Hacker News Reader

Comment by evancoop | Hacker News Reader

Excess deaths will need to become the standard statistic used to compare nations. Otherwise, we'll be debating "deaths FROM vs. deaths WITH COVID" and the differing statistical definitions from nation to nation. Moreover, is a death from lack of access to medical care due to the collapse of a medical system a COVID death? Is a death of despair from isolation and unemployment a COVID death?

My sense is, when the pandemic is analyzed historically, "excess deaths" will be the only salient statistic in the literature.

View on HN

The Economist's excess death tracker is online here: https://www.economist.com/graphic-detail/coronavirus-excess-...

The article was last updated in October 2021, but the data in the charts is current insofar as it's available.

Their methodology is posted on Github here: https://github.com/TheEconomist/covid-19-excess-deaths-track...

They get their data from the World Mortality Dataset (https://github.com/akarlinsky/world_mortality). They claim their modeled baselines fit a linear trend for year, to account for long-term increases or decreases in mortality, and a fixed effect for each week or month up to February 2020.

I haven't dug into the details of the methodology but it appears very sensible. Of course it's very dependent on overall mortality rates having been accurately reported to begin with.

If the Economist keeps this this chart going long enough, we should expect that in countries worst affected by covid, future mortality rates will fall below the expected baseline. The simple (if somewhat morbid) logic being that in those places the most vulnerable people will have already died.

You are in luck, they just updated it today.

It doesn't sound like anyone really knows how many people exist or are born or die in India, so it's not clear that the Economist's attractive presentation of the data could have any real basis in India, without a primary data source.

Evidently the underlying dataset doesn't include China.

I find it strange China isn't in this list.

I do not see any data about India

It's not in the table but there is a mention in the text below:

> In India, for example, our estimates suggest that perhaps 2.3m people had died from covid-19 by the start of May 2021, compared with about 200,000 official deaths.

Excess deaths - and maybe a QALY-loss estimate - are the only way to "score" the pandemic and different countries' response to it, but I suspect there will be a lot of writing on "actual COVID deaths" versus "deaths from hospitals deferring treatment due to COVID surges" versus knock-on deaths attributable to the pandemic, lockdowns, unemployment.

We're getting so much interesting data out of this pandemic, both biological and sociological. Which public health measures were effective? Which ones were acceptable to different populations, and why? How does misinformation flow? How well can governments control the narrative?

There are going to be PhD theses built on this for the next ten, twenty years.

Deaths from hospitals differing treatment are also a great assessment of COVID impact on society. If COVID hardly hit a region, then they didn't need to delay surgeries. If COVID ravaged through the country, likely the hospitals were impacted and had to delay surgeries. All excess death is relevant.

Right, but it's important to distinguish excess deaths caused by covid and covid hostpital impacts vs excess deaths caused by lockdown behavioral changes.

Lockdowns and mandates are something a region can instrument independently, and we now have lots of data to assess the benefits/impact of different degrees and types of lockdowns and mandates in jurisdictions that were impacted by the same number of new covid cases.

I've taken to calling them "Pandemic Deaths", ie deaths caused by the pandemic, regardless of specific source. Makes it a little harder to split hairs.

Then that should include suicides from lockdown inflicted mental distress

You mean it should include a negative adjustment for the fact that suicides fell in 2020 in the U.S.?

https://www.cdc.gov/nchs/data/vsrr/VSRR016.pdf

>Rates for persons aged 10–34 were higher in 2020 than in 2019, whereas rates for persons aged 35 and over were lower.

I highlighted the important part.

> only the 5% increase for those aged 25–34 (from 17.5 to 18.3) was significant.

Hope that helps.

It's super weird that people try to focus on this when the larger failure of policy is that car crash deaths were up more than 18%, resulting in far more absolute deaths. You are talking about a change of 320 deaths while the car crash increment was 3140 additional deaths.

Australia had a severe nationwide lockdown and suffered negative excess mortality in 2020

It does

> Is a death of despair from isolation and unemployment a COVID death?

Since that's driven by the response to COVID, I certainly don't think it should be, because the response is a choice which differs meaningful from a death by COVID infection or comorbidity. Teasing apart the deaths directly caused by COVID from those that indirectly led from our response to COVID is important for making these decisions in the future.

Unfortunately, it's still not a great comparison. Countries with lockdowns had less deaths from things like other illnesses, traffic accidents, et cetera.

The point is to capture all those effects together. People who died in car crashes after 9/11 grounded all the planes have in some sense died from the terrorist attack, even though nothing at the scene suggests it.

