Covid Surge "Peak" Modeling
August 24, 2021 9:51 AM   Subscribe

I've been hearing within the past week that the current wave is expected to peak around September 7th. OHSU modeling shows this.

What kinds of inputs might result in that date? We aren't getting folks vaccinated fast enough, and September 7th is only a couple of weeks away. Are modelers using the assumption that people will change their masking behavior due to the new state mandate? Is there something so fundamental about virus behavior that even with Delta's high transmission rate, the modelers are able to confidently make assumptions about transmission rates decreasing? Are our high numbers dropping just enough where they are just looking at a trend curve?
posted by happy_cat to Science & Nature (7 answers total) 1 user marked this as a favorite
 
Response by poster: *I meant to say that this model is for Oregon.
posted by happy_cat at 9:52 AM on August 24


Best answer: Around the middle of this page (under the header "OHSU COVID-19 Forecast"), there are links to the most recent (8/18) prediction with a couple PDFs that go into more detail on the background & assumptions.
posted by augustimagination at 10:28 AM on August 24 [2 favorites]


I'm not familiar with the specifics of that model, but one input that might imply a peak soon is the evidence from rest of the world. Delta waves started early and have peaked in other countries with similar non-pharmacological interventions and vaccination rates. Assuming the delta wave in the US is about as long as elsewhere suggests the US wave would peak late August/early September. It has suggested this timing for a month or more. The other hint that cases may start going down in, e.g. Oregon, is that they have very recently started going down in Florida, which has had a similarly severe wave, and California, which is next door.
posted by caek at 11:34 AM on August 24 [4 favorites]


What their model is saying is that Delta burns through enough of the susceptible population that new infections fall. Look at their PDF's slide 24 "Model-Herd Chart", the orange "susceptible" going down to below 5%.

What I haven't dug into their docs for is how much they assume vaccinated people and previously infected people will transmit infection. I hope they're not assuming zero.
posted by away for regrooving at 11:58 AM on August 24 [2 favorites]


It's worth distinguishing this type of 'burnout' peak, which takes a couple of months, from the faster week-to-week changes seen in the U.K. and other Delta surges. Fast changes are from people changing their behavior. Whether they maintain the change is another question -- in the U.K. it was brief, the downward slope is not continuing.
posted by away for regrooving at 12:09 PM on August 24 [2 favorites]


Simple extrapolation also gives an early September peak (this is nationwide, not Oregon-specific). If you look at the nationwide totals you can pretty clearly see that it's starting to bend downwards. If you compute (cases this week)/(cases last week) that's been trending downwards for a few weeks, and should cross 1 sometime in early September. (I did this yesterday while avoiding real work; I didn't save the details, but now I'm telling you I did it while avoiding real work.)

Also, states that got this wave earlier have already peaked - Louisiana, Arkansas, Missouri are already on their way down.

(IANAEpidemiologist.)
posted by madcaptenor at 12:24 PM on August 24 [1 favorite]


If you're wondering why delta waves globally seem to be short, sharp waves, there's some evidence that delta has a shorter "serial interval", i.e. there is less time on average between exposure and transmission compared to the original strain. This is bad because it means people are less likely to know they have it while they are still infectious (and therefore infect more people). It also means the population-level wave of cases is narrower (say 3-4 months rather than 4-6 months).
posted by caek at 12:25 PM on August 24 [2 favorites]


« Older Wild swimming in the US   |   My head is hot. My temp is low. What does this... Newer »

You are not logged in, either login or create an account to post comments