Applying convergent parallel mixed methods design to investigate medication adherence among people living with HIV in resource-limited settings: methodological insights from Ghana

Background Suboptimal adherence to antiretroviral therapy (ART) remains a critical challenge in achieving viral suppression among people living with HIV in resource-limited settings. This study demonstrates the application of convergent parallel mixed methods design to investigate adherence barriers among patients who experienced at least one missed clinic appointment but successfully re-engaged in care. Methods We employed a convergent parallel mixed methods design combining quantitative survey data from 288 participants with seven-day pill count adherence measurements across 36 healthcare facilities in Ghana’s Ashanti Region, with qualitative in-depth interviews from 20 participants (10 patients, 10 healthcare providers). Multilevel regression examined associations between demographic factors, self-efficacy, social support, economic factors, and other theoretical constructs with adherence rates, accounting for facility clustering. Thematic analysis of qualitative data identified adherence barriers and facilitators, with findings integrated through joint display matrices. Results Mean adherence rates were 80.29% (SD = 20.55%) for efavirenz-based regimens, 79.00% (SD = 22.73%) for dolutegravir-based regimens, and 78.58% (SD = 23.06%) for zidovudine-based regimens, all falling below the 95% threshold for optimal viral suppression. Multilevel regression revealed significant associations with gender, education, employment, and theoretical constructs, though severe multicollinearity (VIF up to 379.19) limits causal interpretation. Qualitative analysis identified five themes: economic barriers, HIV-related stigma, healthcare facility factors, individual motivation, and social support. Integration revealed convergence for economic barriers and healthcare facility factors. Conclusion This methodological demonstration illustrates rigorous integration approaches for convergent parallel designs. Findings suggest multilevel barriers requiring comprehensive interventions, pending confirmation in larger representative samples with linkage to clinical outcomes including viral load suppression.