Changes to alcohol statistics promised

On Thursday the NHS Information Centre for health and social care will publish its annual compendium Alcohol in England. It will be here.

This publication is generally a signal for wailing, gnashing of teeth and rending of clothes as the statistics reveal another rise in alcohol-related hospital admissions, or yet further evidence that dirt-cheap drink prices are contributing to overindulgence. As regular readers will know, Straight Statistics has questioned both these claims, so a note that appeared last week on NHS IC’s website is interesting.

It announced methodological changes to both these indices that go some way towards meeting the objections made against them. On affordability, NHS IC has already in the 2011 publication introduced one change suggested by Rachel Seabrook of the Institute of Alcohol Studies, who pointed out incontrovertibly that the measure being used to calculate income measured the gross disposable income of the entire country, not the gross disposable income per capita. That change significantly reduced the alcohol affordability index, as I reported earlier. 

This change will continue in this year’s report. But further changes suggested by Rachel Seabrook are only being toyed with, despite NHS IC’s admission that its discussions with the Office for National Statistics suggest they would be beneficial. They include a more accurate way of calculating household disposable income that is not subject to the distortions detailed by her and summarised in my earlier post.

“We will examine these alternatives more closely in future, subject to required resources being available” NHS IC says. Meanwhile, the methods used to calculate alcohol affordability should be regarded as an “interim measure” subject to refinement if deemed necessary. Getting it right doesn’t strike me as especially difficult, or indeed costly. So speaking personally I’d like to see changes sooner rather than later.

Why? The idea that alcohol is historically dirt-cheap underpins the whole “minimum pricing” policy that has captured Holyrood and No 10 Downing Street, so (putting it crudely) consumers are about to pay through their pockets for a duff index. It’s a good example of how misleading statistical indices can lead to poor policy-making.

The alcohol-related hospital admissions statistics are, in my view, just as misleading, and this year will include both a broad measure – the one that in the past has led to headlines about alcohol-related admissions exceeding a million a year – and a narrow measure.  The broad measure sums the alcohol attributable fraction (AAF) associated with each admission, basing the AAF on the “worst-case” diagnosis, including both primary and secondary diagnoses.

That means that if somebody is admitted to hospital with a primary diagnosis unrelated to alcohol but has a range of comorbidities, the AAF for that admission will be that of whichever secondary diagnosis has the strongest link to alcohol. As I and others have pointed out, this method is prey to the increased number of diagnoses hospitals  now attach to their patients in a bid to garner more money through payment by results. What it measures is strongly influenced by coding drift.

This problem is avoided by using the narrow measure which looks at only the primary diagnosis, though this too may not be perfect. For example, admissions caused by alcohol-related violence or accidents will not be included since these causes are not included as primary diagnoses.

This year the NHS IC is promising to provide us with estimates of admission numbers for previous years that would have applied if current coding had applied in those years – that is, it will seek to correct for coding drift. The effect will be to diminish the apparent growth of alcohol-related admissions markedly. It will be interesting to see by how much.

NHS IC will also launch on Thursday a consultation on the methods used to estimate alcohol-related admissions. The current method has served well as a stick to beat the drinker with, and anti-alcohol campaigners will be loath to lose it.

The political argument over alcohol stacks up like this. The industry prefers the narrow measure, for obvious reasons: it is a much smaller number, and is not subject to fashions in diagnosis or to coding changes. The industry’s opponents prefer the broad measure, because it makes the problem look worse, and growing. Of the two the narrow measure, though imperfect, is probably a better way of measuring changes over time.