Observations

The statistical sins of Jeremy Hunt

BMJ 2015; 351 doi: https://doi.org/10.1136/bmj.h6358 (Published 26 November 2015) Cite this as: BMJ 2015;351:h6358
  1. David Craven, Royal Society research fellow
  1. 1University of Birmingham, Birmingham, UK
  1. cravenda{at}bham.ac.uk

Four things the health secretary needs to get right

The health secretary for England, Jeremy Hunt, has been criticised recently over his handling of statistics. Although it can be easy to make statistical errors, his handling of the so called “weekend effect” and new contract for junior doctors has seen some particularly egregious examples.

Weekend mortality—In October Fiona Godlee, editor in chief of The BMJ, wrote to Hunt to pull him up on his conflation of the concepts of correlation and causation and to try to persuade him to stop making this error.1 A study in The BMJ (doi:10.1136/bmj.h4596) had shown a correlation between admission to hospital at weekends and death rates. It found that 11 000 more people died each year within 30 days of admission to hospital on a weekend than on a weekday.2 Hunt repeatedly told MPs and the public that these excess deaths were due to poor staffing at weekends. But the study’s authors did not apportion any cause for the excess or take a view on what proportion of those deaths might be avoidable.

Consultant availability test—Shortly after, on 28 October, Hunt told parliament that, across all specialties, only 10% of hospitals were able to have consultants see patients within 14 hours on weekends. The data he used were not even broken down into weekdays and weekends, making such a claim impossible. The test was that each of the 10 hospital specialties had to see 90% of their patients within 14 hours, otherwise the hospital failed. Moreover, these statistics relied on each specialty in each hospital either guessing or checking a sample of just 10 sets of case notes to give a figure; not all case notes were included. Such a small sample size means that the dataset is highly unreliable, and the insistence that each specialty meets this target separately further compounds this unreliability, so many hospitals that should pass even this rather inappropriate and unrepresentative test will fail because of faulty data. The data simply cannot be used in this way.

Junior doctor contract—Hunt has made numerous contradictory or misleading claims, but we start from his assertion that the new contract is “cost neutral”—that is, that the wage bill will not change. If this were true, then every pound given to one doctor must be taken away from another: so how does this square with his claim of an 11% pay rise? It’s actually an 11% rise in basic pay, compensated for by a big reduction in overtime pay and by reclassification of more hours as part of the standard working week. Also, Hunt claimed that no doctor working within the legal limits would be worse off under the new contract, but what does he mean? Within the limits set by the European Working Time Directive? Or the limits for those doctors who have opted out of the directive? Or does he mean the reduced maximum hours proposed under the new contract? Unsurprisingly Hunt had to backtrack on this somewhat nebulous claim and admit that junior doctors who had opted out would face a pay cut.

Hunt’s next test?—A new study published in The BMJ this week (doi:10.1136/bmj.h5774) shows that babies born at the weekend have a slightly greater risk of death than those born on a weekday.3 This observational study states that it does not show a causal link and that it should not be used as evidence of such. The opportunity for misrepresentation here is clear, and it will be another indictment of Hunt’s sloppy approach to statistical rigour if he seizes it.

Notes

Cite this as: BMJ 2015;351:h6358

References

View Abstract

Sign in

Log in through your institution

Free trial

Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
Sign up for a free trial

Subscribe