Statistics behind the headlines: Have there been 13,000 needless deaths at 14 NHS trusts?

Widespread reports of 13,000 preventable deaths at NHS trusts should be ignored, argues Professor David Spiegelhalter on bmj.com yesterday.

In a feature published yesterday Winton Professor for the Public Understanding of Risk at the University of Cambridge, David Spiegelhalter, challenged media reports that there had been 13,000 ‘avoidable’ deaths as 14 NHS Hospital Trusts.

The Keogh Report, published in July, investigated 14 trusts that had “higher than expected” mortality rates between 2005 and 2012. Before its release there were sensational headlines talking of the large numbers of “needless deaths”.

In the BMJ opinion piece Professor Spiegelhalter explains the fragile basis for these statistics, as well as how they were misconstrued by many media outlets.

As he points out, comparing mortality rates with an expected number of deaths based on a calculation of average national performance will inevitably lead to “around half of all trusts [having] ‘higher than expected’ mortality”.

It would be absurd to label all these as outliers”, the professor argues. Yet the press widely chose to report the statistics by interpreting a higher than expected number of deaths as ‘needless’.

NHS LogoHe explains that the ‘13,000’ is the difference between the total observed and expected number of deaths across all fourteen trusts, over a seven year period, using a particular measure developed by the Dr Foster Unit. However the Keogh report did not feature these statistics, and explicitly stated that it would be “clinically meaningless” and “academically reckless” to attempt to quantify avoidable deaths in this way.

Professor Spiegelhalter concludes by asking what can be done, and reveals that a new national indicator of avoidable hospital deaths is being commissioned by Keogh. In the meantime, he recommends describing possible outliers as ‘above the expected range’, and avoiding talking about “numbers of deaths”.

Comment Here!

comments