We invited all participants of the Communities, Households and SARS-CoV-2 Epidemiology (CHASING) COVID Cohort Study [9] who completed a recent routine follow-up assessment (n=5098) to participate in the DCE. CHASING COVID Cohort Study participants were recruited online using internet-based strategies, including via referral, social media advertisements in English and Spanish, and Qualtrics Panel [9]. Recruitment and advertising strategies were periodically adjusted to increase diversity across racial, ethnic, and age groups. Eligibility criteria included being 18 years or older and residing in the United States, Puerto Rico, or Guam at enrollment. Participants provided informed consent at the baseline assessment and separately for SARS-CoV-2 antibody testing. A total of 4793 (94% of those invited) completed the DCE July 30 to September 8, 2020. A US $5 Amazon gift card incentive was offered to participants completing the DCE.

The DCE was designed and implemented using Lighthouse Studio 9.8.1 (Sawtooth Software) and deployed using Sawtooth’s online survey hosting platform. Participants were asked to consider different combinations of SARS-CoV-2 testing service features in a situation where “...the number of people hospitalized or dying from coronavirus in your community was increasing.” Each participant was presented with five choice tasks, each containing two juxtaposed scenarios comprised of different combinations of the testing features (aka attribute levels) and a “None” option if neither testing scenario was appealing or desirable. Testing service attributes included in the DCE are shown in Table 1 and included: type of test, specimen type, testing venue, and results turnaround time (see also Multimedia Appendix 1). The combinations presented and the order of their presentation to each participant were randomized to reduce bias (see Multimedia Appendix 2).

We estimated zero-centered part-worth utilities for each attribute level and overall relative attribute importance using effects coding in a hierarchical Bayesian model [10]. We used these estimates to simulate changes in uptake of the different testing scenarios that resulted from “swapping” each individual attribute level in Table 1 into the current typical testing option of a polymerase chain reaction (PCR) test using a nasopharyngeal (NP) swab in a doctor’s office or urgent care clinic, with results returned in >5 days. The baseline simulation contained three scenarios: (1) the primary current typical testing scenario; (2) a second, duplicate current typical testing scenario; and (3) a no test scenario. Each attribute level was then individually varied in the duplicated scenario, holding all levels in the other attributes in the duplicated scenario constant. The uptake of each varied scenario was simulated along with the two other original scenarios, and the uptake of the modified scenario was compared to the uptake of the primary current typical testing scenario in the baseline simulation.

We also created a hypothetical testing scenario that optimized preferences across attributes, which included PCR and serology from a saliva sample collected at a pharmacy with same-day results. We then simulated the proportion of participants who would choose this optimized scenario, the current typical testing option, or neither option. For all simulations, predicted uptake of each testing strategy was estimated using the randomized first choice method [11,12], which computes the proportion of participants that would choose each testing scenario based on its total utility, over thousands of draws per participant, assuming that each participant would select the scenario that provides them with the highest total utility summed across attributes. DCE data were analyzed and simulations were conducted using Lighthouse Studio 9.8.1.

This study was approved by the Institutional Review Board at the City University of New York (CUNY) Graduate School of Public Health.

Participants’ median age was 39 (IQR 30-53) years. Out of the 4793 participants, 51.5% (n=2468) identified as female, 45.8% (n=2193) identified as male, and 2.8% (n=132) identified as other gender identities (nonbinary, transgender male, transgender female). Additionally, 62.8% (n=3009) identified as non-Hispanic White, 16.4% (n=788) identified as Hispanic, 9.2% (n=442) identified as non-Hispanic Black, 7.5% (n=361) identified as Asian, 3.9% (n=189) identified as another race or ethnicity, and 0.1% (n=4) had missing information for race and ethnicity. At enrollment, 29.0% (n=1391) of participants resided in the Northeast, 28.3% (n=1358) resided in the South, 23.9% (n=1146) resided in the West, 17.7% (n=850) resided in the Midwest, and 0.2% (n=7) resided in Puerto Rico or Guam; 0.9% (n=41) of participants affirmed US residence but did not provide a zip code with which to assign them a region.

