Международный журнал общественного здравоохранения и безопасности

Comparison of Cox Proportional Hazard Model and Accelerated Failure Time (Aft) Models: An Application to Neonatal Jaundice

Abstract

Folorunso SA and Osanyintupin OD

A primary focus of Survival analysis in medicine is modelling time to surviving of a particular disease. In this paper, survival analysis was carried out on the neonatal jaundice data modeling time to surviving the disease. The data was gotten from collected from University College hospital (UCH), Ibadan, Nigeria. The Kaplan-Meier approach was used to describe the survival functions of the neonatal jaundice patients and Log-rank tests was used to compare the survival curves among groups. Different kinds of models such as Cox Proportional Hazard Model and Accelerated Failure Time (AFT) models like Weibull AFT model, Logistic AFT model, Log-normal AFT model, Log-logistic AFT model and Exponential AFT model are considered to be used for modelling the time to surviving neonatal jaundice. Models selection criteria were used as a guide to unravel the best model for modeling neonatal jaundice. The result revealed that the fitted cox proportional hazard model suggested that there were 0.2708 chances of male neonates having higher median time of surviving jaundice compared to female neonates. Based on the mother's health history, neonates whose mother had illness during pregnancy will have 0.5329 chance of having higher median time of surviving the Jaundice compared to neonates whose mother do not have any illness during pregnancy. The log-logistic AFT model out-performed the other models since it has the lowest AIC and the highest log-likelihood value with 1131.461 and -550.7305 respectively.