Monday, May 08, 2017

The truth is what you want it to be. The danger of dodgy health statistics.

Last week's UK press had lots of commentary about the high rates of alcohol consumption by older people and its cost to the NHS. This is something of an annual event that I first reported back in 2012.

I don't know if there is a problem with older people drinking too much but I am damn sure that neither does the Government if it places any reliance on the numbers churned out by the NHS.

Let me explain why I am suspicious of the data and appalled at the way it is used by the media.

Here is the report from last week's Guardian newspaper

Headline ; Baby boomers warned over alcohol intake as hospital admissions soar

Sub Heading ; Alcohol was the primary or secondary issue in an estimated 1.1m admissions in 2015/16, up 64% from a decade ago

You cannot help but be shocked when you read headlines like this - those lucky old boomers - not only are they a wealthy bunch but they have the time and money to drink themselves into an early grave.

Let's dig a little deeper into the numbers. The Guardian, like much of the media wrote the headline about the combination of primary and secondary causes for admission. If you read the NHS publication that is the source of this data it states that the percentage of hospital admissions where alcohol was the primary reason has changed very little in the last 10 years.

Now let's look at this strange measure of where alcohol was a 'secondary' reason for admission. This is a much larger figure (1.1 million people) but these are not real people they 'virtual patients" . This is where it all gets very complicated. There is a thing called the 'alcohol-attributable fractions' which is a matrix of 63 * 17 cells that links medical conditions, gender and age an their likelihood that if you have one of the conditions it will be alcohol related. So for instance, if you have high blood pressure and female aged 45-64 the NHS thinks there is a 31% chance that it is alcohol related. This document will tell you all about it. Just to make the point about the complexity this is an image of the matrix.

The document explains that the total number of alcohol-related hospital admissions, as described by the indicators, is not a number of actual people or a number of actual admissions but an estimated number of admissions calculated by adding up all of the fractions (i.e. the different illnesses that you have that might be related to alcohol).

Are you still with me? This image should help explain

This shows that 2 alcohol related hospital admission could be made up in a series of ways. Two people who are clearly suffering from an alcohol overdose. Four people with Epilepsy (for the life of me I cannot understand why) or ten people suffering from high blood pressure or indeed 25 people with breast cancer.

Now just to make matters worse, the NHS has a crisis in the way that it codes the reasons you go into hospital. In a report published by Capita about the quality of clinical coding in the NHS it found that :

in 46% of the trusts audited the medical case notes were big and cumbersome, in poor physical condition and disorderly. Add to that 54% of the health trusts had vacancies for staff and because of the pressure 42% of the managers responsible for coding were forced to code themselves.

Do you think I am right to be concerned? The press always reports the headline numbers and not the detail - the methodology is suspect and even more so the quality of the coding.

Is there a problem with alcohol and older people - probably - do these stats help detail the magnitude. Not in my book.

The reason for the rather long and complex blog posting is that I am increasingly worried by the quality of so-called research and the way it is reported. The media is anxious for some catchy headlines and will not bother to check the accuracy of the data which means a lot of people have a very false perception of the reality. Dick Stroud

No comments: