1. Thread: Proactive testing in a partially vaccinated population.

I will start with a disclosure. The work described was done in collaboration with @Color Health and I was paid as a consultant for my efforts. I have no financial stake in COVID tests, treatments, or vaccines.

2. Large-scale proactive testing has been an important COVID control measure, because it identifies those who are presymptomatic, asymptomatic, or paucisymptomatic and allows them to self-isolate.

As vaccination becomes widespread, two questions arise:

3. First, at what level of vaccine coverage is proactive testing no longer necessary?

Second, as we transition to this point, what are best practices for tapering off testing efforts?

I explored these questions with @RS_McGee, @ay_zhou, @jrhomburger, and @hewillia34.

4. In what follows, I’ll provide an overview of our white paper and associated webapp. The white paper is available at https://bit.ly/testing-and-vaccines-memo and the webapp at http://color.com/testing-and-vaccines-model .

5. To answer these questions, we used two very different modeling approaches. The first is based on an analytic approximation for the value of proactive testing that I developed with Ted Bergstrom and Haoran Li last fall.

http://ctbergstrom.com/publications/pdfs/working-frequency-accuracy.pdf

6. The second is the SEIRS+ stochastic network-based modeling framework that @rsmcgee developed and that we used in our previous reports on workplace testing and return-to-school plans.

Threads about those projects: https://twitter.com/CT_Bergstrom/status/1291466354751467520
https://twitter.com/CT_Bergstrom/status/1354082188581605381

7. Using the analytic model, we look at how testing and vaccination can be used together to control disease. The figure below illustrates how the effective reproduction number Re in a workplace or university setting depends on testing rates and vaccination coverage.

8. The black contour lines are isoclines—combinations of vaccination and testing correspond to the same Re. The dashed line indicates the critical value Re=1 at which a sizable outbreak becomes unlikely. Above and to the right of this point, controls are likely to be effective.

9. For example, when R0=2 (left panel), it would be sufficient to test on a semi-weekly basis, or vaccinate half the population, or implement some intermediate combination of less frequent testing and lower vaccine coverage.

10. What we see in this graph is that when vaccination coverage is limited, testing offers considerable benefits. When 40% are vaccinated, for example, weekly testing moves you from yellow (Re>1, outbreaks likely) to blue (controlled).

11. Once vaccination coverage is extensive, testing is unnecessary. You are already in the blue controlled region, and testing doesn’t change Re much anyway (the isoclines have shifted to become nearly vertical).

12. Of course this analytic model simplifies away lots of detail. To account for some complexities of the real world—superspreading, social contact networks, variation from person to person in disease progression, and the role of chance in an outbreak—we turn to the SEIRS+ model.

13. We look at what happens if a single COVID case is introduced into a community of a thousand people, e.g. a large workplace. Because chance plays a major role, we simulate this situation a thousand times for every set of conditions we want to understand.

14. The figure below illustrates what we observe with R0=3. When vaccination uptake is low, outcomes are bimodal. A single introduction either leads to a large outbreak or fails to spark an outbreak at all.

15. At lower vaccination levels, proactive testing helps a lot. With testing, we see major reductions in both the mean number of cases and the size 95th percentile outbreak.

16. As more of the population becomes infected, the average benefit of testing declines. Below, the difference in the mean number of individuals infected with and without testing for three different testing cadences, R0=3, and 10% already immune.

17. From the above figure we see that when vaccination is limited, testing confers sizable benefits and more frequent testing confers larger benefits. Once vaccine coverage is extensive, the benefits of testing almost disappear.

18. Stepping back, we find that the two different models generate concordant results despite very different methodology. That is always reassuring.

19. Testing is beneficial when vaccination is limited, but it becomes unnecessarily—quite abruptly—once vaccine coverage is sufficiently broad.

20. The challenge that decisionmakers face will be in determining when the population has reached the point where testing is no longer needed. Because we rarely have adequate estimates of R0 within an institution, the models have limited predictive value in this regard.

21. In general, titrating testing cadence to vaccination level will not be feasible. The uncertainties are too large, the logistics too difficult, and the period too short during which intermediate levels of testing would be desirable.

22. Instead, we recommend that employers, universities, etc., continue to test at full capacity until they are fairly confident they are beyond the vaccination threshold at which proactive testing can be suspended. Ongoing surveillance testing is recommended even afterward.

23. To illustrate the robustness of the findings, we have developed a webapp that allows users to alter various parameters in the SEIRS+ model. I encourage you to take a look: http://color.com/testing-and-vaccines-model. The full report can be found at https://bit.ly/testing-and-vaccines-memo.