THE POTENTIAL ROLE OF ARTIFICIAL INTELLIGENCE-BASED APPROACH TO SEPSIS MANAGEMENT: A PILOT CASE STUDY

Objectives. Sepsis remains a critical challenge in intensive care units (ICUs), requiring timely and effective interventions. This study aims to explore the integration of a data-driven sepsis risk prediction software application with a large language model (LLM) module that provides personalized recommendations based on clinical cases, thereby improving decision-making in the ICU. Material and Methods. The study presents the analysis of a predictive software application that uses real-time physiological data (heart rate, blood pressure, oxygen saturation, temperature and respiratory rate) to predict the risk of sepsis within a four-hour window. An LLM module uses an enhanced recovery approach, synthesizing insights from 20 recent articles in the literature on sepsis management. A clinical case of a patient with sepsis due to colonic diverticulitis, extracted from the MIMIC-III database and organized as a clinical vignette, serves as a case study. Results. The predictive model allows timely estimation of sepsis risk. The LLM module generated personalized recommendations, including antibiotic therapy and monitoring strategies. Conclusions. Integrating a data-driven prediction tool with an LLM-based module can potentially improve sepsis management in the ICU. This pilot study highlights the value of using artificial intelligence (AI) to provide clinicians with real-time, evidence-based recommendations.