Same with covid, if someone dies from undiagnosed cancer due to lockdowns, they have been affected by covid in one way or another. That's what the "with or from" comment is about, we get past it by looking at the total effect.

> Same with covid, if someone dies from undiagnosed cancer due to lockdowns, they have been affected by covid in one way or another.

If a death would have been prevented by just not having lockdowns, then tallies that it's included in being high shouldn't be used to justify having lockdowns.

It's still a better metric than some guy on science.com effectively saying "I feel this number should be much higher"

Excess deaths per capita, please.

> Excess deaths

Excess deaths are not just due to COVID.

Sure. It doesn't matter at a national population level.

---

Excess deaths is an epidemiological measure and should be used for epidemiological purposes - targeting a public health response.

It seems like quite a few people have confused epidemiology, clinical care, and virology. Each of these fields can suggest data for the others, but can't speak definitively until the studies are actually done.

For example, some have said that omicron is more mild than other variants; but it would be more correct to say that the average experience of omicron in the population is more mild AT THIS TIME.

At this time, many people have been vaccinated or recovered from a previous infection, so without some very particular data, you cannot say that omicron is more mild.

Edit:

> you cannot say that omicron is more mild than another variant to any particular person in a clinical context.

The theory of the matter is that you take what's working and you apply it and discard what's not working. The practice of the matter is that if excess deaths goes up due to drug use increasing during lockdowns, the public health reaction is to increase lockdowns.

> Excess deaths are not just due to COVID.

Can you clarify what you mean by this?

What kinds of things might excess deaths be attributable to, and to what rate / ratio would you expect?

People are now skipping routine cancer screenings, check ups, and altogether avoiding any health care out of fear of catching covid. Its not hard to find stories of people with heart attacks not going to the hospital! Kids are missing routine vaccinations, etc. Drug overdoses are up - more deaths in SF are from ODs than Covid last time I looked. Dementia patients and those in nursing homes truly took a nose dive by completely separating them from human contact (they deteriorate rapidly with no stimulation). I imagine lots of people stopped exercising and put on loads of weight.

Not OP. But it's becoming increasingly clear (to me, at least) that data about reality has been lost. And no amount of "statistics", "modelling" and accounting for confounding variables will get that data back. This applies to masks, lockdowns, vaccine effectiveness and treatment effectiveness.

Even if we do get a reasonable set of that data back via statistical means, I personally wouldn't trust it. Too many things weren't tracked, people had incentives to fudge numbers, data-integrity and duplication issues, etc. As much as I know there are lots of incredibly smart and educated people working on this, I see most of this as people running around throwing spaghetti at a wall to see what sticks, some hoping to get paid for it, others hoping to save lives.

Likewise for your question. We simply don't have data or knowledge about what people did or how they behaved differently that might have caused a non-covid death to be reasonably attributed to covid. Suicide rates, anti-social behavior of children not seeing people's faces, people hiding pro-active tests, dentists visits, hiding from leaving the house and getting depression, the list is huge and specifically unconstrained.

You can complain about everything in the world, but not about masks. Masks work, we've known they work since at least 1919. When worn correctly by large groups of people in risky environments (indoors, poor airflow), they greatly reduce transmission rates.

Everything else, you can say stuff about, vaccines have side effects, we don't know if and how much they reduce transmission, treatments are in flux, but masks are a simple and obvious solution with almost 0 side effects.

From the only randomised controlled study of masks... https://bmjopen.bmj.com/content/5/4/e006577.long

"The rates of all infection outcomes were highest in the cloth mask arm, with the rate of ILI statistically significantly higher in the cloth mask arm (relative risk (RR)=13.00, 95% CI 1.69 to 100.07) compared with the medical mask arm. Cloth masks also had significantly higher rates of ILI compared with the control arm. An analysis by mask use showed ILI (RR=6.64, 95% CI 1.45 to 28.65) and laboratory-confirmed virus (RR=1.72, 95% CI 1.01 to 2.94) were significantly higher in the cloth masks group compared with the medical masks group. Penetration of cloth masks by particles was almost 97% and medical masks 44%."

Things vary wildly across the world but in most of Europe people aren't allowed to use cloth masks anymore, and haven't been for at least half a year.

They have to wear either surgical masks or FFP2.

Masks with a decent seal work and studies with mannequins show this. But in real life, an airborne disease, massive gaps by the side of your nose, wet cloth, rules such as wear it to your restaurant table then take off to eat, it's a pathetic exercise at best.

