Skip to main content

Rebuttal of Mail on Sunday article "Hundreds MORE die unexpectedly at crisis hospitals"

14 December 2016: An article published by the Mail on Sunday on 11 December, said that "hundreds of patients are feared to have died unexpectedly at three NHS hospitals".

It claimed that the number of "unexpected deaths" at United Lincolnshire Hospitals NHS Trust (ULHT) had "almost tripled", citing the Summary Hospital-level Mortality Indicator (SHMI) figures published by NHS Digital.

NHS Digital publishes SHMI together with a clearly defined methodology to support transparency and we welcome the use, and further analysis, of the data. However the article has several flaws.

Firstly, SHMI 'expected deaths' figures are not an actual count of patients, but are instead a statistical construct, which estimates the number of deaths that may be expected at a trust on the basis of average England figures and the characteristics of the patients treated there (including the conditions the patient is in hospital for, other underlying health conditions, age, gender and method of admission to hospital).

There are, however, many other factors which can affect a trust's SHMI, including (but not limited to) the quality of the data upon which the calculation is based; other patient characteristics not listed above (including severity of condition that is not available from hospital data); the organisation of services; the availability of resources both at the hospital and in the local area; and the quality of care.

Secondly, a range of observed deaths can be considered to be 'as expected'. If the observed number of deaths falls outside of this range, the trust in question will be considered to have a higher or lower SHMI than expected.  Although the SHMI values for Northern Lincolnshire and Goole NHS Foundation Trust and Colchester Hospital University NHS Foundation Trust have increased between 2013/14 and 2015/16, their SHMI values remain in the 'as expected' banding meaning that their SHMI is not significantly different from the national average. 

Finally, it's important to realise that there is no suggestion whatsoever that mortality rates have tripled at ULHT. There are several other valid ways the statistics could be presented - overall the crude death rate for ULHT has increased by just under 0.2 percentage points, or around 4 percent, over the period.

In summary, the SHMI is not a measure of quality of care. A higher than expected number of deaths should not immediately be interpreted as indicating poor performance and instead should be viewed as a 'smoke alarm' which requires further investigation. Similarly, a 'lower than expected' or 'as expected' SHMI should not immediately be interpreted as indicating good performance.  We would encourage all trusts to investigate their SHMI in more detail to help them to further understand mortality outcomes for their trust

The SHMI cannot be used to directly compare mortality between trusts and it is inappropriate to rank trusts according to their SHMI.

You can find further guidance info on how to interpret the SHMI here: 


Last edited: 11 April 2018 5:38 pm