This work was accepted for a poster presenation at the 2017 International AIDS Society (IAS) Conference on HIV Science in Paris, France!
Background: CDC has estimated population-level viral suppression (VS) using two data sources. The National HIV Surveillance System (NHSS) laboratory-based estimate (55% of HIV-diagnosed persons in 2013) is from a subset of jurisdictions that vary yearly. The national Medical Monitoring Project (MMP) estimate (42%) is based on a sample of persons receiving HIV care during the first 4 months of the year, possibly excluding some receiving care later. We reconstructed national viral suppression estimates to account for persons receiving care later in the year.
Methods: Using 9 U.S. North American AIDS Cohort Collaboration on Research and Design (NA-ACCORD) cohorts during 2009–2013 (~30,000 patients/year), we assessed timing of HIV care receipt, demographics, and VS status at last test (<200 vs. ?200 copies/mL). We standardized MMP to NA-ACCORD data, using multivariable regression models of care attendance (inside vs. exclusively outside MMP’s sampling period) and VS among NA-ACCORD participants, to yield adjusted national VS estimates with 95% confidence intervals (CI).
Results: An estimated 51% (95% CI: 46–55) of HIV-diagnosed persons in the U.S. achieved VS in 2013. Impacts of MMP’s 4-month sampling period on the national VS estimate were: 1) a 20% (CI: 18–22) underestimation of the number of persons receiving HIV care, causing undercounting of the number of persons virally suppressed; 2) to a lesser extent, lower rates of VS among NA-ACCORD participants receiving care outside versus inside the 4-month sampling period (66% vs. 80%). These findings explained the difference in VS estimates derived from standardized versus unadjusted MMP data (42% [95% CI: 39–46]). VS among HIV-diagnosed persons increased from 43% (CI: 39–46) in 2009 to 51% in 2013 in the standardized population (by age and race in Figure).
Conclusions: This methodology yielded national VS estimates closer to the NHSS estimate than those previously published and can be used to assess temporal trends.