Background: Eswatini still has the highest HIV prevalence globally, and very high HIV incidence among AGYW. Cash transfers linked to school attendance were protective against HIV (Baird et. al 2012), but subsequent studies have not shown impact on HIV incidence (Karim et al. 2015; Pettifor et al. 2016).
Methods: From Nov2015 to April2016, the Sitakhela Likusasa Impact Evaluation enrolled 4389 HIV-negative AGYW aged 15-22 -- 50% of whom were not in education -- in a cRCT of periodic financial incentives for HIV prevention, with HIV incidence as the main outcome. Using a 2x2 factorial design to create 4 sub-arms, 50% of participants were eligible for financial incentives conditional on education enrollment and attendance, and 50% were eligible for periodic raffle prizes conditional on periodic random selection into the raffle, on negative tests for syphilis and Trichomonas vaginalis (TV) if selected, and on being a periodic raffle winner. Education data were collected throughout. The endline survey, 3 years later, included behavioral and risk profile data, and HIV, syphilis and TV testing.
Results: HIV incidence among participants in the education incentive arm was statistically significantly lower compared to those not eligible for the education incentive, 6.34% vs 8.08% (p=0.041); OR: 0.770 [0.599-0.989]; aOR: 0.751 [0.579-0.974]. HIV incidence in the sub-arm offering both the education and raffle incentive was significantly lower than incidence in the control arm (participants not eligible for any of the two incentives), OR: 0.634 [0.442-0.910]; aOR: 0.618 [0.429-0.889].
Conclusions: The financial incentives conditional on education participation significantly reduced odds of HIV infection among AGYW in Eswatini. Raffle incentives on their own did not lead to a statistically significant impact, but it amplified the effect: the combination of both incentives statistically significantly further reduced the odds of HIV infection. Financial incentives can be useful for HIV prevention among AGYW in high prevalence settings.

 HIV incidence over study period, % (n/N)OR [95%CI]P-valueaOR1 [95% CI], adjusted for being in the raffle arm or the education armP-valueaOR 2 [95% CI], being in school, rural-urban residence, region of residence, age, attitude towards risk and assets based social economic status, as well as, in rows 3 and 5 only, for being in the raffle arm or the education arm, as relevantP-value
Education Incentive armEducation Incentive control8.08% (153/1894)1 1 1 
 Education Incentive treatment6.34% (119/1878)0.770[0.599-0.989]0.0410.766[0.596-0.985]0.0370.751[0.579-0.974]0.031
Raffle incentive armRaffle control7.85% (145/1851)1 1 1 
 Raffle treatment6.61% (127/1921)0.833[0.648-1.070]0.1470.827[0.644-1.063]0.1370.823[0.641-1.056]0.126
Randomization sub-armsControl8.84% (80/905)1   1 
 Raffle only7.38% (73/989)0.822[0.588-1.149]0.252n/a 0. 823[0.589-1.149]0.253
 Education only6.87% (65/946)0.761[0.542-1.069]0.115n/a 0.751[0.528-1.069]0.112
 Raffle and Education5.79% (54/932)0.634[0.442-0.910]0.013n/a 0.618[0.429-0.889]0.010
TOTAL7.21% (272/3772) 
[Sitakhela Likusasa Impact Evaluation in Eswatini: HIV incidence by randomization arms and sub-arms]