The link between the output interpretation of the model and its potential physiological or pathophysiological significance improved the interpretability and credibility of the explainable artificial intelligence method. The web-based platform significantly enhanced clinical utility by providing real-time risk assessment, statistical summaries, trend analysis, and actionable insights.

Conclusions: This platform provides real-time dynamic warnings for sepsis risk in critically ill ICU patients, supporting timely clinical decision-making.