Using serial cross-sectional surveys to create a retrospective nested cohort to determine HIV incidence from 20 US cities

Abstract

Objective: To estimate HIV incidence using successive cross-sectional surveys by creating retrospective nested cohorts among MSM, people who inject drugs (PWID), and heterosexually active persons (HET). Design: Cohorts were created among participants who had at least one repeat observation across four surveillance cycles from National HIV Behavioral Surveillance in 20 US cities. Methods: Repeat participants were identified using a combination of date of birth, race/ethnicity, metropolitan statistical area, and gender. The analysis was limited to participants who tested negative for HIV at baseline and were assumed to be at risk between cycles. We calculated person-years at risk from the individual time between cycles and used the total number of seroconversions to estimate incidence and a Poisson distribution to approximate variance. Rate ratios were calculated using age, gender, race/ethnicity, and region. Results: From 2008 to 2019, successive surveys recaptured nested cohorts of 1747 MSM, 3708 PWID, and 1396 HET. We observed an incidence rate of 2.5 per 100 person-years [95% confidence interval (CI) 2.1–2.8) among MSM; 0.6 per 100 person-years (95% CI 0.5–0.7) among PWID; and 0.3 per 100 person-years (95% CI 0.1–0.4) among HET. HIV incidence was higher among younger MSM, black MSM (compared with white MSM), and PWID residing in the South and territories (compared with the Midwest). Conclusion: These estimates are consistent with previously published incidence estimates from prospective cohort studies among these populations. Using repeat cross-sectional surveys to simulate a cohort, may serve as another strategy in estimating HIV incidence.