This self-administered online cross-sectional study was conducted in China and the United States from January to March 2021. The questionnaire was constructed and administered using Lighthouse Studio version 9.8.1 (Sawtooth Software Inc). In the questionnaire, a total of 12 demographic questions and 7 DCE questions were included. First, demographic and socioeconomic information was collected, including age, sex, education level, annual income, and occupation, followed by 1 set of DCEs to investigate the respondents’ preferences for the initial diagnosis of fever during the COVID-19 pandemic using simulated scenarios of different diagnosis and treatment attributes. The questionnaire generally included 7 scenarios, with 1 fixed scenario and 6 hypothetical scenarios with fixed attributes and random levels, where the respondents were required to choose 1 option of 3 in each scenario.

First, the demographic idiosyncrasy of the 2 groups of respondents before and after PSM was presented. Later, the general preferences of the 2 groups of respondents were presented to show population-wide preferences for the initial diagnosis of fever during the pandemic; moreover, to compare the 2 groups of respondents from China and the United States, PSM was utilized to 1-to-1-match the respondents for 5 confounding variables (sex, age, income level, occupations, and educational level), aimed at comparing the preferences without being influenced by the confounding variables and demographic factors.

DCEs are now widely used in the fields of health care and public health [26,27]. The literature indicates that patients’ preferences strongly correlate with their willingness to use diagnosis, treatment services, and follow-up treatment [28]. By consulting several public health experts and reviewing the relevant literature [29-32], this study identified the following 6 attributes concerning diagnoses and treatment services, as well as their corresponding levels: (1) diagnosis and treatment medical institutions, (2) diagnosis and treatment personnel, (3) waiting time, (4) whether to test the nucleic acid of COVID-19 immediately, (5) medical expenses, and (6) reimbursement rate of medical expenses. The detailed attributes and their respective levels are presented in Table 1.

The questionnaire took 5-10 minutes to complete. Upon completing the questionnaire, each respondent immediately got a randomly generated 6-digit code without filling in any personal information. With this code, they received a preset US $0.10 on the Amazon Mechanical Turk (MTurk) platform as a reward. All respondents were required to be at least 18 years old and consented to participate by clicking the “agree to participate in the questionnaire” option before formally starting to answer the questionnaire. Before completing the questionnaire, all respondents were fully informed that this questionnaire was completely anonymous. Once the respondents agreed to take the questionnaire survey, they were informed that they voluntarily agreed to participate in the study and the questionnaire answers would be protected by privacy laws.

In the first part of the questionnaire, respondents were required to provide basic demographic information, including sex, age, educational level, occupation, annual salary, and marital status. In addition to the basic demographic information, respondents were also asked whether they had ever been infected with COVID-19 and whether their acquaintances had ever been infected with COVID-19. The second part asked the respondents to consider a preferred treatment plan among 3 options in a task-choice scenario. Each scenario required the respondents to imagine themselves in a fever state and asked how they would seek health care services. This questionnaire included 6 attributes with a maximum of 6 levels, 7 scenarios per respondent, 3 alternatives per scenario. One example of the task-choice scenario is shown in Figure 1. See Multimedia Appendices 1 and 2 for the English and Chinese questionnaires, respectively.

Internal validity was evaluated using the program developed by Johnson et al [33], which includes stability (with repeated questions), within-set dominated pairs, across-set dominated pairs, transitivity, and attribute dominance (noncompensatory preferences). Multimedia Appendix 3, Table S9, shows information regarding the attributes of the DCE questionnaire, and Multimedia Appendix 1, Table S10, summarizes the test summaries. The internal validity test results and the summary results are shown in Multimedia Appendix 1, Tables S11-S13. According to the relevant research [34,35], our results showed that our questionnaire is efficient.

