Developing a predictive model to identify sub-optimal antiretroviral therapy adherence in Namibia.
Kaonga, Nadi.
2016
-
Abstract: Addressing known and unknown risk factors for non-adherence to antiretroviral therapy (ART) has proved challenging, despite increased access. Work continues to be done to determine how best to optimize adherence, and it remains a priority for HIV/AIDS researchers and implementers. One proposal is to identify sub-groups most at-risk, prior to treatment initiation, and provide resources th... read moreat they may need to maximize their therapy. Therefore, the purpose of this study was to develop and assess a pre-treatment screening tool to identify individuals most at risk ("high-risk") for sub-optimal ART adherence. The screening tool took the form of a diagnostic predictive model. The outcome of interest was calculated using the medication possession ratio (MPR); an MPR below 75% was defined as sub-optimal adherence. Using the Namibia Defaulter Tracing Study Database, a predictive model was developed and evaluated. Multiple imputation methods were used to address missing data. Prior clinical knowledge and univariable analyses helped inform variables for consideration in the model. Multiple logistic regression techniques were used to build the multivariate model. The final model contained the following variables: age (younger being more vulnerable to sub-optimal adherence), gender (males being more likely to have sub-optimal adherence), travel cost, location, religion and beliefs, and food insecurity. Being younger, being male, having higher travel costs, having greater mobility (location), not being Christian, believing God can heal and food insecurity were associated with a greater likelihood of having sub-optimal adherence. Once the model was built, performance measures were assessed. The overall model had modest performance (Brier score: 0.18, c-statistic: 0.66). The predictors can be used to identify a sub-set of the population that is prone to struggling with adherence. It will be worth exploring the relevancy, interactions and challenges faced by individuals with any of the following characteristics in Namibia further and comparing the findings to similar populations elsewhere. Next steps will include verifying the findings with providers in practice, externally validating and updating the model and determining ways to present the information in a meaningful manner for use.
Thesis (M.S.)--Tufts University, 2016.
Submitted to the Dept. of Clinical & Translational Science.
Advisors: Steven Hong, and Jason Nelson.
Committee: Christine Wanke, and Alice Tang.
Keywords: Public health, Health sciences, and Medicine.read less - ID:
- cr56nc55c
- Component ID:
- tufts:20387
- To Cite:
- DCA Citation Guide EndNote
