Friday, 27 May 2011

Drinking and deceit

Nigel Hawkes has a must-read post up at Straight Statistics about the latest chapter in Britain's bogus drinking epidemic. The headline news yesterday was that alcohol-related hospital admissions have topped a million, after doubling in the space of just six years.

The number of admissions reached 1,057,000 in 2009-10 compared with 945,500 in 2008-09 and 510,800 in 2002-03.

Such an enormous leap in the figures should set the alarm bells ringing. A doubling in admissions would be astonishing at any time, but when drinking has been on the decline, it simply defies belief.

But, as the NHS explained when last year's figures came out...

These figures use a new methodology reflecting a substantial change in the way the impact of alcohol on hospital admissions is calculated. Previously the calculation counted only admissions for reasons specifically related to alcohol. The new calculation, for which the methodology is described in the report, includes a proportion of the admissions for reasons that are not always related to alcohol, but can be in some instances (such as accidental injury).

The same note appears on the latest report, but—shamefully—not on the NHS press release that accompanied it. Consequently, it did not get a mention in the news.

Since 2002, the number of diagnostic fields (ie. the category of injury or ailment that people come to hospital to have treated) that are considered 'alcohol-related' has jumped from 7 to 14 and then from 14 to 20. (You can read all about this if you can read the tiny writing at the bottom of table 4.1 in the report).

The significance of this change cannot be overstated. The "proportion" of admissions that are not directly related to alcohol make up three-quarters of the total. It is not only meaningless to compare 2002 against 2009, it is questionable whether most of these cases can reasonably be called alcohol-related at all.

As Nigel Hawkes explains, the figures do not come from doctors and nurses classifying an admission as alcohol-related. Instead, they rely on aggregate data being divided up on a laptop according to a whole set of assumptions.

How is it that almost all the statistics related to alcohol can be moving in the right direction, yet the numbers of alcohol-related admissions keep going up at a dizzying rate?

It’s largely a function of methodology. Alcohol-related admissions are calculated in such a way that if you are unlucky enough, say, to be involved in a fire and admitted to hospital for the treatment of your burns, it will count as 0.38 of an alcohol-related admission – unless you happen to be under 15, when it won’t count at all.

If you drown, it counts as 0.34 of an alcohol-related admission – though most people unlucky enough to drown aren’t admitted to hospital. Getting chilled to the bone (accidental excessive cold) counts for 0.25 of an admission, intentional self-harm to 0.20 per cent of an admission.

These fractions apply whether or not there was any evidence you had been drinking before these disasters befell you.

The one measure that hasn't been twisted and changed is the number of alcohol-related deaths and, as Hawkes says, they fell.

Alcohol-related deaths – that is, those caused by conditions directly linked to alcohol – fell from 6,768 in 2008 to 6,584 in 2009. Much of the fall was attributable to a fall of nearly 250 in deaths from alcoholic liver disease.

That last figure is worth noting, since the claim that liver disease is rocketing is frequently made by temperance crusaders.

If Britain is suffering a drinking epidemic, it is a very peculiar epidemic indeed. It is one that has resulted in an enormous increase in hospital admissions despite a decline in both overall alcohol consumption and excessive drinking. According to the latest figures for 2008/09, it has also resulted in a 3% fall in alcohol-related mortality despite a 12% increase in alcohol-related admissions.

Over the same period, the way alcohol-related hospital admissions have been defined and recorded has changed time and again. It's not difficult to put two and two together here. Any responsible journalist would put the methodological change front and centre of any report.

The irony is that the BBC had, only the day before, exposed the fact that Alcohol Concern Cymru (the NHS-funded Welsh temperance group) had been creating alarm about drinking by inappropriately comparing two different sets of figures:

An alcohol charity claims there a "silent epidemic" of heavy drinking among elderly people in Wales.

