Thursday, 31 May 2012

Lies, damned lies and alcohol statisitcs (revisited)

The National Health Service Information Centre (NHS IC) has released its latest report on alcohol statistics. As ever Nigel Hawkes of Straight Statistics is on hand to try and help us with comprehension. However, even Nigel is confused:

The bottom line is that the NHS IC admits that its old method of measuring admissions attributable to alcohol exaggerated the increase since 2002-03 because it failed to take account of changes in hospital coding practice.

So that's clear. Except: still publishes a table (4.2) based on the old, discredited calculations, as well as another table (4.11) that takes account of the adjustment. Which are we supposed to believe?  According to Table 4.2, total alcohol-related admissions rose between 2009-10 and 2010-11 from 1,056,900 to 1,168,300. But according to the adjusted measure in Table 4.11, they fell over the same period from 1,208,100 to 1,168,300.

Right, alcohol-related admissions have either risen or fallen over the last year. This is very helpful.

Nigel thinks that the measures are all a lot of toss (he actually says "unfit for purpose, adjusted or unadjusted") but this won't stop the usual collection of nannying fussbuckets from making the most of the figures. And guess which figures those folk will use?

There is a serious problem with NHS coding (and not just for alcohol-related conditions) - it relies on turning complex conditions into answers to a check list and is open to manipulation both at the point of collection and subsequently. Yet we are making huge decisions about health priorities on the basis of this sort of dubious, unreliable data.


No comments: