This was a cross-sectional study. Eligible participants were Hong Kong residents aged ≥18 years whose households had received an antiepidemic service bag distributed by the government. A bilingual (Chinese and English versions), open, self-administered, web-based, and population-based survey was designed, covering demographics (age, sex, and residing district); COVID-19 vaccination history (type, date, and dose of the last vaccination received); COVID-19 infection history; signs and symptoms; RAT performed during the 3 campaign days with result; post–positive result actions (reporting to the government’s web-based declaration system, seeking treatment, and isolation); and the number of coliving individuals. Participants could opt to report their coliving individuals’ age, sex, RAT performed and result during the campaign period, and their relationships, up to 5 persons. Due to the simplistic nature of this study, the items were not randomized. Adaptive questioning was used on the same page to display questions relating to COVID-19 vaccination history, details about RAT history during the campaign period, and particulars about the coliving individuals. For those who did not undergo an RAT, they were asked to select at least one of the following reasons for not doing so: recent diagnosis, recently tested, regular testing as part of work requirement, no appropriate time and environment, avoiding isolation if tested positive, avoiding compulsory declaration if tested positive, avoiding sampling discomfort, no confidence to self-test, no symptoms, not worried about getting infected, not believing the campaign was an appropriate antiepidemic measure, and others. There were at most 16 questions for each participant, and a maximum of 5 questions for each coliving individual. No personal identifiers were collected. The survey was tested and refined before fielding. After completing the survey, participants were invited to share their location using the HTML5 Geolocation Application Programming Interface if they were at home or in the workplace. Coordinates outside the territory of Hong Kong were removed.

Web and newspapers advertisements were placed to recruit Hong Kong residents to join as participants. All responses were collected through the bespoke website designed for this web-based survey. The completion of the survey by participants was voluntary. Completeness checking was done using JavaScript before submission. Incomplete responses were not collected. No data were excluded due to atypical time stamps because of the simplistic nature of the survey. An anonymous session identifier was set in the cookies, and the IP addresses of participants were collected. Duplicate entries of the same session identifier were removed.

The study data collected were anonymous. Web-based informed consents were obtained before participants filled out the questionnaire. No incentives were offered upon the completion of the study. This study was approved by the Survey and Behavioural Research Ethics Committee of The Chinese University of Hong Kong (SBRE-21-0685). The conduct of the study was in compliance with the Declaration of Helsinki.

Determinants of undergoing an RAT during the campaign period were identified using univariable and multivariable logistic regression models. Testing rate and positive rate were determined by aggregating both study participants and their coliving participants. The Wilson score method was used to calculate the 95% CI of age-specific testing rate [15]. The testing rate and positive rate were age-adjusted by groups defined by 5-year windows, except those aged ≥85 years were grouped together, using the provisional figures published by the Census and Statistics Department from the 2021 Population Census [16]. The 95% CI of the directly standardized testing rate was computed using the Byar method [17]. The number of prevalent infections in the territory during the campaign period was estimated by multiplying the population size by the age-adjusted positive rate with the 95% CI. Maps were drawn with the QGIS platform (QGIS Development Team) using 2019 District Council Constituency Areas as the spatial unit. There was a total of 452 District Council Constituency Areas, each of which normally containing about 16,599 residents [18]. Secondary outcomes, including determinants of prior COVID-19 diagnosis and reasons for not getting tested, were evaluated using chi-square test and Mann-Whitney U test for categorical and continuous predictors, respectively. Multivariable logistic regression analysis was performed by including variables with a P value of <.05 in the univariable analyses. All statistical analyses were conducted in R statistical software (R Foundation for Statistical Computing). Reporting in this manuscript follows the Checklist for Reporting Results of Internet E-Surveys [19].

Participation is key to a voluntary mass-screening campaign for SARS-CoV-2 infection, the coverage of which would not be known in the absence of an accompanying surveillance mechanism. An evaluation of the Hong Kong campaign was made possible through a separately conducted ensuing population-based web-based survey. The geographical diversity of the participants in this study supported that a diverse sample population had been recruited. The proportion of participants reporting previous COVID-19 diagnosis corresponded well with that recorded (about 16%) in the published government statistics [6]. Although participation in mass screening was voluntary, three-quarters of respondents in our population-based survey had undergone the testing at least once during the 3-day period, confirming the feasibility of implementing self-RAT screening as a complementary means of public health surveillance.

