Q&A: OHSU Economist Shares How He Tackles The Uncertainty Of Omicron
Since early in the pandemic, Peter Graven, a health economist at the Oregon Health & Science University, has been trying to predict how COVID-19 will affect Oregonians and their health care system.
The director of the university’s Office of Advanced Analytics likes to call the forecasts his “best guess,” but says his job is to “strangle” it with real-life medical and epidemiological data until it has a good chance of coming true.
Of late, his weekly predictions have gained new prominence as officials, providers and members of the public prepare for the new, fast-spreading omicron variant of the coronavirus.Since December, his reports have shared one constant: Omicron’s effect on hospitalizations in Oregon will be worse than the worst peak of COVID so far, driven by the delta variant, when people with COVID occupied 1,187 beds on Sept. 1. That's significant because at the time, patients were reportedly dying due to overstretched staff and a lack of beds, even if they never caught COVID.
But beyond that shared trait, Graven's early estimates of omicron swung wildly, jumping from about 2,000 hospitalizations in a draft model to as many as 3,000 the next day, then plunging to about 1,220 the following week, and then up to 1,650 last week.
In an interview on Wednesday afternoon, Graven told The Lund Report that as more data has come in on the omicron’s behavior and severity, the basis for modeling keeps improving. Still, there remain some questions, such as how recent changes in Oregonians’ behavior could delay the spread of omicron.
The following interview has been edited for clarity and brevity.
The Lund Report: Has omicron made it more difficult for you to have confidence in the models you’re using, and has it affected the data quality?
Peter Graven: Yeah … you can think of it as opening up a ball of wax. A lot of things were kind of getting into a pretty narrow range in terms of OK, week after week, we have this many people susceptible; how many of them could possibly get infected next week?
The problem with omicron is (before) you said, “Oh, 18% are susceptible,” and then, now next week, because omicron is suddenly dominant, we’re looking at 75% of the population is now potentially susceptible.
And we don’t have extremely good details – I mean we know that the booster is helping. But then your regular vaccinated folks are pretty much not getting much protection at all. So then you’re basically uncorking the whole state again and saying, “I wonder who’s going to get it,” and, “I wonder how sick they’re going to get when they get it.”
I felt pretty good about where those (Oregon projection) numbers are landing, and just triangulating them as much as possible. Because you really are dealing with a whole new animal in terms of the kinds of protection we thought we had, and the issue is it’s really opening up the vaccinated population is the issue. While the vaccine still works, it works at a certain (reduced) rate. And when you apply that rate to a huge number of people, a slight decimal change compounds itself.
TLR: When you talk about how omicron opens up a whole new range of susceptible people for the vaccinated, non-boostered population, you’re talking about the risk of infection as opposed to severe hospitalization, or…
Graven: Well, but it’s both, right? Let’s say you only have a 1% chance of being hospitalized, of 100 people that’s great, that’s only one person. Of 1,000 people now you’re at 10, and it goes up from there. And if you have a whole bunch more infections, that little percentage chance change really adds up to a lot.
TLR: Doesn’t your data get a little tougher to use because of that distinction between hospitalized with omicron versus hospitalized due to it? Does the special nature of this variant make it more difficult to use that hospitalization data for mapping a curve and making predictions?
Graven: Yes and no. The “hospitalized with” (omicron) has always been there. It’s purely a function of what percent of the general population is infected at any time. So if 2% of the population is infected, then approximately 2% of all your hospitalizations will happen to have COVID, right? And then add that to your percent that were there “because of” COVID. I think we could get to a stage where it’s like 10% or 20% of people are infected. And so that’s certainly higher than before.
I suppose the positive side of that is, then when you look at a high hospital count, you think, well, at least some of those aren’t really there from COVID which would be helpful. The bad part is, of course, they still have COVID and we still have to do all the things to protect them from infecting others. And it does exacerbate conditions. (The patient) could be there for a heart thing or a lung thing, but actually COVID is not helping you at all. When you see big hospitalization numbers in other states, you call them up and talk to them — it’s a disaster.
TLR: You have a new round of projections coming out tomorrow. Right now, is it your guess that it’ll look better or worse in terms of hospitalizations as compared to last week’s projection of 1,650 hospitalized at the end of this month?
Graven: It’s looking about the same (ed. note: here is the release with a report link). Now that we’ve gotten some track record in the U.S., the easiest way is to go look at the states that are ahead of us. If you look at New York, New Jersey, Ohio, you see hospital census per capita that are well over delta levels. That is certainly a possibility for us. I'd love if we could stay between 100% and 150% of delta. But I can tell you, some other states are going above that.
TLR: Given how things have fluctuated in the past, could things look entirely different by tomorrow afternoon?
Graven: I’d like to say, no. Now we should be getting closer to the truth.
TLR: There was an interesting piece in Becker’s Hospital Review on Wednesday, which basically talks about how the surge could last weeks rather than months. Are you still thinking that the peak has moved from February to late January or could it be even quicker than that?
Graven: I think it could move a little earlier. However, on the other hand, for Oregon to do well at this, we want to actually push it back a little. I am starting to see some behavior response. People are adjusting their behavior (to limit the spread of omicron). I can see it, and we absolutely need that.
TLR: How many different data points do you take in in a given week on this stuff?
Graven: I spent most of yesterday translating Danish reports because they have some decent data on vaccination status. I’m scrambling like everybody else in triangulating. It’s pretty much full-time. And the question I still have is, what is omicron’s hospitalization severity for the unvaccinated? Is it any different than delta? We know on average, it's less severe, but is it less severe for the unvaccinated? We see a little bit of a good sign, but I’m not positive. Nobody’s really dialed it in yet.
TLR: It feels like a lot of pressure on you. Until October, the state used to release its own modeling. Now that they’ve stopped doing that, it seems like it’s all on you in Oregon.
Graven: It’s the same thing either way for me. I have my same process every week of doing my, I like to call it my best guess. But obviously as a econometrician and whatnot, mine might look a little bit less like a guess because I use data to kind of strangle it. But, I’m just gonna do my best, and I think my track record gives me some confidence that whatever I’m doing, whatever my process is, it’s been pretty good. And I should just keep doing it.