Previous studies tried to optimize screening strategies according to infection risk levels indicated by infection probability, contact with an infected person, probability of exposure, or presence of COVID-19 symptoms (Table 1). Du et al [20] simulated COVID-19 testing in populations with different severities of clinical symptoms and age groups and compared infection rates and false-positive rates among groups with different strategies. They found that in settings of high infection rate and limited testing capacity, a greater proportion of tests should be allocated to symptomatic individuals. Conversely, in a surveillance scenario characterized by a low infection rate and sufficient testing capacity, the optimal strategy, which involved directing a higher allocation of tests to people aged >50 years, only required 51.5% of available tests [20]. Likewise, the research by Han et al [21] supported that symptomatic testing at health care facilities was more beneficial than asymptomatic testing in the community, until most symptomatic individuals had been tested. It was also suggested that conducting additional tests to screen for asymptomatic infections among household members yielded the greatest benefit after fulfilling the demand for symptomatic testing [21]. Berestizshevsky et al [22] compared total morbidity, testing efficiency, and infection probability under screening strategies such as no testing and no isolation, symptom monitoring combined with random sampling testing, and symptom monitoring combined with “greedy” testing. Testing and quarantine among symptomatic populations using optimized sampling, which is based on community graphs and population risk factors, results in a 30% to 50% reduction in overall prevalence [22]. Kumar et al [23] compared the effectiveness and cost of outbreak control by weekly reverse transcription-PCR (RT-PCR) testing in 100% antigen-negative vaccinated individuals in a high-income country (the United States) and in a low-income country (India). They suggested that in regions with resource-limited vaccination strategies, high-frequency testing is still needed to minimize subsequent outbreaks [23].

Table 1 also shows the results of optimizing screening strategies in different regions such as medium-risk or high-risk regions and domestic or foreign regions. Zhou et al [24] compared the total number of infections and daily nucleic acid test loads among the screening strategies and found that the optimal strategy was to test people from medium- and high-risk areas using nucleic acid tests before they traveled. Wong et al [25] assessed infections based on attendance at an event by applying different strategies: no testing and quarantine in all areas, quarantine of attendees from foreign areas, testing of attendees from mainland China, and testing of all attendees at the event. They found that the strategies of quarantining the attendees from foreign areas and testing all (foreign or local) attendees were effective in controlling the number of infections, and they estimated that the total number of new infections was only 1% higher than the current local prevalence [25]. Baik et al [26] simulated the effectiveness of outbreak control in regions with limited resources, such as low- and middle-income countries. They evaluated 3 screening strategies: using PCR to test symptomatic patients in outpatient settings, community-based screening with rapid antigen tests (RATs), and symptom-driven outpatient diagnostic testing using RATs. This showed that RATs would reduce transmission most efficiently when used to test symptomatic individuals in outpatient settings, and to avoid large numbers of unnecessary isolations, mass testing with lateral flow tests (LFTs) should be considered as a screening tool [26].

COVID-19 screening strategies were assessed and optimized for some specific student and occupational populations (Table 1). Sandmann et al [27] compared the number of infections, number of tests, and duration of isolation between individuals with and without COVID-19 symptoms living in worker dormitories. Testing all the workers was associated with a reduced transmission of approximately 67 individuals per 1000 tests. However, screening workers with COVID-19-like symptoms in isolation only was associated with a higher risk of transmission in the workplace compared with the strategy of screening all workers [27]. Zipfel et al [28] simulated transmission in health care workers using 2 strategies: testing all when positive cases were detected and testing all periodically at predetermined intervals from 1 day to 7 days. The study showed that 38% of hospital-based transmission could be prevented if all staff were tested within 1 day when a positive case occurred, whereas 30% to 78% of transmission could be prevented if daily testing was performed [28].