AAC said the number of over 65s who said they had drunk more than the recommended maximum in the previous week rose from 22% (men) and 7% (women) in 2003/4 to 34% (men) and 17% (women) in 2009.

However, BBC Wales understands that as a result of changes in methodology adopted by the compilers of the Welsh Health Survey in 2006 the two sets of statistics are not comparable.

Quite right too. And yet, in their report yesterday, the Beeb made no mention of the fact that 2002's figures for hospital admissions cannot be compared with 2008's. Instead, it chose to focus on a press release from the NHS's astro-turf front group Alcohol Concern which predicts that admissions will rise to 1.5 million by 2015.

This is shoddy journalism but, in this instance, the NHS has been complicit in deceiving the media. As I mentioned, the NHS press release did not even hint at the change in methodology, nor did it mention the fall in mortality.

To give you an idea of how incompetent/dishonest (delete as applicable) the NHS has been in this matter, let's take the headline figure. That one million figure relates to admissions, not people. This is an important point because, as Hawkes says in his post, some people go into hospital with multiple admissions and most go in not at all. This is basic stuff, but how are hacks supposed to get this straight when the head of the NHS's statistics department is giving them misinformation?

Tim Straughan, chief executive of the NHS Information Centre, said: "Today's report shows the number of people admitted to hospital each year for alcohol related problems has topped 1 million for the first time."

This is frankly pathetic. The behaviour of the NHS and its spokesmen—not to mention Alcohol Concern—demonstrate once again that their desire to lobby for policies, notably minimum pricing, has made them incapable of issuing reliable and credible information. Time and time again, the British public are being deceived on the issue of drinking. Can we believe anything these people say?


Rob said...

Somehow, I'm not the least bit surprised at this. Haven't they been pulling the same kind of underhanded gimmicks with the smoking statistics for years? Here in the States, I get the impression that if a smoker dies in an auto accident it will get counted as a "tobacco-related death."

Furor Teutonicus said...

They have been doing the same for motorcycle "accident" rates for years, as well.

I burned my leg on the exhaust whilst the bike ENGINE was "bench running" in the garage after a rebuild. (I.e, it was not even IN the bike) The hospital put it down as "Motorcycle accident". Which kanckered 15 years of "no claims".

Dick Puddlecote said...

I saw that SS piece this morning, it's truly excellent. Yet again online sources trump the MSM for rigour in reporting.

James Higham said...

Wish I had enough money for a drinking epidemic.

BrianB said...

Nigel Hawkes:

"Of course, much greater contributions to the total are made by conditions such as high blood pressure, which accounted for 383,900 admissions – more than a third of the total. Around a third of admissions for hypertensive diseases in men aged between 25 and 65 are attributed to alcohol, and around a fifth in women. There are far more of these admissions than there are for fires or accidents, so they contribute a huge proportion of the total."

Spot on! I downloaded the NHS report yesterday and rapidly identified this ludicrous statistic of 384k admissions with 'alcohol-related' hypertensive diseases.

I checked with the official numbers at hesonline. In 2009/10 there were 39,739 admissions for high BP (hypertensive disease) in England, regardless of whether alcohol-related or not.

That means the number of admissions for hypertensive disease that were alcohol-related (383,900), as claimed by this report, was nearly ten times higher than all admissions for all types of hypertensive disease!

Bizarre? Yes!

A lie? Well not quite.

Clearly there were something of the order of 1.8 million admissions to English hospitals where high BP was mentioned, ie was included in the list of (up to 20) diagnosis codes recorded in the admission records, but only 39,739 (2.2%) of these were admissions that were directly due to (ie whose primary diagnosis was) that high BP. As for the other 97.8%, well they were poor souls who were admitted for a variety of other reasons, but who just happened to have high blood pressure as well!

And of course, there is absolutely no evidence, none at all, that a single one of the 1.8 millions admissions had high BP that was in any way related to alcohol at all. Not the 'guessed' 383,900. No, not a single one!

