key: cord-0272780-vhsnu2rp
authors: Cardim Falcao, R.; Otterstatter, M.; Ahmed, M. A.; Spencer, M.; Iyaniwura, S.; Janjua, N. Z.; McKee, G.; Irvine, M. A.
title: Impact of routine asymptomatic screening on COVID-19 incidence in a highly vaccinated university population
date: 2021-10-20
journal: nan
DOI: 10.1101/2021.10.18.21265057
sha: 96e5ff69f45c55b48bfdb62c9554537b540836e9
doc_id: 272780
cord_uid: vhsnu2rp

Background: With the return of in-person classes, an understanding of COVID-19 transmission in vaccinated university campuses is essential. Given the context of high anticipated vaccination rates and other measures, there are outstanding questions of the potential impact of campus-based asymptomatic screening. Methods: We estimated the expected number of cases and hospitalizations in one semester using rates derived for British Columbia (BC), Canada up to September 15th, 2021 and age-standardizing to a University population. To estimate the expected number of secondary cases averted due to routine tests of unvaccinated individuals in a BC post-secondary institution, we used a probabilistic model based on the incidence, vaccination effectiveness, vaccination coverage and R0. We examined multiple scenarios of vaccine coverage, screening frequency, and pre-vaccination R0. Results: For one 12 week semester, the expected number of cases is 67 per 50,000 for 80% vaccination coverage and 37 per 50,000 for 95% vaccination coverage. Screening of the unvaccinated population averts an expected 6-16 cases per 50,000 at 80% decreasing to 1-2 averted cases per 50,000 at 95% vaccination coverage for weekly to daily screening. Further scenarios can be explored using a web-based application. Interpretation: Routine screening of unvaccinated individuals may be of limited benefit if vaccination coverage is 80% or greater within a university setting.

By April 2020, nearly all post-secondary institutions in Canada had closed their campuses and moved to online classes to reduce the spread of COVID-19, which had by that time become a global pandemic [1, 2] . With the arrival and availability of highly effective vaccines, universities and colleges have largely returned to in-person classes for fall 2021. The reopening of campuses reintroduces frequent close contact among young adults, which can be a driver of rapid spread of coronavirus. To offset this risk, universities, colleges, and other schools allowing in-person classes 1 . CC-BY-NC 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 20, 2021. ; https://doi.org/10.1101/2021. 10.18.21265057 doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

have implemented protocols, such as mandatory masking, mandatory vaccinations, physical distancing, routine screening, and a hybrid online and in-person class regime. Yet, even with some of these protocols in place, previous reopenings have resulted in spikes in COVID-19 incidence. According to The New York Times, over 700,000 cases of COVID-19 in the US have been linked to university and college campuses by May 26, 2021 [3] . More recently, the number of new infections on US post-secondary campuses seems to be slowly decreasing, and this could be due to vaccination.

In order for universities to plan a safe return to campus, several studies have analyzed the potential for mitigation strategies to control COVID-19 transmission in a university setting [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] . Most of these studies have focused on masks mandates, contact tracing, routine testing, and physical distancing. Some utilized social mixing data from contact surveys [1, 9, 15, 16] , whereas others relied on simulations of daily behavioural contacts among students [17] . Only a few of these studies have considered scenarios that include a highly vaccinated university population [1, 2, 16, 18, 19] .

Many post-secondary institutions implemented asymptomatic testing of COVID-19 during the fall term, yet it is unclear what additional benefit this might offer, especially when some institutions are reporting very high vaccination coverage among students and staff. Another proposed measure, which may pose logistical and ethical challenges, is the routine testing of unvaccinated individuals. In this work, we review previous modelling of COVID-19 transmission in a university setting and highlight the main implications of mass testing to control spread within the context of a vaccinated population. We provide a new probabilistic modelling analysis of routine screening of unvaccinated students on a university campus, based on the example of the University of British Columbia.

Expected numbers of cases per 50,000 across one semester (12 weeks) were derived for various levels of vaccination coverage. This number was chosen to reflect the population size of a large university campus. The rolling seven days mean daily case and hospitalization incidence by age group for the province (24 and under, 25-34, 35-44, 45-54, 55-60, 61-65, 66-70, 71 and over) were obtained for September 15 th , 2021 within BC using the BC Centre for Disease Control (BCCDC) laboratory confirmed and probable COVID-19 case line list. We decided on this date because is near the beginning of term. The hospitalization and case incidence are based on provincial data, and case incidence does vary by region, however, is strongly correlated with vaccination coverage. Rates for vaccinated and un-vaccinated individuals were standardized using population demographic data obtained for the University of British Columbia as an example post-secondary institute and separated into undergraduate, graduate, and staff populations [20, 21] . These rates were then adjusted to generate expected numbers for given vaccination coverage.

We constructed a probabilistic model to determine the expected number of secondary cases averted due to randomly screening unvaccinated individuals within a vaccinated population. The number of secondary cases averted is dependent on the probability of detecting a case during their infectious period and the expected number of secondary cases. The expected number of secondary cases is the vaccine-adjusted effective reproduction number and was calculated incorporating the vaccine coverage, transmission-blocking and onward transmission efficacy of the vaccine, and the pre-vaccination basic reproduction number (R 0 ). The probability of detection is composed of the probability that a randomly selected individual from the population is infected and the probability 2 . CC-BY-NC 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 20, 2021. ;  of detection given that the individual is infected. Assuming a given testing frequency, this probability is further composed of the total probability of detection at a given time point and the probability that the individual is still infectious at that time point across all time points.

The expected secondary cases averted for a population are then multiplied by the population of un-vaccinated to obtain the total expected cases averted across one generation time. This is then multiplied by the semester time-period to obtain the total cases in one semester. Although non-pharmaceutical interventions (NPIs) were not explicitly included within the model framework, their effects can be observed through variation of the R 0 parameter. The model is fully described in the supplementary information.

Parameters Table 1 provides the fixed parameters and the range of the variable parameters that we used for the model. For pre-vaccination R 0 , we considered a range from 5 to 9, allowing for increased transmission due to the Delta variant, the expected increase in contacts in the university population, and the potential impacts of NPIs [22] [23] [24] . For vaccination coverage, we consider that 80% to 100% of the population may be fully vaccinated, given the high vaccination rates in BC, and the self-reported vaccination rate at the University of British Columbia (UBC). As of September 28, a total of 73,779 members of UBC have completed their vaccination status declaration, where 94.78% of those have declared they are fully vaccinated [25] . The vaccination effectiveness consists of the transmission-blocking efficacy v e and the reduction in onward transmission v o . We fixed these parameters to 80% and 40%, respectively [26, 27] . The frequency of testing considered for routine screening of unvaccinated individuals was 1 to 14 days. Lastly, we converted the reported serial interval, with a median of 4 days and ranging from 2 to 10 days [28], into a Gamma distribution Γ(k, θ) with shape k equal to 3.2 and scale θ equal to 0.8 describing the serial interval distribution. Hence, the probability of being infected, but not yet transmitting at time t is given by Γ(t) = Γ(k, θ).

Values 

The number of expected cases and hospitalizations during one semester (12 weeks is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 20, 2021. ;  compared to 99 per 50,000 in a 0% vaccinated university population (See Table A .1 in supplementary for full details). Figure 1 displays the expected number of averted secondary cases due to routine screening of unvaccinated students in one 12 week semester per 50000 students, for a testing frequency of 1-14 days (panel 1a: by vaccination coverage, with a pre-vaccination R 0 = 7; panel 1b: by pre-vaccination R 0 , with a vaccination coverage of 90%). The expected number of secondary cases averted increases with increasing testing frequency and larger pre-vaccination R 0 . However, as vaccination coverage increases there are fewer secondary cases that may be prevented by screening. At 80% vaccination coverage and a pre-vaccination R 0 = 7, and given the overall community incidence rate, daily testing averts only an expected 18 secondary cases within a 12 week semester. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 20, 2021. ; is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

Routine screening of unvaccinated individuals on a post-secondary campus, as a policy to mitigate COVID-19 spread, poses logistical and ethical challenges. To quantify the impact of such a policy, we developed a probabilistic model to calculate the expectation of the number of secondary cases averted due to routine testing of unvaccinated individuals. In addition, we reviewed previous studies quantifying the impact of different NPIs, including mass testing, on the return to campus of a partially vaccinated population. Our results suggest that routine testing of unvaccinated individuals provides limited additional benefit for the mitigation of COVID-19 spread, given the small expected number of secondary cases that could be averted during a 12-weeks semester with impractical testing frequencies [30-32].

To account for the increase in contacts and the highly transmissible Delta variant, we set a high pre-vaccination R 0 in our model. Yet, with vaccination coverage of at least 80% of the university staff and students, only 18 secondary cases are averted in our analysis during a 12-week semester, when screening is done routinely once per day. These results demonstrate the potential limitations of routine testing among unvaccinated individuals as a policy to reduce new cases when vaccine coverage is high and community incidence is relatively low. Our results agree with previous studies that quantify the impact of mass testing as a strategy to mitigate COVID-19 during return to campus when NPIs are in place and there is partial vaccination coverage. These studies [1, 2, 16, 18, 19] find that mass testing provides a small expected reduction in the number of cases for a population with high vaccination coverage. Masking and physical distancing may be efficient strategies to control the spread of COVID-19 in university settings that already have substantial vaccination coverage [1, 2, 16] .

Hospitalizations are underreported in the health authority's laboratory confirmed and probable COVID-19 case line list, resulting in underestimation of expected hospitalizations cases. In a similar way, cases are also under-reported, and given that probability of being infected in our model is based on daily incidence, it should also be underestimated. A possible solution for those would be to include an estimation of an ascertainment fraction, however higher incidence scenarios can also be explored using the accompanying online application.

