Identification of the robust predictor for sepsis based on clustering analysis

Area of Science:

• Clinical Medicine
• Biostatistics
• Critical Care

Background:

• Sepsis presents a significant global health challenge due to its high mortality and difficulty in early diagnosis.
• Lack of specific biomarkers and diverse causes complicate timely sepsis identification.
• Traditional markers like White Blood Cell (WBC) count may not be universally effective, especially in specific patient subgroups.

Purpose of the Study:

• To identify robust risk factors for sepsis using cluster analysis on electronic medical record data.
• To investigate the predictive value of clinical and laboratory markers for sepsis across different patient clusters.
• To improve the screening of potentially overlooked sepsis cases, including those without elevated WBC counts.

Main Methods:

• K-means clustering was employed to categorize 2,490 sepsis patients and 16,916 healthy individuals into 3 and 4 groups based on seven markers: Age, WBC, NLR, Hb, PLT, DNI, and MPXI.
• Logistic regression models were applied to identify sepsis-related features in the overall cohort and within specific clusters.
• Real-world hospital data was utilized for cluster analysis and predictor identification.

Main Results:

• The association between White Blood Cell (WBC) count and sepsis status was insignificant in older patient clusters (K3C3 and K4C3).
• Neutrophil-to-lymphocyte ratio (NLR) and Delta neutrophil index (DNI) demonstrated robust predictive value for sepsis across all subjects and within specific clusters, including older age groups.
• Cluster analysis revealed sepsis predictors that are effective even in the absence of typical inflammatory markers like elevated WBC.

Conclusions:

• NLR and DNI are reliable predictors for sepsis, offering valuable insights beyond traditional markers like WBC.
• Cluster analysis of clinical and laboratory data can uncover robust sepsis predictors, aiding in the early detection of potentially missed cases.
• This approach enhances sepsis screening by identifying individuals who may not exhibit standard sepsis indicators.