Presentation (CTI)

OBJECTIVES: Estimate preferences for COVID-19 pre/post-exposure prophylaxis (PrEP/PEP) among US respondents using a discrete choice experiment (DCE), accounting for expected heterogeneity.
METHODS: In an online, structured survey (May-June 2025), respondents considered a treatment scenario aligning with COVID-19 clinical trial entry criteria and endpoints, followed by 10 choice sets: 8 main sets (balanced, partial-factorial design) plus 2 dominance/contraction tests (excluded to avoid double-counting). Choice attributes include risk-reduction (RR), protection duration, tolerability and drug-drug interactions (DDI). A no-treatment option was included to account for respondents with a strong preference to opt-out of treatment. Four models were fitted to the choice data: Conditional logit (clogit) - population-average preferences; latent class model (LCM) - discrete class heterogeneity; nested logit (nlogit) - independence-of-irrelevant-alternatives sensitivity; mixed logit (mixlogit) - continuous heterogeneity. Pre-specified clogit sub-group analyses used a clinical-rationale interaction model with joint Wald χ²(7) tests; sub-set clogits were fitted for special clinical groups.
RESULTS: In the pooled clogit (n=10,475), respondents preferred higher RR (β=+0.77 for 25% → 75%, standard error (SE) 0.015), no DDI (β=+0.49, SE 0.012) and a well-tolerated medicine (β=+0.33, SE 0.013); however, preferences were heterogeneous. Four preference classes emerged from the LCM (K=4): strong opt-out preference (17%, overlaps with the pre-defined Non-Trader group: no-treatment chosen in ≥4/8 main choice sets), RR-focused (36%), safety-attribute-focused (27%), and safety-conditional (19%). Mixed logit decomposition: Trader vs Non-Trader differences accounted for 59-87% of per-attribute preference variance. In the interaction clogit, all three clinical-rationale groups preferred higher RR, but the strength of preference did not track clinical rationale; no-DDI preference did, with immunocompromised respondents showing the strongest preference to avoid drug interactions and the generally healthy group the lowest.
CONCLUSIONS: Preferences for COVID-19 PrEP/PEP are heterogeneous; latent class and sub-group analyses revealed clinically meaningful preference patterns behind population averages: all engaged participants prefer high RR. Clinical rationale for PrEP/PEP corresponds to no-DDI preference.