key: cord-0300538-lfebo83u
authors: Rabil, M. J.; Tunc, S.; Bish, D. R.; Bish, E. K.
title: Screening for safe opening of universities under Omicron and Delta variants of COVID-19: When less is more
date: 2022-05-07
journal: nan
DOI: 10.1101/2022.05.04.22274667
sha: d488d2685ee8e2150191435271ee50643036bf91
doc_id: 300538
cord_uid: lfebo83u

As new COVID-19 variants emerge, and disease and population characteristics change, screening strategies may also need to change. We develop screening guidelines for the safe opening of college campuses, considering COVID-19 infections/hospitalizations/deaths; peak daily hospitalizations; and the tests required. Our compartmental model simulates disease spread on a college campus under co-circulating variants with different disease dynamics, considering: (i) the heterogeneity in disease transmission and outcomes for faculty/staff and students based on vaccination status and level of natural immunity; and (ii) variant- and dose-dependent vaccine efficacy. Using the Spring 2022 academic semester as a case study, we study various routine screening strategies, and find that screening the faculty/staff less frequently than the students, and/or the boosted and vaccinated less frequently than the unvaccinated, may avert a higher number of infections per test, compared to universal screening of the entire population at a common frequency. We also discuss key policy issues, including the need to revisit the mitigation objective over time, effective strategies that are informed by booster coverage, and if and when screening alone can compensate for low booster coverage.

. With regards to variant prevalences, we discuss two important cases that represent the pandemic progression: 83 ω O = 50%, where both Omicron and Delta variants are in circulation in similar rates, which may represent the U.S. during late 84

December 2021 [34] ; and ω O = 95%, where Omicron takes over as the predominant variant, which was the case in the U.S. 85 starting in early January 2022 [35] . 86 87 We study the impact of pandemic progression (i.e., the transition of the dominant variant from Delta to Omicron) on the 88 performance of screening strategies, under 82% total vaccination coverage (with 64% of the population boosted and 18% 89 vaccinated), and 18% unvaccinated. When both Delta and Omicron variants are in circulation at similar rates (ω O = 50%, see 90 Supplementary Table 4 ), universal screening every 1/2/14 days results in 1,012/4,058/15,458 total infections, and 31/31/142 91 peak daily infections, respectively. If the boosted individuals are excluded from screening, the 1/2/14-day screening strategies 92 yield 7,967/12,168/16,741 total infections, and 45/81/159 peak daily infections, respectively. The strategy that averts the highest 93 number of infections per test is screening the unvaccinated only every 14 days, with 31.6 infections averted per 1,000 tests over 94 no screening (see Fig. 1(a) ); furthermore, this strategy also provides the highest reduction in peak infections per test, reducing 95 the peak by 0.9 infections per 1,000 tests over no screening (see Fig. 2(a) ). 96 Under ω O = 50%, the strategy that averts the highest number of infections per test depends on the booster coverage. When 97 64% of the population is boosted and 18% is vaccinated, if the screening frequency of the unvaccinated is set to every 1/2/3/7/14 98 days, the strategy that averts the highest number of infections per test is screening the unvaccinated and vaccinated every 1/2 99 days, and screening the unvaccinated only every 3/7/14 days, averting 23.2/27.5/29/31/31.6 infections per 1,000 tests (see 100 Fig. 1(a) ), respectively, whereas the strategy that averts the highest number of infections per test with 38% boosted and 44% 101 vaccinated, is screening the vaccinated and unvaccinated every 1/2/3/7 days and screening the unvaccinated only every 14 days, 102 averting 17.6/20.19/20.23/19.3/19.5 infections per 1,000 tests (see Fig. 1 (b)), respectively. 103 When Omicron is the predominant variant (ω O = 95%, see Supplementary Table 5) , universal screening every 1/2/14 104 days results in 8,568/17,327/22,512 total infections, and 61/222/465 peak daily infections, respectively. When screening 105 excludes boosted individuals, 1/2/14-day screening yields 21,578/22,111/22,820 total infections, and 312/380/487 peak daily 106 infections, respectively. Under this scenario, the strategy that averts the highest number of infections per test is the daily 107 universal screening, averting 11.9 infections per 1,000 tests (see Fig. 1(c) ), whereas the strategy that provides the highest 108 reduction in peak infections per test is screening the unvaccinated every 3 days, which reduces the peak by 1.09 infections per 109 1,000 tests (see Fig. 2(c) ). 110 Under ω O = 95%, if the screening frequency of the unvaccinated is every 1/2/3/7/14 days, the strategy that averts the 111 highest number of infections per test is 1/2/3/7/14 day universal screening for both the 64% boosted and 18% vaccinated, or the 112 38% boosted and 44% vaccinated cases, averting 11.4/9.9/7.2/6 and 7.4/5.8/3.8/3 infections per 1,000 tests (see Fig. 1 (c) and 113 Fig. 1(d) ), respectively. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) boosted. The strategy that averts the highest number of infections per test is daily universal screening, averting 9.3 infections 121 per 1,000 tests, see Fig. 1(d) , whereas the strategy that provides the highest reduction in peak infections per test is screening the 122 unvaccinated every 3 days, which reduces the peak by 1 infection per 1,000 tests (see Fig. 2(d) ).

