Scientific studies claiming nCoV has spread far worse than reported - how sound is the methodology?

We already have threads about the ongoing coronavirus in IMHO and MPSIMS, but I just wanted to focus on one specific aspect of the discussion alone, in GQ, while setting aside all the other stuff:
Multiple sources have claimed that the China coronavirus has actually already spread far worse than the official governmental reports. The medical journal *Lancet, for instance, *published a University of Hong Kong study that claimed that as of January 25 (weeks and weeks ago,) there were already over 75,000 people infected in Wuhan alone. (To put it in perspective, even as of February 9, a very long time later, the Chinese government claimed that the number of confirmed virus cases in the entire nation had not yet exceeded 40,000.) The study also claimed that there was or would be a doubling rate every 6.4 days, which would imply that the number of coronavirus-infected people has already spiked into the hundreds of thousands by now.

Now - how sound is such a claim, or its methodology? The attached link above provides all kinds of math and data but to my un-mathematical eyes, it’s like reading Latin. The Lancet, presumably, is a respectable outlet, not some hack rag trying to push sensational stuff for clicks, so maybe it’s valid?

(If there is any error in the article, I would suspect it’s that the authors are assuming the virus to be too easily spread, but that’s my WAG.)

Baby-statistician point of view - looks like perfectly respectable modelling. Basically, they’re making estimates of the transmission rate, combining that with publically available information about the movements of people, figuring out whether the known incidence of infection outside Wuhan is compatible with that transmission rate - if it’s not, go back and estimate again. Then at the end, “if the transmission rate is such and such, then the number of cases in Wuhan right now ought to be…”

This is not necessarily bad news though. Unless the death rate goes through the roof, that makes it more infectious and less deadly than otherwise estimated. A very large pool of people who got it, but didn’t get sick enough to end up in hospital, brings it much closer to ordinary flu in terms of danger to the general population

IANA Epidemiologist, but I know that this proposition can be challenged. R0 is not a fixed constant for all places at all times, and the extreme measures being adopted in the major cities, and greater medical resources (as well as mask allocation etc) mean that the R0 is almost certainly lower there.

Indeed if we take the suggestion that Shanghai, say, would grow exponentially with a lag of only two weeks behind Wuhan as a concrete prediction, then I would say it’s already been falsified.
Yes, China is a secretive nation as everyone knows and points out frequently. But as someone who lives in Shanghai I can say that details of new cases of coronavirus in Shanghai are being widely shared on social media. Plus people live in big residential buildings full of gossips. There is just no way that there could be thousands of people sick enough to be hospitalized being hidden.

Note that these things are not necessarily inconsistent. One is a concrete number of recorded cases, and the other is an estimate.

Right, that’s what I meant - if the *Lancet *article were true, places like Beijing and Shanghai would have become absolute infection war zones by now, and they aren’t.

Note that the paper is saying “if the transmission rate is constant…”. They’re not claiming that it necessarily is; they’re just looking at that case because that’s the one that’s simple enough to model given the data that they have. In a few weeks, it will become apparent if their conclusions were correct, and if they aren’t, they’ll then have enough data to construct better, more detailed models. That’s typical for scientific findings.

How do epidemiologists address the question of people who are exposed and infected, are asymptomatic or minimally symptomatic and therefore never diagnosed, but gain immunity and are thus removed from the pool of potential victims in the population? It seems to me that this as an unknown (at this stage) that could have very significant implications for modeling.

Right, and I also said if we take that proposition as a concrete claim, it’s been falsified.
If we look at the date of the paper, 31st Jan, there were 249 deaths in Hubei by that time. That’s a couple of orders of magnitude more than we’ll see / have seen in Shanghai within the stipulated 1-2 weeks.

It’s fine (and correct) for the paper to use equivocal language and say “If X” but when the media is going to run with it and say Experts believe X, it becomes necessary to point out problems with the claim X.

If a tree falls in the forest and nobody hears, does it make a sound?

If lots of people get sick without displaying symptoms, are they really sick? However, if this were the case, then there would be thousands of cases popping up in random locations all over the world, as asymptomatic people become carriers distributing the virus to others who then show symptoms. (A whole herd of Typhoid Mary’s) The level of travel, especially before the virus became a hot topic, suggests this is not the case.

