Lucia Carratala Castro (Spain)
ISGlobal Tuberculosis Research GroupPresenter of 1 Presentation
“PREDICTING MORTALITY AMONG CHILDREN ADMITTED WITH PNEUMONIA TO A DISTRICT HOSPITAL IN MANHIÇA, MOZAMBIQUE” (ID 1516)
Abstract
Background
Pneumonia is the single largest infectious cause of death in children worldwide. In low- and middle-income countries the in-hospital management of child pneumonia remains a challenge in part due to lack of specificity in the diagnosis. The goal of this study is to identify how to better predict mortality in children under 5 years of age with suspected pneumonia in Manhiça’s District Hospital (Mozambique), a malaria endemic setting.
Methods
In this study we analised the records of 835 children admitted from September 2006 to September 2007 with suspected pneumonia. Primary outcome was mortality during hospital admission and up to 21 days post-discharge. The pre-defined list of candidate predictors of mortality included both clinical variables and complementary test results on admission. We developed 2 models using automated stepwise regression with backward elimination. We calculated Area Under the Curve (AUC) for both models as well as for the Lambaréné Organ Disfunction score (LOD score).
Results
We reported 102 deaths. We identified 11 predictors for the comprehensive model including: age, sex, WAZ, history of seizures, deep breathing, nasal flaring, prostration, hyperpyrexia, wheezing, HIV status and parasitaemia. In the absence of complementary test variables (pragmatic model), cyanosis was found to be a predictor, and deep breathing and hyperpyrexia were dropped from the model. Pragmatic and comprehensive models presented AUC of 0.9 and 0.92 respectively and the LOD score AUC was 0.81 in the same sample.
Conclusions
Both models demonstrated an outstanding discrimination performance in the same sample, but further analysis needs to be carried out before this model can be used in a clinical setting. The LOD score still constitutes a very good predictive score for mortality in this group, similar to what has been previously reported.