Gee, I wonder if anyone thought of testing conditions similar to real life. Nah, nobody did, because everyone is stupid.

https://www.mpg.de/17916867/coronavirus-masks-risk-protectio...

https://www.nature.com/articles/d41586-020-02801-8

Even bad masks poorly worn are better than no masks, and the effect is statistically significant.

Also reducing the areas in which there is unmitigated exposure is a good thing, not a bad thing.

> Gee, I wonder if anyone thought of testing conditions similar to real life. Nah, nobody did, because everyone is stupid.

No need for that, at all.

From [1]:

> A recent British Medical Journal review looked at six, fairly porcine, studies concerning mask-wearing and estimated an impressive 53% reduction in risk. But the single randomised controlled trial estimated the smallest effect: a reduction of about 18% (-23% to 46%) in Sars-CoV-2 infections. The “heaviest” studies, an analysis of US states and a survey of about 8,000 Chinese adults in early 2020, observed rather than experimented and its editorial highlights the risks of confounding variables influencing both wearing masks and infections and the impossibility of disentangling the effects of measures fluctuating simultaneously. Indeed, this review found an identical 53% reduction from handwashing.

There is an introduction of Spiegelhalther and Masters given here[2]. It's well worth reading that BMJ article, btw.

[1] https://www.theguardian.com/theobserver/2021/nov/28/ivermect...

[2] https://statsandstories.net/media1/covid-by-the-numbers

These are all muted by omicron that has an R7-R10.

Sure, wearing a Helmut is safer but if you hit something at 90 km/hr does it matter?

So that you may be identifiable? But you're right, so many of the arguments for masks don't hold up to the context they're used in, whether that's for high Rº or untrained non-medical professionals, type of mask, public use, (probably) aerosolised virus… the list goes on.

That study proves exactly what I said.

It describes an ideal situation where people are wearing well fitting medical grade or better masks with a virus as transmissible as Delta.

> When worn correctly by large groups of people in risky environments (indoors, poor airflow), they greatly reduce transmission rates.

Not if the time in said environment is long. Also masks have good protection against droplets, not much against aerosols.

> Not OP. But it's becoming increasingly clear (to me, at least) that data about reality has been lost.

Are you claiming that death counts are incorrect and data about deaths have been lost?

Not death-counts, but definitely data about the deaths yes. So detailed reasons for the death, specifics about the death that may be pertinent, their covid history prior to the death, how covid did or didn't affect their behavior that may have caused the death, etc.

Example: Are we keeping track of all the people that skipped a routine test because of covid lockdowns? And then, is there a subsequent record of that fact linked to every covid case of death? Or are we stuck modelling the correlation/overlap of those two data points because we track those statistics across populations separately?

Eh, I think we have enough variation in local responses that we'll be able to tease a lot of the data out with a reasonable degree of confidence.

We're living in a world full of counterfactuals, with some places mandating -- and enforcing -- mandates on various public health measures rigorously, and other places not.

The pandemic also has a time dimension. To the extent that excess deaths are from deferred treatment and missed testing, we expect them to get worse over time -- people missing a cancer screening in March 2020 shouldn't have resulted in excess deaths in April 2020, in the extreme. "Real" COVID deaths meanwhile we'd expect to follow the peaks and valleys of the infections.

It will be a lot of work, but that's what grad students are for.

> It will be a lot of work, but that's what grad students are for.

I really get the impression that a lot of these conspiracy theories are really rooted in an inability to understand how statistics work; and just how much can be legitimately derived from the data.

> We simply don't have data or knowledge about what people did or how they behaved differently that might have caused a non-covid death to be reasonably attributed to covid.

Which is somewhat tangential to my question.

Disregard any formal attribution to COVID 19 as a cause of death - direct or indirect.

Parent was suggesting all excess deaths are not just due to COVID (directly or indirectly).

Sensibly allocating excess deaths to different categories is really the question here, and I'm wondering what caveats we should be careful about.

From a terribly abstract perspective, total excess deaths within a population, year on year, prior and then during a global pandemic, should give us some useful baselines.

But the vast majority would be.

That remains to be proven, since authorities have a large incentive to play the COVID-fear card.

What incentive would that be?

And then if it turns out that lockdowns actually cause deaths due to youth suicides, they’ll be swept under the rug and be used to excuse more lockdowns.

An inconvenient truth.

https://www.nbcnews.com/news/us-news/youth-suicide-attempts-...

Reply on news.ycombinator.com