Predicting Mortality in Tanzanian Children With Sepsis Using Point-of-Care Biomarkers.
Sepsis is a leading cause of child mortality worldwide, disproportionately affecting children in resource-limited settings (RLS). Effective risk-stratification tools using readily available data are urgently needed for this population. Therefore, the study objective was to evaluate the performance of point-of-care (POC) biomarkers and clinical characteristics for predicting in-hospital mortality among Tanzanian children with sepsis.
We conducted a prospective observational cohort study of children (28 days-14 years) with sepsis presenting to Muhimbili National Hospital in Dar es Salaam, Tanzania (July 2022-November 2024). POC biomarkers (procalcitonin [PCT], C-reactive protein, ferritin, and lactate) and clinical characteristics were evaluated for their association with mortality. We used the least absolute shrinkage and selection operator (LASSO) regression to construct predictive models of mortality. We evaluated model performance using the area under the receiver operating characteristic curve (AUC) and classification metrics, including sensitivity and specificity.
Among the 755 enrolled participants, 19.6% (n = 148) died during hospitalization. PCT and tested clinical characteristics were significantly associated with mortality (all p<0.001). A multivariable model incorporating PCT, malnutrition, breathing difficulty, and altered mental status demonstrated strong discrimination (AUC 0.87, 95% CI 0.84-0.90), outperforming individual biomarkers and clinical characteristics alone.
A combined POC biomarker and clinical characteristics model was highly predictive of mortality among children with sepsis in Tanzania. Integrating POC biomarkers with easy-to-measure clinical characteristics associated with severity may enable timely risk stratification and inform targeted interventions to improve pediatric sepsis outcomes in RLS.
We conducted a prospective observational cohort study of children (28 days-14 years) with sepsis presenting to Muhimbili National Hospital in Dar es Salaam, Tanzania (July 2022-November 2024). POC biomarkers (procalcitonin [PCT], C-reactive protein, ferritin, and lactate) and clinical characteristics were evaluated for their association with mortality. We used the least absolute shrinkage and selection operator (LASSO) regression to construct predictive models of mortality. We evaluated model performance using the area under the receiver operating characteristic curve (AUC) and classification metrics, including sensitivity and specificity.
Among the 755 enrolled participants, 19.6% (n = 148) died during hospitalization. PCT and tested clinical characteristics were significantly associated with mortality (all p<0.001). A multivariable model incorporating PCT, malnutrition, breathing difficulty, and altered mental status demonstrated strong discrimination (AUC 0.87, 95% CI 0.84-0.90), outperforming individual biomarkers and clinical characteristics alone.
A combined POC biomarker and clinical characteristics model was highly predictive of mortality among children with sepsis in Tanzania. Integrating POC biomarkers with easy-to-measure clinical characteristics associated with severity may enable timely risk stratification and inform targeted interventions to improve pediatric sepsis outcomes in RLS.
Authors
Sorensen Sorensen, Mussa Mussa, Oltman Oltman, Sunzula Sunzula, Mlele Mlele, Msemwa Msemwa, Mathias Mathias, Sun Sun, Baez Maidana Baez Maidana, Mkopi Mkopi, Hooft Hooft, Mfinanga Mfinanga, Manyahi Manyahi, Sawe Sawe, Kortz Kortz
View on Pubmed