The Tory Health Secretary, Jeremy Hunt, provoked the first ever strike by doctors in NHS England last year when he tried to force through a new contract for junior doctors that would have significantly worsened pay and conditions. He justified this on the spurious grounds that:
- There was a weekend effect whereby patients admitted to hospital at weekends had a significantly higher risk of dying (the Department of Health (DH) published references to eight studies which were claimed to prove this);
- Rectifying this effect required more junior doctors to work longer at weekends. This was supposed to be part of the government’s promise to introduce a “seven-day” NHS without any extra staff; and
- This had to be achieved without costing any more.
Hunt’s use of the Tories’ supposed mandate to introduce a seven-day NHS is in itself thoroughly misleading. Hospitals have always operated throughout the week and both junior and senior doctors work at weekends. It is in primary care, GPs’ surgeries, that a five-day NHS operates, and experimental weekend GP services tend not to be much used by patients. But, even admitting Hunt’s seven-day claim, is there actually a weekend effect and are junior doctors’ hours a factor?
Previously, I showed that the DH’s eight studies on the weekend effect included only two independent pieces of work.1 Those studies showing a weekend effect did not try to explain it but suggested that a lack of senior doctors at weekends might be one factor: none referred to a role for junior doctors.
Since the DH’s publication of Hunt’s evidence, the DH itself admitted that it had no evidence that a seven-day NHS would have any effect on deaths or on time spent in hospital. Since the DH’s evidence also, curiously, showed a decreased rate of deaths at weekends, it is conceivable that things might get worse!
Using Hunt’s cited papers, I showed that greater illness among weekend admissions could completely account for increased mortality. Now Professor Sir Nick Black, an adviser to DH and NHS England, has blown Hunt’s case out of the water with more objections to the whole idea of a weekend effect.2
Black shows first of all, referring to his own work in 2010, that methods of calculating hospital death rates (Hospital Standardised Mortality Rates – HSMRs) were flawed.3 HSMR is the ratio of Observed Deaths to Expected Deaths. The observed deaths are not so easy to get wrong as it’s fairly obvious when a patient has died. However, the estimate of expected deaths can be more or less accurate, depending on the completeness of the information available about patients. Ideally, the ratio will be 1:1, i.e. expected deaths will be the same as actual deaths. But, an underestimate of expected deaths will produce an apparent excess of observed deaths, and questions will be asked.
The obvious question, “Did we get our estimates right?”, does not seem to have occurred to Hunt and his advisers. Black describes three problems with the expected deaths calculation.
- First, some patients’ conditions (morbidities) are miscoded. Black illustrates this with a study on stroke patients, published in May 2016 in the British Medical Journal but inexplicably missed by Hunt and his top medical adviser Professor Sir Bruce Keogh.4 This study found that stroke patients admitted as non-emergencies on weekdays (with lower risk of death) were frequently miscoded as new stroke patients (with a higher risk). Their lower actual rate of death resulted in weekend emergency stroke admissions having an apparently increased risk of death. When the coding was correct, the weekend effect disappeared!
- Second, the particular characteristics of each case are not always accurately recorded as a result of delays in doing tests and this can affect estimates of survival, as well as actual survival! It might be expected, according to Hunt’s arguments, that this would be a problem at weekends. Black refers to another study of stroke patients, again published in 2016 in another top medical journal, The Lancet, and again inexplicably missed by Hunt and advisers.5 This study found no weekend effect when comparing the quality of health care associated with different days and times of admission. For your information, the worst time to be admitted was overnight on weekdays.
- Third, patients often have co-morbidities (more than one thing wrong with them) and may not die of the condition for which they were admitted. Other conditions are less likely to be noted or rated for seriousness for weekend admissions which tend to be emergencies. This is important since each condition should contribute to the estimated probability of an individual’s death. If some conditions are not recorded, the expected deaths are underestimated, producing an apparent excess of observed deaths. Black here refers to another 2016 study6 that examined attendances and admissions from all English A&E departments for an 11 month period. Similar numbers attended on weekdays and weekends but significantly fewer were admitted to hospital on weekends (27.5% versus 30%). Weekend admissions tended to be direct from the community, rather than via GPs, and were significantly sicker than weekday admissions. This means that a greater proportion of that smaller number admitted at weekends died within 30 days, not because of poorer care but because they were sicker.
The last point has been confirmed by another 2016 study7 using a new scale of risk of dying based on seven physiological variables. They found that patients admitted from A&E departments at weekends were sicker on average. After adjusting for this, they did not have a greater risk of dying than equally sick people admitted on weekdays.
So the weekend effect does not exist and nor do Hunt’s “11,000 extra deaths per year.” But how many extra deaths occur because of the government’s refusal to fund the NHS and social care adequately?
2Black N. Higher Mortality in Weekend Admissions to the Hospital: True, False, or Uncertain? JAMA 2016:316(24);2593-4