Letters Hospitals put into special measures in 2013

The BMJ report on mortality cuts in hospitals put into special measures in 2013 is misleading

BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h1794 (Published 07 April 2015) Cite this as: BMJ 2015;350:h1794
  1. Michael Campbell, professor of medical statistics1
  1. 1University of Sheffield, Sheffield S1 4DA, UK
  1. m.j.campbell{at}sheffield.ac.uk

The report on the Dr Foster analysis of mortality in hospitals put under special measures is misleading.1 The widely quoted figures are of a 9.5% reduction in mortality in the 11 trusts put into special measures (“Keogh” hospitals) compared with 3.3% nationally.2 The “national” figure is derived from 11 randomly chosen trusts selected with replacement 1000 times. Presumably this is what is meant by “thousands of randomised samples from other English trusts” mentioned in the report.1 The analysis is further complicated by a “broken stick” model being fitted but the break point not being chosen as the point when the trusts were placed in special measures but where the model chose it to be (p 10 of Dr Foster’s report), which is not clear in The BMJ’s report.

Hospital standardised mortality ratio (HSMR) has dropped steadily nationally over the past 10 years, and Dr Foster compared the rate of decline for the last year of this period for Keogh and other hospitals. In my view, the relevant graph to summarise the effect of special measures is that on page 8 of Dr Foster’s report.2 Approximately, the figure shows drops in HSMR over the past two years of 20% for Keogh hospitals and 19% for national hospitals, with Keogh hospitals going from 10% above national hospitals to 7% above.

Although Dr Foster mentions changes in coding after special measures, it would be good practice to give the observed and expected values by quarter separately, as well as the ratio, to show that the increased decline in HSMR in Keogh hospitals was not due to an increase in expected deaths rather than a reduction in observed deaths.


Cite this as: BMJ 2015;350:h1794



View Abstract

Sign in

Log in through your institution