Model-based Respondent-driven sampling analysis for HIV prevalence in brazilian MSM.

Area of Science:

• Epidemiology
• Network Analysis
• Biostatistics

Background:

• Respondent Driven Sampling (RDS) is crucial for studying hard-to-reach populations but yields non-representative samples requiring bias correction.
• Network Model-Assisted (NMA) method offers a way to compute weights for RDS data by modeling contact networks.
• Limited information on population homophily can pose challenges for accurate RDS analysis.

Purpose of the Study:

• To investigate the efficacy of the NMA method for estimating HIV prevalence from RDS data with limited homophily information.
• To assess the NMA method's ability to provide unbiased estimates despite population homophily and differential activity.
• To evaluate the NMA method's performance in the context of preferential recruitment in RDS.

Main Methods:

• Utilized an iterative procedure based on the Network Model-Assisted (NMA) approach.
• Modeled the underlying contact network of the population.
• Applied the NMA method to Respondent Driven Sampling data.

Main Results:

• The NMA method, using an iterative procedure, achieved unbiased HIV prevalence estimations even with strong population homophily and differential activity.
• The method demonstrated limited bias in the presence of preferential recruitment.
• Applied to Brazilian men who have sex with men, the study confirmed high HIV prevalence ranging from 3.8% to 22.1%.

Conclusions:

• The NMA method is a robust tool for estimating HIV prevalence from RDS data, particularly when homophily information is scarce.
• The iterative NMA approach effectively corrects for biases inherent in RDS, ensuring more accurate population estimates.
• The findings highlight the significant HIV burden among men who have sex with men in Brazilian urban centers.