Dueñas Espín, Iván GuillermoJimbo Sotomayor, Ruth ElizabethArmijos, LucianaSánchez, XavierSevilla, CarmenzaRivadeneira, María FernandaTello, BetzabéVivas, LuisDueñas Espin, Ivan GuillermoJimbo Sotomayor, Ruth Elizabeth2023-11-042023-11-04https://repositorio.puce.edu.ec/handle/123456789/4100Background: Risk assessment tools have demonstrated to improve patient care. We analysed neonatal mortality through a longitudinal study that includes all the newborns that died and were registered between 2014-2016 in Ecuador and assessed potential mortality predictors to develop and validate a risk assessment tool. Methods: We developed a univariate description of explanatory variables, followed by a bivariate analysis of neonatal mortality by categorizing survival time into binomial outcomes (before 24, 48, 72 hours and 7 and 15 days). Furtherly, we developed saturated and parsimonious Cox proportional-hazard models using part of our data set, choosing our final model through results of the likelihood ratio test. We validated the model internally (deaths between 2014 to 2016) by using bootstrapping techniques and externally (deaths in 2017). We assessed the performance of our model in both data sets, using the AUC of the ROC. The betacoefficients from the final model were used to create a risk score, giving a specific weight to each predictor. The obtained score will be validated with experts in the next step of this research. Results: We analyzed 3109 newborn children who were born at a median of 32 weeks of gestation (P25:28 to P75:37), with a weight centile for gestational age of 14.9% 3rd centile, 76.6% between 3rd and 97th centiles and 8.5% over 97th centile. The most prevalent disorders presented were asphyxia, prematurity, infections and malformations. Bivariate analyses in the development cohort showed that 6 predictors were statistically significant associated with neonatal mortality. Multivariate analyses showed that gestational age, weight for gestational age centile, Apgar at 5 minutes, type of delivery and comorbidities were strong predictors of mortality at the different time cut-offs. Conclusions: The proposed score showed a good ability to predict neonatal death before 24 and 48 hours; and fair ability to predict neonatal death before 15 days of life.OpenAccess