[Potential biases in epidemiological studies using respondent-driven sampling method: a comparison between its face-to-face and online application.]

Abstract

In epidemiological and social studies on populations without a sampling frame or that are hard to reach, respondent-driven sampling (RDS), under certain assumptions, has the potential to produce asymptotically unbiased and efficient population estimates for these populations. While RDS is typically conducted face-to-face, the online version (WebRDS) has gained attention due to its potential advantages, although its disadvantages have also raised concern. The objective of this study was to contrast these two formats at the level of application and potential biases, where the online version offers improved speed and lower costs but raises concerns about potential biases due to a lack of face-to-face instructions regarding definition of the social network size and peer-recruiting process. Both formats can generate unbiased estimates, however, it's crucial to carefully consider potential sources of bias to meet the necessary assumptions. Therefore, it is important to continue researching the most appropriate analytical approaches to address the specific biases of each modality.