My short message about the 2021-2022 academic year is easy: Don’t be so sure!
I lead the COVID team at Ursinus College, a small college in suburban Philadelphia. We have the experience now of having operated mostly in-person all last year, and felt well prepared for this academic year, despite the Delta variant. I have been in conversation with a range of public health experts and with other Pennsylvania colleges and universities. And as a scientist with a long history of modeling dynamical systems, getting my numerical hands dirty in COVID simulations was irresistible. Having been engaged in this for over a year now, I have learned a few lessons.
1) Vaccines alone will not be enough. The Delta variant has changed the situation for college campuses in ways that don’t apply to the broader national situation. Simulations of what happens in the larger public are not very useful to help understand what happens on a residential college campus. Most college campuses consist of a highly social group of individuals living very close to one another, regularly gathering in uncontrolled, unmasked, crowded party situations. The reproduction number (Ro), often misunderstood as being solely a characteristic of the disease, is a combination of the disease characteristics and the social behavior of the setting. Particularly for a disease with airborne transmission like COVID, the density of individuals and their setting dramatically affect the reproduction number. As a result, Ro is considerably higher on college campuses than in the general population—in our simulations, we use roughly double the number compared to the population at large: ten instead of about five. The goal for containment is, through a combination of mitigation measures like masks, and immunity through vaccination or natural means, to get an effective reproduction number (variously called Reff or Rt) that is less than one. Not surprisingly, that becomes harder to do if Ro is larger. For the general population, vaccination alone will likely be sufficient to quell the Delta variant; on a college campus, even 100% vaccination of students is unlikely to prevent outbreaks1.
2) Size matters. Not surprisingly, the bulk of simulations of COVID at colleges and universities, at least the ones we read about, are done at large institutions for large institutions, which become misleading for planning at small colleges. COVID is not characterized by a consistent transmission process—where one sick person automatically infects five others. In contrast, the average person infects only one other, or maybe even no one else, but a smaller fraction of folks, because of behavior or the course of their particular disease, are extra infectious, leading to superspreader events. It is likely that recently publicized outbreaks at some universities2 arose from these sorts of events. At a large university of, say, 50,000 students, an event that results in 50 positive cases is only a tenth of a percent of the campus. On a small campus of 1,000 students, this same outbreak directly affects 5% of campus, and could easily quarantine another 20% of the student body. Planning for isolation and quarantine space becomes exceptionally challenging, and every faculty member likely has at least one class seriously disrupted by absences.
3) Variability is huge. Transmission of the virus is governed by probability; ventilation patterns, who sits where, who coughs when, timing of gatherings and individual vulnerability are just a few of the barely controllable factors that affect the transmission of the disease. These initial variables are rapidly amplified by the exponential growth that characterizes an outbreak. These characteristics are revealed in simulations, called stochastic modeling, that include the probabilistic nature of transmission. Simulations3 that do not include this variability will reliably give average results, but not suggest the range of results one might expect with exactly the same set of institutional practices. Particularly on a small campus, that range of outcomes can be startling. Our model allows repeated simulations of the same conditions; a set of 20 independent simulations with identical conditions gave results of total infections over a semester that varied by a factor of 15 (in this instance, as low as 5 and as high as 78).
For college administrators, the takeaways are several. First, be prepared with multiple potential effective interventions, because even full vaccination is unlikely to be enough. Second, outbreaks, including lots of asymptomatic cases, can arise very quickly, especially on a small campus, so some sort of testing is essential. Third, be prepared for the unexpected, and don’t take too much comfort from good results for a single semester—it may well just have been good luck! Taking lessons from one or two campuses that “did well” can be similarly misleading. Sadly, following good practices won’t always be enough, so being vigilant and prepared to respond to the unexpected continues to be essential for maximizing chances of success.
3. One recent example of a very flexible model is https://www.medrxiv.org/content/10.1101/2020.07.06.20147702v1 which allows useful adaptation to particular circumstances, but does not reflect the variability that one might expect, particularly on a small campus.