COVID-19 transmission patterns only seem explainable by aerosols
Prof. Jose-Luis Jimenez, Univ of Colorado at Boulder
Version: 1.5; 28-Jul-2020 (I started writing this for Twitter but then decided it’d be too painful to enter and read in that format. The writing is still abbreviated. This seems too important to make it pretty and delay publication. I will add on to this later in response to comments, as needed. Although I think some people knew this already, I have not seen an analysis like this before, let me know if you know of one. Understood that it is “big picture”, but I am just trying to get the ideas across, since there are huge misunderstandings and errors of interpretation that are the basis of the official guidance).
Some additional thoughts on the modes of transmission, which are firming up my confidence on the importance of aerosols. Again hoping for discussion, counter-arguments, other evidence etc. I recommend reading the 11-July-20 Twitter thread first, if you haven’t done so. In terms of why arguments against aerosol transmission are weak, read the MedScape Perspective (which is a more complete and readable version of the 16-Jul-20 JAMA thread).
There are 3 modes of transmission: fomites (touching objects or people directly), aerosols, and ballistic drops. In the 11-July-20 Twitter thread we had already discussed fomites as unlikely to be major, and no major arguments emerged against that (though it is possible, so keep washing your hands). Also a UK SAGE member stated in the discussion that the pattern of infections does not seem to match fomites as being major.
So I’ll focus on aerosol vs ballistic drops (WHO’s “droplets”, image below). Don Milton has published an excellent short paper recently reviewing the terminology.
Aerosols float in the air from tens of seconds to hours, depending on aerosol size and air flows. Ballistic drops fall to the ground in 1-2 m, in a few sec.
One key piece of experimental evidence is the patterns of transmission, so let’s explore whether drops or aerosols are consistent with those. We know that:
(a) Transmission is far more likely indoors than outdoors
(b) A lot of the transmission happens in super-spreader events
(c) Many cases from contract tracing are consistent with “close contact”
(d) Taller people are more likely to contract COVID-19
(e) No long range aerosol transmission has been reported
(r) Transmission patterns in hospitals are not consistent with measles-type disease
(g) Average R0 is ~2-3 with a lot of dispersion (many low values and a few high values)
(If I have missed an important pattern, pls email me with the details and refs.)
Let’s discuss them one by one:
(a) Transmission is far more likely indoors than outdoors.
A lot of transmission happens in extended close contact (ECC). There is ECC both indoors & outdoors. Drops are ballistic, there is no time for dilution or UV to remove the virus, they are little affected by indoors or outdoors, their infectivity should be similar in both cases. Aerosols are carried by the wind, there is incredible dilution outdoors, also more time for UV light to destroy virus (which is very quick). If we run the Skagit choir case in the aerosol transmission estimator (“choir” sheet in https://tinyurl.com/covid-estimator), it reproduces the infection rate. Now move the exact same choir outdoors (“outdoors” sheet), infection drops from 83% to 0.4%.
In the real-world: "The vast majority of transmission seems to be through close contact with an infected individual, primarily in an indoor setting." Or there has been a lack of spikes associated with the Black Lives Matter demonstrations in the US. Only aerosols can explain this.
(b) A lot of the transmission happens in super-spreader events.
Key events can ONLY be explained with aerosols. “Extended close contact” is needed for transmission, per CDC: “within 6 feet of an infected person for at least 15 minutes.” For the Skagit Choir case that we studied, 53 people out of 60 present were infected from the index case in 2.5 h singing. They were aware of COVID-19 and hand-washing recommendations, didn’t shake hands, use hand sanitizer, limited opportunities for fomites (plus fomites unlikely to be major per CDC, see above). Most of the time they were singing in fixed positions, there was nobody within the 2 m landing zone of index patient, they took a couple of 10 min breaks. It is physically impossible for the index patient to have extended close contact with so many people and have enough drops land on them during the breaks. Assuming the index patient was talking to 2 people at a time, that person would have needed 6.6 hrs of break time to have 15 min with each pair to infect 53 people. Privacy limits the release of some info that makes drops even more unlikely, but we can share that info with WHO or CDC. Maybe a few infections could be due to fomite or drops, but the overwhelming majority had to be aerosols. I suspected aerosols when I started working on COVID-19, but I wasn’t sure. The choir case convinced me that it can definitely be transmitted by aerosols, at least in some cases. Similar for other well-studied cases by Yuguo Li, Guangzhou restaurant, buses etc. Or this nursing home in Canada where the ventilation system failed, and every single one of the 226 residents was infected (I am not aware of a published investigation on it). “Contortionist thinking” (B. Nazaroff) is required to explain these w/o aerosols.
