Would 2009 H1N1 (Swine Flu) ring the alarm bell?

post by AllAmericanBreakfast · 2020-04-07T07:16:11.367Z · score: 33 (11 votes) · LW · GW · 13 comments

Contents

  Introduction
  Stock Market Trends in 2009
  Factor 1: Transmissibility
  Factor 2: Harm
  Factor 3 - spread
  Factor 4: Institutional response
None
13 comments

Introduction

In order to assist in anticipating and responding to the next pandemic, I created an "alarm bell" questionnaire [LW · GW] to suggest whether or not one should behave as if the world is about to experience a stock market-crashing pandemic.

To evaluate whether it is correctly calibrated, my intention is to investigate whether the "alarm bell" would have rung during historical pandemics, and whether the stock market would have crashed before it rang, after it rang, or not at all. I have performed the first test, based on the 2009 H1N1/swine flu pandemic. The stock market did not crash, and in my analysis the alarm bell did not ring - but see the comments for caveats.

This is because 2009 H1N1 was much less deadly than COVID-19, and it also did not shut down the world economy. The "alarm bell" rings if 13 of the 16 below criteria are operative, but only 11 of the 16 criteria were ever met.

This is one point of evidence in favor of this alarm bell being correctly calibrated. I would also like to investigate this question for Ebola, MERS, SARS, HIV/AIDS, the 1968 Hong Kong flu, and (as far as it's relevant) the 1918 Spanish flu.

Stock Market Trends in 2009

The stock market was at a low point in 2009 around March 6th, but had begun a strong upward trend that would last the rest of the year by mid-March, when the first cases of swine flu were confirmed in Mexico. Although the New York Times reported stock market tremors related to swine flu on April 27th, 2009, looking at the data for the whole year, the daily fluctuations for the year look pretty much like a random walk (April 27th highlighted in red):

Factor 1: Transmissibility

Yes, as with all H1N1 flu.

Yes, as with all H1N1 flu.

Yes. According to a fact sheet released by the state of New Hampshire, "People with H1N1 flu virus infection should be considered potentially contagious one day before the onset of symptoms and as long as they are symptomatic, and possibly up to 7 days following the onset of illness."

A paper published on the 14th of May 2009 estimated that the "reproduction ratio was less than 2.2 – 3.1 in Mexico." For comparison, R0 for COVID-19 is estimated at 1.5-3.5. I count this as yes as of on or before the 14th of May 2009.

Factor 2: Harm

This is one of the trickiest ones, because in early days when case rates are not known accurately, it's hard to predict. It's important to distinguish between confirmed case fatality rate and overall case fatality rate, and global vs. age-related CFR.

One recent scholarly global CFR estimate for COVID-19 is 0.51%. An estimate from the 7th of February put it at 0.18%-2.8%. So I will consider adding a global, non-age-based CFR threshold to the criteria.

This project is also not about estimating the true value of CFR; it's about normalizing a heuristic for deciding when to behave as if a new illness could be as serious as COVID-19.

So we have to ask what our threshold for evidence is. I can think of two possibilities.

Given that it was predictable that there were many more undetected cases that lowered the true CFR, I'm going to disqualify the "rough division" from meeting the criteria and require a scholarly CFR estimates.

That 0.4% June scholarly global CFR estimates for swine flu were within the 0.18%-2.8% range of the Feb. 7th global CFR range for COVID-19. However, that Feb. 7th paper was pretty skeptical of these figures. I'd like to see if there was another COVID-19 estimate prior to Feb. 20th that presented an estimate with more confidence.

I provisionally count this as a no, but may change this later.

From an NCBI paper: "The acute symptoms of uncomplicated infections persist for three to seven days, and the disease is mostly self-limited in healthy individuals, but malaise and cough can persist for up to 2 weeks in some patients. Patients with more severe disease may require hospitalization, and this may increase the time of infection to around 9 to 10 days." I count this as a "no," because it is only malaise and cough that persisted for 2 weeks, not the need for hospitalization.

By CDC data, upper bound of US hospitalizations divided by number of cases is less than 1% of cases requiring hospitalization. I count this as a "no."

Because flu was already known to be Tamiflu-resistant in 2008, by early January 2009 the CDC was already recommending Relenza (zanamivir) and other alternate treatments for H1N1. I count this as a "no."

Factor 3 - spread

The WHO's 61st pandemic update put the world death toll at "at least 2185" on the 23rd of August 2009.

The first cases of what would later be confirmed as swine flu were diagnosed in Mexico in early March when 60% of the small town of La Gloria, in Veracruz, was sickened. It was confirmed in its 10th country, the Netherlands, in April 2009. It wasn't confirmed in a set of countries totaling a world population of at least 1 billion people until it was found in China on May 1st, 2009.

The first community outbreaks in the US were confirmed on April 25th, 2009.

News of Mexico shutting down parts of its economy were reported in Reuters on April 29th, 2009. I'll count this as "lockdown."

