Today I’ve frequently found myself overrun by the flow of new information, new data on the unfolding pandemic. It’s not clear to me whether that is more a reaction to an accelerating of new information or some cognitive fatigue or degeneration due to everything that has happened over the last three weeks. But my reading today drives home for me just how much we and really the whole world are still flying largely blind as we navigate one of if not the greatest global crisis of our lifetimes.
We’ve talked a lot about the failure of testing in the United States. But the specific issue in the United States only touches one of the known unknowns. There is an increasing body of evidence suggesting that there is a very large population of infected individuals who never become symptomatic or have such mild illness that they’re not showing up in our numbers. This new study published two days ago in Science uses a series of data analysis and modeling methods to estimate that prior the shutdown of Wuhan on January 23rd fully 86% of infections were undocumented.
The study is clearly written, clear enough that I could follow what authors were saying they were doing with their data and models despite having no ability to judge the merits of their methodology. But this is a study by highly credentialed researchers and published in a top scientific journal. So we can assume it is a very serious grappling with the available data.
Here are the key points, as I understood them.
From January 10th to 23rd, as the outbreak spiraled out of control in Hubei province, only 14% of the cases were being detected and documented. The travel restrictions and clampdown on Wuhan came on January 23rd. From that point, much more aggressive testing and much higher treatment seeking (from press attention, public health messaging) dramatically increased the percentage of documented infections. From January 24th through February 3rd the authors of the study estimated that 65% of infections were documented. The percentage was basically the same for January 24th through February 8th. So before January authorities were detected 14% of cases and the couple weeks after that they were finding 65%.
There are several critical implications of these numbers.
The first is that while undocumented, relatively mild or asymptomatic cases were less efficient at spreading the disease they were responsible for most of the force of the outbreak. The authors estimate that in the pre-clampdown breakout period (Jan. 10-23rd) 86.2% of infections were from these undocumented cases. Chinese authorities ability to document dramatically more infections combined with greater awareness in the population and protective/social distancing activity shifted things dramatically. A key quote: “While the estimate for the relative transmission rate, μ, is lower than before 23 January, the contagiousness of undocumented infections, represented by μβ, was substantially reduced, possibly reflecting that only very mild, less contagious infections remain undocumented or that individual protective behavior and contact precautions have proven effective.”
So the less symptomatic you are, the less contagious you are. Finding a ton of people who were kind of sick but still circulating and getting them isolated had a decisive impact on the outbreak. Those who remained undetected were less able to spread the disease to others.
At the risk of stating the obvious, to get a handle on this you need to be able to test on a massive scale, not simply for clinical purposes but for surveillance and quarantine purposes. That is not only to reduce the reach of the outbreak and the number of fatalities but just as much to be able to calibrate the pros and cons, the weighing of different kinds of risk in how much you shut down society and economy to achieve your purpose. There’s simply no way everyone on the planet can stay in their houses for eighteen months. But without knowing more than we do we really don’t know where to strike the balance.
Another major takeaway is that the prevalence of undocumented cases has significance implications for mortality. Even after Chinese authorities moved toward more intensive testing and surveillance the authors estimated that a third of infections were going undocumented. One of the studies authors, Jeffrey Shaman, told StatNews said, “I think there are many more than the [nearly] 70,000” confirmed cases in Hubei province.
Mortality is a fraction of the total number of people who die from the disease divided by the total number of people who contract the disease. If the actual number of people infected goes up, the percentage chance of dying goes down. A separate study, released over the weekend in preprint and not yet peer-reviewed, estimated that in the initial outbreak in Wuhan the risk of dying from COVID-19 after developing symptoms was 1.4%. As StatNews notes https://www.statnews.com/2020/03/16/lower-coronavirus-death-rate-estimates/, if the study on undocumented cases is generally accurate that would suggest a still lower fatality rate taking into account all who become infected.
My point here isn’t to say the fatality rate is higher or lower. The risk of dying for COVID-19 that these two studies together suggest isn’t that different from what many epidemiologists suspect. It’s the more general point that there are still a huge number of basic things about this disease we don’t know. Obviously it’s terrible if the fatality rate is higher than we think and great if it’s lower. But there are more basic things that aren’t clearly “good” or “bad” but have dramatic implications for what we do to combat it. A big driver of that uncertainty is the lack of testing in the United States. But it’s not just that. These uncertainties exist globally. And that is a big, big part of the problem.
As a global community we are still something like pilots in a plane in which most of the instruments aren’t working and flying in foggy weather. We don’t really know our altitude, the weather ahead or much about our direction. But we have to make decisions now regardless.