The boosted proportion further impacts hospitalizations (see Supplementary Table 5 In this section, we study the impact of further customizing the screening frequencies. We first consider that the screening 133 frequency can be customized for each vaccination category included in the screening population, under 82% vaccination 134 coverage (64% boosted, 18% vaccinated). When ω O = 50%, the strategy that averts the highest number of infections per test is 135 screening the unvaccinated every 14 days (see Fig. 3 (a)), leading to 31.6 infections averted per 1,000 tests (see Supplementary 136 Table 9 ). The same strategy also provides the highest reduction in peak infections per test (see Fig. 4 (a)), i.e., a reduction 137 of 0.9 infections per 1,000 tests. When ω O = 95%, the strategy that averts the highest number of infections per test is daily 138 screening of the unvaccinated and vaccinated, and 2-day screening of the boosted (see Fig. 3 (b)), resulting in 12.4 infections 139 averted per 1,000 tests, with a 4% improvement compared to the most effective universal screening strategy (i.e., the strategy 140 that screens the entire population with the same screening frequency), see Fig. 3 (b) and Supplementary Table 9 ; whereas the 141 strategy that provides the highest reduction in peak infections per test is screening the unvaccinated only every 3 days (see 142 Fig. 4(b) ), reducing the peak by 1.09 infections per 1,000 tests. 143 We next consider fully customized screening, where the screening frequency can be customized for each group (faculty 144 versus students) and each vaccination category included in the screening population, under 82% vaccination coverage (64% 145 boosted, 18% vaccinated). When ω O = 50%, the strategy that averts the highest number of infections per test is screening 146 only the unvaccinated students every 14 days (see Fig. 3 (c)), leading to 32.5 infections averted per 1,000 tests, with a 2.8% 147

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; improvement compared to the most effective strategy that customizes the screening population only. The same strategy also 148 provides the highest reduction in peak infections per test (see Fig. 4 (c)), with a reduction of 0.93 infections per 1,000 tests, 149 which provides a 3.3% improvement compared to the most effective strategy that customizes the screening population only. 150

When ω O = 95%, the strategy that averts the highest number of infections per test is the daily screening of the unvaccinated and 151 vaccinated students, 2-day screening of the boosted students, and the unvaccinated and vaccinated faculty, and 3-day screening 152 of the boosted faculty, resulting in 12.5 infections averted per 1,000 tests, with a 5% improvement compared to the most 153 effective universal screening strategy (see Fig. 3(d) ); whereas the strategy that provides the highest reduction in peak infections 154 per test is screening only the unvaccinated students every 3 days (see Fig. 4 (d)), reducing the peak by 1.13 infections per 1,000 155 tests and providing a 3.7% improvement compared to the most effective strategy that customizes the screening population only. Next, we study the impact of vaccine effectiveness, which is imperfect, decaying over time, and variant-dependent (see 169   Table 3 ), over several outcomes. When ω O = 95%, even when the entire population is vaccinated (but not boosted) and is 170 screened under perfect compliance, 1/2-day screening leads to 13,755/21,496 total infections, and 70/115 hospitalizations under 171 vaccine effectiveness values reported for Omicron (see Supplementary Table 7 ). If the vaccine effectiveness for Omicron were 172 as high as for Delta (see Supplementary Table 8) , the same level of mitigation efforts would yield 123/327 infections, and 173 1/2 hospitalization(s). In this scenario, the number of infections and hospitalizations would be 690/4,079/9,068, and 4/17/37, 174 respectively, under 3/7/14-day screening (see Supplementary Table 8 ). The discrepancy between the two scenarios is less 175 striking, but still significant, when the proportion of Omicron is lower, i.e., ω O = 50%, or when the entire population is boosted 176 (see Supplementary Table 8) .

