Predicting viral suppression in anti-retroviral therapy patients using transition models and mixed effects models
Loading...
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Namibia
Abstract
Background: Acquired Immune Deficiency Syndrome (AIDS) which is caused by the human immunodeficiency virus (HIV) is a leading cause of death, a limitation to development in Africa, as well as a health threat to millions worldwide. The main purpose of antiretroviral therapy (ART) is to keep people living with HIV in good health. Progression of health in HIV/AIDS patients on ART is characterised by an increase in CD4 cell counts as well as a decrease in viral load (HIV RNA) to undetectable levels. Monitoring and evaluation (M&E) of HIV occurrence and responses allow countries to track the epidemic and their prevention and control efforts. This study assesses the use of viral load in monitoring HIV/AIDS progression and predicting viral suppression.
Objective: The main purpose of this study was to estimate prediction models and determine factors that influence viral load and CD4 cell counts in ART patients using transition autoregressive model and the mixed effects model.
Methods: The models were applied to secondary data from ART patients drawn from a retrospective cohort design. The participants of the study were aged fifteen years or older (referred to as adults), who initiated ART treatment at selected health facilities in Erongo region, Namibia, between January 2010 and December 2015, giving a total population of 8068 patients. The data was analysed in three parts, namely, the descriptive analysis, the linear mixed effects model and the transition autoregressive model. Both models assume that the future values of a variable depend on its historical values as well as on other covariates.
Results: The findings of this study points to a viral suppression prevalence rate of 71% in Erongo region. The findings provide additional support to the concept that demographic (sex, age and weight) and clinical (follow up time, adherence and clinical stage) characteristics are the determinants of viral load as well as CD4 cell counts. The Log Likelihood, AIC and BIC show that the Transition Autoregressive model is the best predicting model for viral load as well as CD4 cell counts compared to the Mixed Effects Models.
Conclusion: Based on the AIC and BIC, it was shown that the transition model and the mixed effects model have almost the same predictive accuracy at the first few visits of a patient; however, for longer time series the transition model offers somewhat better predictions. Since the transition model is convenient in practice and needs less historical information compared to the mixed effects model, HIV studies may use this model to predict the future viral load. In the final analysis, this study recommends the use of the Transition Autoregressive model in modelling change and event occurrence in longitudinal studies such as HIV studies, specifically in predicting viral suppression in the future.
Description
A mini thesis submitted in partial fulfilment of the requirements for the Degree of Master of Science (Biostatistics)
Keywords
Viral suppression, Transition Autoregressive mode, Mixed Effects models