Background: Women with human immunodeficiency virus (HIV) often have bacterial vaginosis (BV). The goal of this analysis was to assess how BV prevalence changed over time and across U.S. regions in enrollment cohorts of the Women's Interagency HIV Study. Methods: In a multisite study, BV was diagnosed retrospectively when pH and two of three other Amsel criteria were met. Prevalence was determined across four recruitment waves: 1994-5, 2001-2, 2011-2, and 2013-5. Generalized estimating equation multivariable logistic regression models assessed changes in visit prevalence across waves after controlling for HIV disease severity and other risks. Results: Among 4,790 women (3,539 with HIV and 1,251 without HIV), BV was diagnosed at 7,870 (12%) of 64,444 visits. Baseline prevalence across enrollment waves was 15.0%-19.2%, but declined in all cohorts, with prevalence in the initial cohort falling to 3.9% in the 1994-5 cohort after up to 21 years of continuous observation. Prevalence varied within U.S. regions. HIV status was not associated with BV. Conclusion: BV prevalence decreased with time in study. Prevalence varied across sites, but was not uniformly increased or decreased in any U.S. region. Clinical Trials.gov identifier: NCT00000797.