Estimating local covid prevalence
March 28, 2023 7:52 AM Subscribe
At this point in the pandemic the best measure we have of prevalence is wastewater concentrations. How do I convert that to asymptomatic positivity rate?
What I really want to know is: the likelihood that someone in a group around me has covid, eg, 1 in 10, or 1 in 50, or 1 in 200. This informs my choices of where to go and what to do.
"Case counts" and such are useless at this point, since the rise of rapid tests and the wind-down of public reporting and facilities that has happened. Bob Wachter used to report asymptomatic test positivity rates from UCSF hospitals, which is arguably the closest proxy for the likelihood that some random person near you has covid. However, I don't live in SF, and that rate is not available for my local area. What I do watch now is wastewater.
How do I convert wastewater (copies/mL) to an infection or asymptomatic rate?
What I really want to know is: the likelihood that someone in a group around me has covid, eg, 1 in 10, or 1 in 50, or 1 in 200. This informs my choices of where to go and what to do.
"Case counts" and such are useless at this point, since the rise of rapid tests and the wind-down of public reporting and facilities that has happened. Bob Wachter used to report asymptomatic test positivity rates from UCSF hospitals, which is arguably the closest proxy for the likelihood that some random person near you has covid. However, I don't live in SF, and that rate is not available for my local area. What I do watch now is wastewater.
How do I convert wastewater (copies/mL) to an infection or asymptomatic rate?
Best answer: I'd be really surprised if that had a simple answer, unfortunately. I don't think these metrics are comparable across cities/regions, and I suspect too that they'll be affected by some extrinsic stuff (e.g. maybe heavy rain or drought?). I think "it's going up or down in this region, in a general sense, over the last several weeks" might be the best you can glean from this type of data. I think this is more or less what's recommended by the CDC (link).
The gold standard for estimating prevalence as far as I'm concerned is what the UK Office for National Statistics put together, which was a population based sample, but I believe they're discontinuing this project. This made it possible to get a back of the envelope estimate of what fraction of infections were becoming officially detected cases, and how that was changing over time. The most recent estimate I saw for this fraction was in December (see Trevor Bedford's reasoning in this thread). It wasn't perfect, of course, because there's an ocean in the way, but it was something.
If I felt so moved, one thing I might try in the current environment is obtaining hospitalization rates; using our best estimates of risk of hospitalization to estimate what the number of total infections could be, given those rates; and then comparing that curve to wastewater data, allowing for some time lag. Then maybe you could get some kind of intuition for how these things align in your region. But the problem with this is that the risk of hospitalization has gone down at the same time that our ability to count cases has gone down, and as Bedford notes, that probably means that even this lower risk is an overestimate. Using an inflated hospitalization risk in this context will map to underestimating total infections. To be clear, the lower risk associated with infection for most people is super good news, but if you're trying to avoid infection, it's something you need to be mindful of.
But mostly I think the ship has sailed on having any kind of real time sense of exposure. I assume every indoor space other than my home has COVID in the air, unless I know that everyone in the space has tested recently, and I roleplay accordingly.
posted by eirias at 8:34 AM on March 28, 2023 [8 favorites]
The gold standard for estimating prevalence as far as I'm concerned is what the UK Office for National Statistics put together, which was a population based sample, but I believe they're discontinuing this project. This made it possible to get a back of the envelope estimate of what fraction of infections were becoming officially detected cases, and how that was changing over time. The most recent estimate I saw for this fraction was in December (see Trevor Bedford's reasoning in this thread). It wasn't perfect, of course, because there's an ocean in the way, but it was something.
If I felt so moved, one thing I might try in the current environment is obtaining hospitalization rates; using our best estimates of risk of hospitalization to estimate what the number of total infections could be, given those rates; and then comparing that curve to wastewater data, allowing for some time lag. Then maybe you could get some kind of intuition for how these things align in your region. But the problem with this is that the risk of hospitalization has gone down at the same time that our ability to count cases has gone down, and as Bedford notes, that probably means that even this lower risk is an overestimate. Using an inflated hospitalization risk in this context will map to underestimating total infections. To be clear, the lower risk associated with infection for most people is super good news, but if you're trying to avoid infection, it's something you need to be mindful of.
But mostly I think the ship has sailed on having any kind of real time sense of exposure. I assume every indoor space other than my home has COVID in the air, unless I know that everyone in the space has tested recently, and I roleplay accordingly.
posted by eirias at 8:34 AM on March 28, 2023 [8 favorites]
You may find the People's CDC weather reports useful for an idea of the level of spread in your area.
But like eirias said above, it's safest to assume that all public indoor spaces have covid in the air.
posted by Lycaste at 8:55 AM on March 28, 2023 [3 favorites]
But like eirias said above, it's safest to assume that all public indoor spaces have covid in the air.
posted by Lycaste at 8:55 AM on March 28, 2023 [3 favorites]
Yeah, at this point just assume it's everywhere and you should be masking indoors/taking whatever other precautions you might take.
That said, I read somewhere that what you should be paying attention to is how many people you know are getting covid. I haven't had anyone around me come down with it since late December, weirdly enough. But also I haven't been doing any shows since the last one ended in January, which could certainly explain that as well.
posted by jenfullmoon at 9:00 AM on March 28, 2023 [1 favorite]
That said, I read somewhere that what you should be paying attention to is how many people you know are getting covid. I haven't had anyone around me come down with it since late December, weirdly enough. But also I haven't been doing any shows since the last one ended in January, which could certainly explain that as well.
posted by jenfullmoon at 9:00 AM on March 28, 2023 [1 favorite]
Best answer: This is a really important scientific and public health question that there is not one answer to. Wastewater surveillance is great for identifying trends and hotspots. It is not well translatable to case counts, especially asymptomatic ones. At best, we can compare wastewater concentrations between areas or look at trends. Even that is complicated by diversity of settings, collection methods, and analysis methods. There are modeling studies that use fancy math to predict specific things (like hospital admissions or viral load among ICU patients) using wastewater concentrations - but they are short-term and very setting and method specific. Since scientists don't have a good count of asx infections in an area, they can't model that very well.
