<p>Dataset.</p>

A mixed logit model was used to estimate the marginal utilities, relative importance (RI), HIVST profile utilities and predicted uptake (compared to opt-out). Between December 2024 and May 2025, 340 participants were enrolled with mean age 19.8 years (standard deviation (SD): 2.6), and 65.9% of them were not in school. In the main effects model, participants preferred; facility-based over community-based distribution model (β = 0.14 (95% confidence interval (CI): 0.06, 0.23)), primary over secondary delivery channel (β = 0.09 (95% CI: 0.00, 0.17)), and unassisted over assisted testing (β = 0.46 (95% CI: 0.37, 0.54).

The sample type didn’t significantly influence the preferences of participants to use HIVST (β = 0.05 (95%CI: -0.04, 0.13). The level of assistance emerged as the most important attribute (RI: 60%), followed by facility-based distribution model (RI: 18.9%) and primary delivery channel (RI: 12.2%). There was a significant interaction between delivery channel and level of assistance (β = 0.66 (95% CI: 0.07, 1.24)).

Primary delivery was preferred over secondary delivery when testing was unassisted (β = 0.41 (95% CI: 0.11, 0.71), but not when testing was assisted (β = -0.25 (95% CI: -0.56, 0.06). AGYW preferred HIVST models combining primary delivery with unassisted testing while facility-based distribution also increased preference. This highlights the importance of considering both individual attributes and their interaction when designing HIV delivery models.</p></div>