Our Work

Discrete Choice Experiments to Determine HIV Testing Preference in Tanzania (Duke-CFAR)

Project Overview

This project was implemented in conjunction with an NIMH-funded R21. CFAR and R21 funding were used to demonstrate the feasibility of using a form of stated preference (SP) survey research known as Discrete Choice Experiment (DCEs) in the context of HIV testing in a resource-poor setting. The DCE method, sometimes referred to as conjoint analysis, is used to determine which characteristics (or attributes) of good or service most influence respondents' choices or decision-making. Practically this method can be used to identify HIV testing strategies that combine diverse attributes in configurations that are not currently available.

One aim of the CFAR component of the study was to use qualitative methods to prioritize characteristics of HIV testing options with respect to their expected influence on testing decision; the results were used in to design and implemented the DCE among randomly selected community members as part of the R21. CFAR funds were further used to recruit, and conduct additional DCE surveys with, two high-risk populations in the study area: female barworkers (FBW) and male Kilimanjaro mountain porters (KMP). The goal was to compare the HIV testing preferences of high-risk populations to those of the general population.

To-date, the joint project has generated 3 published manuscripts:

Ostermann J, Njau B, Mtuy T, Brown D, Muehlbacher A. One size does not fit all: HIV testing preferences differ among high-risk groups in Northern Tanzania. AIDS Care. 2014, in press.

Ostermann J, Njau B, Brown DS, Mühlbacher A, Thielman N. Heterogeneous HIV testing preferences in an urban setting in Tanzania: Results from a Discrete Choice Experiment. Plos One. 2014. DOI: 10.1371/journal.pone.0092100

Njau B, Ostermann J, Brown DS, Mühlbacher A, Reddy E, Thielman N. HIV Testing Preferences in Tanzania: A Qualitative Exploration of the Importance of Confidentiality, Accessibility, and Quality of Service. BMC Public Health. 2014:838. doi:10.1186/1471-2458-14-838.

Additional manuscripts are in preparation. In addition, an R01 application was scored (6th percentile) that seeks to test the relationship between stated preferences and revealed preferences (i.e., actual testing decisions) in a randomized controlled trial.

In "HIV testing preferences in Tanzania: a qualitative exploration of the importance of confidentiality, accessibility, and quality of service", published in BMC Public Health, we describe our use of in-depth interviews (IDIs) and focus group discussions (FGDs) to identify which characteristics of HIV testing options most influence testing decisions. IDIs and FGDs with diverse community members, including men and women who had previously tested for HIV and those who had not, were used to develop a concept map of preference-relevant features of testing options, and to prioritize features with respect to their expected influence on testing decisions.

Using the results of the qualitative work, we developed a DCE for HIV testing. In two pilot studies and iterative pre-tests, we narrowed the number of HIV testing features evaluated in the DCE to five characteristics: distance to testing, confidentiality, testing days (weekday vs. weekend), method for obtaining the sample for testing (blood from finger or arm, oral swab), and availability of HIV medications at the testing site. The DCE was programmed into iPads and administered to 486 community members enrolled using cluster-randomization and Expanded Programme on Immunization (EPI) sampling methods.

Mixed logit analyses of DCE choice data identified distance to testing as the most important attribute to participants, followed by confidentiality and the method for obtaining the sample for the HIV test. In "Heterogeneous HIV testing preferences in an urban setting in Tanzania: Results from a Discrete Choice Experiment", published in PLoS ONE, we highlighted significant variation in preferences among participants, and the potential benefits of tailoring HIV testing interventions to match the preferences of specific sub-populations.

We subsequently used respondent-driven sampling methods to recruit 162 female barworkers and 194 male mountain porters and administered the same DCE survey. Seed participants were recruited from barworkers presenting for a health check-up at a municipal health center and from climbing companies and a porters union. In "One size does not fit all: HIV testing preferences differ among high-risk groups in Northern Tanzania," published in AIDS Care, we compare HIV risk characteristics and preferences of these two high risk groups to those of the randomly selected community sample.

Barworkers and porters exhibited significantly elevated risk profiles, and bivariate analyses of survey responses and mixed logit analyses of DCE choice data demonstrated significant variation in testing preferences between groups. On average, distance was less important to high-risk groups than to randomly selected community members. Barworkers were more reluctant to test at home and had a stronger preference for venipuncture than other female community members. Porters placed more value than other males on the availability of HIV medications at the testing site. Important

Our study demonstrated that rigorously designed and implemented DCEs can identify distinctly different sets of HIV testing preferences across and within diverse HIV risk groups in a resource-poor setting. Our R01 application seeks to help to address persistent low testing rates by identifying the stated HIV testing preferences of FBW and KMP (Aim 1), by identifying HIV testing options that are expected to better match the preferences of these populations than existing options (Aim 2), and by linking a preference-informed HIV testing intervention to actual rates of testing in the context of a pragmatic RCT (Aim 3).




Project Status


Share This