(c) many cases from contract tracing are consistent with “close contact”
(c1) The close contact situation
Now we get to the critical point of this thread. Often we hear: “a lot of transmission happens during close contact, which is explained best by drops.” But the logic is flawed, and repetition doesn’t make it correct. Both aerosols and drops are coming out of the mouth and nose of the infected person. Drops can be found ONLY close in front of the infected person. But aerosols are also most concentrated there, in the expiratory plume in front of infected person, and are diluted quickly with distance, given typical indoor wind speeds of 5 cm/s plus some momentum from the exhaled flow. A susceptible person will breathe the highest aerosol dose under close contact, much higher than if the air is diluted onto the entire room. A person breathing out smoke aerosols visualizes this plume. The graph below (Nielsen & Liu, 2020) illustrates the transition from the close contact to the room scale. We deal with the close contact next, and later (in c2) with the room scale.
Detailed physics-based modeling by Yuguo Li, using measured amounts and size distributions of expired particles and very well-established physics, shows that the exposure at close contact when talking is dominated by aerosols. I use talking and not coughing because it is the most relevant for a/presymptomatic transmission in the community. Drops are only competitive at less than 20 cm distance. Typical US conversation happens at 45-90 cm. At 50 cm distance, aerosol exposure is x100 more important. At 1 m distance, aerosols are over x2000 more important. This figure shows the exposure to drops (red) and aerosols (black) as a function of distance from the speaker. Note the logarithmic Y scale.
This is solid work with realistic assumptions. There are uncertainties, but not the x100-2000 that would be needed to make drops competitive.
Influenza virus has been shown to be most concentrated in smaller particles (Yan et al., 2018), which would favor the aerosol route even more. This is thought to be due to the bubble bursting-like mechanism of respiratory particle formation, similar to what is observed for the ocean surface (e.g. Kim Prather’s work).
So if close contact dominates, aerosols likely dominate! Being generous by a factor of 100-2000, aerosols are still competitive with drops. But definitely aerosols cannot be discarded as negligible based on the fact that many infections happen at close contact.
(c2) Consistency with the room scale transmission
Now we transition to the room scale. The expired flow is eventually mixed into the room by air currents, depending on ventilation and thermal gradients. This leads to a lot of dilution. An example is smoke diluting into a room, as in the picture below. So for a virus that is infective through aerosols, but much much less than measles, it is a challenge to build up a concentration in the room which is comparable to the concentration in the breathing area of the close contact situation. But is it possible, and is it consistent with the data?
How much larger is the dilution of aerosols, compared to the close contact situation?
Let’s do a simple order-of-magnitude estimate comparison of the dilution in the close contact situation (that we know leads to infection w/ people talking) and the room situation (that we know led to widespread infection for the choir). The schematic below from Yuguo Li’s paper gives us an idea of the dimensions.
Let’s assume that the expiratory flow (0.6 m3 / h when talking) is diluted into an approx. 35 cm diameter cylinder that contains most of the exhaled breath. Assuming an indoor air speed of 0.1 m/s and a distance of 1 m, the exhaled air will spend ~10 seconds in that volume. So that’s a dilution rate of ~60. We know that can lead to infection. So inhaling the exhaled breath of an infected person, diluted x60, for 15 min (per CDC) often leads to infection. Maybe easier to understand, the susceptible person is inhaling 0.15 m3 (150 liters) of air in 15 min. of close contact, and that volume is 59 parts “clean air” and 1 part exhaled air from the infected person. So then we can estimate the infectious dose as inhaling 2.5 liters of undiluted exhaled breath from the infected person.
Now let’s do the same calculation for the Skagit choir. Assuming an 800 m3 room, 1.5 h duration (less than the actual duration,