Toward Pan-Age Sepsis Detection: Leveraging Pediatric Transcriptome Data to Enhance Diagnostic Model Generalizability in Heterogeneous Populations

Sepsis poses a threat to individuals across all age groups, from infants to geriatrics, due to its non-specific predisposition. Patients across different age groups demonstrate marked differences in physical characteristics, immune responses, and the progression rate of diseases, introducing additional complexity to the diagnosis of sepsis. In this study, we investigated the critical role of the age factor in sepsis diagnosis at the pan-age level. We performed feature engineering and model construction on 2,375 pan-age blood transcriptome samples from 17 cohorts and separately constructed two independent sepsis diagnostic models, Pediatric Sepsis Diagnosis Model (PSDM) and Adult Sepsis Diagnosis Model (ASDM). Results demonstrated that PSDM overall outperformed ASDM at the pan-age level (AUC= 0.965 vs AUC=0.890), especially in the diagnosis of pediatric sepsis. This study underscores the importance of age as a significant factor in constructing sepsis diagnostic models, revealing the differences in the pathogenesis of sepsis among different age groups and providing insights for the future development of pan-age precision medicine.