A Comparative Analysis of Two Data Collection Years and Sampling Methods of the National HIV Behavioral Surveillance (NHBS) System Conducted in Washington, DC.

摘要

The National HIV Behavioral Surveillance (NHBS) System is a national study that collects information on individuals in major metropolitan areas considered to be in the three populations at highest risk for HIV acquisition: men who have sex with men (MSM), injection drug users (IDUs), and heterosexuals at high risk of HIV acquisition (HET). IDU and HET data collection years are implemented using respondent-driven sampling (RDS), a form of chain-referral sampling used to access hard-to-reach populations, whereas MSM data collection years use venue-based sampling (VBS), a method of recruiting individuals from randomly selected venues frequented by the population of interest. Recommendations for both sampling methods are to present data as weighted in order to maintain limited generalizability. However, these guidelines are not strictly followed: most, but not all, RDS studies report weighted prevalence estimates and no VBS studies report weighted prevalence estimates.;In the District of Columbia (DC), NHBS-HET-1 was conducted in 2006-2007 and enrolled 750 eligible individuals; NHBS-MSM-2 was conducted in 2008 and enrolled 500 eligible individuals. The goals of this study were to examine the effects of weighting on these two data collection years of NHBS and to describe differences in behavioral and biomedical outcomes among the participants in NHBS-HET-1 and NHBS-MSM-2. Analysis of data collected using RDS revealed the importance of presenting data as weighted. We found minimal differences in results generated by SAS, Stata, and R when weighting data, and thus they can be used interchangeably. When NHBS-HET-1 data were dichotomized by gender, no differences were found between use of a univariable and a bivariable weight, indicating their interchangeability. Similar to findings from NHBS-HET-1, data collected using VBS during NHBS-MSM-2 displayed the importance of presenting weighting data. Based on our findings, we recommend that all studies that sample using RDS or VBS present data as weighted. Future, larger-scale studies are needed to validate results from this research.;A multivariable logistic regression comparison of HET and MSM revealed significant differences: the HET population was older, less likely to be white, had lower education and annual income, and were less likely to be HIV-positive. When HIV-positive individuals were compared to HIV-negative individuals in a multivariable logistic regression model, HIV-positive individuals were significantly more likely to be MSM, older, non-white, ever incarcerated, and have had 4 or more sexual partners in the previous 12 months. Comparisons revealed significant differences between HET and MSM that are potentially a factor of the recruitment method and eligibility criteria.