Questionnaires were distributed via multiple international online panel providers (for data collection in the United States) and recruited volunteers (for data collection across China) from January to March 2021 [36,37]. Specifically, MTurk was used for data collection in the United States, and stratified sampling by age and geological locations was used for data collection in China [38]. MTurk was found to provide census-level sampling data during ongoing social events [39-41]. In total, 10,921 respondents participated in the survey, but only 9112 (83.4%) finished all the questions. Therefore, a total of 9112 respondents were included in the study, with 5411 (59.4%) respondents from China and 3701 (40.6%) from the United States. According to the rule of thumb [42], the sample size of a DCE depends on the number of choice tasks (t), the number of alternatives (a), and the number of analysis cells (c). According to the equation

when considering the main effects, c equals the largest number of levels for any attribute. For this study, c=6, t=7, and a=3; thus, a minimum of 143 respondents were required.

Sociodemographic data were analyzed using descriptive statistics of frequency and percentage. Comparisons between the respondents from China and the United States were conducted using the chi-square test, and the results were presented as frequencies and percentages. Statistical significance was set at P<.05 (2-tailed). All the results were analyzed using STATA version 14.0 (Stata Corp), except the latent class models (LCMs) and assessment of internal validity of the DCE, which were analyzed using Lighthouse Studio version 9.12.1.

For the DCE, a mixed logit model (MXL) was first used to quantify the preferences of the respondents for the attributes and levels of an initial diagnosis of fever during COVID-19 in their trade-off in general. After using the MXL, we dummy-coded all the attribute levels, with the levels with the lowest model parameter β as the reference level in each attribute, by which we could enhance the interpretation of the preference weights by specifying the difference between 2 random coefficients.

PSM was performed to minimize the confounding bias of respondents from the United States and China that arose from the discrepancy of the demographic characteristics in both groups [38]. Specifically, the covariates were identified through the pairwise Pearson correlation matrix, and the final set of covariates for PSM was decided by minimizing the residual confounding factors as much as possible, where a logistic regression model was conducted to estimate the propensity scores for each group of respondents. Later, we conducted 1-to-1 matching without replacement so that a candidate respondent in the United States could be matched to only 1 respondent in China, after which the distribution of the covariates between the 2 groups would be the same [43]. Finally, a total of 2480 respondents, with 1240 (50%) from China and the other half from the United States, were matched from the total 9112 respondents, with the covariates being sex, age, occupation, educational level, and annual income. The flowchart of the PSM is shown in Multimedia Appendix 3, Figure S1.

The MXL was used to quantify the preference importance and weights of the various attributes of the DCE in the respondents' trade-offs. Additionally, the utility that the coefficients and SD used measured the levels of each attribute. The attribute cost was transformed into a continuous variable. Other parameters were assigned with a normal distribution, and we generated 1000 Halton draws for each population. We assumed that the attribute levels with P<.05 were statistically significant. We calculated each attribute preference's general estimated weight to identify its importance. The formula is:

LCMs were used to explore the preference heterogeneity among the populations from the United States and China; this study also presented an LCM analysis, which divided the respondent population from the United States and China into a fixed proportion. Moreover, the number of latent groups was identified using the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) [44]. In this study, 3 groups of the respondent population from China and 3 from the United States were identified and included in further research. In the study [45], we compared the models with 2-5 classes according to the AIC, the BIC, and the consistency information criterion (CAIC). Multimedia Appendix 3, Tables S3 and S4, show the AIC, BIC, and CAIC values of different classes in China and the United States.

The willingness to pay (WTP) is a measure used to capture the upper limit of the amount of money that people are willing to sacrifice to obtain the benefits of a particular medical service, diagnosis, and treatment plan—that is, the highest amount of money that respondents were willing to sacrifice when they chose their preferred diagnosis and treatment service in this study. Our study analyzed the WTP of the respondents to determine the homogeneity or heterogeneity caused by the cost in the choice of treatment options. We estimated the WTP:

where β_cost is the coefficient on the cost parameter and vx₀ and vx₁ are the coefficient before and after a change in the level of attribute x, respectively. For each reference attribute, vx₀ was considered 0.

