Estimating HIV prevalence and associated factors among Multidrug resistant tuberculosis patients in treatment centres in Uganda for the period 2017-2019; A comparison of methods

Abstract

Background: Multidrug resistant tuberculosis (MDR-TB) is a major clinical challenge, particularly in patients with human immunodeficiency virus (HIV) co-infection. The main objective of this study was to determine the prevalence of HIV among the Multidrug-resistant tuberculosis patients in Uganda and to determine which binary data analysis model was more appropriate. Methods: This was a cross sectional study where all the MDR-TB patient’s data was accessed from all the MDR-TB treatment centre records in Uganda from 2017 up to 2019. The study employed the binary regression models of Logistic regression, Log Binomial regression and Modified Poisson regression were used to assess the association between the exposure and the outcome variables at 5% level of significance (p < 0.05) as well as accounting for missingness using multiple imputation. To compare the methods the study used AIC, and BIC statistical data techniques. Results: Descriptive statistics were conducted reporting frequencies and proportions for the categorical variables. The overall prevalence of HIV among the MDR-TB patients was 46.7% (95% CI: 44.2, 49.3), 47.1 (95% CI: 43.9,50.3) among males, 46.1 (95% CI: 41.8,50.5) among females. Comparing children, adolescents, and adult was 48.3 (95% CI: 35.4,61.3), 45.8 (95% CI: 38.4,53.5), 46.8 (95% CI: 44.0,49.6) respectively. In terms of a better model, based on Akaike Information criteria (AIC) and Bayesian Information criteria (BIC), the estimates of binary logistic model were (AIC = 1427.79, BIC = 1442.66), that of modified Poisson model were (AIC = 1662.44, BIC = 1677.31) and of log-binomial model were (AIC = 1427.79, BIC = 1442.66). In complete case all the models showed the factor associated was MUAC, based on log-binomial and modified Poisson with multiple imputation, the two factors associated with HIV prevalence were weight and mid upper arm circumference (MUAC). MDR-TB patients weighing 50 kgs and more were significantly less likely to have HIV compared to those weighing less than 30 kg (adjusted risk ratio, ARR=0.84, 95%CI: 0.71-0.99). Also, MDR-TB patients MUAC “red” (acutely malnourished) were less likely to be HIV positive compared to those with MUAC “green” (ARR=0.60, 95%CI: 0.44, 0.82). Conclusions: The findings of this study revealed that the prevalence of HIV among MDR-TB patients was 46.7%. Binary logistic and log-binomial models offered better fits (with equal AIC and BIC values) than the modified Poisson model. However, due to the estimated high prevalence of HIV among the MDR-TB patients (46.7%), log-binomial regression with multiple imputation was used and the two factors associated with HIV prevalence were weight and MUAC. The risk decreased with increasing weight and it was lower among acutely malnourished (MUAC “red”) MDR-TB patients. This study suggests when dealing with HIV/MDR-TB coinfection data there is need to pay keen attention to missingness and being Severely malnourished has less chances of the patient being coinfected with HIV/MDR-TB. The patients with HIV/MDR-TB coinfection should be treated as special category and attention be given to support the nutritional needs of the clients and treatment so as to have improved desirable outcome.