Research to improve differentiated HIV service delivery interventions: Learning to learn as we do

Many low-and middle-income countries with high HIV prevalence are working to modernize their approaches to HIV treatment service delivery.Under the umbrella term of "differentiated service delivery" (DSD), health systems are evolving away from traditional models of care, in which encounters were both frequent and exclusively at the health facility, to models that vary the location (e.g., to venues in the community), reduce the frequency (e.g., quarterly, biannually), and expand the roles of pharmacists, nurses, lay healthcare workers, and patients themselves (e.g., for peer drug distribution) in service delivery [1].Diverse practices from increasing antiretroviral medications dispensed at each visit (and thereby reducing visit frequency), to community groups where one member picks up medications for all, and many others are considered DSD models.DSD models, however, are unified by reducing unnecessary barriers to seeking care for individual patients as well as relieving frail health systems of clinically unnecessary patient encounters-which, in turn, enables reallocation of resources to reach even more patients.The role of research in the global scale-up of DSD models is a topic of sometimes heated discussion.Although some suggest that rigorous research is required before fully investing in scale-up of DSD, that view imposes, in our view, an unreasonable standard for innovation.After all, the status quo treatment models with highly standardized encounter intervals, irrespective of an individual's clinical condition and treatment history, emerged when antiretroviral therapy (ART) was restricted and many patients had low CD4 levels and relatively advanced immunosuppression.Today, because of treatment for all and a growing time on treatment, most patients are healthy, and practices targeting the sickest individuals represent an unnecessary burden for the typical patient.On the other hand, adoption of strategies without accompanying research fails to recognize the contributions that rigorous causal knowledge can play.Different models make different demands on the health system and also reduce barriers to care through different mechanisms (i.e., simply dispensing more medications at each visit addresses opportunity costs but does not create the social support that is hypothesized to play a role in models with patient groups)-which implies that their roles may differ across contexts.Understanding the comparative implementability and effectiveness (especially in combination) of DSD models through research will inform strategies for continued scalability and anticipate sustainability.Investing in implementation-based evidence during scale-up of DSD will strengthen progress toward evidence-based implementation.