A date with death


You may perhaps have noticed some stories in the media about weekend staffing in NHS hospitals. This was triggered when Jeremy Hunt, the Health Secretary, announced his plans to forcibly alter Consultant contracts to enforce weekend working.

In justification for these measures, he quoted some statistics, and made some additional statements.

These were:

“You are 15% more likely to die in hospital if you are admitted on a Sunday compared to a Wednesday”

“6000 people lose their lives every year because we don’t have a proper 7 day service”

“This is the reason the government will introduce mandatory 7 day contracts for consultants”

“This will not increase the number of hours worked by any individual doctor”

He then went on to cast some aspersions on our professionalism, and our sense of vocation.

If you want to hear this from his own lips, the link is here:


He doesn’t give the source of his data, but it wasn’t too hard to find a paper in the Journal of the Royal Society of Medicine with some similar statistics. It’s a nicely designed and well written paper.

If you’re interested in reading the paper and coming to your own conclusions, it’s freely available online here:


What does the paper really say..?

Statement 1: “You are 15% more likely to die in hospital if you are admitted on a Sunday compared to a Wednesday”

Jeremy Hunt’s first statement is a safe one, and the paper backs him up. In fact, he got his figure slightly wrong – the headline finding is that being admitted on a Sunday is associated with a 16% increased risk of death in hospital in the next 30 days compared to a Wednesday admission.

But what does this mean, and what might be causing this interesting finding?

First, it’s important to look at the detail. The comparison being made is in death rates in hospital in the 30 days after admission. But perhaps the 30 day time window is not relevant – perhaps the extra deaths that underlie this statistic occur almost immediately after admission? Well, fortunately this question is also answered in the paper. They present an analysis of deaths occurring in hospital by the day the death occurred (in contrast to Jeremy Hunt’s figure, which related to day of admission). And here, the outcomes are neatly reversed. If you’re in hospital on a Saturday or a Sunday, you are significantly less likely to die than if you’re in hospital on a week day. The paper offers a plausible explanation for this – fewer medical procedures occur on weekends, leading to fewer complications, and hence fewer deaths. In other words, weekends keep the patients safer because there are fewer doctors around to kill them. Food for thought.

It’s something of a stretch to imagine that staffing levels on the day of admission have an influence on death rates for 30 days after that day. Only 2 days in 7 are weekend days – so there should be plenty of time with full staffing levels to get things under control. It’s a stretch, but it’s not impossible.

The authors of the paper acknowledge that this type of study, looking back at old data, cannot definitively explain what caused the effects they recorded. But they do offer some possibilities. One of these is that reduced staffing, including lack of senior doctors, tests, senior nurses, radiographers etc has an adverse effect on the admissions occurring on weekends (although we’ve already seen that fewer actual deaths occur on weekends). It’s not hard to imagine that urgent procedures might be delayed, or diagnoses might be missed, because the hospital is running with fewer staff, and this might affect outcomes a few days or weeks down the line.

But there are other possibilities mentioned too, and they are worth thinking about. For instance, although the authors tried their best to compare like with like, there may be differences in the type of patient coming in on a week day and a weekend.

Let’s start by considering elective admissions. These are people coming into hospital for non-urgent procedures that will be almost certainly performed on a week day. If there are 4 patients on a Monday operating list, who will come in on the Sunday? The more complex or difficult case, the person with major comorbidities. And if a consultant has an operating list on a Monday and another on a Friday, which list will have his complex cases? The Monday list, as he doesn’t want the complex cases to go through the difficult post-op period over the weekend. So the elective patient admitted on a Sunday is the most difficult case from the most complex operating list, and has the highest mortality, but this is nothing to do with staffing levels when he is admitted, which of course is before he has had his surgery.