Van Pelt et al [29] analyzed the number of RT-PCR tests required to identify each true-positive case and the true-positive rate in a student population using strategies such as symptom screening only, nucleic acid test screening of symptomatic students, and nucleic acid test screening of all students. Conducting RT-PCR testing for all students and retesting those with initially negative results can effectively identify cases with a correct rate of 86.9% [29]. Leng et al [30] further classified the student population by age and simulated the implementation of daily mass screening or screening after the occurrence of positive cases. The study found that mass antigen screening among students significantly reduced the likelihood of not attending class but often required a large number of tests [30]. Similarly, Colosi et al [31] analyzed strategies such as routine testing, symptom-based testing, screening, and quarantine when a positive case was detected and found that weekly testing of 75% of unvaccinated students would reduce the number of cases by 34% in primary schools and 36% in secondary schools.

To address the risk of outbreak caused by population movement, researchers compared the timing of screening of traveling people (eg, before, during, and after travel; Table 1). Chowell et al [32] simulated the impact of testing at different times before and after boarding a cruise ship on the cumulative number of infected cases. The study found that testing before boarding, daily testing on board, and maintaining social distancing significantly reduced the possibility of onboard transmission [32]. Johansson et al [33] found that PCR testing on the day of departure and isolation at the destination reduced the risk of transmission. Testing on the day of departure reduced the risk of transmission when traveling by 44% to 72% [33]. Steyn et al [34] assessed the transmission potential of SARS-CoV-2 and the number of infections by simulating PCR or LFT screening at different times, such as the day before departure, the day after arrival at the destination, and the fourth and fifth days after arrival. It was found that the combination of testing and home isolation could reduce the risk of community outbreaks to approximately 0.01, and using daily LFTs or a combination of LFTs and PCR testing could reduce the risk to levels comparable with or lower than those using PCR testing alone [34]. Kiang et al [35] evaluated the cumulative number of days of infection and the number of infections when travelers used PCR or antigen testing at timings such as 3 days before departure, the day of departure, and 5 days after arrival at the destination. The results indicated that nucleic acid testing 3 days before departure reduced the risk of infection during the travel, and the cumulative number of infection days was reduced from 8357 to 5401 days [35]. Jen et al [36] compared the morbidity, missed tests, and proportion of asymptomatic and presymptomatic individuals in travelers with different strategies such as PCR testing on arrival and quarantine for 5 days and quarantine for 14 days without testing. It was found that >82% of the cases would progress from the presymptomatic phase to the symptomatic phase during the 5-day quarantine period, and the quarantine time with 2 PCR tests depended on the risk, testing and quarantine strategy, and vaccination status of the country of departure [36]. On the basis of previous screening strategies for travelers, Lee et al [37] analyzed transmission potential, infection rates, and subsequent infections for strategies such as testing before traveling, after arrival, or at the end of quarantine and 14 days of quarantine without testing, with a consideration of vaccination factors. It was found that at an incidence rate of 0.4 and a time-dependent reproduction number of 16, testing with a sensitivity of ≥98% and specificity of ≥97% both before traveling and on arrival ensured lower expected transmission in vaccinated than unvaccinated individuals with a quarantine of 14 days [37]. For entry-exit pandemic prevention, Dickens et al [38] analyzed the number of unidentified infected persons using the strategies of isolation only; predetection combining entry testing and isolation; and predetection combining entry testing, isolation, and daily testing. The results showed that the risk of transmission was greatly reduced by adopting predetection, which combined entry testing and isolation. During the isolation period, if an RAT was performed every 3 days, only 3% of the infected individuals were unidentified at 7 days and 0.7% at 14 days [38].

