Friday, 17 October 2014

Were there really 9.6 million alcohol-related hospital admissions last year?

Mark Wadsworth has spotted that the number of alcohol-related hospital admissions have risen rather sharply in the last couple of years. In fact 'risen sharply' is an understatement. 'Rocketed into the stratosphere' might be a better way of putting it.

BBC, 26 May 2011: The number of alcohol-related hospital admissions in England has topped 1m for the first time, according to official statistics.

From The Daily Mirror, yesterday:

Heavy boozers are putting the NHS under “intolerable strain” and risk sparking a health crisis which will cost the country billions, a charity claimed yesterday. Alcohol Concern said 9.9 million NHS admissions in England – including hospital patients and clinic and A&E visits – were related to alcohol last year...

The Office for National Statistics is the usual port of call when looking up alcohol-related hospital data. Their latest figures for England tell us the following:

In 2012/13, there were an estimated 1,008,850 admissions related to alcohol consumption where an alcohol-related disease, injury or condition was the primary reason for hospital admission or a secondary diagnosis. Of the estimated 1,008,850 alcohol related admissions:

65% (651,010) were due to conditions which were categorised as partly attributable chronic conditions

6% (60,830) were for conditions categorised as partly attributable acute conditions

The figure of 1,008,850 admissions is considerably higher than it was a decade ago for various reasons, but it is lower than it was in 2011/12, 2010/11 and 2009/10.

Similar data from Scotland show that there were 35,926 alcohol-related discharges in 2012/13. Feel free to look up the figures for Wales and Northern Ireland, but it's quite obvious that the total number for the UK is nowhere near 9 or 10 million. It is an order of magnitude lower at just over one million. To put that in context, England's NHS deals with 125 million hospital admissions every year and alcohol-related admissions make up 1.4 per cent of the total.

There are various ways of inflating the number of alcohol-related admissions, such as widening the range of 'alcohol-related' illnesses and including admissions which are only partially related to alcohol. However, these techniques have all been exhausted and the ONS figures includes the widest range of admissions that can conceivably be described as alcohol-related.

The majority of admissions are not wholly, or even necessarily mainly, attributed to alcohol use. Most relate to chronic diseases such as hypertension and breast cancer. These figures are not calculated by doctors and nurses making assessments of patients. Instead, the system of alcohol-attributable fractions is used. This assumes that a certain percentage of admissions for each disease were caused by drinking. Chronic illnesses (which typically require many visits to hospital to treat) make up the majority:

Of the 1,008,850 admissions in 2012/13,

- around 711,840 admissions were for reasons that are partly attributable to alcohol consumption (i.e. the attributable fraction associated with the diagnosis (either primary or secondary) most strongly associated with alcohol consumption was less than 1)

- over half (57%) of these partly attributable admissions were for hypertensive diseases (ICD-10 codes I10 – I15), accounting for approximately 404,650 admissions. Admissions with other partly attributable diseases, injuries or conditions were much lower in comparison

- second highest condition in this category was cancer (ICD-10 codes C00 – C15, C18 – C22, C32 and C50 ) with 83,510 admissions (Table 4.1).

It is worth noting the various conditions that people are admitted for and the way they are categorised because the unwary newspaper reader might assume that all, or most, of the alcohol-related admissions are injuries, accidents and overdoses that take place on a Friday or Saturday night in 'Binge Britain'. That's hardly surprising when even the Morning Advertiser uses photos like this to illustrate the story.

But where does the 9.6 million (some papers reported 9.9 million) figure come from? The source is the temperance group Alcohol Concern who have been working their buddies in the pharmaceutical industry to produce a nifty website which supposedly allows users to see how many alcohol-related admissions there are each year in each area of the country.

They explain their methodology as follows:

The inpatient admissions and A&E attendances data in this map are for 2012/13. Estimates for outpatient attendances are based on benchmarks from the Birmingham Heavy Drinkers Project (1997 to 2004), The General Lifestyle Survey (2009) and the number of high risk drinkers taken from Local Alcohol Profiles (LAPE) (2005) estimates.

No more details are available but they have clearly derived estimates based on some (fairly old) data and some unexplained assumptions.

You would only bother coming up with estimates from a computer model if the real figures were not available. But here's the thing. The ONS has detailed hospital admission data for exactly the same areas that Alcohol Concern make guesstimates for. And what a difference there is between the ONS's figures and Alcohol Concern's estimates.

In Barnsley in 2012/13, for example, the ONS says there were 900 alcohol-related hospital admissions (600 were partly attributed to alcohol, 300 were wholly attributed to alcohol). Alcohol Concern says there were 46,992.

The difference between 900 and 46,992 is non-trivial to put it mildly.

To take another example from my neck of the woods, Alcohol Concern reckons there were 128,922 alcohol-related hospital admissions in West Sussex in 2012/13. The ONS says there were 14,210.

Alcohol Concern reckons there were 52,092 admissions in Brighton and Hove. The ONS says there were 4,640.

Alcohol Concern says there were 48,745 alcohol-related hospital admissions in Westminster. The ONS says there were 3,360.

These are massive discrepancies and Alcohol Concern make no attempt to explain why their figures are ten to fifty times higher than the ONS's. On the contrary, their press release implies that theirs are the official figures.

Since the ONS is a reputable organisation using official NHS records and a transparent methodology, I am inclined to think that their figures are much closer to the truth than those of a partisan pressure group.

Still, it got an enormous amount of newspaper coverage so well done Alcohol Concern. But be careful—one day a journalist might actually bother doing some basic fact-checking.


JD said...

>>But be careful—one day a journalist might actually bother doing some basic fact-checking.

To be fair some do, but their editors or sub-editors don't like it. Facts spoil the story.

Chris Oakley said...

The official figures are based on a flawed methodology that inflates the number of admissions and that ought to be reviewed.

The Alcohol Concern numbers are dishonest in the extreme and ought to be ignored by anyone with any respect for truth for reasons that you eloquently explain above. It is telling that they have been highlighted by a nasty little rag with a reputation for being economical with the truth.

BrianB said...

No, the flaw isn't in the methodology - I can't see anything in the "nifty website" that has any methodological basis whatsoever.

The simple truth is that the 9.6 million figure is the combination of Hospital Admissions and A&E Attendances and Outpatient Attendances. It's the equivalent of adding apples, paper clips and spiders together and claiming that the total is a measure of fruit!

In your West Sussex example, Chris, they actually counted 14,159 hospital admissions in the total of 128,922, so not far off the ONS figure (and as likely to be accurate as ONS). This number was dwarfed, though, by 73,672 A&E attendances and 42,090 outpatient attendances!

Having worked with NHS statistics for my whole working life, I can assure you - and Alcohol Concern - that no-one in the NHS would ever dare to add inpatient and outpatient (A&E is outpatient too) statistics together. Once would be a terrible mistake, to repeat it would be a sacking offence. To place such stated numbers in the public domain - eg on a “nifty” interactive web-site - is tantamount to fraud; plain and simple.

To be fair to AC (hard though it is), they do, in their website, use the description "Admissions & Attendances" , so you can rest assured that the man on the Clapham omnibus won't be confused between the two - can't you?

But shame on the newspaper headlines that refer to 'admissions' only. Don't present-day journalists ever check anything given to them in press releases?

No need to answer: of course they don't: and Alcohol 'Concern' know that they don't: and they know, therefore, that they can get away with this: and, because of this knowledge, it is fraud!

Don't get me started on "attributable fractions" - yet another statistical fraud borrowed from the tobacco smoke-hating liars!