Assessment of SARS-CoV-2 Screening Strategies to Permit the Safe Reopening of College Campuses in the United States

Key Points Question What screening and isolation programs for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) will keep students at US residential colleges safe and permit the reopening of campuses? Findings This analytic modeling study of a hypothetical cohort of 4990 college-age students without SARS-CoV-2 infection and 10 students with undetected asymptomatic cases of SARS-CoV-2 infection suggested that frequent screening (every 2 days) of all students with a low-sensitivity, high-specificity test might be required to control outbreaks with manageable isolation dormitory utilization at a justifiable cost. Meaning In this modeling study, symptom-based screening alone was not sufficient to contain an outbreak, and the safe reopening of campuses in fall 2020 may require screening every 2 days, uncompromising vigilance, and continuous attention to good prevention practices.


eAppendix. Model Description
We developed a dynamic, compartmental model using a modified "susceptible-exposed-infectedrecovered" (or SEIR) framework. The model portrays the epidemiology and natural history of infection in a homogeneous population of at-risk individuals as a sequence of transitions, governed by difference equations, between different health states (or "compartments"). The flow diagram ( Figure S1, below) illustrates the modifications we made to the basic SIR framework: • Addition of regular, repeated screening with a test of imperfect sensitivity and specificity.
• Removal of infected individuals from the transmitting population based on either screening test findings or the development of COVID-defining symptoms.
• Removal (and return) of uninfected individuals from the transmitting population based on "false positive" screening test findings.
• Importation of additional new infections from exogenous sources (e.g., infections transmitted to students by university employees or members of the surrounding community.
Compartments. We defined a total of 8 model compartments, divided into three pools: • Active transmission and testing pool. Everyone is in this pool at time 0. All transmission of infection takes place between individuals in this pool. This is also the pool in which screening for infection takes place.
• U: Uninfected, susceptible individuals • E: Exposed, asymptomatic, non-infectious • A: Infected, asymptomatic Individuals in the Exposed compartment are assumed to be neither infectious nor symptomatic.
(We also assume that these individuals will invariably test negative, if screened.) Note that, without testing, individuals in these three compartments are indistinguishable from one another.
• Isolation pool. Individuals in this pool are assumed to be isolated from the active transmission pool and from one another. It is assumed that transmission is not possible within this pool.  Testing is implemented in the model as a constant rate, governed by parameter . This means that students are screened, at random, an average once every 1/ cycles. This does not reflect the possibility of pulsed or scheduled screening at regular intervals. Note that there is a lag of one cycle between the time that a test is conducted and the time that a student receiving a positive test result is moved to isolation; the model is specifically designed to capture this delay. This captures the time to transport the sample to the lab, obtain the result, locate the student, and effect the transfer to isolation.

Governing equations
 Uninfected (t+1) = Uninfected (t) -New Infections -New FPs + Returning FPs -Exogenous Shocks  Exposed (t+1) = Exposed (t) -New Infectious + New Exposeds + Exogenous Shocks  Estimating Key Rate Parameters 1) : rate of symptom onset for infected individuals. We assumed that 30% of all infected individuals would eventually develop symptoms. In the absence of a screening program, this implies that  / (+) = 0.3. Assuming a mean recovery time of 14 days and computing all rates per 8-hour cycle yields  = 1/(3* 14 days) and we solve for  = 0.0102.
2) β: rate at which infected individuals contact susceptibles and infect them. The effective reproductive number Rt = β / (+

eFigure 2. Expected Daily Occupancy of the Isolation Dormitory Under Worst-Case Assumptions
Worst case assumptions include R t = 3.5, 25 exogenous shocks each week, and a 98% specific test. The panels show results of screening at different frequencies: (a) daily screening; (b) screening every 3 days; (c) weekly screening; and (d) symptom-based screening (i.e., symptom-based detection). In Panels a and b, the effect of exogenous shocks (25 per week) is visible in the scalloped borders; this is less evident with less frequent testing (and symptom-based screening) where the number of true positive cases masks the comparatively small impact of exogenous shocks.  Best case assumption include Rt = 1.5, five exogenous shocks each week, and a 99.7% specific test. The panels show results of screening at different frequencies: (a) daily screening; (b) screening every 3 days; (c) weekly screening; and (d) symptom-based screening (i.e., symptom-based detection). In Panels a and b, the effect of exogenous shocks (5 per week) is visible in the scalloped borders; this is less evident with less frequent testing (and symptom-based screening) where the number of true positive cases masks the comparatively small impact of exogenous shocks.