National HIV prevalence estimates for sub-Saharan Africa: controlling selection bias with Heckman-type selection models

Abstract

ObjectivesPopulation-based HIV testing surveys have become central to deriving estimates of national HIV prevalence in sub-Saharan Africa. However, limited participation in these surveys can lead to selection bias. We control for selection bias in national HIV prevalence estimates using a novel approach, which unlike conventional imputation can account for selection on unobserved factors.MethodsFor 12 Demographic and Health Surveys conducted from 2001 to 2009 (N=138 300), we predict HIV status among those missing a valid HIV test with Heckman-type selection models, which allow for correlation between infection status and participation in survey HIV testing. We compare these estimates with conventional ones and introduce a simulation procedure that incorporates regression model parameter uncertainty into confidence intervals.ResultsSelection model point estimates of national HIV prevalence were greater than unadjusted estimates for 10 of 12 surveys for men and 11 of 12 surveys for women, and were also greater than the majority of estimates obtained from conventional imputation, with significantly higher HIV prevalence estimates for men in Cote d'Ivoire 2005, Mali 2006 and Zambia 2007. Accounting for selective non-participation yielded 95% confidence intervals around HIV prevalence estimates that are wider than those obtained with conventional imputation by an average factor of 4.5.ConclusionsOur analysis indicates that national HIV prevalence estimates for many countries in sub-Saharan African are more uncertain than previously thought, and may be underestimated in several cases, underscoring the need for increasing participation in HIV surveys. Heckman-type selection models should be included in the set of tools used for routine estimation of HIV prevalence.

A Simultaneous Equation Approach to Estimating HIV Prevalence With Nonignorable Missing Responses

“…If data are missing because HIV positive individuals are more likely to decline to test (con-2 th , To appear in the Journal of the American Statistical Association ditional on observed characteristics), then the assumption of missing at random is violated and hence conventional methods, including imputation or analysis based only on non-missing observations, will generate biased results (e.g., Heckman, 1990;Puhani, 2000;Vella, 1998;Janssens et al, 2014). In addition, because imputation-based models do not acknowledge that there is uncertainty surrounding the relationship between participation in testing and HIV status, confidence intervals based on this approach are likely to be too narrow when non-participation is common (Hogan et al, 2012).…”

“…Selection bias occurs if HIV prevalence among those who participate in testing differs from those who do not. In many contexts the extent of non-participation is substantial; for example, 37% of eligible male respondents failed to participate in testing in the 2004 Malawi Demographic and Health Survey (Hogan et al, 2012).…”

National South African HIV prevalence estimates robust despite substantial test non-participation

“…One such variable is the identity of the interviewer who attempted to enrol each potential participant into testing: supervisorassigned interviewers vary in their ability to persuade invited individuals to participate, but the HIV status of participants is unlikely to be associated with the identity of the interviewer assigned to them. [5,9,10] Past analyses of HIV prevalence in Africa using interviewer identity as a selection variable have found varying levels of bias, from none to an almost doubling of HIV prevalence. [5,9,11] In SA, a recent analysis of a full-population cohort in rural KwaZulu-Natal Province found significant selection bias.…”

“…[5,9,10] Past analyses of HIV prevalence in Africa using interviewer identity as a selection variable have found varying levels of bias, from none to an almost doubling of HIV prevalence. [5,9,11] In SA, a recent analysis of a full-population cohort in rural KwaZulu-Natal Province found significant selection bias. [12] We conducted a selection model analysis on the most recent SA national HIV prevalence survey to determine whether existing estimates are affected by selective survey non-response.…”

“…[3] There has been increasing concern about response rates in nationally representative surveys potentially allowing for biased HIV estimates. [5] Consent to test among those eligible to be interviewed has been <70% in all four SA surveys.…”