Risk prediction model for neonatal mortality among neonates hospitalized with sepsis, Bahir Dar, Ethiopia.

Neonatal sepsis is a significant public health challenge across the globe, particularly, in low-resource settings. While early identification of at-risk patients is a recommended approach to reduce sepsis- related deaths, there is limited evidence from low-resource settings. Therefore, this study aimed to develop prediction models to estimate neonatal mortality among neonates with sepsis. The study was conducted among 975 neonates diagnosed with sepsis in the NICUs of two large hospitals in Bahir Dar city. Data were analyzed in R software, and regression analyses were performed. Receiver operating characteristic curve (ROC) and calibration plot were used to assess model performance, and internal validation was performed using bootstrapping technique. Sepsis accounted for 27.9% of neonatal deaths. Key predictors of mortality were low birth weight, late initiation of breastfeeding, gestational age, fifth-minute Apgar score, tachycardia, respiratory distress, and convulsions. The original and nomogram models demonstrated an AUC of 81.3% and 81.2%, respectively. In conclusion, the high incidence of sepsis-attributed mortality highlights the need for effective predictive tools. The nomogram, based on key clinical predictors such as low birth weight, late initiation of breastfeeding, prematurity, fifth-minute apgar score, tachycardia, respiratory distress, and convulsions, demonstrated very good discrimination, and calibration performances, These findings support the use of the model as a practical tool for early risk stratification and clinical decision-making in neonatal intensive care settings.