The case for altruism in institutional diagnostic testing

Abstract

Amid COVID-19, many institutions deployed vast resources to test their members regularly for safe reopening. This self-focused approach, however, not only overlooks surrounding communities but also remains blind to community transmission that could breach the institution. To test the relative merits of a more altruistic strategy, we built an epidemiological model that assesses the differential impact on case counts when institutions instead allocate a proportion of their tests to members' close contacts in the larger community. We found that testing outside the institution benefits the institution in all plausible circumstances, with the optimal proportion of tests to use externally landing at 45% under baseline model parameters. Our results were robust to local prevalence, secondary attack rate, testing capacity, and contact reporting level, yielding a range of optimal community testing proportions from 18 to 58%. The model performed best under the assumption that community contacts are known to the institution; however, it still demonstrated a significant benefit even without complete knowledge of the contact network.

Figure 1

(A) Example of a contact network representing members of the institution (large, purple nodes) and their contacts in the periphery (small, orange nodes). Here we have 10 institution members who make an average of 2 contacts within the institution and 2 contacts outside the institution ($\text{variance}=3$ for both distributions). (B) Flowchart of compartments and possible state transitions.

Figure 2

(A) Modeled cumulative cases over time at CMU under 5 different proportions p of peripheral testing; (B) cumulative cases on day 40 as a function of the proportion of tests deployed to the periphery, with the minimum at 45% peripheral testing.

Figure 3

Cumulative cases on day 40 as a function of the proportion of tests deployed to the periphery under different values of (A) the initial prevalence in the periphery, $V_{0\ast }$; (B) the secondary attack rate among institution members, ${\mu }_{\rho }$; (C) the tests-per-person-per-day ratio, c, and (D) the proportion of contacts traced $\omega$.