Predicting the next wave of covid?
The covid pandemic seems to come with continual surprises, small and big, from toilet paper shortages to emerging variants. But, perhaps, there might be some predictability arising from the chaos. Maybe.
Amongst all the many graphs of the pandemic (new cases, hospitalizations, deaths, national and global), one chart stands out as strikingly periodic: new cases, worldwide. In that one chart, which happens to be the chart with the highest numbers and the most data, there is a cycle of peaks coming every 4 months or so: April 2020; August 2020; December 2020; April 2021; August 2021; and now heading upwards to a likely 6th peak at the end of December 2022.
Is this meaningful?
I asked Francisco Arroyo Marioli, an economist with the world bank who is deep in statistical data and who has worked on defining epidemic waves. “Indeed this 18-week pattern seems to show up and is interesting to point out,” he emailed me. “However, I would hesitate to jump to conclusions.”
It’s unclear whether past epidemics really had ‘waves’ at all, for example, or, if so, how they worked. Most of our instinctive ideas about disease waves comes from seasonal illnesses, like seasonal flu — but how and why that periodicity happens is also under debate by scientists (and only happens for viruses that have been around a while, where natural immunity is high enough that the weather becomes a more important factor for viral spread).
Most epidemiological studies are done on much smaller scales, attempting to look for patterns within units that have some semblance of similar conditions. Every place, after all, has a different climate, different national policies, different border controls and mask wearing restrictions, different vaccination rates and background health issues. This has created a messy patchwork of complex conditions around the globe, with highly-variable experiences of covid-19.
In Canada, where I live, there was no August 2020 peak. In India, there have been only two apparent peaks, the first coming very late in the game in September 2020 (it was initially speculated that there might have been high levels of natural immunity at play in that country), and the second devastatingly high in May 2021. In Israel, where vaccination was achieved quickly, April 2021 passed with very low rates of infection, but August/September 2021 saw cases bounce back up. On a national level, it is messy and seemingly unpredictable. But could the international numbers be more stable?
It’s a human inclination to see patterns, even where none exist — we see a face on the Moon, shapes in clouds and stars, a trend on the roulette table. It’s called apophenia. There’s also a tendency to seek simplicity in complex systems. So, I’ll approach this observation with caution. But, still, there is truth to the idea that the more data you have, the more any intrinsic pattern is to emerge more clearly. And, sometimes, in some systems, the more actors at work the more predictable the statistical outcome.
This is true, for example, for molecules in air — it’s impossible to predict the movements of any one given molecule, but we can predict how their bulk action creates things like temperature and pressure. The science fiction author Isaac Asimov famously made use of this idea when inventing ‘psychohistory’, the fictional branch of science that uses the statistically-significant mass of people in the far distant future (one quintillion people) to predict bulk societal properties and events like upcoming wars. This comparison glosses over a lot, of course. Humans, unlike single molecules, have free will — or I like to think they do.
Is it possible that despite the messy progress of covid-19 around the globe, and the patchwork of climates, living conditions and policy responses, something about the virus’s interaction with humanity is averaging out to a statistically predictable pattern?
“From a statistical point of view, 6 cases [waves] are interesting but not enough for forecasting,” says Marioli. “One can always speculate of course, but proper forecasting would require more backup.”
Consider this my speculation.