Our model does not account for the potential homophily of unvaccinated individuals, which can lead to clusters of cases among the unvaccinated population. Clusters of cases of unvaccinated individuals can potentially be prevented when routine testing unvaccinated individuals, and rapidly isolating the COVID-19 positive cases. Moreover, over-dispersed clusters of cases, a known property of COVID-19 [33-36], is not considered in our model. Over-dispersion would likely increase subsequent secondary cases and the number potentially averted through screening of unvaccinated individuals. However, previous work [1, 2] have found that lower levels of vaccine coverage or efficacy result in right-tail events with a comparatively high number of total cases, for example, values being 5-fold the median (see Figure 1 of Junge et al. [2] ). However, they also found that for large vaccination coverage, right-tail events are less likely and with a lower number of cases.

As future work, developing a detailed simulation of COVID-19 dynamics within a university setting could help to more precisely quantify the individual and combined benefits of vaccination and various NPIs, including routine testing. Social contact structure, including potential homophily of unvaccinated individuals and over-dispersion of cases, would be useful additions and could provide a more robust estimation of the number of secondary cases averted. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

For a university campus with high vaccination coverage and low community transmission, the benefits of routine testing of unvaccinated individuals are limited. Other mitigation strategies such as masking and physical distancing may play a more important role in controlling COVID-19 infections compared to mass testing as previous studies have shown [1, 2, 16] .

. CC-BY-NC 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

A.1 Probabilistic model

The expected number of secondary cases averted due to rapid testing an unvaccinated individual in a population with vaccine coverage p and pre-vaccination R 0 is given by: E (secondary cases averted|p, R 0 ) = P (detection of primary case) E (secondary cases|p, R 0 ) , where P (detection of primary case) is the probability of testing the primary case, and the expected number of secondary cases that would have arisen due to this detected case is E (secondary cases|p, R 0 ). The expected number of secondary cases in a post-vaccination environment is equal to:

where v e is the transmission-blocking efficacy of the vaccine, and v o is the reduction on onward transmission due to vaccine.

The probability of detecting a primary case can be further subdivided into two terms: the probability of being infected, and the probability of being detected given that an individual is infected. Given a testing frequency of d days, the probability of detecting a case is modelled by a Poisson process with rate 1 d . Then, the probability of a primary case not been yet detected by time t is given by: 1 − e −t/d . The probability of being infected, but not yet transmitting at time t is given by Γ(t). Thus, we obtain: is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 20, 2021. ;  given that the probability of being infected is equal to the incidence on the unvaccinated population, I unv (t), and for a population of size N we have N (1 − p) unvaccinated individuals, the total expected number of secondary cases averted by routine test is given by: is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 20, 2021. ; https://doi.org/10.1101/2021.10.18.21265057 doi: medRxiv preprint

Estimating data-driven COVID-19 mitigation strategies for safe university reopening

Safe reopening of university campuses is possible with COVID-19 vaccination

Tracking Coronavirus Cases at U.S. Colleges and Universities

COVID-19 transmission in a university setting: a rapid review of modelling studies COVID-19 transmission in a university setting: a rapid review of modelling studies COVID-19 transmission in a university setting: a rapid review of modelling studies

Entry screening and multi-layer mitigation of COVID-19 cases for a safe university reopening

Estimating the number of COVID-19 cases being introduced into British Higher Education Institutions during Autumn 2020

Title: Evaluation of SARS-CoV-2 transmission mitigation strategies on a university campus using an agent-based network model Running Title: SARS-CoV-2 modeling on a university campus

Repeat SARS-CoV-2 testing models for residential college populations. Health Care Management Science

High COVID-19 transmission potential associated with re-opening universities can be mitigated with layered interventions

SARS-CoV-2 infection in UK university students: lessons from

A model of COVID-19 transmission and control on university campuses

SARS-CoV-2 Seroprevalence in a University Community: A Longitudinal Study of the Impact of Student Return to Campus on Infection Risk Among Community Members. medRxiv : the preprint server for health sciences

The impact of University reopenings on COVID-19 cases in Scotland *. 2021

Secondary Household Covid-19 Transmission Modelling of Students Returning Home from University

Contacts and behaviours of university students during the COVID-19 pandemic at the start of the 2020/2021 academic year

Efficacy of SARS-CoV-2 Repeat Testing to Control Spread in Residential College Populations

Simulation-based what-if analysis for controlling the spread of Covid-19 in universities

Threshold analyses on combinations of testing, population size, and vaccine coverage for COVID-19 control in a university setting

Benefits of Surveillance Testing and Quarantine in a SARS-CoV-2 Vaccinated Population of Students on a University Campus

data retrieved from PAIR UBC

University of British Columbia Annual Enrolment Report 2020 and 2021

The COVID-19 epidemic

The reproductive number of the Delta variant of SARS-CoV-2 is far higher compared to the ancestral SARS-CoV-2 virus

Transmission dynamics and epidemiological characteristics of Delta variant infections in China

Update: UBC's COVID-19 Rapid Testing Program

Effectiveness of BNT162b2 mRNA vaccine against infection and COVID-19 vaccine coverage in healthcare workers in England