Hacker News Reader

Comment by yosito | Hacker News Reader

I'm really confused about how air pollution can be measured in deaths. I get how air pollution can shorten lifespan, but how do you attribute a specific death, other than extreme cases, to air pollution? When someone claims that 10 million deaths a year should be attributed to air pollution, what exactly do they mean, in a nutshell?

View on HN

> And while none of these estimates is meant to suggest a single cause of mortality, such as a gunshot wound or a dose of poison in your morning tea, the calculus for air pollution is the same as for obesity or smoking: take the problem away, and the number of premature deaths will fall by many millions.

They mean that if air pollution was not there throughout our lifetimes up until now, we would have 10M fewer deaths per year. Longer lives translates into fewer deaths per year.

When diseases like COPD and lung cancer are extremely prevalent, we may not notice the deaths as caused by air pollution, but they are. Everybody should get rid of their gas stoves, but we instead have full on sale pr campaigns for them.

> Longer lives translates into fewer deaths per year.

Are you sure about that?

I mean your technically right to question because the only thing which can increase deaths are births ultimately as everything else 'just' moves them forward, but I think it was quiet clear what they meant.

Think of how many lives we could save if we could just reduce births!

All else being equal, then very clearly yes. You could perhaps try to incorporate predictions of more indirect effects of longer lives (like increased reproduction rates and population) to argue that it could somehow cancel out, but that would take a lot of work to be convincing.

I think their point was that if everyone lives to 150, the rate of deaths per year will eventually converge to the same as if they live to 50. Lengthening the pipe doesn’t mean more water flows through it per minute.

Not true.

If there are 9 billion people (just to keep the numbers round), and they all live to 50, then every year 9 billion / 50 = 180 million people die (and the same number are born). But if they all live to 150, then every year 9 billion / 150 = 60 million people die (and the same number are born).

Only if the birth rate proportionally drops, which seems unlikely.

On the other hand, if the birth rate remains steady then the percentage of people alive that die in a given year decreases, which is maybe the more relevant metric as it seems like there’s less death around.

>just to keep the numbers round

That's not "just keeping the numbers round", that's putting a huge assumption on a complex system (births = deaths, no change in overall population over time).

The 9 billion was just to keep the numbers round. 8 billion would be a more accurate number, but it's not divisible by 3, and therefore not divisible by 150.

What you seem to be quibbling with is the assumption that the population was steady. That's an assumption that could be argued with, but it's not the "keeping the numbers round" part.

Fair enough

When the population is growing, yes. A bit of hand-wavy calculation to illustrate the phenomenon: if people die at 80, then people dying today are those born in 1931, if they died at 85 it would be people born in 1926, which were less numerous => less death. (Of course, not every one die at the same age and all, but you get the idea)

> Longer lives translates into fewer deaths per year.

Of course that's not true. Everyone dies, regardless of individual lifespans.

We are not talking about steady state dynamics, we are talking about specific numbers of births each year in the past, and an intervention that would shift the distribution of life spans.

The way it was stated in the article was correct.

My paragraph of text was also correct, but your response to me was incorrect.

> Everybody should get rid of their gas stoves, but we instead have full on sale pr campaigns for them.

Huge +1 to this btw. If you are in the position of owning a gas stove, make sure to always run the range hood vent whenever using the stove.

In California, at least, an evacuation hood is not required by law, even for a gas stove.

How does longer lives translate into fewer deaths per year? I understand how extending everyone’s life say 1 year starting tomorrow means fewer deaths (i.e. zero) next year, but after that it’s back to the same rate in deaths per year.

Understand that I’m in favor of people living long and healthy lives, but even if everyone maxed out at 120 years of lifespan the deaths per year will be equal to the number of people reaching 120 years old, and that is determined by how many people were born 120 years before.

> I understand how extending everyone’s life say 1 year starting tomorrow means fewer deaths (i.e. zero) next year, but after that it’s back to the same rate in deaths per year.

This is precisely what is meant by fewer deaths per year, it's now zero instead of the normal rate.

Similarly, if everybody's life is shortened by 1 year by something, there is an increase in death rate followed by a return.

I'm not sure why this concept elicited so many contrary comments. You already understand the concept, clearly, but I'm not 100% clear on what you are objecting to with your comment, or what meaning you are trying to communicate with this:

> but even if everyone maxed out at 120 years of lifespan the deaths per year will be equal to the number of people reaching 120 years old, and that is determined by how many people were born 120 years before.

Everybody still dies, but we only lived to 120 seems like a good outcome. But right. Effort that you said you are in favor of people living to 120, so what is the purpose of the rest of the paragraph?

Trying to correct the autocorrect in the last paragraph:

Everybody still dies, but we lived to 120 seems like a good outcome. But even you said you are in favor of people living to 120, so what is the purpose of the rest of the paragraph?

Go forth and learn!

https://www.who.int/data/gho/data/indicators/indicator-detai...

Well that was a fun detour. The clincher is here, a Global Burden of Disease Study in 2010:

https://sci-hub.mksa.top/10.1016/S0140-6736(12)61766-8

That is where they draw a line from air pollution to risk, with lots of statistics to cancel out other factors. Its beyond my depth.

----

This page was useful:

ttps://www.who.int/data/gho/data/indicators/indicator-details/GHO/ambient-air-pollution-attributable-dalys-(per-100-000-population) -> Metadata

> Burden of disease is calculated by first combining information on the increased (or relative) risk of a disease resulting from exposure, with information on how widespread the exposure is in the population (in this case, the annual mean concentration of particulate matter to which the population is exposed). This allows calculation of the 'population attributable fraction' (PAF), which is the fraction of disease seen in a given population that can be attributed to the exposure, in this case the annual mean concentration of particulate matter. Applying this fraction to the total burden of disease (e.g. cardiopulmonary disease expressed as deaths or DALYs), gives the total number of deaths or DALYs that results from ambient air pollution.

