As a family physician on the bucolic shores of the Big Island of Hawaii, I have largely been watching the Covid-19 pandemic evolve from the sidelines. With 145 cases total here in mid-August since the pandemic began, we probably have one of the lowest per capita rates of disease of any county in the U.S. That said, like most physicians, I have been watching closely. What most caught my interest back in April were the reports from some of the seemingly worst places to catch Covid-19 — prisons and meat processing plants — showing that well over 90% of those testing positive were displaying no symptoms whatsoever. How could this reconcile with a disease we were told had only a 40% asymptomatic rate? The question literally kept me up some nights; and finally I decided that it must have to do with “cafeteria settings”: all these places shared the common feature of large congregate gatherings in big spaces with shared air.
Like many of the theories that have been cooked up by non-epidemiologists in the past six months, this one had holes in it. That was made clear to me last month in a disturbing outbreak in my home state. An airline training, with reports suggesting a median participant age in the 30s, held in a very large room, with appropriately distanced desks, and masks used during interactive sessions, resulted in at least 6 of the 8 positive cases being significantly symptomatic, with 1 of the 8 students hospitalized, and another dying. Why did this congregate setting lead to such severe disease?
Studying outliers can get an amateur statistician into trouble. However, when apparent outliers are not anomalies but rather quite common, it is time to adjust our perspective. We have been able to see for months now that some settings, despite being crowded with high-risk populations, regularly lead to largely asymptomatic outbreaks — while others lead to severe disease clusters. I did not see a single report in the mainstream media on the potential importance of this concept until just this past week, when on August 8 the Washington Post ran the article, “Forty percent of people with coronavirus infections have no symptoms. Might they be the key to ending the pandemic?”
While I part company with the conclusions of this article, I fully agree that studying how asymptomatic outbreaks differ from severe outbreaks might point us towards the most important ways we can soften future waves of this pandemic. Covid-19 cases are going to happen. Manipulating them toward being mild cases could be our goal. Understanding the environments that have been associated with the highest rates of asymptomatic Covid-19 could allow Americans to gather together to work and study without the level of illness and death we have seen thus far.
This might sound like common sense, but actually it is a long walk from the conventional wisdom about this virus. To get there, we need to agree on some points:
Assumption #1: Asymptomatic Covid-19 is more than just bad epidemiology.
Let’s get our terms straight. For our purposes, a truly asymptomatic patient never develops any Covid-19 symptoms at all. A presymptomatic patient has no symptoms at the time of initial inquiry (i.e., a large-scale testing or a contact tracing) but later goes on to develop symptoms. A good study follows a population for a couple weeks to make this distinction. Presymptomatic people fill hospital beds and morgues just like symptomatic people; they might be grouped in with asymptomatic people by epidemiologists focused on viral transmission, but in studying outcomes they need to be differentiated from the truly asymptomatic. The better the data collection for Covid-19 outbreaks in this regard, the stronger the conclusions we can draw.
Assumption #2: Asymptomatic Covid-19 is neither very common nor very rare.
That data seems to be telling us that asymptomatic cases of covid-19 are a normal part of the disease spectrum. At the beginning of this pandemic, reports of overwhelmed hospitals in Wuhan and northern Italy were so terrifying that we rather lost track of the frequency of asymptomatic disease. However, even early in the pandemic, when entire towns (like Vo, Italy), or random populations (as done in Iceland, albeit without follow-up), were tested, an unexpected picture emerged: perhaps 40-45% of SARS-CoV-2 infections presented entirely without symptoms. The CDC pegs this figure at 40%. More recent data from a large study of over 5000 Covid-19 cases in northern Italy suggested that number might be closer to 70%, although with the caveat that only fever and respiratory symptoms were counted, and we know that some significant minority of Covid-19 infections have different symptoms.
Unsurprisingly, you can manipulate the asymptomatic rate by changing the population you study. Multiple reports have shown that younger, healthier people are more likely to have asymptomatic Covid-19. In the Mission District of San Francisco testing, for instance, people with medical comorbidities were exactly half as likely (38% vs 76%) to have asymptomatic Covid-19 than those without comorbidities. That study out of Italy showed an asymptomatic rate falling as age increased, from around 80% asymptomatic cases under 40 years old, to 70% in the 40s and 50s, all the way down to 35% for those over 80; and critically ill cases increased with age group as the asymptomatic rate fell.
A woman gestures, left, while undergoing a COVID-19 test at Garfield Square Monday, April 27, 2020, in the Mission District of San Francisco. (AP Photo/Ben Margot)
(AP Photo/Ben Margot)
The bottom line is that we could reasonably expect about half of a given randomly tested population to be asymptomatic. If the population is a nursing home, we would expect the number to be far lower; a summer camp, far higher. When we start to see substantial exceptions to this rule, it’s time to ask: “What was different about this exposure?”
Assumption #3: Aerosolized spread is important for SARS-CoV-2 transmission.
We emit a mix of respiratory droplets, from tiny to large, every time we cough, sneeze, speak, or even breathe. The tiny ones are termed “aerosols.” Somewhere around 5 microns and smaller — small enough that hundreds can fit on the proverbial head of a pin — can stay in the air for prolonged periods of time and float like bits of dust in the sunshine. Larger particles, termed “droplets,” tend to settle down to the ground swiftly, especially those bigger than 50 or 100 microns. Research has shifted from indicating most spread occurs via large droplets and direct contact, like getting coughed on or touching a door knob, to the belief that at least some substantial proportion of SARS-CoV-2 transmission is aerosol-based, and probably most super-spreader events are, as well. This became so clearly evident to a group of prominent scientists that they wrote a well-circulated letter to the WHO, objecting to its glacial speed in recognizing this likelihood.
This question is still hotly contested, but I think the right side of the argument is finally gaining ground, as evidenced by three pieces recently run on the same day exhorting us to pay attention to aerosol spread. This Atlantic article does a good job summarizing the arguments on both sides. I will say this: some of the apparently maddening inconsistencies of this pandemic make a lot more sense when you accept that Covid-19 is a disease spread by the few to the many — perhaps 10-20% of people lead to 80% of cases — and the most obvious means of major transmission events with few infected people spreading the virus to many has to be aerosolized spread.
Assumption #4: All aerosols are not created equal.
Less controversial within this complicated realm is that some aerosols are going to be more virulent than others. Different sized aerosols host different amounts of viral particles (the assumption being that larger aerosols usually carry more virus), and the qualities of these aerosols can affect the infectiousness of the viral particles aboard (for instance, an evaporating aerosolized droplet might have too high a salt concentration for the virus to stay active.) Temperature and humidity affect the viability of those viral particles, although again in complex and variable ways. Put simply, the precise environmental conditions in which you are exposed to SARS-CoV-2 are likely to greatly influence the types and numbers of aerosols you inhale, and thereby the severity of your exposure.
Assumption #5: More virus, worse disease.
There is general consensus in the infectious disease community that larger SARS-CoV-2 exposures lead to more severe disease. It stands to reason that a massive exposure is both more likely to overwhelm the innate immune responses and to infect more cells initially, leading to a more rapidly growing viral load. The common sense flip side to this is that milder exposures probably lead to less severe disease, and the ultimate version of “less severe disease” is an asymptomatic case. If you have to get exposed to a virus, “less severe” is good.
Assumption #6: Location, location, location — where a virus goes affects host immune response.
Smaller SARS-CoV-2 particles are thought to evade the upper airway and its defenses when inhaled, settling directly into the lungs. Larger droplets in direct contact type exposures get trapped in the upper respiratory tract and have to run the gamut of its many immune defenses before they can begin a possible journey into the lungs. (This figure nicely depicts the concept for visual learners.) Now, it’s understandable to think that direct inhalation into the lungs would be a bad thing – we don’t want this virus causing lung disease and putting people into ICUs on ventilators!
However, a couple of studies tell an interesting story. A small but meticulously detailed study followed 9 PCR-positive Covid-19 patients for 2 weeks while quarantined in Hong Kong. Despite a median age of 58, 6 of the 9 were completely asymptomatic — and half of those six had the classic ground glass appearance to their chest CT that identifies a viral pneumonia! A larger study of 37 asymptomatic covid-19 patients followed in a Chinese hospital showed similar results: totally asymptomatic, but 57% had abnormal lung CTs consistent with a viral pneumonia. What is the most logical explanation for a virus evading host immune response until it settled into the lungs? Direct inhalation of tiny aerosolized SARS-CoV-2 particles.
Assumption #7: Some crowded places make for mild disease.
This was, for me, the real eye-opener as April went into May. We started getting news of huge clusters of Covid-19 in congregate settings after large-scale testing in places like prisons, meat processing plants, and cruise ships. That congregate settings lead to clusters was no surprise. They tend to be crowded. People are forced to interact with each other in close proximity. Large numbers increase the odds that there will be an infected “super-emitter” present — those people with Covid-19 who spew a lot of virus and spread it to many others. The basic equation for exponential spread of Covid-19 appears to be, super-emitter + active covid-infection + time with large group of people in enclosed space = super spreader event, so this was not surprising. What was surprising was that so many of these congregate setting outbreaks involved an asymptomatic rate far, far higher than the age and/or comorbidities of the affected population might have predicted.
