International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
The effect of HIV prevention products on incentives to supply unprotected commercial sex in South Africa
Matthew Quaife, Fern Terris-Prestholt, Maria Cabrera, Sinead Delany-Moretlwe, Robyn Eakle, Peter Vickerman

Last modified: 28 March 2017


Objectives: Female sex workers (FSWs) face a high risk of HIV acquisition. Because of this, new HIV prevention products are often developed and trialled among FSW populations. FSWs face different risks and incentives from other high-risk populations because they work in a competitive market environment. In addition to navigating risk in sexual relationships, FSWs face high powered incentives and market forces. In many places, agreeing to supply unprotected sex can often result in large price premiums, yet the impact of new products on the economics of sex work is not generally considered. This study explores whether financial incentives for sex work might be affected by the introduction of new HIV prevention products. Furthermore, as self-reported measures of sexual risk may be subject to substantive reporting biases, this study explores whether stated preference methods can be informative in eliciting risk behaviour data.

Methods: This study used a repeated discrete choice experiment (DCE) to explore how the introduction of an effective HIV prevention product may affect FSW preferences for pricing, type of sex-act and client characteristics. All respondents were asked to choose between two hypothetical commercial sex acts over ten choice sets. Acts were described by their price, condom use, sex type, and whether a client was thought to have an STI and/or be HIV positive. Respondents could choose to opt-out of providing either act. Later in the survey, the DCE was repeated among self-reported HIV negative respondents – these respondents were asked to choose as if they were now using a product which fully protected them from HIV acquisition. The DCE was developed through qualitative focus group discussions and a literature review, and was piloted among FSWs. A multinomial logit model was used to explore the data, before extending analysis with interaction terms to a random parameter logit model. A latent class model was also used to further explore heterogeneity.

Data: This study collected primary DCE data from 203 FSWs in urban Ekurhuleni Municipality, 30km South-East of Johannesburg. 122 respondents self-reported HIV negative status, and we restrict analysis to this sample. We used a respondent-driven sampling process to locate FSWs, and data collection was carried out using tablet computers and Open Data Kit software.

Results: Consistent across all specifications, condom use and the type of sex provided were the most influential attributes in both DCE framings. When asked to respond as if using a fully efficacious HIV prevention product, the utility provided by condom use diminished by around 25%. Price was not a significant driver of choice in current practice, however strongly influenced choices under full HIV protection. Participants appeared to understand the framing as client HIV status became non-significant under the protected framed DCE. HIV knowledge and household income did not affect preferences towards act price, however, exposure to violence from partners or clients greatly increased the utility provided by act price. In a two-class latent class analysis, class membership for the group who found price influential with product use was predicted by household hunger and experience of violence from partners or clients, both signals of substantial structural vulnerability to HIV risk. After integrating simulations into a dynamic infectious disease model, these results suggest that some of the epidemiological impact of new HIV prevention products may be mitigated through impacting the market for protected and unprotected commercial sex. Comparison of stated and self-reported revealed preferences for current choices show broad concordance.

Discussion: Before their introduction it is difficult to predict the impact of new HIV prevention products. In addition to changing risk behaviours in non-commercial relationships these products might also change the commercial sex market structure, specifically modifying incentives for FSWs to bear risks from client demand for unprotected sex. Results suggesting that condoms are valued less under HIV protection are in line with concerns in the literature that risk compensation may occur after the initiation of HIV preventive products. Because products (to date) only have partial HIV efficacy, and do not protect against other STIs or pregnancy, their total impact may be substantively reduced if not consistently used in parallel with condoms. Finally, the influence of structural vulnerability on preferences for act price reinforces social epidemiological evidence that exposure to HIV risk is more complex than a simple decision to negotiate condom use. Where FSWs face substantial income or food insecurity, the salience of a need for income may lead to the provision of riskier acts for greater financial reward.

Conference registration is required in order to view papers.