Validation of self reported measures of adherence to ART and factors associated with adherence in Jinja, Uganda

Made, Felix
Journal Title
Journal ISSN
Volume Title
Background: Good adherence to ART prolongs survival and improves quality of life in people living with HIV/AIDS. Adherence is commonly assessed using self-reported measures, but these tend to over-estimate adherence. Viral load testing is the gold standard for measuring ART adherence but it is unaffordable in resource limited settings. Therefore, the aims of this small sub-study were to validate self-reported measures of adherence and to find factors associated with adherence to ART in Jinja, Uganda. Methods: This study was a secondary analysis of data collected from a cluster randomized equivalence trial which was carried out to compare facility based ART care versus home based care. In the main study, 1453 participants aged 18 and above were enrolled. A total of 1276 men and women qualified for this sub-study. Receiver operating characteristic (ROC) was computed to see how well two self-reported measures of adherence predicted virological failure. The two self-reported measures were firstly a visual analogue score (VAS) where participants rated the number of doses that they had taken in the past month on a scale from 0 (meaning no ART taken) to 100 (meaning that all required doses had been taken) and secondly an adherence score based on the number of pills missed in the three days before the visit. Logistic regression models were fitted with survey estimator to find factors associated with virological failure. Tobit models were fitted to find factors associated with self-reported adherence measures, since these were restricted to the range of 0-100% and censored. We then compared associated factors among the three different outcome measures. Results: There were 914 women and 362 men in this study. Home based care had larger number of patients (754) than facility based care (522). The median age of the patients was 38 years (IQR 32.0-44.0). Most of the participants were either married (518) or single (456). The majority of the trial participants had primary school education (n=713) and very few achieved tertiary education. A large number of participants had CD4 cell counts of less than 50 cells/mm3 (n=351), and very few of the patients in the trial had CD4 counts greater than 200 cells/mm3. The median CD4 count of the study participants was 116 cells / mm3 (IQR 43.0-167.0). A very large number of the patients were either in WHO clinical stage II or III (Stage II: n= 595; Stage III: n=577). A total of n=1079 (84.56%) and n=197 (13.44%) participants had no virological failure and failure respectively. The ROC methods showed that the iv self-reported adherence measures estimated virological failure with a sensitivity that ranged between 35-65%. Female patients had lower odds of experiencing virological failure (odds ratio: 0.7; 95% CI: 0.485, 0.968; p=0.033). The odds of virological failure decreased with each one year increase in age (OR: 0.95; 95% CI: 0.928, 0.979; p=0.001). Participants who found adherence reminders very useful were less likely to experience virological failure (P=0.001). Conclusion: This study show that self-reported measures are not good predictors of ART adherence since approximately only a half of the Jinja participants with virological failure were predicted by such measures. None of the factors associated with virological failure was also associated with both of the self-reported adherence measures. Viral load testing should be encouraged in place of self-reported adherence measures to ART. In addition, alternative methods of measuring adherence such as electronic medication monitoring, pharmacy refills and drug level detection should be investigated.
A research report submitted to the Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, in partial fulfillment of the requirements for the Degree of Master of Science in Epidemiology and Biostatistics Johannesburg, 14th December 2016