Finally, you might want to look at the way in which the 'alcohol attributable fractions' are worked out for each condition. The majority of these seem to have been based on one meta-analysis, carried out in 2000 at the University of Milan (Corrao et al), of various studies conducted over the period 1966 to 1996! It beggars belief that risk factors calculated from a heterogeneous set of epidemiological study samples, recorded between 15 and 45 years ago, can be seen to somehow represent the inherent risks in the drinking habits of the population of England in 2009/10!

I've said this many times before, and I will go to my grave saying it: Meta-analysis is cheating! It's an insidious method of trying to make strong statistical (causal) chains out of incredibly weak (and insignificant) links.

It is, of course and not surprisingly, a methodology well loved by the innumerate creeps who infest the public health professions nowadays.

But this is just another application of the tried and tested anti-tobacco template is it not? This report is just as dishonest as Comrad - er sorry - Konrad Jamrozik's 2005 classic piece of numeric prestidigitation in calculating the number of deaths due to passive smoking.

And we all remember what that led to!

Dan said...

"Lies, Damned Lies and Statistics"

A powerful lobbying tool for legislation to suit a minority's "wish list" and VERY dishonest.

Take a road accident where a car runs into the rear of one at red lights. Breathylisers used, the waiting car's driver shows over the limit, the "crash" driver not. Who gets arrested? What shows on the statistics?

Sadly we do not have MP's with savvy, Life experience or understanding, well able to see what's put to them and through ploys.

Anonymous said...

And so we see a constant accumulation of UNTRUE AND FRAUDULENT statistical misinformation.

We commenters do not have the funds to contest these lies. Who does? It really amazes me that there are indubitably people who are smokers (wealthy celebrities, for example), who must be as angry as the rest of us about this vicious law, but who are not prepared to do us all a favour and contest it. If I possessed, say, £20 000 000, it would not bother me to risk £1 000 000. Or would it? It may be true that 'the more you acquire, the more you want to acquire'.

My real point is that there seems to be no way in out democracy for false assertions to be contested. Studies conducted by this university or that cannot be contested. They are FACTUAL, regardless of their veracity.

In the end, one way or another, when politicians pass a law which is based upon statistics, the truth of the statistical conjectures must be confirmed.

As regards the smoking ban, statistical conjectures were accepted without question, by a committee appointed for that purpose - the purpose being, to accept the statistical conjectures as fact.

The Government is actually stupid or fraudulent. It is stupid if it accepts statistical conjectures as fact without proper scientific scrutiny, or it is fraudulent if it deliberately sets up scientific scrutiny which is tainted. I read somewhere that scientists are just as fallible as everyone else - offer them enough money and they will produce the result that you want.

The serious idea is that epidemiology is not scientific. It is an educated guess.

No law should be based upon an educated guess.

Anonymous said...

A bit late to comment on this really, I suppose, but I will.

In the DT Business section of 29th May 'Lansley takes on Supermarkets over alcohol'.

We can see how Alcohol Concern et al are following the program set out by ASH. The article says that they have withdrawn from the 'responsibility deal' (supermarkets not putting alcohol promotions at the front of their shops) 'because, they said, Gov is not being tough enough'. Sounds a lot like the initial ban on smoking at the bar, to me, which ASH also said 'was not tough enough'.

What surprised me reading that was the inference that Alcohol Concern has any involvement with supermarkets' policies at all. What is Alcohol Concern? It is just a pressure group with a few employees, paid for with grants. It has no legal or parliamentary standing whatsoever.

I wonder if supermarkets have learned from the tobacco experience? If they have any sense, they will have. If they have any sense, they will refuse to have anything to do with these pressure groups and deal only directly with the Gov. No fake charity involvement at all. (In fact, I suspect that this is what has happened).

I just thought that it might be of interest.

PS. Won't Lansley be far, far too busy trying to defend his NHS reforms? Who, then, in the Health Dept is pushing this stuff?