Covid-19 May Take Longer Than 14 Days to Spread
As Covid-19 spreads worldwide, public health authorities have reconsidered nearly everything about the virus. Changes in policy have addressed everything from basics of how it spreads, to the value of protection measures like face masks. One thing that largely hasn’t changed, though, is guidance on Covid-19’s incubation period and the time course of its spread. Both the Centers for Disease Control and Prevention and the World Health Organization estimate this number at 5–14 days.
Their figures, however, are based on a variety of assumptions — as well as data which is often distressingly thin and biased. New data and empirical observations are beginning to suggest that the time course of Covid-19’s spread, and perhaps the virus' incubation period itself, may be much longer in certain populations and situations— perhaps up to 30 days. If that’s the case, many of the models used to plan reopening, monitor disease progress and coordinate testing could be dangerously wrong.
As the Covid-19 virus began to spread in the United States in early 2020, authorities quickly worked to determine its incubation period. Even today, both the CDC and WHO prominently cite a single study by Laurer et al., originally published near the start of the virus’ spread in March of 2020, titled The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application.
In fact, almost everyone cites this study. According to Pubmed, it has been cited over 325 times since its publication — and that’s just in the official scientific literature. The study, though, was not actually a study at all. It wasn’t based on patient data gathered by its authors, but rather on “News reports and press releases from 50 provinces, regions, and countries outside Wuhan, Hubei province, China.”
Already, that should already raise red flags: public health authorities should be basing their recommendations on carefully gathered and vetted scientific data, not on early reports cobbled together from Chinese newspaper articles. Initial data like this was fine in the first days of the pandemic. Months in — with hundreds of thousands of cases in the United States alone — it’s a totally insufficient source to back up such an important statistic.
Deeper analyses of the virus’ incubation period have turned up concerning gaps, too. A literature review of incubation period data released in May (currently a pre-print) looked at over 44,000 articles published about the coronavirus. The authors found only 29 reputable articles, of which only 18 were deemed worthy of further review. One was the Lauer study. Others had tiny sample sizes of less than 100 patients. And all were conducted in the early stages of the pandemic, when comparatively little was known about the virus.
Even as health authorities confidently proclaimed that Covid-19 has a maximum incubation period of 14 days, a trickle of reports provided contradictory evidence. Journalists for Reuters, for example, reported on a case from China that appeared to have a 27 day incubation period. A study from China early in the pandemic found that the range was 0 to 24 days (the WHO dismissed this, saying it was due to second exposures). And a recent report shared the story of a fishing crew, who became infected with the virus after 35 days at sea.
Perhaps the most compelling suggestion of a longer time course for Covid-19 infections, though, comes from empirical data around reopenings in the United States. When Georgia and Arizona announced that they would largely reopen their economies around May 1st, 2020, many experts predicted an unmitigated disaster. The Atlantic called Georgia’s reopening an “experiment in human sacrifice.”
When weeks later the expected surge in cases had failed to materialize, many commentators began to declare victory. Fox News called the reopening a “success”, and even CNN published a hopeful article about businesses in Georgia beginning to return to normal operations. Other states, including my home state of California, began their own reopening processes in mid to late May, perhaps emboldened by mostly positive data from neighboring states.
But the surge and ensuing bloodbath came. It just came later than expected. By June 1st — about 30 days after Georgia’s shelter in place order expired, and Arizona began to lift its own restrictions — both states saw a massive spike in new cases. As I write this, both states have been forced to roll back aspects of their reopenings, and Georgia is running dangerously low on ICU beds to treat sick patients.
Nearly everywhere that states have reopened in America, the same pattern has emerged. There has been about 30 days of relative calm, followed by an exponential rise in new cases of the virus. This empirical observation — repeated in several places around the US — supports the idea of an incubation period longer than 14 days, or some other statistical or psychological factor that’s increasing the time course of the virus’ spread, and needs to be better accounted for.
Why would Covid-19 have a longer incubation period or time course than was originally projected? There are a variety of reasons — some medical, some mathematical, and some behavioral--and experts are divided on which is correct.
For one, new data has emerged about the virus itself. While earlier coronaviruses like SARS-COV were primarily respiratory viruses, mounting evidence suggests that Covid-19 may actually be a vascular disease with respiratory components. If the virus begins in the blood stream — or another reservoir in the body, like fat tissue — and doesn’t spread to the respiratory system until later in the disease course, it might not be picked up initially by current testing, which largely relies on nasal swabs.
Patients could be exposed on day 0, harbor the virus in their body for weeks, and not begin to show the classic respiratory symptoms (cough, sore throat, etc.) until later in their disease. If they were tested 7–14 days after exposure, they might see a negative result, because the virus hadn’t yet spread to their respiratory system, and might not appear on a nasal swab. But after another few weeks, they might begin to show respiratory symptoms and positive results upon nasal swab testing.
It’s also possible that the disease incubates at different rates in different populations. A pre-print article from April 2020 raises this concerning possibility. According to the authors, who looked at cases in 21 cities in China, older people can have a much longer incubation period than the official average.
Crucially, the authors were unable to establish a maximum incubation period for the population they evaluated. Even 17 days (the maximum period studied) may not have been the longest incubation period.
