Survival Analysis of Dialysis Patients under Parametric and Non-Parametric approaches

Abstract

Dialysis is a recommended way of treatment for end stage kidney diseases and it provides a life saving procedure. Transplantation can also be useful source but it is restricted by financial limitations especially in developing countries like Pakistan. Censoring is an important part of the survival data which causes insensitivity to the usual procedures of analysis. A little work has been done in literature regarding the estimated survival time of dialysis patients in Pakistan. So, this study has estimated the median survival time of male/females patients separately by parametric and non-parametric approaches. Moreover, comparison of survival time to patients (<=50 years and >50 years) was also compared. Frequently, in modeling the survival data, most of the time we have no prior information about the theoretical distribution of survival time is available, that’s why, and non-parametric methods are commonly used. The significance of this study is the fitting of probability distribution of real life time data of dialysis patients which is not done before. It is very laborious job to fit an appropriate distribution of the data. We find that the probability distribution of our real life time data is weibull distribution. Finding suggested that the Kaplan-Meier method and weibull model based on Anderson-Darling test provided a very close estimate of the survival function in both genders and age groups. On the average survival time in males is relatively high but not statistically different from females.