The Epidemiology of Clinical Retention Among HIV-Infected Persons in North America

Embargo until
Journal Title
Journal ISSN
Volume Title
Johns Hopkins University
Background: Clinical retention is central to the HIV care continuum and a determinant of improved individual- and population-level HIV outcomes. The goals of this dissertation are to improve retention surveillance by quantifying measurement error due to laboratory-measure proxies for encounters and to examine retention disparities by demographic, HIV risk, and geographic characteristics, using data from 2000-2010. Methods: We analyzed data from the North American AIDS Cohort Collaboration on Research and Design, the largest North American HIV cohort collaboration. Clinical retention was defined using the Institute of Medicine indicator: ≥2 encounters, >90 days apart, within one calendar year. Discordance between laboratory-based and encounter-based retention measures was evaluated using logistic regression with GEE and inverse probability weights for confounder adjustment. Relative times and cumulative incidences of first retention discontinuation after ART initiation by demographic and HIV risk factors were analyzed using weighted Cox regression. Geographic differences were assessed using modified Poisson and logistic regression with GEE and cluster detection methods. Results: We identified significant retention disparities by measurement method, patient characteristics, and geography, even adjusting for confounders and clinical practice differences. Misclassification of encounter-based retention by laboratory-based measures was 19% overall, which remained stable over time. Among individuals initiating ART, the cumulative incidence of retention discontinuation was 74% and adjusted cause-specific hazard ratios (HR) were lower for females (HR: 0.81, vs. males), but higher for Black (HR: 1.18, vs. non-Black) patients and individuals with injection drug use as HIV risk (IDU) (HR: 1.35, vs. non-IDU) (p<0.05, each). The South and West (adjusted Risk Ratios [RR]: 0.95 and 0.89, respectively) lagged the Northeast and Midwest (Ref. and RR: 1.03) in improved retention over the study period (p<0.05, each). Conclusions: Clinical retention improved within all groups over time, yet disparities by important characteristics persisted. Agreement between encounter-based and laboratory-based metrics was strong, but laboratory measures were imperfect proxies. Public health interventions to address poor retention in high-risk populations are needed, and more accurate surveillance of care outcomes will be essential in monitoring HIV policy benchmarks if, indeed, we are to continue to make progress toward them.
HIV/AIDS, Clinical Retention