There's a lot of dichotomous thinking about #COVID risk on #airplanes.
Some believe it's completely safe, others completable dangerous.
I minimize flight travel and wouldn't fly without a fit-tested high-quality mask (N95 or elastomeric respirator). Here's why. 🧵
1/16
Field research from @sri_srikrishna found that across 3 models of aircrafts, they had an air cleaning rate of 10.9-11.8 air changes per hour (ACH).
A U.S. operating room should have 15 ACH, so flights are pretty good, right?
Wrong. I'll explain why.
2/16
10-12 air changes per hour (ACH) on a flight sounds good, even overkill, right?
Actually, no.
If the air has good mixing, the best case scenario is that each air change is still imperfectly efficient, cleaning out about 2/3 of the air each air change.
Uh oh.
3/16
The imperfect efficiency means that the air cleaning must exceed the rate at which the air is being contaminated with airborne infectious virus laden aerosol particles.
The air cleaners must work faster than the people breathing into the air, or virus can accumulate.
4/16
This is a question of the density of the people in the room. If the room has hardly anyone in it, a low air cleaning rate can prevent virus from accumulating in the air. If the room has many people in it, a high air cleaning rate is needed to prevent viral accumulation.
5/16
Accordingly, the 2023 ASHRAE standards indicate per-person air cleaning rates, rather than the number of times a room's air should be cleaned per hour.
This is where the risk of airplanes is revealed.
6/16
ASHRAE only deals with "buildings," not planes (confirmed by phone). As consumers, we can draw approximate inferences by examining their standards for similar "rooms."
A lecture hall/auditorium should clean the air at a rate of 50 cubic ft per minute (cfm) per person.
7/16
If we consider the typical density of people in such settings, 50 cfm/person in a lecture hall is equivalent to about 50 ACH (it's a coincidence both are "50").
If 50 ACH is the minimum standard for air cleaning in a lecture hall & airplanes are similar to lecture halls in terms of seating and density, then airplanes may be cleaning the air at <25% of the standard we would expect.
That means virus accumulating in the air.
9/16
Alternatively, if we assume 11.7 ACH on a plane, that the cabin volume is 6,978 cubic ft, and there are 100 people on the plane, then it would clean at 1,361 total cfm.
It should clean at 5,000 cfm.
That's about 25% of the minimum expected rate. Virus accumulation.
10/16
We are all often in rooms with poor air cleaning rates. However, the risk increases as fewer people are masked, duration increases, and the number of infectious people increases.
Masking is low, flights can be long, and we're entering a U.S. surge.
11/16
Moreover, the ASHRAE standards focus on far field transmission (preventing viral accumulation in the "room") not on those in near proximity.
Good air cleaning can somewhat reduce risk from those nearby, not completely, but poorer air cleaning increases that risk too.
12/16
So, (A) avoid flight travel, or (B) wear a fit-tested high-quality mask throughout.
13/16
These are my recommendations for high-quality N95 masks that fit most people. I have worn an elastomeric (reusable) mask called a @flo_mask on flights (profile picture). No COIs.
PMC COVID-19 Tracker, Dec 11, 2023
The surge continues.
Today:
🔹1.2 million daily infections
🔹1 in 41 infectious (2.5%)
In 4 weeks (Jan 8):
🔹1.6 million daily infections
🔹1 in 30 infectious (3.3%)
1/
A few key methodologic updates. 1) Biobot correct levels downward for the past two weeks, so you might notice that this week's estimates seem similar to last week's or marginally lower.
2) Our forecasting model uses a combination of historic data (situation past several years) and current data (past 4 weeks). In the historic model, we switched from using mean-type data to median-type data. This avoids overestimating levels based on the BA.1 surge and allows us to predict accurately a little faster, rather than predicting high and waiting for the current 4-week's data to correct it.
3) The forecasts depend a lot on the most recent week's data. To the extent Biobot is accurate or inaccurate in real-time, this leads to divergent forecasts.
You'll see the forecasts differ considerably (1.3 to 1.9 million daily infections) in 4 weeks.
2/
However, they mostly agree on the peak. It could be as early as Jan 1 or as late as Jan 15. It's a moot point. Transmission will be similar across that timespan and the weekly reports lack the precision to say whether it will peak on the 4th or 9th, for example. Early Jan will remain bad.
