In 23 days, the DCE survey link was clicked 3319 times. 2505 respondents failed to meet the study eligibility criteria or quit before eligibility could be established, and 11 respondents were removed after deduplication of identical phone numbers. The 803 eligible participants completed 4738 choice tasks. Alternative “A” (choice on the left side) and alternative “B” (choice on the right side) had comparable probabilities of being selected (47% vs. 45%, respectively); the “opt-out” alternative was selected in 8% of all choice tasks.
Participant characteristics are presented in Table 1. Overall, study participants were relatively young (median age: 24 years old), mostly single (86%), mostly self-identified as gay (78%), and most had previously tested for HIV (68%). Distribution of education, income, and disclosure of MSM identity to health care providers aligned closely to the quota sampling scheme. Participants who never received an HIV test were more likely to be less than 28 years old (79% vs. 69%), have only elementary or middle school education (21% vs. 15%), and have income less than 217 USD per month (28% vs. 19%) compared to men who had ever received an HIV test.
Single-Item Assessment of Testing Preferences
Single-item assessment of testing preferences (i.e., participants reporting their HIV testing preferences independently for each attribute) are presented in Table 2. Regarding test location, participants most preferred to test at home (34%), followed by testing at gay community-based organizations (25%), and local health departments (18%). The most popular pricing models were free testing (55%) and incentivized testing (24%). Two out of three participants preferred walk-in testing (66%), while only one in five preferred appointment-based testing. The majority of individuals preferred to test anonymously (75%), not be required to disclose their same-sex sexual activities (60%), and to be tested by a trained health professional (60%). Finger-prick testing was preferred over venous blood testing (48% vs. 27%), but one in four participants was indifferent to the test type (25%). Participants who had never tested before had a greater preference for testing at home (45% vs. 29%), self-testing (33% vs. 15%), finger prick testing (54% vs. 46%), and real-name testing (22% vs. 14%).
Overall Design Attribute Effects
Supplementary Table S2 shows participants’ attribute-specific HIV testing preferences based on results of the MNL analysis. In contrast to the single-item assessment, home was the least preferred testing location (β = − 0.10, p < 0.01). Notably, participants expressed slightly stronger preference for free testing over testing with monetary incentives (β = 0.32 vs. β = 0.23). Preference ranking for all other parameter estimates were in the order as expected (e.g., stronger preference for cost of $7.50 USD test vs. $15 USD test).
Table 3 shows results of the MXL analysis. Each attribute contained at least one level with statistically significant standard deviation estimates of the coefficient, thus implying substantial heterogeneity of preference weights across respondents for all attributes .
Table 4 shows results of the MXL-I analysis. The MXL-I model extends the MXL model to explore preference heterogeneity by testing experience. The log-likelihood ratio test indicated that the MXL-I model fit significantly better than the MXL model (p < 0.001, χ2 = 42.3, 13 DF). The MXL-I analysis indicated that preference for testing at home was significantly stronger among test-naïve men, compared to previous testers (β = − 0.58, p < 0.001). Participants with testing experience expressed significantly stronger preference for testing at the health department (β = 0.41, p < 0.001). In addition, naïve testers were significantly more likely to choose the opt-out choice, compared to men with testing experience (β = 0.23, p < 0.05).
A sensitivity analysis was conducted to examine how sociodemographic may have influenced preferences of naïve and experienced testers. The sensitivity analysis entailed re-running the MXL-I analysis among sub-samples stratified by age, sexual orientation, and income. Results of the sensitivity analysis indicated that naïve testers’ stronger preference for home testing and weaker preference for testing at the health department (compared to experienced testers) was consistent within each stratum of age, sexual orientation, and income levels (results available upon request).
Scaled HIV Testing Preferences
Figure 3 illustrates scaled HIV testing preferences, enabling direct comparisons between testing levels and attributes. Larger values indicate stronger preference for a specific testing characteristic. Results showed that switching from real-name testing to anonymous testing was as influential on participants’ stated preferences as changing from $7.50 USD out-of-pocket testing to free testing, or changing the test administrator from a lay-person to a health professional (gain of approximately 6 points on rescaled scale).
Table 5 presents the relative importance of HIV testing attributes as percentages, stratified by HIV testing history. Overall, cost/incentive was the most important attribute (34.0%, 95% CI 31.2–36.7%), followed by anonymity (20.8%, 95% CI 18.1–23.3%), and test administrator (20.4%, 95% CI 7.0–30.4%). Testing location was of modest importance (9.2%, 95% CI 5.7–12.4%), but disclosure of MSM activity (7.9%, 95% CI 5.3–10.4%), test type (3.9%, 95% CI 1.0–6.7%), and appointment scheduling (3.7%, 95% CI 1.0–6.2%) were of limited importance. The relative importance of attributes was generally similar between participants with different testing histories.