Many of these reported clusters had somewhat limited data quality. Many did not provide follow-up to confirm the asymptomatic cases were not simply presymptomatic, like the slightly over 90% of the positives that were asymptomatic at the St Joseph, MO, meat processing plant; the 95% at an Oregon seafood processing plant; the prison populations from a four state review that found an asymptomatic rate among 3277 covid-19 positive inmates to be 96%. Sometimes they were incomplete, as in the Tyson meat processing plants in northwest Arkansas which described a 95% asymptomatic rate amongst 481 positives from whole facility screening, but left out the status of the 212 workers who were diagnosed with covid-19 outside the facility and likely had a much higher symptomatic rate. While these reports are tantalizing, holes in the data make it hard to draw reliable conclusions from them.
FILE – In this May 7, 2020, file photo, workers leave the Tyson Foods pork processing plant in Logansport, Ind. Federal recommendations meant to keep meatpacking workers safe as they return to plants that were shuttered by the coronavirus have little enforcement muscle behind them, fueling anxiety that working conditions could put employees’ lives at risk. Major meatpackers JBS, Smithfield and Tyson have said worker safety is their highest priority. (AP Photo/Michael Conroy, File)
(AP Photo/Michael Conroy, File)
However, several reports have provided more exact figures and included follow-up of the positive cases, and had asymptomatic numbers in accord with these other mass asymptomatic outbreaks. A cruise to Antarctica that tested everyone on board found an 81% rate of asymptomatic positives at the end of their quarantine; a surprising rate given that Antarctic cruises have an average age well above a typical population. In the infamous Diamond Princess cruise, carrying an even older population with a median age in the 60s, the initial reports of a 52% asymptomatic rate were revised upwards to 74% in a pre-print study modeling all the probably missed asymptomatic cases in the early phase of limited testing. In two correctional facilities that followed prisoners for two weeks after PCR testing, Montgomery County (PA) prison posted a 95% asymptomatic figure, and 99% of positive inmates at the Bledsoe County (TN) Correctional Complex never developed symptoms.
Why are these numbers so striking? If I were cherry-picking this data out of hundreds of reports that included the asymptomatic proportion, it could just be random chance that a handful were so high – but these are some of the only reports we have on asymptomatic rates from Covid-19 clusters in congregate settings. If the average rate of asymptomatic carriage of this virus is really anywhere around 50% in a typical population, the odds of random chance leading to hitting the 99% asymptomatic mark would be like flipping 99 heads out of 100 coin tosses! The question is: why would congregate spread be so different than community spread?
Assumption #8: Virus transmission style varies by setting.
Think about household spread of a virus: lots of really close, prolonged, contact, and many shared surfaces. Then think about typical community spread: shaking hands with business associates, paying the cashier for your purchases, taking meals with friends. Both settings favor large droplet, direct contact type of transmission, although sharing of close range aerosols can still be common. On the other hand, think of congregate settings, whether cruises, ski lodges, churches, prisons, large restaurants, or huge factories. There is still plenty of hand-shaking and close contact, but with a limited proportion of the total group. However, there is a lot of time spent being around other people more so than closely interacting with them. Recall the “super-emitters” who release disproportionate amounts of viral particles when they sing, talk, or even breathe, in aerosol clouds that might stretch a couple dozen feet. In congregate settings, we have the opportunity to breathe the same air as dozens, or hundreds, of people in close proximity at the same time. This is the first chapter in any “How to” guide of starting a Covid-19 cluster. The most obvious difference between community spread and congregate spread is that, with the latter, it’s usually in the air.
Assumption #9: The exceptions are just as important as the rule.
It’s tempting to absorb this data and conclude, as I had last month, that congregate settings are probably, all in all, pretty safe places to get Covid-19 — maybe not a cramped New York City bar, where the ratio of patron lung volume:room air volume gets distressingly high, but more like a cruise ship cafeteria, or a church, or a factory. In theory, a super-emitter might breathe out a good bit of viral particles in that setting, but they would be diluted within a large indoor space, and most people would have a nice, low inoculum that could predispose to an asymptomatic case of Covid-19.
The nightmare scenario of that airline training in Hawaii was my reminder that gathering people in a spacious venue was not enough to avoid severe disease. It also mirrored the lessons of a much larger, formal report of one of the first published Covid-19 clusters, the oft-discussed Seoul, South Korea office outbreak. In that super-spreader event, of the nearly 100 office workers, mean age of 38, who tested positive in a mass screening and were quarantined and observed for 2 weeks, 96% were found to be symptomatic. Again, the deviation from an expected mean of roughly 50% symptomatic disease all the way up to 96% is striking – we are talking about flipping an awful lot of tails in our 100 coin tosses!
Another “exception to the rule” of mild disease in congregate settings is the Georgia summer camp outbreak in June recently reported by the CDC. Of the 160 campers and staff testing positive for whom there was data on symptomatology, 74% were symptomatic. Contrast that to the 18% symptomatic rate in that northern Italian study in those under 20 years of age.