The respondents provided informed consent before filling in the questionnaire and agreed to participate in screening and to the use and publication of their data in journal papers. The questionnaire was completely anonymous, and the answers were protected by privacy law. During the process of filling in the questionnaire, all respondents could withdraw from the survey at any time. The study was conducted according to the guidelines of the 1964 Declaration of Helsinki and was approved by the Jinan University Medical Ethics Committee (JNUKY-2021-004). All procedures performed involving human respondents were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

A total of 9112 respondents from China and the United States were included in the final analysis, the demographic characteristics of whom are shown in Table 2. Of these respondents, 5411 (59.4%) respondents were from China and 3701 (40.6%) respondents were from the United States. After PSM, 1240 (22.9%) respondents from China and 1240 (33.5%) from the United States were matched, and no apparent differences were found between the 2 groups of respondents (P>.05 for all sociodemographic factors), as shown in Table 2.

After PSM, of the 1240 respondents from China, 1188 (95.8%) were between 18 and 60 years old and 706 (56.9%) were female. Of those from the United States, 1182 (95.3%) were between 18 and 60 years old, 705 (56.9%) were female, and 18 (1.5%) had a postgraduate degree.

The comparison of relative attribute importance between China and the United States before and after PSM is shown in Figure 2. After PSM, respondents from China attached the most importance to the types of the medical institutions (39.9%), followed by the reimbursement rate (34.3%), and the waiting time was the least essential attribute (6.5%). For respondents from the United States, the reimbursement rate was the most important attribute (34.6%), followed by the waiting time (25.3%).

The MXL results depicting the levels of each attribute of respondents' preferences in China and the United States for an initial diagnosis of fever during the COVID-19 pandemic before and after PSM are shown in Tables 3 and 4, respectively. Respondents from China strongly preferred going to a fever clinic (utility coefficient=0.974) or the emergency department (utility coefficient=0.973) compared to a network consultation. In contrast, US respondents preferred private clinics (general practices) the most. The more negative correlation of cost for the Chinese respondents showed that they cared more about the cost than the US respondents did. In addition, both populations showed a similar preference for immediate COVID-19 nucleic acid tests with a high reimbursement rate, which indicates that people consistently prefer low-consumption treatment plans.

A WTP greater than 0 indicates that the WTP can ensure a change in the reference level, while a WTP less than 0 indicates the patients are willing to pay to avoid a change in the reference level. Through the analysis, it was evident that the Chinese respondents preferred hospital emergency and fever clinics, for which they were willing to pay US $0.14 (reference level: online consultation US $0) and US $0.14 (reference level: online consultation US $0) to receive services from these 2 types of medical institutions. At the same time, they are willing to pay US $0.09 (reference level: paramedic US $0) for the treatment provided by doctors. Compared to the WTP of the respondents from China, the respondents from the United States were more willing to pay US $0.10 (reference level: online consultation US $0) and US $0.08 (reference level: online consultation US $0) for treatment in private hospitals and fever clinics, revealing a preference discrepancy with China. Both US and Chinese respondents were willing to pay US $0.15 and US $0.05 for immediate COVID-19 nucleic acid testing (Tables 5 and 6). They were willing to pay a certain amount for a shorter waiting time and a higher reimbursement rate, indicating that a diagnosis service with an immediate COVID-19 nucleic acid test, shorter waiting time, and lower cost is more acceptable for respondents. Specifically, the interaction test indicated that cost and reimbursement rate have a significant interaction. This indicates that the effects of choice obtained with the 2 attributes vary together (Multimedia Appendix 3, Tables S5-S8).

After comparing the AIC and BIC, we determined 3 classes for respondents from China and 3 for those from the United States. The segmented sizes of the 3 classes of respondents from China were 870 (70.2%), 270 (21.8%), and 100 (8.0%), respectively. The US respondents’ segmented sizes were 269 (21.7%), 139 (11.2%), and 832 (67.1%), respectively.

Figure 3 shows the heterogeneities of attribute importance of different classes of respondents from China and the United States, and Figure 4 shows preference weights stratified by group and class. Class 1 of respondents from China ranked reimbursement rate and claims as the first important attribute, while classes 2 and 3 thought that the importance of the types of clinics is the most critical factor affecting their medical preference. Meanwhile, classes 1 and 3 of respondents from China considered the waiting time the least important, while class 2 of respondents from China ranked the immediate COVID-19 test as the least important attribute.

