Review of sampling hard-to-reach and hidden populations for HIV surveillance

Adequate surveillance of hard-to-reach and 'hidden' subpopulations is crucial to containing the HIV epidemic in low prevalence settings and in slowing the rate of transmission in high prevalence settings. For a variety of reasons, however, conventional facility and survey-based surveillance data collection strategies are ineffective for a number of key subpopulations, particularly those whose behaviors are illegal or illicit. This paper critically reviews alternative sampling strategies for undertaking behavioral or biological surveillance surveys of such groups. Non-probability sampling approaches such as facility-based sentinel surveillance and snowball sampling are the simplest to carry out, but are subject to a high risk of sampling/selection bias. Most of the probability sampling methods considered are limited in that they are adequate only under certain circumstances and for some groups. One relatively new method, respondent-driven sampling, an adaptation of chain-referral sampling, appears to be the most promising for general applications. However, as its applicability to HIV surveillance in resource-poor settings has yet to be established, further field trials are needed before a firm conclusion can be reached.