Intended for healthcare professionals

Analysis

Increased mortality associated with weekend hospital admission: a case for expanded seven day services?

BMJ 2015; 351 doi: https://doi.org/10.1136/bmj.h4596 (Published 05 September 2015) Cite this as: BMJ 2015;351:h4596

Re: Increased mortality associated with weekend hospital admission: a case for expanded seven day services?

Dear Sir,

I read the latest response from Pagano and Freemantle regarding the emergency admissions subgroup analysis with great interest- as expected the offset 'weekend effect' persists, as it did in the 2012 paper, albeit to a lesser extent.

I would contest however the authors have still failed to properly adjust the case mix. The paper clearly delineates a cohort of patients, in the top quintile of predicted risk of death, that is markedly different between the weekday and the weekend. By the authors own survival model this quintile has an average risk of death of 7.88% at 30 days. On a Sunday 29.4% of patients admitted were in this top quintile, on Saturday 24% and every weekday 20%.

For simplicity let us assume the risk of death in the bottom four quintiles is 0 (in reality it is <1%). Using the actual admission numbers reported, with a 7.88% predicted risk of death, we would expect there to be approximately 43, 012 deaths in each weekday cohort, which gives an absolute death rate of 0.016 deaths/patient admitted. For patients admitted on Saturday this is approximately 24,446 deaths (0.02) and Sunday 21,764 (0.023).

This totals an expected 261271 deaths- which is 89% of the total actual deaths reported, representing the lions share of any mortality contribution.

If you were then to calculate the relative risks of the above you would extrapolate a ratio of Wed to Saturday hazard ratio to be 1.23 and a Wed to Sunday ratio to be 1.43. The fact that the apparent actual hazard ratios are 1.1 and 1.15 respectively should be applauded as a testament to the great care received at the weekend.

The authors would argue that they have corrected for this already by excluding deaths within three days of admission and rerunning the analysis. With such a large effect, a subgroup with mortality four times the population average at thirty days, not three, this seems hardly adequate.

Could it be that the authors have unwittingly delineated the true cause of the internationally observed 'weekend' effect?

It is unfortunate the authors conclusions neglect this vital characteristic of their patient population, which quite clearly explains the apparent discrepancy.

Yours faithfully,
Dr Dominic Pimenta
Core Medical Trainee, London

Competing interests: No competing interests

03 November 2015
Dominic R Pimenta
Core Medical Trainee, London
None
London