For the classes of respondents from the United States, classes 1 and 2 ranked cost as the first important attribute, while class 3 attached the most importance to the reimbursement rate and claims. For these 3 classes of respondents from the United States, the least important attributes were waiting time, types of staff, and types of clinics, respectively. Cost had overwhelming importance in class 2 compared to the other 2 classes.

The study found that respondents from China and the United States had distinctly different preferences for attributes regarding the initial diagnosis of fever during the COVID-19 pandemic. Types of medical institutions were the most important factor for Chinese respondents, while those from the United States thought that cost was the most important factor when seeking medical services for fever during the pandemic. In addition, both populations highlighted the importance of the reimbursement rate. These heterogeneities and homogeneities may result from differences in the medical systems, health care services provided, COVID-19–mitigating strategies, and medical insurance systems between China and the United States.

DCEs can be used to understand people's underlying psychological situations, and the influencing factors and weights of choice preferences can be obtained through the hypothetical medical choice [46]. DCEs are also widely used in a series of aspects such as epidemic prevention, control, and supervision [47-49]. To the best of our knowledge, this is the first study to explore the preferences for health care services for the initial diagnosis of patients with fever during the COVID-19 pandemic. Our previous work found that respondents in China had a rather considerable basic knowledge of the detection methods of SARS-CoV-2 and the types of testing kits, even if they have no experience in contracting the virus or undergoing screening tests [50]. Nevertheless, considering that during the pandemic, everyone in China seeking health care services for fever, together with those accompanying them, would be screened by the SARS-CoV-2 nucleic acid test [51], the Chinese respondents placed the least importance on the necessity of immediate nucleic acid tests, while for the US respondents, COVID-19 tests accounted for 12.3% of the relative importance.

According to Caldow et al [52], patients prefer medical services provided by doctors, which is consistent with our results that doctors are preferred over nurses and paramedics for diagnosing fever [52]. During the pandemic, to relieve the consultation pressure of fever clinics in hospitals and prevent cross-infection, online fever clinics, an internet-based clinic system, were utilized in China, where a study enrolling more than 60,000 patients found that online fever clinics may efficiently ease patients' worry and clinicians can educate patients who are suspected of having COVID-19 to isolate and protect themselves [53]. Online and telephone consultation services have the characteristics of convenience and rapidity and can transcend distances to achieve preliminary medical services. In the case of future outbreaks, public health guidelines and policymaking may incorporate these 2 services into the first step of medical services to quickly divert different patients to curb the population contact transmission of infectious diseases. Although the respondents did not readily choose and prefer online clinics, Zhao et al [54] found that during the pandemic, many patients had trouble obtaining offline health care services and relied heavily on the internet for health information [54].

Chinese respondents regarded the medical institution type as the most important preference factor, especially public medical institutions, while respondents from the United States preferred private medical institutions, as shown in Table 4. The phenomenon may be attributable to the different medical and health service systems of China and the United States [55].

In China, health care providers include hospitals, primary health care institutions, and specialized public health institutions, where government-owned public hospitals and health institutions remain dominant in providing health care services, accounting for around 60% of all hospitals [56]. Nevertheless, the US health care system is more a combination of multiple systems operating individually but synergically, where nongovernment entities play a vital role in building the industry standard, fostering health accessibility, improving the quality of life, and controlling costs at various levels. With more than 6000 hospitals registered in the United States, only about 200 hospitals are owned by the federal government. The majority (more than 5000) of the hospitals are community hospitals, which include nonfederal hospitals, short-term general hospitals, and other special hospitals [57].

In this sense, it would be reasonable to assume that Chinese citizens would choose government-owned health institutions, while US citizens would prefer private-owned hospitals, which corresponds to the notion that government-owned hospitals predominate in China and community hospitals prevail in the United States. This fact consistently correlates with our results in Table 3, which show that Chinese respondents preferred fever clinics (designated public hospitals specializing in managing feverish patients during the pandemic), while the US respondents tended to choose private clinics.

According to LCM results, all respondents (from both China and the United States) were more willing to accept the diagnosis and treatment scheme with immediate nucleic acid testing, lower treatment costs, and higher reimbursement rates. In the LCM results, class 3 in China showed a preference for private medical institutions, and class 2 in the United States was more willing to receive telephone consultations than to travel to medical institutions; these findings differed from the overall performances of the Chinese and US groups.

