In our analysis of children with infections in acute-care hospitals, we employed a pediatric sepsis surveillance method that successfully identified children with sepsis with a high risk of mortality. We found that in-hospital mortality was strongly associated with the diagnosis of sepsis and the number of organ dysfunction by the method.
Previous studies have identified children with sepsis using the pediatric Sequential Organ Failure Assessment (pSOFA) score or IPSCG definition [18, 19]. However, their applications have been mostly limited to children admitted to pediatric intensive care units (PICUs) because these approaches require detailed clinical data, such as the Glasgow Coma Score and vital signs that are largely missing or not recorded accurately enough to be used outside of PICUs. The pediatric sepsis surveillance method has an advantage of screening children with infection in wider settings because the method only requires variables that are essential and commonly available in electronic medical records from variables based on the pSOFA score [10]. To our knowledge, this is the first study to apply the surveillance method in pediatric populations in acute-care hospitals at this scale. The sensitivity and specificity of sepsis diagnosis were 0.84 and 0.87 in our study, suggesting that the surveillance method could effectively discriminate in-hospital mortality in children with infections.
However, we should interpret the results with caution, as a high sensitivity and specificity may not necessarily mean a high positive predictive value (depending on the likelihood of mortality in a cohort). The positive predictive value could decrease in a milder-risk cohort, such as children presenting in emergency departments in high-income countries, while it could increase in a high-risk cohort, such as children in low-income countries or those admitted to PICUs with hospital-onset sepsis [20].
Regarding the prediction of mortality, two studies have reported that pSOFA score showed strong discriminatory ability, with AUC values of 0.83 and 0.94 [18, 19]. In our study, the AUC value for in-hospital mortality based on the number of organ dysfunction was 0.88, indicating a comparable discriminatory ability for death, despite the simplified criteria used in the sepsis surveillance method. Furthermore, an AUC of 0.94 was reported in a study using the maximal daily pSOFA scores. We assessed organ dysfunction in a window period before and after two days of blood culture, potentially suggesting that the prediction of mortality may be feasible in the early stage of sepsis based on organ dysfunction, which can be calculated automatically using electronic medical records.
We evaluated the criteria for organ dysfunction in pediatric sepsis, with a particular focus on hepatic and coagulation dysfunction, and lactate. The discriminatory ability remained unchanged after removing bilirubin from the criteria and exhibited slight improvement upon the incorporation of ALT. This observation can be supported by the fact that isolated increases in bilirubin levels are rarely observed in pediatric sepsis, and ALT was included as a maker of hepatic dysfunction in the IPSCG definition [10, 17]. Our findings warrant future studies using larger datasets to optimize the criteria for organ dysfunction, which would enable more accurate identification of children with sepsis who had a high risk of mortality.
The present study is associated with several limitations. Due to limited clinical information in the database, we could not compare the performance of the IPSCG definition, the pSOFA score, and the surveillance method in predicting mortality. However, the sensitivity and specificity for mortality in our study were in line with previous publications that used other scores [18, 19]. Second, our findings do not necessarily support using this method for clinically diagnosing pediatric sepsis. As this method was developed for surveillance purposes, the diagnosis of sepsis was based on organ dysfunction that occurred within two days before and after blood culture. Since early recognition and timely treatment are recommended in the management of pediatric sepsis, timeliness be more prioritized in the clinical setting [21]