Details:
The real-time model (purple) anticipates the highest surge levels. This assumes that Biobot real-time reports are accurate, but they were substantially corrected for the past two weeks, and there were some issues with real-time accuracy during the summer wave. The turtle model (green) discount’s the most recent week’s data as an aberration, assumes transmission should be corrected upward a little, and predicts a steady rise with peak around January 1. The cheetah model (yellow) says that because last week’s data were corrected downward, this week’s estimate should be too, so it’s much more conservative on the next several weeks. The average of all models (red) guides forecasted numbers for the next four weeks. A month from now, we will see about 1.6 million new U.S. cases per day (range of 1.3 to 1.9 million across forecasting models), with 3.3% of the U.S. population or 1 in 30 people actively infectious.
Zooming out, you'll see that we're in a very bad place historically. With the divergent forecasts, it's merely a matter of whether this is the 2nd biggest U.S. COVID surge or 4th biggest.
The #LongCOVID cases resulting from these infections may top 400,000/week.
PMC COVID-19 Tracker, Dec 4, 2023
The U.S. surge is worsening faster than anticipated.
Today:
🔹1.2 million daily infections
🔹1 in 38 infectious (2.6%)
In 4 weeks (New Year's Day):
🔹1.8 million daily infections
🔹1 in 26 infectious (3.9%)
1/
You'll note the diverging forecasts. Biobot #wastewater levels increased more than anticipated this week, and they updated last week's numbers upward too.
The cheetah forecast (yellow) assumes this week's levels will also get revised upward like last week's. The turtle model (green) ignores this week's data as an aberration. The real-time forecast (purple), which assumes all real-time estimates are accurate, is barely visible behind the red line. The red line is the composite average of all 3 forecasts.
You'll notice the blue line (wastewater levels) and red line (forecast) overlap marginally. Biobot reports levels for Nov 29, and we carry them forward in the forecasting model through Dec 4 (today). Transmission is accelerating so much, usually this 5 day lag isn't even visible graphically.
2/
This table shows how additional social contacts increase risk today (Dec 4).
🔥10 people (daycare, team meeting) = about a 1 in 4 chance someone has infectious COVID
🔥🔥30 people (large K-12 class) = over a 50% chance someone has infectious COVID
We're entering the 8th pandemic wave, likely surging to >2% infectious (>1 million cases/day) in a month.
Today's numbers:
🔹 1.41% (1 in 71) are infectious
🔹 >670,000 C0VID cases/day
🔹>34,000 #LongCovid cases/day
1/
Note that the different forecasting models show high convergence.
December 11 by the Numbers:
🔹 2.25% (1 in 44) likely to be infectious
🔹 >1 million anticipated C0VID cases/day
🔹>50,000 resulting #LongCovid cases/day
2/
Zooming out to the full #pandemic, there is no debate we're in an 8th U.S. C0VID wave, likely entering a "surge" in my view. That's not a word I take lightly.
There's more transmission than during 54% of pandemic days.
#MaskUp #VaxUp 😷💉
Today's Numbers:
🔹 1.27% (1 in 78) are infectious
🔹 >600,000 C0VID cases/day
🔹>30,000 #LongCovid cases/day
We will pass the late-summer wave's peak in just over a month.
1/
The different forecasting models reach a strikingly similar conclusion about where we'll be in a month: very bad.
November 27 by the Numbers:
🔹 1.76% (1 in 57) are infectious
🔹 >800,000 C0VID cases/day
🔹>40,000 #LongCovid cases/day
2/
Forecasting nuance:
Alt Model #1 (turtle) thinks the current real-time numbers are an underestimate, and it ignores the most recent week's data. Alt Model #2 (cheetah) accounts for recent errors in the real-time numbers; with low error, it maps on very closely to the real-time (red) line. The black line shows the composite used for reporting estimates. Note, everything converges in 4 weeks.
Zooming out to the full pandemic, you can see that we are entering the 8th wave.
Today, there is more transmission than during 50.6% of pandemic days. It's a coin toss as to whether any particular day of the pandemic has had more or less transmission than today. 😷💉 3/