Development and validation of risk score to predict in-hospital mortality among severely malnourished children under the inpatient treatment center: A follow up study

Abstract Introduction: The prognosis of severely malnourished children is determined by various factors. The care provided to the child throughout the management is one of the most influencing factors for the child’s status after treatment. There is no evidence-based model that shows the risk of mortality among severely malnourished children. Thus, this study aimed to develop and validate a model to predict mortality among children with severe acute malnutrition.Method: We developed a prediction model using a retrospective cohort of 677 children with severe acute malnutrition admitted to the Amhara region referral hospitals. A stepwise multivariable analysis was done to develop the model. The accuracy of the model was evaluated using the area under the receiver operating characteristic curve (AUC) and calibration plot. All accuracy measures were internally validated using the bootstrapping technique. To improve the clinical applicability, a simplified risk score was developed to classify children with severe acute malnutrition at high or low risk of mortality. The clinical impact of the model was evaluated using a decision curve analysis across various threshold probabilities. Result: Child’s Age, the form of malnutrition, HIV/AIDS, heart failure, provision of antibiotics, and folic acid and vitamin A supplementation remained in the prediction model. The AUC of the model was 0.81 (95%CI: 0.76-0.85). The decision curve analysis indicated that the model provides higher net benefit across ranges of threshold probabilities. Conclusion: The model can guide clinicians in identifying severely malnourished children with a high risk of mortality. As the model showed easily obtainable predictors of mortality, it provides a considerable opportunity for care providers to avert hundreds of mortalities among those children. The model needs to be externally validated in another context before utilizing it for clinical decision-making.