Recognizing that different campus populations may exhibit different characteristics in their screening compliance, we 178 further investigate the impact of screening compliance. For 82% vaccination coverage (64% boosted, 18% vaccinated) and 179 ω O = 95%, increasing the screening compliance of universal screening from 75% to 90% reduces the infections from 8,568 to 180 6/23 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.04.22274667 doi: medRxiv preprint

As new COVID-19 variants emerge, the challenges for effective mitigation of the pandemic in closed communities evolve. 183

Considering the Spring 2022 academic semester and the co-circulating Delta and Omicron variants in this period, the results 184 from the extended compartmental model in this study suggest that routine screening continues to play a key role in the safe 185 opening and operation of universities. However, allocating the limited screening resources in the most effective manner 186 requires extensive planning, considering the specific transmission and disease dynamics of the circulating variants, as well as 187 the vaccination coverage, the imperfect, waning, and variant-dependent immunity from vaccination, and the level of natural 188 immunity in the population.

As expected, the benefits of routine screening increase as the screening coverage is expanded and screening is more frequent. need to be screened, and at a higher frequency, in order to maximize the infections averted per test. Several factors, including the 202 higher reproduction number of, and the lower vaccine efficiency against, Omicron, and the waning vaccine-induced immunity 203 against both variants, contribute to this finding. Furthermore, we observe that as the proportion of Omicron (versus Delta) 204 decreases, the screening frequency that is needed to maximize the efficiency also decreases.

Another key finding is the need for the decision maker to revisit their mitigation objectives as new variants, having different 206 characteristics, emerge. Our results show that when Omicron is the primary circulating variant and screening resources are 207 limited, it might be better to focus on minimizing the peak infections, instead of the total infections, where the latter requires 208 aggressive screening that may not be resource-feasible, or practical, for most universities. We show that screening only 209 unvaccinated individuals (that is, customizing the screening population) is the most efficient strategy in terms of the peak 210 infections averted per test under various Omicron proportions and boosted coverage rates. This finding signifies that when 211 a variant with a higher reproduction number is the dominant strain, allocating the available testing resources to the most 212 7/23 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.04.22274667 doi: medRxiv preprint vulnerable provides the most efficient response to the pandemic, by "flattening the curve." We need to emphasize, however, that 213 the most efficient strategy, i.e. the strategy that maximizes infections (or peak infections) averted per-test, may not (and in most 214 scenarios will not) minimize the total number of infections. Since increasing the screening coverage and/or frequency always 215 reduces the total number of infections, universities may need to choose their strategy based on sequentially increasing the 216 screening coverage and/or frequency, in the most efficient way, until the expected number of infections is reduced to a tolerable 217 level.

There have been significant discrepancies among U.S. colleges and universities regarding routine screening. For example, 219 some universities conducted universal screening [38], whereas some others customized the screening population based on 220 vaccination status, but still used the same screening frequency for all the screening population [39, 40], yet some others did 221 switch, at some point, to screening the faculty and students with different screening frequencies. For instance, at the beginning 222 of the Spring semester, Boston University required the faculty to be routinely screened once a week, and the students twice 223 a week [41] . Stanford, on the other hand, required students to be screened weekly but exempted the faculty from routine 224 screening at some point in the semester [42] . Comparing screening strategies with varying degrees of customization, our 225 findings demonstrate that customizing both the screening population and the frequency based on vaccination status may avert 226 slightly more infections per test over universal strategies, or strategies that customize the screening population only, especially 227 when it is feasible to screen the unvaccinated at higher frequencies, and when both Delta and Omicron are in circulation at 228 similar rates. In this case, the most efficient strategy calls for screening the vaccinated and the boosted less frequently than the 229 unvaccinated, and perhaps not screening the boosted at all. This finding is significant, as it implies that, through customization, 230 a less aggressive strategy (that screens a smaller portion of the population) can provide higher per-test efficiency than universal 231 screening.

Full customization, which considers both the vaccination status and faculty versus student groups further increases the 233 infections averted per test, over customization based on vaccination status alone. While the improvement is small, the main 234 message is again that the decision maker can achieve higher per-test benefits with less screening. In particular, when full 235 customization is considered, the most efficient strategy (for infections averted per test) recommends the faculty to be screened 236 either less frequently than the students, or not at all. In terms of the peak number of infections averted per test, on the 237 other hand, customizing the screening population already provides a highly efficient strategy, and further customizing the 238 screening frequency does not offer significant benefits: screening only the unvaccinated remains the dominating strategy. 239

Overall, our results suggest that allowing customization of both the screening population and the frequency based on population 240 characteristics may indeed yield more efficient strategies; equally important is the fact that these customized strategies may also 241 lower the required testing resources during the semester. This is because in many cases efficient customized strategies call for 242 some groups to be either tested with very low frequency, or not at all, and this may even reduce the logistical complexity of 243 screening, which was contrary to our initial intuition about customized strategies.