Our survey results showed that those participating in the RAT campaign were more likely to be older, female, and vaccinated against COVID-19. These characteristics were similar to participants in a previous PCR-based voluntary screening campaign in 2020, which showed that participation was associated with perceived efficacy of the campaign in controlling the epidemic, perceived susceptibility to COVID-19, and their trust of the government [20]. About one-third of nontesters did not believe the RAT campaign could control the epidemic. Such a low perceived efficacy might have prevented some citizens from participating in the campaign. The mandatory reporting of positive results had lowered the willingness of a certain proportion of people to participate. The higher odds of being vaccinated against COVID-19 and receiving more doses among testers could be a result of one’s perceived susceptibility. Trust in the government that the policies are efficacious in controlling the epidemic could contribute to their engagement in vaccination and screening [20]. On the contrary, distrust in the government could also contribute to self-regulation to prevent infection and protect one’s own interests, leading to passive compliance with antipandemic measures [21]. Separately, in the recent epidemic waves in Hong Kong, children were more likely to be asymptomatic and be infected through household transmissions rather than exogenous acquisition from schools [22], which might have prompted household members living with children to be tested to prevent transmission to the younger members if they tested positive.

Differentiating the epidemic situation between the time of the 2020 PCR-based screening campaign and the 2022 RAT-based one, a greater proportion of population had already been infected prior to the latter campaign, which may have affected the participation rate as some residents may not consider it necessary to undergo testing. The presence of symptoms was one of the indicators of SARS-CoV-2 infection, which prompts one to have testing performed [23]. This testing process was educated publicly to encourage people to get tested when they are presented with symptoms; on the other hand, people without symptoms may not be interested in taking the RAT. Although the figures in the previous study cannot be directly compared with findings from this study due to methodological differences, it is worth noting that the participation rate of the previous PCR-based screening campaign was only 47%, and one-quarter of nonparticipants noted they did not have time for screening [20]. With three-quarters of respondents having participated in the self-RAT screening and just 1% concerned about spending time on it, the convenience and acceptability of an RAT-based voluntary screening campaign over a PCR-based one was highlighted.

As the official number of locally reported cases was just 5819 during the campaign period [6], the reported cases might have accounted for about 7.65% (95% CI 6.47%-9.14%) of all infections given the 1% positive rate; the rest being not reported despite statutory requirement or not detected because of either the low sensitivity in picking up early infections or that no screening had been performed. As only 62% participants who tested positive declared their results to the government, the number of reported cases could be underestimated. Since a higher proportion self-isolated after receiving a positive result, they were willing to take precaution to prevent onward transmissions in the community, although they did not declare the results to the authority. The high mobility of the younger working population and older adults were not reduced much during the epidemic, which predisposed them to the risk of infection [24]. Evidently, the Omicron wave has rapidly receded after over a million people have reportedly been diagnosed in the preceding 2 months. The size of the residual burden has, however, remained high and could easily be underestimated if statutory reporting statistics alone is used for epidemiologic surveillance. The high vaccination uptake rate and its protective effect might have played a role in minimizing the population risk. RAT mass screening has contributed to the assessment of the epidemiologic situation in a receding Omicron wave in Hong Kong.

Our population-based survey carried some limitations, notably self-selection bias with older and health-conscious adults and those testing positive being attracted to join the survey. The uptake rate may, therefore, be overestimated. In the analysis, we have performed age-adjustment to better reflect the situation in the population. The generalization of the results to the entire Hong Kong population should be cautioned due to the use of nonprobability sampling. Similar to other population-level surveys, recall and social desirability biases were inevitable. The survey was rolled out a week after the campaign to minimize recall bias. We assured participants of the anonymous nature of the survey to ensure the accuracy of the test result reported and compliance. By including proxy participants in the household, duplicate entries from the same household may have happened. We removed entries with the same session identifier to minimize duplicate records. As multiple brands of RAT were distributed and used with different sensitivity and specificity levels, their performance was unlikely to be perfect, so even if all participants were sampled and interpreted and reported the results correctly, the true infection status of all individuals might not have been determined definitively. It should also be noted the positive predictive value could be low in places where the prevalence is low [25]. Previous studies have, however, showed that RAT had a low false-positive rate [26] but an adequate sensitivity to identify asymptomatic and high–viral load cases [27]. The 1.03% positive rate found in this study was similar to the estimated daily point-prevalence on the last day of the campaign (0.76%, 95% CI 0.32-1.56%) [28], demonstrating the reliability of the results from this population-based survey. In conjunction with its low cost, RAT is well positioned to be used should mass screening be adopted as a cost-effective intervention in the public health control of COVID-19 [29]. As a perfect reporting rate of positive results is unlikely to be achieved, an accompanying survey would be needed and could be a feasible and appropriate means to estimate the actual prevalence in the community.

In a receding Omicron wave in 2022, a large proportion of residents in Hong Kong self-performed an RAT during the “voluntary universal rapid testing” campaign promoted by the government. RAT could be a useful adjunct not just for clinical diagnosis but also as a tool for public health surveillance and self-detection of infection. Accompanied with an information system, isolation facilities, and supporting services, voluntary mass RAT screening could support the estimation of the residual population burden and for supplementing risk assessment.