ABSTRACT
Background: Children with hemato-oncological diseases or following stem cell transplantation (SCT) are at high risk for life-threatening infections; sepsis in this population constitutes a substantial proportion of pediatric intensive care unit (PICU) admissions. The current pediatric prognostic scoring tools to evaluate illness severity and mortality risk are designed for the general pediatric population and may not be adequate for this vulnerable subpopulation. Methods: Retrospective analysis was performed on all PICU admissions for sepsis in children with hemato-oncological diseases or post-SCT, in a single tertiary pediatric hospital between 2008 and 2021 (n = 233). We collected and analyzed demographic, clinical, and laboratory data and outcomes for all patients, and evaluated the accuracy of two major prognostic scoring tools, the Pediatric Logistic Organ Dysfunction-2 (PELOD-2) and the Pediatric Risk of Mortality III (PRISM III). Furthermore, we created a new risk-assessment model that contains additional parameters uniquely relevant to this population. Results: The survival rate for the cohort was 83%. The predictive accuracies of PELOD-2 and PRISM III, as determined by the area under the curve (AUC), were 83% and 78%, respectively. Nine new parameters were identified as clinically significant: age, SCT, viral infection, fungal infection, central venous line removal, vasoactive inotropic score, bilirubin level, C-reactive protein level, and prolonged neutropenia. Unique scoring systems were established by the integration of these new parameters into the algorithm; the new systems significantly improved their predictive accuracy to 91% (p = 0.01) and 89% (p < 0.001), respectively. Conclusions: The predictive accuracies (AUC) of the PELOD-2 and PRISM III scores are limited in children with hemato-oncological diseases admitted to PICU with sepsis. These results highlight the need to develop a risk-assessment tool adjusted to this special population. Such new scoring should represent their unique characteristics including their degree of immunosuppression and be validated in a large multi-center prospective study.