Since the health system reform was enacted in 2009, more private-owned hospitals and health institutions have been established to provide the general population with equal access to basic health care [58-60]. According to a report by the China Statistical Information Center, from January to February 2021, the outpatient volume in China reached 960 million people, including 510 million in public hospitals and 90 million in private hospitals [61].

Despite its uneven distribution of medical resources in urban and rural areas, China is gradually beginning to promote internet hospitals. Telephone and online consultations have grown rapidly during the COVID-19 pandemic. Internet hospitals enable patients to consult doctors from large university-affiliated public hospitals for treatment through other internet hospitals [62]. During the pandemic, the application of internet hospitals in China, together with a remote drug delivery platform, has helped maintain constant health care services and provide for those in need, specifically those with chronic and mild diseases. This has been considered a potent tool to optimize medical resource distribution by relieving offline hospitals and catering to those in need without contracting the virus [63].

However, telephone and online consultations are still in their early stage of development. Because of the lack of standardized diagnosis and treatment standards, poor operation, and management issues, these 2 consultation schemes cannot wholly replace the conventional diagnostic process. Compared to traditional face-to-face medical schemes, the number of patients they serve is still small [64-66]. Nevertheless, the low selection rate of these 2 consultation modes may be attributable to the respondents having no idea of how these 2 types of consultation work and how they provide health care services [67].

COVID-19 is inherently not a disease that can solely be diagnosed via symptoms and signs, and the internet hospital is still in its exploration stage, where various issues remain unsolved. Thus, not being preferred does not necessarily mean that the internet hospital is suboptimal in guiding patients toward effective medical treatment.

Additionally, for respondents from the United States, cost was the most important factor affecting preference. This may be caused by the differences in the treatment costs and medical reimbursement systems in China and the United States.

In the United States, only around 30% of the population is covered by the public financing system, mainly via Medicare and Medicaid, and around 54% of the population receives private health insurance [68]. Nevertheless, one-sixth of Americans are uninsured, and high out-of-pocket expenditure still may put a heavy burden on some of those receiving insurance, hindering timely health care and medications.

Nevertheless, in China, the Information Office of the State Council of the People's Republic of China announced that the cost of COVID-19 in China would be covered by the national free treatment policy [69,70]. Therefore, it is reasonable that the respondents from China do not attach the maximum importance to the cost of diagnosis and treatment is the most important factor. This may be due to the national free treatment policy, which helps eliminate the burden of treatment costs on the public.

COVID-19 patients with severe and nonsevere complications were admitted to hospitals at an average cost of US $20,292 according to data provided by the Kaiser Family Foundation in the United States, and this is about 8.5 times the average cost in China. At the same time, although there is a medical insurance system in the United States, even after Medicare reimbursement, the average out-of-pocket expenses of patients may exceed US $1300 [71]. Experts have recommended that new federal legislation should be established to expend federal funds on emergency responses, hiring and training of personnel, and distribution of diagnostic tests, therapeutic approaches, and vaccines at different levels and, most importantly, to expand the coverage rate of medical insurance for diagnosing, treating, and following up patients with COVID-19 [72]. During COVID-19, a wide range of previously unavailable telehealth services were covered by Medicare and Medicaid, allowed by the Section 1135 waiver in the United States [73]. Moreover, the havoc COVID-19 caused on the economy resulted in a 15% unemployment rate in May 2020 [74], which increased the rate of the uninsured and enrollment in Medicaid, and hence some have addressed the need to put forward “Medicare for All” [75]. Nevertheless, scholars have addressed that ensuring effective government action with sensible private sector regulation may be a preferable option over turning to insurance to pay for COVID-19–related medical costs [76]. Generally, we believe that both expanding the coverage of insurance to those uninsured or with low insurance for COVID-19–related health care costs and ensuring the federal mandate for free access to COVID-19 testing and treatment can effectively motivate those potentially infected to undergo testing and proper treatments.

