# New insights: How much does air exchange really matter for Covid-19 transmission?

### A lot. But super-shedding needs to be considered more carefully.

SARS-CoV-2 hangs out in the air.

It’s been a while since I thought about the math around this, but a new “study of studies” inspired some thinking that I want to share with you.

#### The approximate math on air exchange.

Let’s work some numbers to discover why air exchange (indoor ventilation) can make a huge difference on whether someone catches Covid-19 or not.

We’ll consider two scenarios. In each case, a person with Covid-19 is hanging out a small office (1,000 cubic feet). Each person exhales 15,000 infectious particles per minute. They are there for an hour. You enter the room at minute 59. (Are we having fun yet?)

Room #1: The only turnover in the air comes from the door opening and closing (negligible).

Room #2: A ventilation system effectively replaces the air six times per hour (i.e., every 10 minutes). (For the sake of simplicity, we’ll imagine that the exchange happens all at once, exactly every 10 minutes.)

Clearly, we all want to enter Room #2. In Room #1, there are 885,000 (59 x 15,000) infectious particles by 59 minutes, thanks to our Covid-positive friend. But in Room #2, by 59 minutes there have been 6 air exchanges. Because the last exchange was completed 50 minutes into the hour, only 135,000 infectious particles are in the room by minute 59. That’s 85% less compared to the “bad” room.

Now let’s say that in order to have a 95% chance of getting infected, you need to inhale 1,500 viral particles. With regular breathing (0.5 liters per breath, 20 breaths per minute), it would take under 5 minutes to inhale the infectious dose in Room #1. After 10 minutes, you’d have inhaled double the infectious dose of viral particles. In Room #2, it would take 31.5 minutes to reach the infectious dose. After 10 minutes, you’d have inhaled less than 1/3 of the infectious dose. (The real math on this is more random than this, but I’m making it simple for us.)

Thought experiments like these are the underpinning of the movement to increase indoor air quality.

#### A study of studies shows that isolation and air turnover matter.

A new study of studies looked at Covid-19 transmission in hospitals and in homes. No surprise, SARS-CoV-2 concentration in the air was inversely related to the number of air exchanges per hour.

On top of that, there were **three key findings **from the 93 studies that this paper combined**:**

In

*hospitals*, the air in rooms where an infected person spent most of their time (their “primary” room) had higher concentration of viral particles than secondary rooms (other rooms they spent time in).In

*homes*, the viral particle concentrations in primary and secondary rooms were similar.*Some types of places have better air exchange than others*. In general, homes had fewer air changes per hour (0.35 per hour) compared to retail spaces and classrooms (~2 per hour). Gyms (~5 per hour) were about as bad as intensive care units (but better than stores and schools). Hospital-based infection isolation rooms were the best, at 12 air exchanges per hour.

Because hospitals had so much better air turnover, the amount of virus in the air was generally *lower *than homes where someone with Covid-19 was isolating. That means that living with someone with Covid-19 is actually more likely to get you infected than visiting a Covid-19 patient in the hospital on a per minute basis.

That said, **I take issue with the authors’ conclusion that primary and secondary rooms at home were basically similar in terms of Covid-19 risk**. The issue comes down to the distinction between *means and medians.*

*Aside on means and medians, if you want.*

(For those of you who just had a low-key panic attack, you just may have PTSD from math class. But I promise this is not as painful as you remember. Math is actually cool.)

The mean is the average value. Add up all the values and divide by the number of data points. The mean of 5, 10, 12, 18, and 30 is 15 (75 divided by 5 equals 15).

The median is the

*middle*value in the set. Half of the values are smaller, half are larger. So, the median of 5, 10, 12, 18, and 30 is 12 (~~5~~,~~10~~,**12**,~~18~~,~~30~~).Notice that if I swap out 30 and replace it with 3,000, the mean goes up to 609 (3,045 divided by 5 equals 609) but the median stays the same (

~~5~~,~~10~~,**12**,~~18~~,~~3,000~~)!

Turns out that in both homes and hospitals, the mean detected coronavirus concentrations in the air *far, far, far, far *exceeded the median in the primary isolation rooms. In other words, there were a few measurements that were so off-the-charts high, that they pulled the average way up. Of course, as you just saw above in the “*aside,”* medians can be “resistant” to change resulting from massive outliers.

Why does this matter? The means were pulled up by “super-shedders.” Each person with Covid-19 likely has a peak “super-shedding” window, perhaps lasting somewhere between 6-72 hours depending on the case. During that peak, the amount of virus exhaled is not merely double of triple, or even quadruple; it might be 100, 1000, 10,000, or even 100,000 times greater than the other days of infection. For many people, 99% of the virus they will *ever exhale *probably happens in a one or two day period.

To demonstrate how this works, I made a hypothetical graph showing the amount of virus that eleven Covid-19 patients exhale in a day (organized from the least contagious on the left to the most contagious on the right). In this example, the individual #6 is the median (5 patients have lower viral loads, 5 have higher), coming in at 600 units per day. But because of individual #11 being in super-shedding mode (that one person accounts for 66.6% of all the virus exhaled by the entire group), the mean is 1,409. Ten of eleven people are shedding *below the mean*. If you think about this, you realize how much super-shedding probably matters.

Let’s say that a person must exhale 1,500 units per day to be very contagious during a brief interaction. If so, it’s likely that 99% of the spread within our eleven-person cohort is coming from *just one person* on the day these measurements were taken.

That means if we want to limit Covid-19 spread in the future, we might be able to isolate people *just when they are shedding like crazy—like individual #11. *The rest of the time, it might be okay to let people go out of their homes (in good masks, low density settings, and certainly not in prolonged contact with severely immunocompromised people; the implied social contract will require people to be careful with things like public transit).

Can we identify the people who really need to isolate for a few days? Yes, yes we can.

#### Putting it all together.

We’ve known much of this for a long time. Better air exchange was already known to be associated with fewer sick days in schools well before the Covid-19 pandemic! And we’ve long understood that super-shedding is a massive driver of Covid-19 spread.

The fantasy—which we could achieve today, given what we know—is a future in which isolation periods will be shorter but more intense. I believe we can decrease 90%-99% of flu, RSV, and Covid-19 transmission with a combination of upgraded air exchange/ventilation systems, semi-quantitative rapid tests that indicate if/when you are super-shedding, and making sure people have access to high-quality masks for when they are outside of their super-shedding window. Ideally, we’d live in a world in which people isolated for their whole illness. But in the one we live in, I know we can do better than the new guidelines being considered by the CDC.

While rapid tests and good masks are keys to this, improved air exchange in high-risk settings (like schools and stores) and high-*stakes* settings (healthcare) has an advantage: it requires no effort from individuals.

Yes, we’ve got a theme this week at *Inside Medicine*: Friction-free public health.

*Questions? Ideas? Math corrections? Please contribute in the Comments section and join the conversation!*

***Hi all***

Just a quick note. I’m working clinically today so I probably won’t have time to respond to you comments and questions until this evening. But keep them coming and I’ll be back later! -Jeremy

Can you please cite any studies you've found that investigate the correlation between antigen test positivity and infectivity? Do they really capture the super-spreading window or do they lag? Does "faintly" positive indicate low infectivity?

We know people are infectious when presymptomatic and symptomatic, and we know that the antigen tests are now often taking 3-5 days after symptoms begin to turn positive, so I would be concerned about missing that window if we rely on antigen tests.

When we really want to know, we use a home NAAT test made by CUE. What are your thoughts on the superiority of these and other highly sensitive tests to predict ihow contagious someone is?

Thank you!