Results turnaround time had the highest relative importance (30.4%), followed by test type (28.3%), specimen type (26.2%), and venue (15.0%; see Table S1 in Multimedia Appendix 3). Participants strongly preferred rapid receipt of results, with progressively weaker preference for slower test results. Within test type, participants showed a strong preference for testing scenarios that detect both current and past infection (see Table S2 in Multimedia Appendix 3). Participants most preferred testing scenarios that use cheek swab specimens and least preferred scenarios that require a deep NP swab. There was a preference for at-home self-collection of specimens using kits received and returned via mail; testing in a doctor’s office or urgent care clinic was the least preferred testing venue. Participants chose neither testing option in only 3.6% (861/23,965) of the choice tasks.

Simulating changes in SARS-CoV-2 testing uptake by varying attribute levels individually, we found the largest marginal increases in testing uptake from including immediate test results (+47%) or same-day test results (+43%), with more modest increases for results 48 hours after the test (+36%) and 5 days after the test (+16%), compared to results in >5 days (see Figure 1). Testing scenarios that offered both PCR and serology also substantially increased marginal uptake (+43%), whereas serology testing alone slightly decreased uptake (–3%). Among specimen types, oral specimens (cheek or spit [+42%]) had the largest increase in uptake over NP swab, followed by finger prick (+39%), urine (+38%), shallow nasal swab (+36%), and blood draw (+25%). Though smaller in magnitude, we found increases in uptake for testing venue alternatives to a doctor’s office or urgent care clinic, with the greatest increases for the home testing venues (receiving and returning the test kit in the mail [+15%] and receiving the kit in the mail and returning to a collection site [+14%]), followed by pharmacy (+13%), drive-through community testing site (+13%), and walk-in community testing site (+2%).

We also simulated the proportion of participants that would pick the current typical testing scenario versus a scenario with multiple more preferable features: both PCR and serology using a saliva specimen collected at a pharmacy with same-day test results. Simulated uptake of this hypothetical scenario was 97.7% compared to 0.6% for the current typical testing scenario, with 1.8% opting for no test when presented with these two choices.

Our participants preferred faster test results from less invasive specimens collected at home that provide comprehensive information about current and past infection. From a public health perspective, faster test results are more actionable [13], and at the individual level, delayed test results can provoke anxiety in other diagnostic settings [14-16]. Participants tended to favor specimen collection venues that could be construed as more convenient (pharmacy or drive-through testing site) or better able to facilitate social distancing (home), compared to a walk-in clinic or doctor’s office, where one might be more likely to come into contact with infectious individuals. Our venue-related results are in line with findings from the HIV and sexually transmitted infection literature, where at-home specimen collection for diagnostic testing has high acceptability and reliability [17,18], and with other recent findings indicating high willingness to collect at-home specimens for a SARS-CoV-2 research study [19]. The strong preference for both PCR and serology may be related to a belief that antibodies confer immunity against subsequent infection [20,21] and a general desire to get the most utility out of a single specimen.

Our findings suggest that expected advances in SARS-CoV-2 testing technologies, such as rapid, at-home saliva tests, will be highly acceptable and used when they become available, particularly in communities with increasing deaths or hospitalizations. Some preferred tests for SARS-CoV-2 (eg, at-home rapid antigen tests) may be less sensitive than gold standard diagnostic tests (PCR via NP swab). Nevertheless, these findings are significant from a public health standpoint since it’s possible that widespread and frequent use of a less sensitive SARS-CoV-2 antigen test could detect much greater numbers of people with active infection—and more quickly—than the current typical testing scenario [22]. Indeed, our data suggest that NP swabs may be a deterrent to testing, which could be addressed by adding serology or relying on saliva specimens.

Limitations of this study include the omission of other attributes that may influence testing preferences, such as frequency of testing, cost, facility wait times, or distance. In addition, the majority of our participants had already completed at-home self-collection of a dried blood spot specimen for our study. Though the venue attribute had the lowest relative importance, this prior experience may have influenced their preferences for venue in the DCE.

To the extent that it is possible to align public health strategies to deliver testing services with the preferences of those being targeted for testing, greater uptake and engagement may be achieved. Additional research is needed to increase SARS-CoV-2 testing uptake in ways that are aligned with the public health goals of the pandemic response, including preferences for engaging in public health interventions following a positive test, such as isolation and contact tracing [3].