Identification, tracing, and isolation of those infected are vital for containing the community spread of COVID-19. Nevertheless, if those infected remain untested and no universal nucleic test programs are conducted, then the infected citizens may still be constantly spreading the virus in the community. Therefore, identifying potential COVID-19–infected patients in the community and encouraging them to undergo testing and quarantine is important for curbing COVID-19 spread in the community.

In China, large-scale community nucleic acid testing can be implemented to identify potentially infected people due to the state's attitude toward early diagnosis, early isolation, and early treatment of COVID-19. The willingness of the public to participate in screening tests depends on their awareness of the risks and benefits. The governmental entities stipulate that residents in controlled areas must participate in community nucleic acid testing and those who do not participate without good cause may suffer legal consequences and difficulties in daily life [51]. This is in line with the general policy of “dynamic zero COVID-19 strategies” for preventing and controlling the former COVID-19 pandemic in China [77].

For the United States and other noncentralized countries where large-scale community-wide mass screening tests are difficult to implement, it is difficult to identify those infected in the community unless they undergo screening tests voluntarily. However, some people still avoid nucleic acid testing as they are afraid of testing positive for the disease and other factors [78]. Although the Centers for Disease Control and Prevention (CDC) and other institutions have issued a series of guidelines on coping with COVID-19 [79], according to Park et al [80], the psychological pressure of contracting COVID-19 may reduce the rate of public compliance with official health guidelines. Therefore, it is recommended that the public be encouraged to participate in screening tests, either nucleic acid tests or antigen tests, in a number of ways, including broadcasting public messages by medical and health experts, providing information on social media, and distributing small gifts to those who participate in testing, to make every resident aware of the importance and obligation of nucleic acid screening [81,82]. Improving the participation rate in those screening, testing, and detecting of those infected may help control the spread of COVID-19. We advocate the use of publicity campaigns in the media, the spread of rigorous scientific information, the promotion of culturally sensitive psychological counseling, and other related services to account for different needs and to encourage the public to be willing to participate in screening tests. Regarding the stigma and discrimination related to being diagnosed as COVID-19 positive, it is necessary to provide social support to relieve the potential stigma and social unrest. To make this possible, multidisciplinary teams comprising experts from clinics, social sciences, government entities, communication, and the media are needed [83].

There are limitations of our research. First, the nature of this cross-sectional study inherently led to reporting bias, information bias, and confounding bias. In the study, we used a closed-end, self-administered questionnaire to prevent missed data and used online an panel platform (MTurk) to prevent selection bias, as a previous study proved the census-level quality of survey data collected via MTurk [61,84]. Nevertheless, selection bias may still exist. PSM was used to control the confounding effects when directly comparing the 2 cohorts of respondents. However, as various factors may underlie the respondents' decision-making in the 2 distinctly different countries, the scale differences may not be completely accounted for, and the results should be interpreted with caution. Moreover, as the DCE asked participants to make choices between hypothetical scenarios, which may not reflect real-world situations and hypothetical bias may exist, and we did not investigate external validity in the study, the results of the study should be interpreted cautiously. Our questionnaire involves the treatment modes of telephone and online consultations, which some respondents may not have experienced and may have led to selection bias. Moreover, we did not include questions on past experience with online/telephone consultation as well as previous experience with COVID-19 testing or treatment, so how such underlying factors may influence preferences could not be distinguished and need future exploration. In addition, we did not include the investigation of how different ethnicities and residence locations (urban and rural) may affect the respondents' preferences, which can be further explored in a future study. The significant interactions between cost and reimbursement rate render interpreting these 2 attributes difficult, so the WTP should be interpreted cautiously.

Improvements in the availability of COVID-19 testing, medical professional skills, and designated health care facilities may help boost potential health care seeking during COVID-19 and prevent unrecognized community spread of SARS-CoV-2 in China and the United States. Moreover, to better prevent future waves of pandemics, identify undiagnosed patients, and encourage them to seek health care services to curb the pandemic, it is suggested that the hierarchical diagnosis and treatment system be improved in China and that the United States focus on reducing medical costs and raising the reimbursement rate of medical insurance. Second, online and telephone consultations may serve as patients' primary medical services, which may triage suspected and nonsuspected patients of infectious diseases, reducing the possible cross-infection during the pandemic.