Abstract
Objective: The aim of this study was to establish a statistical model for prediction neonatal deaths.
Methods: A case-control study was carried out in the State of Maranhao, Northeast Brazil. The database used in this study included 446 neonates (146 cases and 297 controls). The CART method (Classification and Regression Tree) was used to construct the Tree Classification. Several trees were developed containing explanatory variables related to maternal condition and newborn infant. The tree that presented the smaller cross-validated classification error was selected. It was calculated the sensitivity, specificity and accuracy. The predictive ability of the tree to differentiate survivors and non-survivors was determined by the ROC curve. The classification tree was developed in the program R-2.7.0-WIN.
Results: The tree presented an accuracy of 88.2%, sensitivity 80.7%, and specificity 91.7%. Moreover, it was identified four risk groups. Neonates with problems and birth weight < 2500g with probability of death was 91%. Second group, neonates with problems at birth and Apgar score < 7, with a probability of 70%. The third group focuses neonates with congenital anomaly with probability of death of 57%. Finally, in the fourth group are those whose gestational age was < 37 weeks and probability of 53%. The area under the ROC curve was 0.915 [95%CI=0.87;0.95].
Conclusion: The classification tree demonstrated good discriminatory power. It was proved to be a useful tool to determine which neonates will need more care and it can be used in the medical routine.
Article PDF
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Ribeiro, V., Santos, A. & Queiroz, L. 455 Classification Tree Applied to Neonatal Mortality. Pediatr Res 68 (Suppl 1), 233 (2010). https://doi.org/10.1203/00006450-201011001-00455
Issue Date:
DOI: https://doi.org/10.1203/00006450-201011001-00455
This article is cited by
-
Prediction of neonatal deaths in NICUs: development and validation of machine learning models
BMC Medical Informatics and Decision Making (2021)