Factor 4: Institutional response

I don't find reports of major city lockdowns. The most significant travel restriction I can find was of Mexican travel to Japan. I personally don’t count 2009 Mexico as a “major economy” but that is a controversial and not pre-declared analysis decision. For future analyses I’ll use the 2009 Mexican fraction of world GDP as my cutoff for “not a major economy.” The WHO was recommending against travel restrictions early on. I count this as a no.

The WHO declared swine flu a "public health emergency of international concern" on April 25th, 2009.

There were multiple stories on page A1 in the New York Times by the last few days of April. There may have been earlier front-page swine flu news, but I'm not sure based on what's coming up on their digital search.

Yes, there were mask shortages.

13 comments

Comments sorted by top scores.

comment by avturchin · 2020-04-07T14:55:26.410Z · score: 6 (4 votes) · LW(p) · GW(p)

Death rate estimation you cite is from 2013, which is 4 years after the event, so it can't be used in time of the event. First time I heard about H1N1 in 2009, it was reported that 60 people of 1000 had died in Mexico, which implied 6 per cent death rate, and really alarmed me at the time, so I started buying food and masks.

comment by AllAmericanBreakfast · 2020-04-07T19:15:11.799Z · score: 5 (3 votes) · LW(p) · GW(p)

Thanks for pointing that out. I’ll see if I can update with an earlier estimate.

comment by Zachary Robertson (zachary-robertson) · 2020-04-07T14:11:36.935Z · score: 5 (4 votes) · LW(p) · GW(p)

2.4 Should be yes. Clearly there wasn’t a vaccine.

4.2 Should be yes. Mexico was clearly on lockdown. You shouldn’t get to cherry pick. It’s worth mentioning that CDC was also recommending shutting down school for 14 days.

2.1Only in retrospect do we know what the mortality rate truly is. It was 1.9% was reported in Mexico at one point.

It seems like H1N1 would’ve triggered this simply because if you’d been following the checklist as news came in, you would’ve noted there wasn’t a vaccine ‘check’ and then saw Mexico went on lockdown ‘check’. There wouldn’t have been this hindsight. Going through this list I conclude this would trigger false alarms.

I’m assuming the checklist was designed to have 16/16 for COVID19 given the cutoffs in the questions so it seems worth pointing out it’s also possible to have the opposite occur where a disease doesn’t ring the bell simply because its different than COVID19 in a few key ways. In the grand scheme this alarm bell seems designed to overfit to our current pandemic.

Obviously there should exist an alarm bell, but it seems too easy to subjectively wiggle the question answers. This is a prototype, the idea is nice though!

Source: https://www.wired.com/2009/06/apocalypse-not-behind-swine-flu-hysteria/

comment by AllAmericanBreakfast · 2020-04-07T16:14:51.386Z · score: 2 (2 votes) · LW(p) · GW(p)

You’re right that the wiggle room is a bug for historical testing purposes.

Even in this case, I disagree with two of your points. A treatment was available from the beginning. The original question isn’t whether a vaccine “and” treatment is available, it’s “or.”

I also don’t consider Mexico a “major economy,” so the travel restriction to Japan wouldn’t count. Others may disagree! In the future for testing purposes, I’ll use the size of Mexico’s fraction of world GDP as my threshold since I admit Mexico could be categorized either way.

I think you’re right that there was evidence of 1%+ mortality in early days. I’m not sure when that came down.

An important question is whether the criteria should be met all at once in order to count. My inclination is “yes.” If so, swine flu probably scores even lower, given for example that the death toll didn’t hit 2,000 until August.

I think ambiguity is a feature for practical use. “Is Mexico really a major economy? Do we really have enough evidence to say there’s a 1% death rate?” These are exactly the kinds of debates this alarm bell is meant to provoke. The criteria are primarily meant to give people an excuse and motivation to start a debate, not to end it.

I’ll keep all this in mind and perhaps overhaul it when I’ve done some more testing. Thanks for the feedback!

comment by Zachary Robertson (zachary-robertson) · 2020-04-07T19:00:04.759Z · score: 3 (3 votes) · LW(p) · GW(p)

The original question said “or”. There was not vaccine or a treatment. You could’ve said neither/nor, but you phrased it this way for a reason I thought. Also, not having a vaccine is bad, so I’d be annoyed if you changed the criteria so both had to be met.

Also Mexico is 15th largest by nominal GDP so I’m not sure what you mean by “major” either. I’m trying to point out that your list gives false alarms. You can claim that this alarm is meant to spark debate, I’m just saying it’s original purpose is questionably achieved. In fact, saying there’s room for debate threatens to ruin the entire premise of the alarm no? At that point we’re just discussing if I think your classifier is better than my classifier.

comment by AllAmericanBreakfast · 2020-04-07T20:34:14.022Z · score: 2 (2 votes) · LW(p) · GW(p)

Those are fair critiques. In 2009, Mexico was 1.5% of world GDP. I think for future efforts, I'll arbitrarily choose 2% of world GDP as my cutoff for "major economy."