As to the statistical part, you are simply working with bad data if the people reported as infected is not a complete reporting.

Well no, that’s not an apt comment. The “unheard sound” is something real and significant - unknown immunity.

If someone has never been infected, they are not immune and are susceptible to future infection. If someone has been infected, but were minimally symptomatic and never diagnosed, then they have antibodies against the virus and are not susceptible to future infection. It’s equivalent to having been vaccinated without knowing it, and the proportion of the population that’s immune to infection is surely highly significant for modeling.

It seems to me that different viruses could have very different properties in terms of the number of people who get infected but are minimally symptomatic and never diagnosed. It’s not a question of “bad” data, it’s a form of ascertainment bias intrinsic to the methodology.

I’m sure epidemiologists consider this in their methodologies, I’m interested in how they think about it.

Actual infection is measured by the presence of antibodies, which means that the body has been exposed and had to fight the intruder, not just sat in the same classroom which might be enough to get them thrown into quarantine. The number of people who have antibodies cannot be less than the number with symptoms [assuming there is no confusion, eg cold <=> flu] and can be much more. I expect this sort of thing happens more well after the event, but could give a better correction factor to create a total infection count for modelling and measuring morbidity and mortality.

On a quick search two recent influenza outbreaks where this was done scored 63% and 75% being asymptomatic, so take the number of people showing symptoms and triple-quadruple it for actual number infected.

The Lancet is one of the most respected medical journals in the world. It is also the journal that published Andrew Wakefield’s foundation paper for the anti-vax movement, linking MMR and Autism. A decision explained by the editor with the observation that the Lancet is in competition with other entertainment media.

IOW, part of the criteria for inclusion in the Lancet is “how interesting the claim is”. And that is, unfortunately, particularly a problem with epidemiology.

If that bothers you :), here’s an article in NEJM: https://www.nejm.org/doi/full/10.1056/NEJMoa2001316 “In the early period of exponential growth, we estimated the epidemic doubling time and the basic reproductive number”

In contrast to the Lancet, criteria for the NEJM is “new and unverified”.

There’s also a difficulty in that it takes time to do real scientific studies, and to disseminate the information from them, but no time at all to just toss off a soundbite. So the good information will always be lagging behind the bad. Those same scientists who wrote that Jan 31 paper probably are updating their models right this moment, based on 10 days of additional data. But we don’t yet know what they’ve found (or are finding) from those revised models.

No, I think I have a valid point (there’s a surprise).

Are we talking about a form of infection which allows the “not obviously sick” to pass on the virus so it becomes a lethal infection in some others? In that case, the spread is just as high-velocity, the problem is surprise new clusters of the disease where nobody expects it (this has not happened AFAIK)

Or perhaps we’re talking about a less strong version which is not easily distinguished from the common cold for anyone further it infects? In that case, the non-lethal low grade victims are irrelevant. By the time they have gotten over their version, the wave of infections is long past and moved on to others. It just means the overall mortality of the disease is lower than originally thought, because a lot of people contracted it and did not even get seriously ill.

What I think I read there is the second wave immunity question? What if when an outbreak happens, a significant number of the population are immune? Sort of like “If you had cowpox, you won’t get smallpox”? I believe there’s a mathematical model for that. The first bubonic plague swept across Europe in 1348-50 and supposedly killed about one third of the population. It then petered out because most of the remaining population was immune. It returned every generation or so when the newer population (those too young to have had it, not immune) reached a tipping point. Off the top of my head, that number is dependent on the percentage of non-immune, the ease with which the disease spreads, and population density. (and sometimes, public health standards).

What we see - bubonic plague, Spanish flu, or AIDS, most diseases start out strong - but the ones that infect the most people are the less lethal strains, since that host survives longer to infect more people, and the subsequent waves seem to be less and less lethal.

I can’t speak for other parts of the world, but I have seen no empirical evidence that is the case here in the United States. “Far worse” is something that would definitely make it a prominent news item. Heck, they trumpeted the handful of cases that we do have here.

It depends what is meant by “start out”. Many, perhaps most, pathogenic viruses are the result of jumping a species barrier.
And in viruses like this they may become more virulent as they adapt to the new host’s physiology, and only later is there are clear trend of less virulence, as those strains are the most enduring.
This is what we saw with Spanish flu, for example.