Predicting mortality in pediatric sepsis: A nationwide data analysis using a pediatric sepsis surveillance method.

INTRODUCTION: A sepsis surveillance method using electronic health records is increasingly used to describe the epidemiology of adult sepsis. However, its application in pediatric populations has been limited, and real-world epidemiology of pediatric sepsis remains unknown. We aimed to determine whether this surveillance method could identify children with sepsis at high-risk of mortality. PATIENTS AND METHODS: From a nationwide inpatient database in Japan, we included children who underwent blood culture and received antimicrobials for ≥ 4 days between 2014 and 2021. We stratified these children into those with sepsis or without sepsis by the presence of organ dysfunction. We evaluated the discrimination for in-hospital mortality by the sepsis diagnosis and the number of organ dysfunction. RESULTS: Among 6553 eligible children, in-hospital mortality was 7.2 % in 875 children with sepsis and 0.2 % in 5678 children without sepsis. Sepsis diagnosis discriminated in-hospital mortality with a sensitivity of 0.84 and a specificity of 0.87. The area under the curve for mortality based on the number of organ dysfunction was 0.88 (95 % confidence interval:0.84 to 0.93). In-hospital mortality incrementally increased with the number of organ dysfunction; zero, 0.2 %; one, 3.4 %; two, 12.7 %; three, 20.9 %; four, 33.3 %; and five, 50.0 %. CONCLUSIONS: The surveillance method effectively identified children with sepsis with high-risk of mortality and demonstrated strong discrimination of mortality.