When an outbreak occurs, the appropriate timing of screening facilitates the identification of infected individuals and helps control the spread of the disease. Several studies have explored the impact of different screening timings on the spread of the pandemic (Table 1). Quilty et al [39] compared the impact of daily antigen testing of close contacts over 1, 3, 5, 7, 10, or 14 days on the spread of the epidemic. It showed that quarantining for 7 days with an antigen testing on the last day or daily antigen testing for 5 consecutive days without quarantine was effective in reducing the potential for secondary cases [39]. Wells et al [40] evaluated the impact of screening timing, such as at the start or end of isolation and during the period of isolation, on the risk of continued transmission after isolation. It was suggested that PCR testing at the start and end of isolation could reduce the risk of continuous transmission and shorten the isolation period from 14 days to 7 days. However, testing only at the start of isolation had no notable effect on reducing the risk of transmission and shortening the isolation period [40]. Foncea et al [41] simulated the screening of close contacts at different timing, such as days 1, 2, and 3 after exposure to an infector, and compared the expected days of infection during outbreaks. This suggested that antigen testing should be performed on days 1 and 3 for epidemic prevention and control. Two tests were sufficient to effectively prevent infection, and the effectiveness was equivalent to a 14-day isolation period when personnel compliance was 80% to 90%. If an additional test (PCR or antigen) was performed, it was equivalent to the 14-day isolation period effectiveness when personnel compliance was 90% to 100% [41]. Peng et al [42] conducted a similar study to assess the risk of spread after the end of the quarantine period for PCR testing at different timings, such as the days 1, 2, 4, 5 and 6 during the quarantine period. The results showed that PCR or antigen testing at different timings reduced the quarantine period to different degrees without increasing the risk of transmission. Combining testing with shorter quarantine periods is more cost-effective in terms of both time and expenses compared with a 14-day quarantine. For instance, using 3 highly sensitive RT-PCR tests along with a 6-day quarantine yielded a similar risk of transmission as the traditional 14-day quarantine. [42].

Given the huge demand for sampling and testing, it is important to optimize the screening procedure, and the pooling strategy (ie, collecting multiple samples in a pool for testing) has been frequently used. For example, the National Health Commission of China and the Centers for Disease Control and Prevention of America have issued guidelines for pooled sample tests [47,48]. Three issues related to the pooling strategy have been studied: scenarios for adopting the pooling strategy, pool size, and test procedure.

In general, the decision to adopt a pooling strategy is determined by comparing the average number of pool tests per person with a baseline number of 1 (Multimedia Appendix 3 [49-83]). A study demonstrated that when the prevalence exceeded 0.1, the average number of tests per person was >1 for pooled testing with a pool size of 32; such pooled testing was no better than the individual test [49]. The average number of tests per person varied with the pool size. Choosing an optimal pool size can minimize the average number of tests per person. In this case, when the prevalence was <0.07, the pooling strategy could save the need to perform more than half of the tests. When the prevalence was close to or >0.3, the number of pooled tests was close to or exceeded that of the individual test [50]. A prevalence of 0.3 may be considered as the threshold for performing pooled testing.

The optimal pool size can be calculated based on expected positive rate and detection accuracy with the objective of minimizing the number of tests, and the number of tests decreases with decreasing prevalence and increasing pool size. If the accuracy was 100% and the prevalence was 0.001, 0.005, or 0.01, the optimal pool sizes were estimated to be 32, 15, and 10, respectively [51]. There is a certain upper threshold value for the pool size that is limited by testing accuracy (Multimedia Appendix 3).

To estimate the optimal pool size, the prevalence (positivity rate or infection rate) must be assumed. It is also assumed that all individuals are independent of each other, and that the probability of infection is uniform. However, in practice, the prevalence remains unknown until the test results are available (Multimedia Appendix 3). Pikovski and Bentele [52] considered the prevalence to be a random variable uniformly distributed between the expected maximum and minimum values substituted in the calculation of the optimal pool size. An optimal pool size of 4, 3, or 5 was acceptable when the prevalence was uniformly distributed between 0 and 0.3 [52]. In addition, there is heterogeneity and correlation in the probability of infection among people. Fewer tests are needed when individuals in the same pool for testing are homogeneous in terms of age, sex, and other risk characteristics [53-55]. Libin et al [56] considered that combining pools with several families for testing was more conducive to home isolation. The propagation dynamics simulation found that a family-based pool size of 32 and testing volume of 50,000 per day could achieve the weekly testing of the entire population in Belgium [56]. A larger optimal pool size is required considering the correlation of individuals in the pool [57], or even the social graph [58], in which an edge represents frequent social contacts between 2 persons. Furthermore, Augenblick et al [59] showed that if the pool size could be adjusted to be optimal with the infection rate at any time, screening with a high testing frequency could quickly reduce the infection rate. The final number of tests may decrease despite the high testing frequency owing to the increasing optimal pool size, that is, “frequency gain” [59].