Even for more established wastewater surveillance usages, like polio monitoring, we cannot say "a concentration of X equals Y infections". We can say, "there is polio in this area, this is the type of polio and there is likely more polio in this area than that area, and at this time than that time".
I am an infectious disease epidemiologist; I do not currently work on COVID.
posted by quadrilaterals at 9:03 AM on March 28, 2023 [15 favorites]
Even for more established wastewater surveillance usages, like polio monitoring, we cannot say "a concentration of X equals Y infections". We can say, "there is polio in this area, this is the type of polio and there is likely more polio in this area than that area, and at this time than that time".
I am an infectious disease epidemiologist; I do not currently work on COVID.
posted by quadrilaterals at 9:03 AM on March 28, 2023 [15 favorites]
Statistics Canada released a study yesterday March 27 as part 2 of the Canadian COVID-19 Antibody and Health Survey (CCAHS-2).
Some of the results are surprising :
Between April and August 2022, 98% of Canadians had antibodies against COVID-19 and 54% had antibodies from a previous infection
These antibodies were acquired through either vaccination, a previous infection, or both
.
Over this period, 53.9% or about 16.4 million Canadian adults living in private dwellings had antibodies indicating a past infection, which is more than 20 times higher than the 2.6% observed from November 2020 to April 2021.
41.7% of Canadians with antibodies from a past infection never previously tested positive nor suspected an infection.
---
My unprofessional conclusion is that the study shows it's everywhere.
posted by yyz at 10:26 AM on March 28, 2023
Some of the results are surprising :
Between April and August 2022, 98% of Canadians had antibodies against COVID-19 and 54% had antibodies from a previous infection
These antibodies were acquired through either vaccination, a previous infection, or both
.
Over this period, 53.9% or about 16.4 million Canadian adults living in private dwellings had antibodies indicating a past infection, which is more than 20 times higher than the 2.6% observed from November 2020 to April 2021.
41.7% of Canadians with antibodies from a past infection never previously tested positive nor suspected an infection.
---
My unprofessional conclusion is that the study shows it's everywhere.
posted by yyz at 10:26 AM on March 28, 2023
The most recent (and final) ONS infection survey publication was for the week ending 7th March 2023. It found that an estimated 1 in 40 people in England would test positive for covid in that week. It was also 1 in 40 in Scotland and Wales, and 1 in 70 in Northern Ireland .
That is with zero mitigation and almost zero masking in the population.
There are currently around 8,000 people admitted to hospital with covid in England and around 170 on ventilators.
posted by knapah at 2:12 PM on March 28, 2023
That is with zero mitigation and almost zero masking in the population.
There are currently around 8,000 people admitted to hospital with covid in England and around 170 on ventilators.
posted by knapah at 2:12 PM on March 28, 2023
Remember that wastewater is an "over all / average" measurement, but it sounds like what you care about is the "these people right here in front of me" rate. With infectious diseases, by nature this is not evenly distributed.
I'm teaching 2 college courses - same class, same building, same days, essentialy equivalent students, yet in one class I've had about 20% come down with COVID and in the other about 1%, over the past 8 weeks.
But to answer your question (since I want the same info) I use the current wastewater count, and then use that number (viral copies / mL) and go back to September - December 2022 data, and find the equivalent case count at that time, because it's my belief that people were testing and reporting back then.
As an example, for San Diego the data shows about 1.5M copies/mL, which corresponds to a Nov/Dec 2022 case rate of about 13 per 100000.
posted by soylent00FF00 at 4:54 PM on March 28, 2023 [4 favorites]
I'm teaching 2 college courses - same class, same building, same days, essentialy equivalent students, yet in one class I've had about 20% come down with COVID and in the other about 1%, over the past 8 weeks.
But to answer your question (since I want the same info) I use the current wastewater count, and then use that number (viral copies / mL) and go back to September - December 2022 data, and find the equivalent case count at that time, because it's my belief that people were testing and reporting back then.
As an example, for San Diego the data shows about 1.5M copies/mL, which corresponds to a Nov/Dec 2022 case rate of about 13 per 100000.
posted by soylent00FF00 at 4:54 PM on March 28, 2023 [4 favorites]
13 per 100,000 per day (I assume; that is, 91 per 100,000 per 7 days) sounds quite low to me, given that per knapah the most recent ONS estimate is 1 in 40 over a week, or 2500 per 100,000 per 7 days. It’s hard to imagine why England would have 27 times the virus we do, when neither place is surging. I think these numbers aren’t really commensurate, and the ONS numbers map more closely onto the outcome the OP is interested in than do any public health department numbers (which depend on voluntary testing and so are unlikely to capture asymptomatic people).
posted by eirias at 5:18 PM on March 28, 2023
posted by eirias at 5:18 PM on March 28, 2023
« Older I want a tool to track "Things I Should be doing" | Advice for a late May itinerary to Costa Rica Newer »
This thread is closed to new comments.
It also seems likely that as people have greater immunity, the produuce (and thus shed in wastewater) less virus than people with naive immune systems.
I think the wastewater data are meant for understanding trends, not incidence.
posted by If only I had a penguin... at 8:30 AM on March 28, 2023 [3 favorites]