Understanding what might explain these differences is crucial. For ethical reasons, we’re just not going to get a randomized, controlled human trial exploring different building environments and the disease severity they cause. We should not dismiss the importance of what we already know: 96% of workers in a South Korea office got ill from their exposure to this virus, while 90+% of meat processing workers had no symptoms. We don’t have any real suggestion the SARS-CoV-2 in meat processing plants had mutated to a benign variant. It’s also a huge reach to claim that this part of Asia had missed out on a hypothetical mild cold-type coronavirus outbreak to bolster citizens’ immune response; or that South Koreans have some terrible genetic weakness to covid-19 — after all, South Korea has had an exponentially milder covid-19 epidemic than the US. We don’t have reason to think South Korean call center workers are older or less healthy than American meat processing plant workers. We just are left with this massive disparity: for some reason, the people in that call center were roughly 10X more likely to develop symptomatic disease from their exposure than the meat processing plant workers in the US, and almost 25X more likely than our examples in US prisons. Again – why?
Assumption #10: It’s all about the air we breathe.
If we can accept that the populations experiencing these mild outbreaks were not somehow inherently gifted at fending off the SARS-CoV-2 virus, and that some benign mutated strain was not magically affecting prisons and meat processing plants while the communities around them were suffering like everywhere else, then we are left with a different explanation: there must be some unique factors of the exposure itself that can explain these discrepancies.
That camp in Georgia promises air conditioning in all of its cabins – and was dinged by the CDC for not “opening windows and doors for increased ventilation in buildings.” The tragic airline training in Hawaii? An enclosed air conditioned room by the Honolulu airport tarmac. Reports of the South Korea call center office space mention an “evaluation of ventilation system” after the (presumably sealed) office building was temporarily closed.
It’s easy to look at poorly ventilated buildings or air-conditioned rooms and say, “These environments probably favor the spread of Covid-19.” However, we might acknowledge the nuances of how this virus behaves in different environments, and ask instead, “What kind of Covid-19 do they tend to spread?”
While air conditioning has drawn a lot of attention as a possible contributor to Covid-19 transmission, there is probably a huge range in the likelihood of any given HVAC (“Heating Ventilation Air Conditioning”) system leading to severe disease spread. Filters very likely play a role – and high-level HEPA filtration might be why there have been so few reports of airline clusters despite the concerning pre-conditions of crowded indoor space shared for prolonged periods of time. There are many factors in play besides filters, though. Venting in more outdoor air almost certainly helps disperse virus-laden aerosols. UV lights might be effective at inactivating SARS-CoV-2. Ceiling height, outside relative humidity and temperature, fan use, vent locations — there are so many variables which could, in theory, be the difference between a fatal cluster and a largely asymptomatic one.
Assumption #11: Some of these assumptions will need updating.
That August 8 Washington Post article highlighted the recently published work of Dr. Monica Ganhdhi and colleagues — suggesting an association between mask use and asymptomatic disease — in its exploration of the possibility that asymptomatic Covid-19 cases might hold essential clues towards limiting the damage caused by this virus. Their idea strikes me as plausible; mask-wearing certainly could alter the way viral particles are inhaled into our respiratory tracts. However, I suspect irregular or absent mask use in the highly asymptomatic places like prisons and meat processing plants. In the example of the Antarctic cruise ship, while true that the published case report mentions that all passengers were given masks, it was not until the 8th day of the cruise, at which point passengers were confined to their cabins; I would expect that most spread had already happened by this point, and had been mask-less. However, what matters is that data collection for events with dramatic rates of symptomatic or asymptomatic disease is commencing, and that it be shared rapidly with public health decision makers.
Surely, with a team of public health workers at my disposal — and I realize I might be descending into fantasy now given our current state of affairs — I would be interrogating every one of these outlier outbreaks. The goal would be to inquire into the true severity of their outbreaks, and to learn every possible detail about the building HVAC systems in which they took place: filter size, outside air ventilation, temperature, humidity, heating type, use of A/C, fans, and so on (and, yes, mask use, too). This data can be run through regression models to determine which variables are most closely related to the disease severity of the outbreaks. Given what we have seen so far, I would expect to find that certain ventilation systems strongly correlate with better outcomes than others. In other words, combine a contagious super-emitter with the right indoor environment and we expect to see a lot of mild or asymptomatic disease as a result; combine that same super-emitter with the wrong environment and we see hospitalizations and death.
At this point in the U.S. pandemic, there is virtually no appetite left for lock-downs, although some epidemiologists continue to call for them. An effective vaccine for Covid-19 is probably many months away, and that might be optimistic. Available treatments have only made modest inroads into limiting mortality. As the nation gears up for the opening of schools and yet more inter-mingling, our best option left might be to admit that some amount of spread is inevitable, and aim to minimize its health consequences. The easiest route to that end? Learn the lessons of asymptomatic Covid-19 — let people gather, but in the right places for mild disease.