Did 28,000 people die from air pollution in 2012? Part 3

 

COMEAP needed to estimate how many percent excess mortality would be caused by 10 micrograms per cubic metre of PM2.5 air pollution. To do this they just asked themselves what they thought the answer was. The details of this 'elicitation' process are described below.

The Committee on the Medical Effects of Air Pollutants (COMEAP) needed to estimate how many percent excess mortality would be caused by 10 micrograms per cubic metre of PM2.5 air pollution. To do this they just asked themselves what they thought the answer was. Their answers range from -1% (i.e. air pollution prolonged life) to 17% excess mortality. The details of this ‘elicitation’ process and the meaning of excess mortality are described in the text.  This graph represents their consensus view: individual estimates are shown at the end of the article. (SOURCE COMEAP 2009 Page 160.)

Continued from PART 2

In Part 1 we saw that is untrue that air pollution annually causes 28,000* deaths of otherwise healthy people.

In Part 2 we described the qualitative effects of air pollution.

We now look at how the Committee on the Medical Effects of Air Pollutants  (COMEAP) estimate the effects of air pollutants, and how they have chosen to communicate their estimates.

What I have written below is the result of staring at reading two extremely technical documents, an activity which I invite you to share:

The factual basis…

The COMEAP studied hundreds of scientific papers but singled out one for special mention: the American Cancer Study (ACS) [COMEAP 2009 Page 2: Point vii]

The ACS uses measurements of soot particles called PM2.5 as a ‘proxy’ for all the other forms of air pollution.

It does this because PM2.5 s are easy to count and the concentration of the other air pollutants is more difficult to quantify, but they are typically correlated with PM2.5s.

So we don’t know which part of the pollution is actually the active cause of any effect – a shortcoming the experts consider exhaustively in their 2009 report.

The ACS finds a correlation such that there is a 6% rise in all-cause mortality for a 10 microgram per cubic metre increase in PM2.5 concentration.

I haven’t been able to read the ACS but extraordinarily, COMEAP report no uncertainty in this 6% figure.

Perhaps it is because of this that they decide to make their own estimate. And to do this they used a procedure called elicitation.

What is elicitation?

Wikipedia defines elicitation as:

the synthesis of opinions of authorities of a subject where there is uncertainty due to insufficient data or when such data is unattainable because of physical constraints or lack of resources. Expert elicitation is essentially a scientific consensus methodology… [it] allows for … an “educated guess”.

Each expert on the committee was asked:

  • “What do you think is the chance that the correlation is not a 6% increase in mortality for a 10 microgram per cubic metre increase in PM2.5 concentration, but some other figure?

The experts were asked in turn:

  • What do you think is the chance that the true figure is greater than 0%?
  • What do you think is the chance that the true figure is greater than 1%?
  • What do you think is the chance that the true figure is greater than 2%?
  • etc. until they arrived at the figure of 17%, beyond which possible correlations were discounted as highly improbable.

The results are shown in the chart at the head of this article and in the chart at the end of the article.

And the results…

The results of this elicitation form the basis of all the mortality estimates that you have read in the previous two articles.

The results are literally, an educated guess. None of the fancy maths changes this.

The average estimate was 6% and experts considered that the true answer was ‘95% likely’ to be in the range between 0% and 15%

Since the level of PM2.5s in the UK is typically just under 10  microgram per cubic metre, it is this figure of 6% of the roughly 500,000 deaths per year that directly links to the estimate of 28,000* deaths per year.

The range of expert opinion is that the true figure could lie between 0 and 55,000: every one of the seven experts considered it marginally possible that there was no effect of pollution i.e. that the excess mortality was 0% or less.

Although the experts considered a plethora of studies – and their report is exhaustive and exhausting – in the end, this is just their opinion.

How to express their results?

COMEAP consider at length how to express the results of their elicitations. Mostly they considered that the effect on life-expectancy (a few months for most people in the UK) is clearest.

However they do specifically endorse the use of mortality to express the effect of air pollution. But they note:

… the result expressed in terms of attributable deaths or additional deaths may easily be misunderstood or misrepresented. This calculation is not [their emphasis] an estimate of the number of people whose untimely death is caused entirely by air pollution, but a way of representing the effect across the whole population when considered as a contributory factor to many more additional deaths.

...consequently we consider it unrealistic to view air pollution as the sole cause of death in a number of cases equal to the population attributable deaths.” [COMEAP 2010 Page 3]

Regarding the actual estimates COMEAP explicitly recommend that the excess mortality estimates should always include the full uncertainty estimate – i.e. including the possibility that the excess mortality may be zero.[COMEAP 2009 Page 3 Point xiii]

And my conclusion…

I agree with COMEAP that the expression of the effects of air pollution as excess deaths per year can be “easily misunderstood”. In fact I think it is nearly universally misunderstood.

And given that their recommendation that the full limits of uncertainty – which include the possibility of no effect – are generally not quoted – I think this gives quite the wrong impression.

The possibility that air pollution – mainly from traffic – might be killing 28,000* people a year – 17 times more than are killed in car accidents – is horrific.

But in fact it appears that air pollution of itself kills nobody.

Rather, when we are near the end of lives – which are now longer than they have ever been in human history – our lungs do not function as well as they might otherwise have done.

As a result, our demise – from whatever cause – is hastened.

Somehow this reality is not quite as scary as COMEAPs vision.

Other Figure

The graph as the top of the page was compiled from the responses of 7 experts. These graphs show the cumulative probabilities ascribed to a particular sensitivity coefficient by the each of the 7 experts.

The graph as the top of the page was compiled from the responses of 7 experts ‘A to G’. These graphs show the cumulative probabilities ascribed to a particular sensitivity coefficient by each of the 7 experts. Looking at any particular line, the graph tells you, in the opinion of that expert, the likelihood that the coefficient exceeds any particular % value. (SOURCE COMEAP 2009 Page 159)

Footnote

* COMEAP 2010: Page 5: Point 18 estimate 29,000 deaths per year, but this is sometimes reported as 28,000. Given the large uncertainty in the figure, I have taken the lower estimate throughout.

 

 

Tags:

One Response to “Did 28,000 people die from air pollution in 2012? Part 3”

  1. Did 28,000 people die from air pollution in 2012? Part 2 | Protons for Breakfast Blog Says:

    […] sense of science « Did 28,000 people die from air pollution in 2012? Did 28,000 people die from air pollution in 2012? Part 3 […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: