Statistics cloud air pollution issues
It is time the Canterbury District Health Board explained the statistics behind assertions that smoke from woodburners kills more than 100 people a year, writes Dave Saville.
"Lies, damned lies, and statistics!" is a phrase that makes statisticians cringe, but is often appropriate in relation to the usage of statistics by politicians and lobby groups.
This came to mind when I read an article in Saturday's Weekend Press, which cited a Canterbury District Health Board report (Housing, Home Heating and Air Quality: a Public Health Perspective) which estimated air pollution in Christchurch caused 158 premature deaths each year, and smoke from woodburners was responsible for 78 per cent (thus, 123 premature deaths).
Given the current debate on whether citizens replacing earthquake-damaged homes are entitled to full replacement, including reinstatement of woodburners if these existed, it is timely for the health board to commission its report's authors to write an article for The Press which sets out in layman's terms the statistical evidence behind this statement.
During the clean air hearings in the 1990s, I was called as an expert witness to comment on the evidence available at that time.
This evidence was what I call "soft evidence", consisting of a correlation between the daily death rate due to respiratory disease and the daily air pollution (as judged by PM10 readings in St Albans).
That is, the death rate due to respiratory disease was higher on days when air pollution was higher. On this basis, it was asserted about 20 deaths due to respiratory disease were "associated" with air pollution, the implication being these lives could be saved if pollution was reduced.
One problem with this sort of evidence is the presence of "lurking" variables.
In this case, air pollution is associated with "coldness", as measured by the average evening temperature.
Cold nights mean we crank up our woodburners, which increases pollution.
However, there is also a direct effect of coldness on the death rate due to respiratory illness, as alluded to in the previous paragraph of the Press article, where the health board report is quoted as saying "premature death, respiratory disease and mental health problems were all consequences of cold homes".
That is, the correlation between daily death rate due to respiratory disease and daily air pollution is the result of the direct effect of the air pollution on the death rate, the direct effect of coldness on the death rate, and the effect of other, unknown variables which correlate with air pollution.
With this sort of soft evidence, involving several correlated explanatory variables, it is impossible to assign "cause and effect".
To make another analogy, you could prove there is a correlation (association) between per capita consumption of kiwifruit and the incidence of cancer.
However, no-one would suggest kiwifruit causes cancer. The two variables simply both increase with time.
Back to the correlation between daily death rate due to respiratory disease and daily air pollution. A question which was raised: were these respiratory disease patients exposed to the air pollution?
For example, did an asthma sufferer die when out in the polluted air? Or did a terminally ill patient die of respiratory disease while confined to their bed either indoors at home or in the hospital? It appeared the latter was more common.
With this in mind, the medical fraternity was encouraged to embark upon the collection of "hard evidence", based upon a more detailed analysis of the cause of death in each case. In some cases air pollution could be the clear cause of death, in other cases it could be implicated, and in other cases it could be entirely ruled out.
Has this data collection been attempted during the last decade or so?
An implication taken from the health board report of smoke from woodburners being responsible for almost 123 deaths per annum, is that if woodburners were to be moved entirely, 123 lives would be saved annually.
This appears to be a misrepresentation of the truth.
For example, a person dying of lung cancer would not survive to live a long and happy life (if woodburners were removed).
If this many lives could be saved annually, it should show up in the medical statistics, since pollution due to logburners has reduced considerably over the past two decades. However, as the Association for Independent Research has pointed out, there does not appear to be any such reduction in the death rate.
(One possible reason is that presented by rheumatologist Peter Moller at the health board meeting - that "logburner smoke does not kill people".)
The concerned scientists who asked me to get involved in the 1990s were Pat Palmer, John Hoare, Sir David Hay and colleagues in the association. Like most Christchurch folk, I was keen to see an improvement in air quality and could see obvious advantages both for myself and more especially for friends who suffered from asthma.
However, as a professional statistician, I was also keen to see the scientific evidence interpreted in a reasonable manner.
The topic is an emotive one but I feel this needs to be set aside when the strength of the evidence is being judged. Since the 1990s, I have not kept up with any changes in the evidence, so would be keen to see the evidence dusted down and presented to allow a proper public debate.
The topics discussed above were highly relevant to the original hearings, but if "harder evidence" has been collected, it would be interesting to read it. In conclusion, I repeat my suggestion that the health board commission its report's authors to write an article for The Press which sets out in layman's terms the statistical evidence on this topic.
Dave Saville is principal biometrician at Saville Statistical Consulting, Lincoln.