Modelling testing frequencies required for early detection of a SARS-CoV-2 outbreak on a university campus

Early detection and risk mitigation efforts are essential for averting large outbreaks of SARS-CoV-2. Active surveillance for SARS-CoV-2 can aid in early detection of outbreaks, but the testing frequency required to identify an outbreak at its earliest stage is unknown. We assess what testing frequency is required to detect an outbreak before there are 10 detectable infections.

Methods

A dynamic compartmental transmission model of SARS-CoV-2 was developed to simulate spread among a university community. After introducing a single infection into a fully susceptible population, we calculate the probability of detecting at least one case on each succeeding day with various NAT testing frequencies (daily testing achieving 25%, 50%, 75%, and 100% of the population tested per month) assuming an 85% test sensitivity. A proportion of infected individuals (varied from 1–60%) are assumed to present to health services (HS) for symptomatic testing. We ascertain the expected number of detectable infections in the community when there is a > 90% probability of detecting at least 1 case. Sensitivity analyses examine impact of transmission rates (R _{t = 0} = 2, 2.5,3), presentation to HS (1%/5%/30%/60%), and pre-existing immunity (0%/10%)

Results

Assuming an 85% test sensitivity, identifying an outbreak with 90% probability when the expected number of detectable infections is 9 or fewer requires NAT testing of 100% of the population per month; this result holds for all transmission rates and all levels of presentation at health services we considered. If 1% of infected people present at HS and R _t=0 =3, testing 75%/50%/25% per month could identify an outbreak when the expected numbers of detectable infections are 12/17/30 respectively; these numbers decline to 9/11/12 if 30% of infected people present at HS. As proportion of infected individuals present at health services increases, the marginal impact of active surveillance is reduced. Higher transmission rates result in shorter time to detection but also rapidly escalating cases without intervention. Little differences were observed with 10% pre-existing immunity.

Conclusions

Widespread testing of 100% of the campus population every month is required to detect an outbreak when there are fewer than 9 detectable infections for the scenarios examined, but high presentation of symptomatic people at HS can compensate in part for lower levels of testing. Early detection is necessary, but not sufficient, to curtail disease outbreaks; the proposed testing rates would need to be accompanied by case isolation, contact tracing, quarantine, and other risk mitigation and social distancing interventions.