Abstract
Discrete-time approach is used in survival data analysis when only the time interval in which the event of interest has occurred is known or when this event occurs in a discrete - time scale. The work presented in this paper is motivated by the analysis of HIV/AIDS follow-up data collected in Burkina Faso during the 5-YEAR Global Fund program implemented to fight AIDS, Tuberculosis and Malaria. The research question that motivated the work is the likely existence of different mortality risk profiles of people infected with HIV/AIDS, depending on their characteristics and health status at the beginning of their care. In order to answer these questions, we considered a binary tree regression approach for survival data analysis since such a model owns the ability to handle interaction effects between the outcome covariates without a tight specification of such effects during the model statement step. This helps to prevent specification and interpretation errors. The fitted model resulted in splitting patients into three disjoint subgroups, corresponding each to a specific hazard profile.
Highlights
From 2003 to 2007, the Global Fund supported health institutions in Burkina Faso to promote the access of HIV infected persons to Antiretroviral Therapy (ART), considered as a therapeutic advance in the fight against HIV/AIDS (Kouanda et al, 2008)
Our present work proposes a semiparametric modelbased binary tree for the analysis of the follow-up data of the HIV infected people who were included in the Global Fund program, with the aim of analyzing the correlation between the survival of the people who had access to this therapy, and their health status at the time of admission to the program, as well as the means by which these patients arrived at the program, the characteristics of the patients and the attributes of care facilities
We proposed a tree-based approach for the analysis of discrete time-to-event data
Summary
From 2003 to 2007, the Global Fund supported health institutions in Burkina Faso to promote the access of HIV infected persons to Antiretroviral Therapy (ART), considered as a therapeutic advance in the fight against HIV/AIDS (Kouanda et al, 2008). Health data were recorded during clinical visits. An evaluation of this program was done in 2008 in order to assess the efficiency of the program and it involved the analysis of follow-up data gathered during the program execution. Contributions in tree-based methods for discrete-time survival analysis include (Bou-Hamad et al, 2009) and (Schmid et al, 2016). Both methods consider that timeto-event data are observed jointly with covariates that describe individuals and consider that hazard of event occurrence is a function of time and covariates. In Bou-Hamad et al (2009) approach, hazard probability has been modeled as follows: for all x D(X) and time index t, log h(t | x) h(t | x) t
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
