A Joint Modeling of Longitudinal and Survival Data With Application to HIV-Infected Patients under HAART Follow-Up: a Case of Mekelle General Hospital, Ethiopia

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Abeba university


Despite tremendous progress in the control of the global HIV epidemic, the burden of HIV is still severe in Sub-Saharan Africa. Longitudinal and survival data frequently observed together in practice and useful for analysis of HIV related data. The separate analyses of longitudinal and survival endpoints may not be adequate and could lead to ine cient estimation or biased results. Joint modeling approaches correct for this bias by accounting for the association between the two responses. The main purpose of this study was to jointly model and analyze longitu- dinal and survival endpoints with application to retrospective cohort data of 469 HIV-infected patients under HAART follow-up in Mekelle General Hospital, Tigray, Ethiopia. The analysis consists of exploratory data analysis and tting three di erent models namely; a linear mixed e ects model for the longitudinal data, a semi-parametric survival model for the time-to-event data and a joint modeling of the two responses via shared random-e ects approach. The results of both the separate and joint analyses are consistent. However, the use of a joint analysis compared to independent models shows a reduction in the standard errors which indicates that more adequate and e cient inferences can be made by using joint model estimates. The esti- mated association parameter ( ) in the joint model is -0.138 (with 95% CI: -0.196 􀀀 -0.079) and statistically signi cant (p 􀀀 value < 0:0001). This indicates that there is strong evidence of association between the e ect of the longitudinal biomarker to the risk of death. The results indicates that higher initial values of CD4 cells is associated with a better survival. Further- more, patients with lower initial weight, being male, late WHO clinical stage, being ambulatory and bedridden were associated with higher risk of death. Future extension of this research could possibly be to account for missing data and attempt should be given to health workers and data clerks working with patients under HAART to improve the quality of the data records of patients. Keywords: HAART, HIV/AIDS Data, Joint Modeling, Longitudinal Data Analysis, Survival Data Analysis



HAArt, HIV/Aids Data, Joint Modeling, Longitudinal Data Analysis, Survival Data Analysis