Tuesday 2 June 2015

The latest smoking ban miracle in detail

This is a guest post by Chris Oakley about the latest smoking ban miracle.

When I read Chris’s post on the latest smoking ban miracle and the journal article it related to, something rang a bell. The laughable methodology and unjustifiably precise claims were all too familiar but so too was the name of the second author, Christopher Millet. I have written about this man once before in a post for Liberal Vision. Back in 2013. Millet, an anti-tobacco activist who believes that films containing smoking scenes should be 18 rated, conspired with Stanton Glantz to produce one of the most outrageously flawed papers that I have ever had the misfortune to read. [See here for my take on it - CJS.] Their article appeared in Pediatrics possibly because, in common with other medical journals, its acceptance standards are so low as to be practically non-existent.

Sadly, Millet escaped any censure and has gone on to contribute to yet another piece of policy based chicanery in an allegedly highly respected but apparently no more scientifically rigorous journal.Once again, the trick involves cherry picking time frames, taking advantage of unexplained variations in the data, ignoring significant contra indicators, being vague about methods and results, emphasizing irrelevant or misleading “sensitivity tests” and generally avoiding good scientific practice as much as possible whilst hiding behind a camouflage of pseudo-scientific gibberish. It is slightly more sophisticated than Millet’s last effort but not much.

The emotive and somewhat implausible claim is that the English smoking ban has led to 11,000 fewer children being admitted to hospital with lung infections every year. I have accessed the annual NHS data on the incidence of acute admissions for lower respiratory tract infections (LRTI) and upper respiratory tract infections (URTI) in the age group 0-14 and used ONS population data to produce the figure below, which shows population normalised incidence of admissions for LRTI.



I note that Been, Millet et al. omitted the data from 1999-2001 on the grounds that they only had both population and admissions data from 2001. I find that puzzling as I have population data from 1971. These two unspectacular years would almost certainly impact their model so I am suspicious.

The rise in admissions since the enforcement of the smoking ban should suggest to any sane, unbiased and even slightly enquiring mind that factors other than smoking are driving up admissions and that the ban is an irrelevance. Being hopelessly biased and questionably sane, the activists ignore logic and somehow find “evidence” to support their cherished legislation. It is hard to say how because with flagrant disregard for good scientific practice, they don’t reveal their method in any detail but my educated guess is that it may have something to do with the odd pre-ban peak in 2005-06 which they do not even mention, let alone explain.

The results section of the paper doesn’t contain actual results or a clear explanation for what passes for them. However, through the statistical smokescreen we glean that the authors are claiming that they can predict what would have happened without the ban. They do present a picture that illustrates that claim.


The data to the left of the dotted line representing the ban illustrates the difficulty in such predictions. The big winter peak is consistent but the shape of each cycle varies. Some years show a rapid fall and a small second peak, others a more gradual decline into the summer months. It is hard to see how any honest and competent scientist could derive an accurate predictive model from this data without knowing what causes these variations and allowing for it, or using a lot more cycles.

The data to the right shows the author’s model superimposed on reality. The most striking thing about it is the odd shape of the model data. The consistent high shoulder (circled in orange) is absent from any of the pre-ban data. The actual data for the first two years post-ban are remarkably similar in shape to the first two years pre-ban and I see no plausible explanation for why they should morph into the author’s creation. Been et al. are effectively claiming that a cyclic incidence pattern with significant annual variation in shape should suddenly adopt a hitherto unseen and more consistent shape because of a smoking ban not being introduced (for that is the counter-factual). The concept is ludicrous.

The authors and their chums in the press then add to the overall ridiculousness by going on to talk about falls in admissions when they actually mean unexplained rises that are, in their opinion smaller than they should have been according to an unexplained model made up by activists for smoking bans. Inexplicably, this approach survived peer review, editorial review and the hawk eyed scrutiny of the BBC health editors who allowed yet another ludicrous and hyperbolic press release to be regurgitated uncritically with added exclusive comment from a pressure group.

No comments: