Identification of sepsis phenotypes with artificial intelligence using basic clinical data and laboratory parameters

Title: Identification of Sepsis Phenotypes with Artificial Intelligence Using Basic Clinical Data and Laboratory Parameters Introduction and Objective: Sepsis is a life-threatening condition characterized by organ dysfunction resulting from a dysregulated immune response to infection. It remains a major challenge in intensive care units with high morbidity and mortality rates. Sepsis phenotyping aims to better understand patients' clinical characteristics, treatment responses, and prognoses, potentially enabling personalized treatment approaches. To identify sepsis phenotypes using artificial intelligence techniques based on complete blood count and hemodynamic monitoring data, and to analyze the impact of these phenotypes on clinical outcomes. Methods: This retrospective study utilized the MIMIC-IV database, containing data for over 65,000 adult patients treated in the intensive care units of Beth Israel Deaconess Medical Center between 2008 and 2019. Patients diagnosed with sepsis according to Sepsis-3 criteria were included. Demographic information, vital signs, laboratory results, and comorbidities were analyzed. Clustering analysis was applied to identify sepsis phenotypes. Differences between phenotypes were evaluated using Kruskal- Wallis and χ2 tests, while 28- and 90-day survival analyses were performed using Kaplan-Meier curves and Cox regression analyses. Results: Machine learning techniques identified three distinct phenotypes. Laboratory parameters, such as mean platelet and leukocyte counts, were influential in determining these phenotypes. In particular, it was found that patients with high platelet counts (Phenotype HP) had a higher mortality risk at intensive care unit admission, whereas patients with normal platelet counts (Phenotype NP) had the lowest mortality rates. Conclusion: This study revealed significant differences in mortality, morbidity, and clinical outcomes among sepsis phenotypes identified using early hemodynamic findings and basic blood count parameters. The phenotype classification obtained through machine learning techniques may be utilized in patient risk assessment and personalized treatment approaches. The integration of artificial intelligence in sepsis management emerges as a pioneer for innovative approaches in this field and presents a critical turning point for future research. Keywords: Sepsis, Phenotyping, Artificial Intelligence, Machine Learning, Intensive Care