Saving lives by making alcohol more expensive

The BBC is excited about the possible benefits of a 50p per unit minimum alcohol price in England:

The deaths of 50,000 pensioners could be avoided over the next decade if minimum alcohol pricing is rolled out in England, according to new research. The BBC's Panorama programme commissioned the research from statisticians at Sheffield University.

That's a lot of deaths avoided - 5,000 each year. I compared it with ONS statistics on alcohol-related deaths, which can be downloaded here.  According to those data, there were 3,651 alcohol-related deaths in England in 2010 among people aged 55 and over (the State Pension age in the UK is currently 65 for men, and is in the process of increasing from 60 to 65 for women, with future increases to 66 and then 67 for both sexes).

So superficially the research is claiming that well over 100% of alcohol-related deaths can be avoided.  However, the official figures include only deaths obviously related to alcohol.  The analysis here uses a broader definition of alcohol-related deaths, based on the dubious method of Alcohol-attributable fractions, and gets a number (for 2005) of 9,005 deaths among people aged 55 and over, compared with 3,347 in the ONS data.  Of those 9,005 deaths, 5,803 are in people aged 65 and over.  If you use this number, and scale it up a bit to correspond to the slight increase in ONS figures, you find that the research is claiming that a mere 80% or so of alcohol-related deaths would be avoided.

I hesitate to commit myself before seeing what the Sheffield statisticians have actually done, but my initial reaction is that their numbers are a big overestimate.  It's disappointing, but not surprising, that the BBC should report this research uncritically.

I note that the NHS itself says that a 40p per unit minimum price could reduce alcohol-related deaths in England (across all age groups) by 900 a year.  It cites what "David Cameron is reported to have said".

I believe that alcohol is somewhat poisonous.  But I see no reason to use bad statistics to prove it.
__

There's one caveat, which is that the published methodology tells us that the ONS data are age-standardized, i.e. scaled to a European-standard age distribution.  In reality, there are more old people in England than in the standard, so actual deaths are higher.  (It's unclear to me whether the higher number using the AAF methodology is age standardized.)

__

Update: I watched the Panorama programme Old, Drunk and Disorderly? this evening.  The 50,000 figure was quoted a few times, but no information was given about how it had been arrived at.  It was juxtaposed with an estimate of 300 lives a year being saved by the introduction of a 50p minimum price in Scotland.  Scaling according to populations that would correspond to about 3000 lives a year in England, across all age groups.

Papers and reports from the Sheffield University group are listed here, not including the research reported by Panorama.