List Randomization for Eliciting HIV Status and Sexual Behaviors in Rural KwaZulu-Natal, South Africa: A Randomized Experiment Using Known True Values for Validation

Citation

Haber, Noah; Bärnighausen, Till; Harling, Guy; Cohen, Jessica R.; Mutavedzi, Tinofa; Tanser, Frank C.; Gareta, Dickman; Herbst, Kobus; Pillay, Deenan; & Fink, Günther (2018). List Randomization for Eliciting HIV Status and Sexual Behaviors in Rural KwaZulu-Natal, South Africa: A Randomized Experiment Using Known True Values for Validation. BMC Medical Research Methodology, 18, 46. PMCID: PMC5968464

Abstract

Background: List randomization (LR) is a survey method intended to mitigate biases related to sensitive statuses and behaviors for true/false questions by shrouding patient responses in a longer list of questions. We piloted this method for HIV-related responses in the low-education, high HIV prevalence setting in KwaZulu-Natal to test LR-elicited answers against known truth, using researcher-known HIV status and test refusal data.
Methods: Participants were recruited from the participation list from the 2016 round of the Africa Health Research Institute demographic surveillance system, oversampling individuals who are HIV positive. Participants were randomized to two groups: the list block arm (60%) and the direct questionnaire arm (40%). In the list block arm, participants were given blocks of 5 true/false questions, one of which was the sensitive item (e.g. I used a condom during my last sexual encounter), and the remaining 4 are non-sensitive (e.g. I prefer mangoes over strawberries). In the direct questionnaire arm, participants were administered the non-sensitive questions first, and then directly asked the sensitive questions. We compare mean responses to true HIV and HIV status data collected through the demographic surveillance system.
Results: A total of 483 participants were enrolled in the study. 262 were randomly assigned to the list block arm and the remaining 221 to the direct questionnaire arm. LR estimated the percent of the sample population who was HIV negative to be 56% (95% CI: 40% to 72%), compared to direct questionnaire which estimated 47% (95% CI: 39% to .54%), while the true percent was 26% (95% CI: 22% to 30%). 15% of the population had refused HIV tests (95% CI: 12% to 18%). LR method estimates suggest 55% (95% CI: 37% to 73%), while direct questioning yielded 13% (95% CI: 4% to 8%), while the true population prevalence was 15% (95% CI: 12% to 18%).
Conclusions: In this context, the list randomization method performed poorly. List randomization is an inherently complex method to administer, introducing a wide variety of opportunities for bias and error in this setting.

URL

http://dx.doi.org/10.1186/s12874-018-0507-9

Reference Type

Journal Article

Year Published

2018

Journal Title

BMC Medical Research Methodology

Author(s)

Haber, Noah
Bärnighausen, Till
Harling, Guy
Cohen, Jessica R.
Mutavedzi, Tinofa
Tanser, Frank C.
Gareta, Dickman
Herbst, Kobus
Pillay, Deenan
Fink, Günther

PMCID

PMC5968464

ORCiD

Haber - 0000-0002-5672-1769