Measuring Retention in Antiretroviral Therapy Programs—a Synthetic Review of Different Approaches for Field Use in Low- and Middle-Income Settings

Purpose of Review Retention measures in antiretroviral therapy programs are important, but there is wide variation in their calculation and relatively little systematic discussion of their relative advantages and limitations. Recent Findings We extracted and compared distinctive approaches to quantifying retention through a systematic search in PubMed and undertook a purposive selection of articles published in peer-reviewed journals and policy documents. We also created a simulated dataset and code examples to help illustrate observations about each metric. Among identified retention approaches, were metrics based only on proportions of either visits alone (constancy) or visits and appointments (visit adherence), which are simple and most accessible in settings using only paper records and registries. However, they are generally appropriate for patients with similar potential follow-up times and do not incorporate all available information. Survival analysis techniques such as Kaplan-Meier and competing risk approaches offer more nuanced retention measures over time, and can combine individuals with different potential follow-up times into one summary, but have trouble capturing the dynamic nature of retention. Newer approaches, including multi-state models and trajectory analyses, enable more nuanced examination of retention but are analytically difficult to carry out and do not yield one single summary. Summary Simple analytical approaches are more widely useable but may miss important gaps in retention. Use of complex analytical approaches might be limited by requirements of electronically available data, data management requirements, and analytic capacity. Overall, efforts to evaluate retention may benefit from informed selection of one or more approaches to meet a range of objectives.