As the vaccine-induced immunity in the population wanes over time and/or new variants, which are more resistant to the 245

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.04.22274667 doi: medRxiv preprint available vaccines, emerge, as was the case at the beginning of 2022, important policy questions arise on whether on-campus 246 screening would be sufficient for infection control, and how these efforts should be adjusted based on the booster coverage at the 247 start of the semester. Our results indicate that when both Delta and Omicron are in circulation at similar rates, aiming for both 248 aggressive screening and high levels of booster coverage may be redundant; screening alone may even be able to compensate 249 for a lack of an appropriate booster coverage at the start of the semester. Ideally, integrating moderate levels of booster coverage 250 and screening can provide a highly effective, yet not aggressive, mitigation effort to keep both the number of infections and 251 hospitalizations under control. When Omicron is the primary circulating variant, however, integrating boosters and screening 252

is key for effective mitigation, as none of these efforts would be sufficient, on its own, for controlling the infection, even 253 when implemented at an aggressive level. Accordingly, our results suggest that both adequate booster coverage and routine 254 screening are essential for a safe opening of university campuses, considering the diminishing vaccine effectiveness over time 255

and new vaccine-resistant variant threats. From a practical perspective, integrated screening and booster efforts work especially 256 well towards creating an academic environment that is conducive to in-person learning, because the system is unlikely to be 257 overwhelmed by a large volume of students and/or faculty missing from the classroom due to an active infection or isolation 258 orders; this can also reduce the potential testing fatigue. vaccination/booster coverage are maintained. Our analysis also indicates that, although higher screening compliance leads to 263 fewer infections, its impact is not substantial, as long as the compliance is kept at a reasonable level (e.g., 75% in our study). 264 

• Routine screening excludes subjects who are symptomatic (symptomatic testing is conducted separately), or who are in 294 isolation, at the hospital, or who have tested positive for, and recovered from, the infection (i.e., "recovered and known" 295 subjects). We study routine screening, with the specific strategy dictating the screening population (i.e., vaccination 296 categories, or faculty versus student groups included in routine screening) and the screening frequency of each vaccination 297 category, or faculty/student group. The screening strategy can be universal across all groups and vaccination status 298 categories, or customized. We study various screening policies, presented below in increasing level of customization, see 299 positives are corrected the next day (through additional testing). We assume perfect compliance for all isolation orders and face 303 masking policies, and model imperfect compliance for routine screening.

Setting and Parameters: We simulate the infection spread in our hypothetical college of 24,000 (22,500 students and 2,500 305 faculty members), with 135 students and 9 faculty members (0.6% of each group) having undetected, asymptotic SARS-CoV-2 306 infection at the outset, and some individuals arriving on campus as vaccinated or boosted. The study period is an 80-day Spring 307 2022 academic semester that starts in January 2022. Table 2 reports the key parameters, along with corresponding references, 308 and the details are provided in the Supplementary information. we compute the basic reproduction number per group, and vaccine effectiveness per vaccination status, as weighted averages 315 of their respective values for each variant, that is, as a function of ω 0 -see Table 3 , which provides the key parameters, the 316 weighted average formula, and the computed parameters for the ω 0 = 50% case, considering a 3:1 ratio between the R0 values 317

for Omicron and Delta [45, 46, 47] . While these numbers may seem high for the general population, they are more relevant for 318 the college campus setting [48] . We study the effectiveness of various universal and customized screening strategies (see Table 1 ), obtained by varying the 325 screening frequency(ies) (every 1,2,3,7, 14 days, or no screening) for each vaccination status, see Table 2 . Different strategies 326 may require different numbers of tests, representing scenarios with different testing capacities or testing kits.

The compartmental model is coded in C++, and the results are analyzed in Microsoft Excel, through various plots. Our analysis 329 does not involve any statistical tests, therefore, we do not report statistical significance levels. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. 

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.04.22274667 doi: medRxiv preprint 

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.04.22274667 doi: medRxiv preprint 

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.04.22274667 doi: medRxiv preprint

How colleges are dealing with high COVID case counts on campus