Dynamic models for time-varying outcomes: An application to the 2015-2017 patient cohort on antiretroviral therapy at Luderitz hospital, Namibia

Patients' adherence to a prescribed medication regimen is one of the most significant barriers to successful antiretroviral therapy (ART). In addition, adherence to ART is one of the key determinants of Human Immunodeficiency Virus (HIV) disease progression, while non-adherence severely compromises treatment effectiveness and leads to unsuppressed virus. The extent of the impact of poor adherence on resulting health measures is often unknown, and typical analyses ignore the time-varying nature of adherence. The main objective of this study was to model time-varying outcomes of patients, while accounting for data missingness and measurement error using dynamic models, with application to a cohort of 154 adult patients initiated on ART between January 2015 and December 2017 at the Luderitz hospital. The outcome variable of this study was viral load which was measured at scheduled follow-up visits of patients. Baseline CD4 count, baseline weight, age at start of ART and gender were the non-dynamic covariates which were measured at the ART initiation, while adherence to ART and weight at follow up were the dynamic covariates measured at follow up visits. This study used mixed effects model and Generalized Estimating Equations (GEE) to model longitudinally measured viral load as a function of the dynamic as well as non-dynamic covariates. To account for missingness in the outcome variable as well as potential measurement error in covariates, a Simulation Extrapolation Inverse Probability Weighted Generalized Estimating Equations (SWGEE) model that incorporates missing and measurement error was used to model the data. The study found that adherence was good in female patients as compared to male patients. Furthermore, the study found that patients with a good adherence rate achieved viral suppression within 12 months of treatment unlike non-adherent patients. In conclusion, viral load of patient’s on ART differ across the patients’ baseline demographic and clinical characteristics.

Grantor dc:publisher

University of Namibia

Year dc:date.issued

2020

Author dc:creator

• Gabriel, Lineekela

• Antiretroviral therapy
• Dynamic linear models
• Time-varying outcomes
• Viral load
• Non-dynamic covariates
• Dynamic covariates

Language dc:language.iso

en

Handle dc:identifier.uri

http://hdl.handle.net/11070/2950

OAI identifier oai:identifier

oai:repository.unam.edu.na:11070/2950