Structural Equation Modeling to Detect Predictors of CD4 Cell Count Change due to Long Term Antiretroviral Therapy Administered to HIV-Positive Adults at Felege Hiwot Teaching and Specialized Hospital, Bahir Dar, Ethiopia

Background The relationship between predictors and the variable of interest was estimated using a structural equation model which is used to predict latent variables. The main advantage of the SEM is the ability to estimate the direct and indirect pathways of the effect of the primary independent variable on the outcome, given sufficient sample sizes. Despite not directly modeling the mediated pathways, GLMMs excluding mediating variables performed well with respect to power, bias and coverage probability in modeling the total effect of the primary independent variables on the outcome. In longitudinal studies, data are collected from subjects at several time points. The main purpose of longitudinal analysis is to detecting the trends or trajectories of the variables of interest. Methods A longitudinal study was conducted on 792 adults living with HIV/AIDS who commenced HAART. Structural equation modeling was used to construct a model to detecting predictors of CD4 cell count change. The procedure was illustrated by applying it to longitudinal health-related quality-of-life data on HIV/AIDS patients, collected from September 2008 to August 2012 monthly for the first six months and quarterly for remaining study period. Results The result of current investigation indicates that CD4 cell count change was highly influenced by certain socio-demographic and clinical variables. Out of all the participants, 141 (82%) have been considered 100% adherent to antiretroviral therapy. Structural equation modeling has confirmed the direct effect that personality (decision-making and tolerance of frustration) has on motives to behave, or act accordingly, which was in turn directly related to medication adherence behaviors. In addition, these behaviors have had a direct and significant effect on viral load, as well as an indirect effect on CD4 cell count. The final model demonstrates the congruence between theory and data (x2/df. = 1.480, goodness of fit index = 0.97, adjusted goodness of fit index = 0.94, comparative fit index = 0.98, root mean square error of approximation = 0.05), accounting for 55.7% of the variance. Conclusions The results of this study support our theoretical model as a conceptual framework for the prediction of medication adherence behaviors in persons living with HIV/AIDS. Implications for designing, implementing, and evaluating intervention programs based on the model are to be discussed.