Lessons Learned from Data Harmonization across the Seek, Test, Treat, and Retain (STTR) Studies

Introduction: Harmonizing data from a range of HIV-related studies facilitates the large sample sizes needed to investigate short and long-term effects of substance use on HIV care cascade and other clinical outcomes.

Methods: The Seek, Test, Treat, and Retain (STTR) cohort consortium is a unique combination of 21 studies of participants with criminal justice involvement and vulnerable populations particularly related to substance use focused on HIV treatment cascade outcomes. These studies collected information across 11 domains such as demographic characteristics, criminal justice involvement, HIV risk behaviors, HIV and/or hepatitis C infections, laboratory measures of CD4 cell count and HIV viral load, mental health, socioeconomic status, health care access, and substance use. We harmonized this data to allow complex analyses of outcomes over diverse populations.

Results: Lessons have been learned, including the need for flexible harmonization approaches across variable types, ranging from simple practical approaches to collapse similar items with different response categories to modern psychometric approaches using item response theory to calibrate instruments with different measurement properties and precision onto a single metric. Creating harmonized variables of varying specificity was essential, with more specific definitions generally resulting in a smaller number of eligible subjects. A strength of STTR has been diversity including both domestic and international studies, intervention and observational studies, and studies at every step of the HIV care cascade, including “Seek” studies among those without diagnosed HIV. This diversity has resulted in different subsets of studies in every analysis, highlighting the need for a nuanced approach to inclusion/exclusion study criteria. Including studies outside STTR (such as the CNICS study) has enabled clinically relevant analyses that would not have otherwise been feasible by using STTR studies alone. One study example that is only feasible due to harmonized data across studies is the impact of reducing but not stopping illicit opiate use on HIV viral load over time (e.g., decreasing opioid use was associated with a 47% lower viral load (p=0.02)).

Conclusions: Data harmonization has allowed us to examine clinically important outcomes that would otherwise not be feasible, such as demonstrating the benefit of opiate reduction even without cessation. Including additional studies outside STTR, when needed, enabled better understanding of the differences between vulnerable populations and the larger HIV-infected population in care across the U.S. Combining individuals from many studies allowed outcomes and risk factors to be better evaluated among unique populations, such as transgender individuals.