I don't think we've ever had a vaccine in the early days of a novel illness, so it's not useful as a way to distinguish more from less dangerous illnesses. I might even remove it entirely and just focus on whether treatments are available.

This heuristic is a non-authoritative plan for guiding conversation and research efforts. Part of that plan is the "alarm bell" criteria, which is akin to basic guidance you might give people on whether they should pull a real-life fire alarm. Other people might have criteria that are more loose or strict for whether they should pull the fire alarm. That's fine. But the concept of having a fire alarm you can pull, a dialog about when you pull it, a set of common sense notions for when it's OK, and what to do when somebody does, I think is valuable. I'm trying to help organize that effort.

These criteria are going to have to be vetted with historical data, refined over time, and eventually replaced. But it's a place to start.

comment by ChristianKl · 2020-04-08T07:20:06.873Z · score: 2 (1 votes) · LW(p) · GW(p)

Why shouldn't Mexico count? Why should it be different if a disease starts in China then if it starts in Mexico?

comment by AllAmericanBreakfast · 2020-04-08T07:36:26.109Z · score: 1 (1 votes) · LW(p) · GW(p)

Well, it’s just easy for me to imagine someone saying “major new illness in China? Sounds bad for the global economy.” It’s harder to imagine them saying that about Mexico.

I’m assuming that it’s easier for a disease to spread within a country than to cross international borders. So it’s more likely to wreak havoc in the economy of a country it’s in already than one it hasn’t yet reached. A virus in Mexico has access to 2% of world GDP. In China it has access to 16%. Much more destructive potential.

GDP follows a power law so there are a few economies - the USA, China, the EU, and Japan, where it matters greatly whether a pandemic is spreading.

From one point of view, maybe spread doesn’t matter. Either a disease has the transmissibility to spread worldwide and evade containment, like COVID, or it doesn’t, like Ebola.

If that were true, only R0 and case fatality rate would be needed to adequately predict the consequences of a disease. It wouldn’t matter which countries the disease was in.

That would be an interesting alternative model, with the virtue of simplicity.

For this model, I tried to pick factors that I think the public and epidemiologists would consider important, including spread and whether it was in bigger more prosperous countries. That may or may not matter for predicting the stock market, but I think it does matter for starting a conversation about a future pandemic, which is the whole point of this project.

comment by ChristianKl · 2020-04-08T10:31:39.180Z · score: 5 (3 votes) · LW(p) · GW(p)

The main reason why COVID-19 is bad for the market and for us in the West is because it wasn't easily containable.

Whether or not the first cases happens in some Chinese provience or in Mexico doesn't matter much for that.

comment by Lukas_Gloor · 2020-04-07T16:56:42.490Z · score: 4 (2 votes) · LW(p) · GW(p)
>The death rate from swine flu was 0.02%, hitting the young harder than the elderly. I count this as a no.

This is not quite the right way of looking at it! I think you'd have to look into what experts thought during the early months of the Swine flu outbreak. I haven't researched this but I've read that early best estimates for Swine flu fatality were at least a factor of 5 higher than the true infection fatality rate, if not even higher. (The IMO misguded folks who think the IFR for Sars-CoV-2 could be as low as the flu's are constantly pointing this out, failing to flag that this is far from a universal trend among outbreaks – e.g., early estimates of Sars-1 fatality turned out to be underestimates.)

That said, it seems plausible that even with my proposed adjustment, the numbers would still remain below the thresholds you list under "harm." It depends on how much credence experts put on the higher end of the range during the early months of the Swine flu outbreak. I don't know just how high the highest estimates were that still came from credible experts.

comment by habryka (habryka4) · 2020-04-07T17:13:09.022Z · score: 3 (2 votes) · LW(p) · GW(p)

Fixed the image. The problem was that you used a link to the page where imgur embeds the image, and not an actual link to the image itself. To get the image link, right-click on the image and click "Copy Image Address". 

comment by AllAmericanBreakfast · 2020-04-07T17:27:40.410Z · score: 1 (1 votes) · LW(p) · GW(p)

Thanks so much for your help!

comment by Lukas_Gloor · 2020-04-07T17:04:56.747Z · score: 2 (1 votes) · LW(p) · GW(p)
I would also like to investigate this question for MERS, SARS, the 1968 Hong Kong flu, and (as far as it's relevant) the 1918 Spanish flu.

I'd be very interested in analyses of those (esp. if you look at it from the limited perspective people had in the early stages of those outbreaks). I feel like I completely missed it at the time, but the more I hear about SARS-1 the more I feel like the alarm bells should have gone off like crazy (and that probably happened in Asia but the way I remember it, reporting on SARS in the West felt no different from reporting about bird flu or Swine flu – but probably I didn't play close attention because I was really young).