As philosopher Nassim Nicholas Taleb famously said, “Never cross a river that’s on average four feet deep.” Average numbers for the virus’ incubation period may miss populations or individuals whose symptoms take much longer to appear. The CDC’s confident assertion that “97.5% of persons with COVID-19 who develop symptoms will do so within 11.5 days of SARS-CoV-2 infection” is based on a single study (Lauer, again) on a biased sample of travelers. It presumably uses a bell curve to project from this limited data to make generalizations about the whole population. This methodology should not be trusted.
As Business Insider points out in a recent article, differences in the virus' spread like these may be compounded by reporting delays, as well as the effects of increased testing. As the article points out, “as coronavirus cases are detected much earlier in the course of an illness, there may be an even longer lag” between infection, testing, symptoms and deaths than in earlier states of the pandemic.
Earlier testing largely focused on patients who were very ill. They may already have been far along in their disease course, leading to an underestimation of how long they had been sick, and a corresponding underestimation of their incubation period. This phenomenon is known as “lead time bias”.
According to epidemiologist Eleanor Murray, “When you start identifying people at earlier stages of a disease, it looks like they survive longer (or have the disease longer) compared to when you identify based on severe symptoms.” Again, this could have lead to an underestimation of Covid-19’s incubation period, based on early testing data.
There could be psychological factors at play, too. Societies could reopen, and initially citizens could be diligent about measures like wearing masks and social distancing. As infections fail to spike after the first 14 day period, they could start to become lax and abandon these protection measures, causing a spike a week or two later.
Data from polling by Gallup confirms that Americans relaxed their social distancing practices in May and June. But rates of mask wearing (thought to be among the most effective ways to prevent Covid-19’s spread) remained high and unchanged during the same period. Behaviors may factor into patterns of coronavirus’ spread. But they don’t appear to account for all aspects of it.
It’s also possible that cases indeed begin to increase immediately after reopenings, but that they don’t show a measurable spike until later. The first part of an exponential growth curve looks like a flat line; periods of apparent calm may actually represent the first part of a trend towards uncontrolled spread.
Epidemiologist Justin Lessler of Johns Hopkins, a co-author on the Lauer study, supports this explanation for the long delays between reopenings and spikes in Covid-19 cases. He acknowledges that “There are a lot of potential biases in the early estimates” of incubation periods, but “If a generation takes 6 days, we will see 150 cases at day 6, 225 day 12…1139 day 30. So it takes a bit for a bi[g] resurgence.”
Any — or all — of these explanations could be correct. But for whatever reason, if we did get the timing wrong on Covid-19’s spread, what impact would this have? A massive one. Essentially, we’d have to throw out — or radically modify — much of what we know about the virus, and many of our predictions around reopening economies (at least in the US).
Let’s begin with quarantine periods. The CDC currently recommends a quarantine period of 14 days from a confirmed Covid-19 exposure. If the disease can actually take longer than 14 days to develop in a substantial number of patients, this period wouldn’t be sufficient. Someone could contract the disease, enter quarantine, leave it after 14 days, and still develop symptoms days or weeks later.
Timelines for reopening would have to be reconsidered, too. When Georgia and Arizona reopened in early May and cases failed to spike after 14 days, many states started to announce their own reopenings for later in the month. Again, my own state of California began our reopening process about 2-3 weeks after Georgia.
Many of these reopening plans are based on observing case counts over successive 14-day periods. The term “14-day” appears in the CDC’s official document on reopening America over 21 times. While authorities are often cagey about their specific reasons for making recommendations, this 14-day observation window is likely based on the disease’s official median 4–5 day incubation period, plus a “8- to 10-day lag between the date of symptom onset and the date the case is reported” as described in the CDC’s document.
If the incubation period of the disease is longer than the official 4–5 day median value, or cases don’t appear until later due to other factors, these 14-day observation windows would be far too short. Following the CDC’s recommendations would cause states to reopen much too quickly, miss delayed spikes in new cases, and have to roll back their plans. Indeed, that’s exactly what we’ve observed in the real world.
If patterns from other states hold up, then one region in particular may have major challenges ahead: New York. The state entered Phase 2 reopenings beginning in early June, but its most populace city, New York City, did not enter this later phase until closer to the end of the month.
As I write this in mid July, articles have begun to appear heralding the state and the city’s success against the virus. These may be premature. If New York follows the patterns of Georgia, California and numerous other states, then the impact of reopenings in New York wouldn’t be felt until late July. And the impact of New York City’s reopening wouldn’t begin to be felt until the end of the month, and wouldn’t truly spike until into early August.
For reopenings to be sustainable worldwide — and for testing to play a larger role in containing the disease going forward — we need to have a better sense for its incubation period, and the overall time course of its spread. If the time between exposure and symptoms is longer than 14 days — even for a small subset of the population — this could have major ramifications for all aspects of managing the pandemic, from quarantine periods, to testing recommendations, to reopening timelines.
Scientists should delve into the virus’ incubation period using new data, rather than relying on preliminary findings from early in the pandemic. We should also consider other methods of testing for the virus. If Covid-19 is indeed a blood disorder, it’s possible that earlier blood tests for active infection could detect the disease before it appears in nasal swab tests.
Public health authorities should also stop using 14-day windows to evaluate reopenings. Slower reopenings (along with aggressively enforced mask-wearing) have allowed countries like Germany to reopen safely without the spikes that we’ve seen in America.
We’ve already reconsidered masks, routes of transmission, and treatment protocols. It’s time to reconsider the disease’s incubation period and the time course of its spread, too. We need to make sure our initial assumptions were indeed correct. Millions of lives potentially hang on getting these basic parameters of the disease right.