Is hospital care actually worse on the weekend?

The HiSLAC study ran two point prevalence surveys of hospital specialists in English acute NHS trusts, one on Sunday 15 June 2014 and one on Wednesday 18 June 2014, then linked the results at trust level to Hospital Episode Statistics covering every adult emergency admission in financial year 2013-14. Of 141 eligible acute trusts, 127 agreed to participate and 115 (91%) contributed data to the point prevalence survey. Of 34,350 clinicians surveyed, 15,537 (45%) responded (Aldridge et al., 2016).

The weekend staffing gap it documented was large. Substantially fewer specialists were present on Sunday (1,667, or 11% of the 14,532 specialist responders) than on Wednesday (6,105, or 42%). Median specialist intensity on Sunday, defined as self-reported specialist hours per ten emergency admissions between 0800 and 2000 hours and adjusted for each trust's survey response rate, was 48% of the Wednesday figure (IQR 40 to 58). The specialists who were on site on Sunday worked about 40% longer per person than their Wednesday counterparts and still delivered less than half the intensity. In 104 of the contributing trusts (90%), the Sunday to Wednesday ratio was below 0.7.

Weekend admission was associated with higher mortality: an adjusted odds ratio of 1.10 (95% CI 1.08 to 1.11, p<0.0001) against weekday admission. The result that received far less attention is what happened when the authors put the two halves of their own study together. There was no significant association between a trust's Sunday to Wednesday specialist intensity ratio and its weekend to weekday mortality ratio (r -0.042; p=0.654). Their own reading is that this cross-sectional analysis did not detect a correlation between weekend staffing of hospital specialists and mortality risk for emergency admissions.

Aldridge and colleagues were careful about what that null means, and readers should be equally careful. They noted that the absence of an association in a preliminary cross-sectional study does not allow anyone to discard the hypothesis, and they warned against attributing the weekend effect to a single component in a complex system. Their null is an ecological comparison between trusts, which has limited power to detect staffing effects that operate inside a single trust. It is evidence that weekend specialist cover did not track weekend mortality across English trusts. It is not evidence that specialists do not matter.

Meacock et al. (2016) looked at where the denominator comes from. In a retrospective observational study of all 140 non-specialist acute hospital trusts in England from April 2013 to February 2014, covering 12,670,788 accident and emergency (A&E) attendances and 4,656,586 emergency admissions (940,859 direct admissions and 3,715,727 via A&E), they found that hospitals admitted 30.0% of weekday A&E attendances and 27.5% of weekend attendances. A higher bar for admission at weekends means the admitted weekend population is on average sicker than the admitted weekday population.

That shows up directly in the mortality numbers, depending entirely on who is counted. Measured against admitted patients only, weekend admission carried odds of death of 1.055 (95% CI 1.042 to 1.068) in the A&E records and 1.054 (95% CI 1.040 to 1.069) in the admission records. Measured against all A&E attendances, the weekend odds ratio was 1.010 (95% CI 0.997 to 1.022), which is not statistically significant. Part of what the weekend effect measures is therefore an artifact of the denominator rather than a difference in what happens to comparable patients.

Walker et al. (2017) tested a related question with patient-level records: are weekend patients measurably sicker when they arrive? Across 503,938 emergency admissions for 257,596 individuals at four Oxford University Hospitals sites between 1 January 2006 and 31 December 2014, 24,383 (4.8%) admissions ended in death within 30 days. Crude 30-day mortality was 4.7% for weekday admissions (18,313 of 385,647) and 5.1% for weekend admissions (6,070 of 118,291), p<0.0001.

Adjusted for standard case-mix factors, the relative risk of death within 30 days compared with a Wednesday admission was 1.08 (95% CI 1.03 to 1.14) for Saturday and 1.09 (95% CI 1.03 to 1.14) for Sunday. Adding 15 routine hematology and biochemistry results taken at admission moved those to 1.07 (95% CI 1.01 to 1.13) and 1.05 (95% CI 1.00 to 1.11). On the authors' accounting, routine test results explained 33% of the Saturday excess (95% CI 21 to 70), 52% of the Sunday excess (lower 95% CI 34), and 87% of the public holiday excess (lower 95% CI 45). Hospital workload showed no independent association with mortality.

This became a political argument in England before it was a settled scientific one. The most cited anchor for the claim, Freemantle et al. (2015), is classified by the BMJ under Analysis rather than Research, and it was corrected on 29 March 2016. The correction was not a numerical fix. Its substance was that co-author Bruce Keogh, then national medical director of NHS England, was a long standing proponent of improving NHS services seven days a week, and that this should have appeared in the authors' conflict of interest statement. Two of the seven authors sat in NHS England's Medical Directorate while the government was making the political case for seven-day services. That context belongs with the citation.

There is also a limit on what this review can vouch for. The Freemantle figures were not verified at source for this page, so the only magnitude carried here is second hand: Aldridge and colleagues describe their own 10% figure as lower than the 11% to 16% relative risk increase quoted by Freemantle and colleagues, who included elective admissions in their analysis and used 30-day mortality following admission. Different populations and different outcome windows produce different summary figures, which is one reason this literature looks noisier than it is.

So the disagreement is not about whether weekend admission is associated with higher mortality. Across large English datasets it reliably is. The disagreement is about why. Meacock et al. (2016) show a large share of the gap moving when the denominator changes. Walker et al. (2017) show a third to a half of it accounted for by how sick patients already are on arrival. Aldridge et al. (2016) show that the trusts with the thinnest weekend specialist cover were not the trusts with the worst weekend mortality. None of these findings rules out a care-quality contribution, and none of them establishes one.

Every figure here comes from the English NHS and describes acute hospital inpatients. Aldridge and Meacock use national English data from 2013-14, and Walker covers four Oxford hospitals through the end of 2014. All of it predates the seven-day services rollout, later workforce changes, and the pandemic, so treat these as historical baselines rather than a description of current performance. They say nothing about weekend staffing outside healthcare and nothing about non-acute settings.

Keep the scales straight, because this literature mixes them freely. Aldridge and Meacock report odds ratios, so an odds ratio of 1.10 means 10% higher odds of death, not 10% more deaths. Walker reports relative risks. The Aldridge staffing result is a correlation coefficient across trusts, a different quantity again. Treating 1.10, 1.055, and 1.08 as though they were the same measurement of the same thing conflates three different quantities.