Autor(es): Szwarcwald, Celia Landmann ; Junior, Paulo Roberto Borges de Souza ; Damacena, Giseli Nogueira ; Junior, Aristides Barbosa ; Kendall, Carl
Data: 2012
Origem: Oasisbr
Assunto(s): Brazil; Complex sample; Design effect; FCSW; HIV prevalence; RDS; HIV; Análise de Dados; Profissionais do Sexo; Estudos transversais / estatística & dados numéricos; Análise de Variância; Gráficos; Brasil /epidemiologia
Fundação Oswaldo Cruz. Instituto de Comunicação e Informação Científica e Tecnológica em Saúde. Rio de Janeiro, RJ, Brasil.
Fundação Oswaldo Cruz. Instituto de Comunicação e Informação Científica e Tecnológica em Saúde. Rio de Janeiro, RJ, Brasil.
Fundação Oswaldo Cruz. Instituto de Comunicação e Informação Científica e Tecnológica em Saúde. Rio de Janeiro, RJ, Brasil.
Ministério da Saúde. Programa Nacional de DST e Aids. Assessoria de Avaliação. Brasília, DF, Brasil.
Tulane University. Department of Community Health Sciences. Center for Global Health Equity. New Orleans, LA, USA.
Background: Respondent-driven sampling (RDS) is a chain- referral method that is being widely used to recruit most at-risk populations. Because the method is respondent driven, observations are dependent. However, few publications have focused on methodological challenges in the analysis of data collected by RDS. Methods: In this article, we propose a method for estimating the vari- ance of the HIV prevalence rate, based on the Markov transition prob- abilities and within recruitment cluster variation. The method was applied to a female commercial sex workers study carried out in 10 Brazilian cities in 2008. Both the inverse of network size and the size of the city were considered in the estimation of overall sampling weights. The study included a behavior questionnaire and rapid tests for HIV and syphilis. Results: About 2523 interviews were conducted successfully, excluding the seeds. Results show a positive homophily between recruits for those HIV+; HIV- recruiters selected HIV+ recruits 4% of the time; HIV+ recruiters selected other HIV+ recruits 19.6% of the time, about 5 times higher. The prevalence rate was estimated at 4.8% (95% confidence interval: 3.4 to 6.1), and a design effect of 2.63. Conclusions: Using statistical methods for complex sample designs, it was possible to estimate HIV prevalence, standard error, and the design effect analytically. Additionally, the proposed analysis lends itself to logistic regression, permitting multivariate models. The stratification in cities has proved suitable for reducing the effect of design and can be adopted in other RDS studies, provided the weights of the strata are known.