Test accuracy, including screening sensitivity and specificity, also affects the optimal pool size and the upper limit of the pool size. Bish et al [60] found that the optimal pool size would modestly increase when the sensitivity of the pooled sample test decreased. In the PCR test, the sensitivity would decrease and the specificity would increase in a pooled sample test owing to the dilution effect [61], which needs to be considered in the calculation of the optimal pool size [62]. The maximum pool size recommended in previous studies varies from 8 to 30 [63-65]. The sensitivity of the individual test and the influence of the dilution effect of pooling on sensitivity are related to the specific techniques of sampling and testing (such as sampling tools, sample processing reagents, detection instruments, and standardization of operation); therefore, test accuracy is an important determinant for the selection of pool size.

For the original Dorfman pooling approach, each individual in a positive pool is tested separately. If a pooled sample is negative, then all individuals in the pool are regarded as negative. Several suggestions were made to optimize the pooling approach, which are summarized in Table 2.

First, sequential pooling may be used. A positive pool is divided into several subpools, and the samples in the positive subpools are tested individually [55,66]. Binary pooling divides people to be screened into 2 pools, and the positive pool is divided and pooled again until all positive individuals are found [67]. In the nested pooling strategy, the samples in a positive pool are divided into smaller pools with an optimal number of stages, and the optimal pool size of each stage is calculated according to various measures such as the predicted number of positives and time limit [68,69]. Ng et al [70] conducted simulations of a household-based sequential pooling approach to optimize a universal testing scheme in Hong Kong. They showed that the household-based sequential pooling approach could rapidly screen people in high-risk groups for COVID-19 infections and quarantine those who tested positive [70]. Although these approaches reduce the number of tests required, the operability of such strategies for time-critical epidemic control should be carefully assessed.

Second, repeated testing of the same pool of samples may be conducted to reduce false negatives of the pooled test [71]. Litvak et al [72] conducted a second pooled test after reordering and recombining the samples in the negative pools. For the sequential pooled test, some researchers allocated a part of the samples to 2 subpools to improve accuracy [73].

Third, a copy-link optimization strategy may be used to accurately link the results of the pooled test to the individuals in the pool. The primary “copy-link” strategy is matrix pooling [84]. Samples are arranged in the form of a matrix, with each row and column forming a pool, and each sample is tested once in the row pool and once in the column pool. Research on matrix pooling for COVID-19 has only mathematically simulated the number of tests and the accuracy of the pooled sample test. The Dorfman pool test may be more economical when the prevalence is extremely low, whereas matrix pooling may be more economical when the prevalence is relatively high [74]. Žilinskas et al [75] broadened the concept of matrices by dividing each sample into 2 pools to create as many links as possible between pools. Zhou and Zhou [76] applied the copy-link strategy in designing the Pentagram minipool test. Mutesa et al [77] expanded the 2D matrix to a 3D or multidimensional hypercube, where the number of copies of each sample was split into different planar slices of the hypercube. The subsamples on 1 planar slice of the hypercube were tested in a pool together. Investigators have demonstrated the feasibility of this “hypercube testing strategy” in the laboratory, and field trials are underway in Rwanda and South Africa [77]. Wu et al [78] improved the current hypercube testing strategy by calculating the prevalence, edge, and dimension because every edge had a best performance range, and hypercube pooling with edge=3 may not be the optimal strategy in different outbreaks. Daon et al [79] used a Bayesian model to determine the best combination of pool size, detection steps, repeat detection, and split sample detection to maximize the mutual information between the infection status and testing results. However, this is limited to a simulation analysis.