----

And the overall methodology is something called DALYs - Disability Adjusted Life Years

https://www.who.int/data/gho/indicator-metadata-registry/imr...

https://www.who.int/healthinfo/statistics/GlobalDALYmethods_...

That second link was a good basic read on what DALY means - its deaths due to a cause, weighted by "lost years" (so if life expectancy is 92 and someone died at 30, that would be weighted higher than someone dying at 80 for the same cause) PLUS years of healthy life lost even if it does not result in death - which is sort of a weighting based on severity of disease, where like say a person living with a disease severity of 0.2 would mean each year counts as 4/5th a year of a healthy life.

Not an expert in the area at all, but it seems to me that if a particular factor shortens 10 lives by 7 years apiece, and life expectancy is 70 years, then in a statistical sense it has "caused" one death, and that this is the case even if no one exactly died from it, in the acute sense that a car crash or firearm causes a death. I'm not sure if that's what's going on here, but even a slight harm multiplied across national or planetary population scales could easily get you to a number like 10M/year.

I don't think these studies are just taking the average life expectancy (let's say 70 years) and considering each 70-year reduction of life expectance to be 1 death. I think it's more like each time a person dies in a given year where event A occurs who would not have been expected to die in that year if event A hadn't occurred, that counts as 1 death caused in that year by event A.

Isn't that just about the same, in aggregate? 60 years from now, 7 people die at age 60 instead of 70, "caused" by air pollution. Smoothing it out, that's one death per year from now until then.

I suppose it works out to be the same if the cause of death is distributed evenly among all ages, but I suspect that's rarely the case. If something causes a lot of people to die at 90% of average life expectancy you wouldn't want each one to only count as one tenth of a death.

I think you’re right in that, especially for things that generally kill you towards the end of a normal life span, counting deaths by attributing the thing that wins the race to kill you doesn’t make much sense. If you’re murdered, eaten by a bear, or contact some rare infection and die early, it’s not like some other cause of death was just a few years off. It is however an easier to digest statistic than some of the others like DALY.

What you want is some statistic about how much life was lost in absolute terms of years and how much quality of life was lost due to air quality related diseases.

It can't. Anyone credible is using an age adjusted metric of some sort. YPLL or something like that. Of course none of the activists, journalists or editors use those numbers because they are boring and don't grab headlines like "a bajillion deaths" does.

Right. As a thought experiment, assuming a 75 year lifespan, what if air pollution causes EVERYBODY to die one day earlier than they would have otherwise?

Is the death rate 100%? Or is it .00365% (1 / (75 * 365))?

But an age-adjusted metric can give non-intuitive results. Thanos' snap, randomly killing 50% of everybody, would have had a 25% lives-lost-equivalent.

How you calculate death rate has real consequences in how you evaluate something like Covid.

Here’s the 2017 study mentioned by name in the article: https://www.thelancet.com/journals/lancet/article/PIIS0140-6...

I don’t pretend to understand fully, but these seem to be some sentences explaining their methodology:

> Ambient PM2·5 was the fifth-ranking mortality risk factor in 2015. Exposure to PM2·5 caused 4·2 million (95% uncertainty interval [UI] 3·7 million to 4·8 million) deaths and 103·1 million (90·8 million 115·1 million) disability-adjusted life-years (DALYs) in 2015, representing 7·6% of total global deaths and 4·2% of global DALYs, 59% of these in east and south Asia.

> Attributing deaths and disability-adjusted life-years (DALYs) to ambient air pollution requires spatially and temporally resolved estimates of population-weighted exposure, specification of a theoretical minimum risk exposure level (TMREL), estimation of relative risks across the exposure distribution, and estimates of the deaths and DALYs for diseases linked causally to air pollution. We combined estimates of exposure and relative risk to estimate the population-attributable fraction (PAF), the proportion of deaths and DALYs attributable to exposure above the TMREL. The numbers of deaths and DALYs for specific diseases were multiplied by the PAF to estimate the burden attributable to exposure. A more general description of the methods used to estimate the PAF and attributable burdens in GBD 2015 has been reported previously;3 here, we present details specific to air pollution.

While I don’t know enough about this methodology to defend this study, I would just point out that no matter what you consider to be a cause of death, no matter how “direct” you think that cause is, is still “merely” a death of someone who would have died later anyway. This mode of argument gets used all the time to minimize deaths of e.g. COVID-19 (“most of the deaths are old people with multiple comorbidities”), but this is a weak mode of argument precisely because it could apply just as well to literally any situation. We generally wouldn’t apply this mode of argument to minimize, for example, a serial killer at a nursing home.

You can do comparisons. If you can make good comparisons between places and the only major factor between them is pollution you can just compare average death per 100k citizens and then draw conclusions from there. You can also do the same for places that did not have a coal power plant and then when they did. Or the reverse!

Note: this is vastly overly simplified and there are other ways to do this and you need to do major and complex error calculations, but this will give you the highly abstracted view.

One approach that comes to mind: https://en.wikipedia.org/wiki/Micromort

What would really help is if they reported lost person-years. 90 year-olds dying vs 30 year olds would both be "premature deaths," but these are very different types of tragedy!

They're making it all up.

There are more tactful ways of saying this, but at the end of the day, it's functionally equivalent to pulling a number out of their hat.

Reply on news.ycombinator.com