How about emergency admissions? Is there any reason to imagine that emergency admissions on a Sunday will be different to those on a Monday? Again, it seems there is. Imagine two identical people who are both pottering in their garden on a Sunday afternoon. They are both the same sex and age, both have the same comorbidities. One suffers a major stroke, and is unable to speak, paralysed down one side of his body, and collapses. An ambulance is called and he is rushed to hospital. The other notices that his left hand is a bit clumsy, so he gives up in the garden. He considers seeing a doctor, but it’s Sunday, so instead he goes to bed early. When he wakes up on Monday morning and the hand is still not right, he sees his GP and is sent to hospital on Monday. These patients have similar risk profile, and the same diagnosis (stroke). They both became ill on the same day (Sunday). But the sicker of the two is admitted on Sunday, and the less sick is admitted on Monday. There are other reasons, aside from patient behaviour, why weekend admissions may be different. Consider a patient turning up at A+E with a terminal diagnosis. On a weekday, she might be diverted to a hospice. On a weekend, with a skeleton service running, she will be admitted to hospital.

Finally, it’s worth taking a moment to think about the rather dull topic of coding. Every time a person is admitted to hospital, their admission is converted into a series of codes. These include demographic information about a patient, information about their other illnesses (comorbidities), information about the illness that has brought them to hospital, procedures carried out, and the outcome of the admission (e.g. death or discharge). This information is recorded during the admission by the clinical staff, and converted into codes by “coders” who get their information from the medical records. The codes are important for the hospital, as this is how NHS hospitals get paid. But this information is also used in research studies, such as the article linked above, to improve their modelling. For example, a person coming in with a heart attack who has had several heart attacks before, is a heavy smoker, has diabetes and high blood pressure, and is 95 years old, has a higher mortality risk than a 35 year old with no risk factors. If coding is not done properly, you can unknowingly be comparing a sick patient with a less sick patient, and the research findings become much less reliable. An important example of this is the Leeds paediatric cardiac surgery unit, where the department was shut down due to higher than expected mortality, but the cause was actually poor or absent clinical coding. Patients were not being flagged as high-risk cases, and therefore the death rates seemed higher than they should be. http://www.telegraph.co.uk/news/health/news/9980869/Sir-Bruce-Keogh-admits-that-inaccurate-data-led-him-to-suspend-childrens-heart-surgery.html

Is it possible that record keeping, and therefore clinical coding, is worse on weekends, and that leads to an apparent higher mortality rate for those admissions? Yes.

Statement 2: “6000 people lose their lives every year because we don’t have a proper 7 day service”

Here, Mr Hunt has confused association with causation, which is a schoolboy error. For instance, there is an association between Nobel Prize Winners per capita in a country and Chocolate consumption (link below) but that doesn’t prove that one causes the other.


There is a link between Sunday admission and increased likelihood of death in hospital in the following 30 days, but the reasons for this are unproven.

And they are not necessarily “extra” deaths. For example, another explanation for the difference would be if the less sick people choose to wait until Monday to come to hospital and the sickest ones don’t wait, or the sickest pre-operative patients are admitted on Sundays so that their operations can be done on Monday morning. If the difference is caused by comparing sick patients (coming in on Sunday) with less sick ones (coming in on Monday) then there may be no extra deaths.

In the Leeds cardiac surgery unit, poor coding caused death rates to appear “double the national average”. In fact, once the data was corrected, the death rates were within the normal limits. So something banal like incomplete or inaccurate coding of patient data can imply “extra deaths” that were not really extra at all.

Statement 3: “This is the reason the government will introduce mandatory 7 day contracts for consultants”

So, according to Mr Hunt, ensuring more consultant cover on weekends is going to reduce mortality. Before allocating blame for these (possibly imaginary) unnecessary deaths to the lack of weekend consultant cover, how about finding a way to test this hypothesis? Surely, it would be easy to take a specific condition (say, stroke), and look at how staffing levels affect death rates.

In fact, someone did think of this. Here is the study…


It’s a big study, looking at nearly 57,000 patients in 103 stroke units. And the presence of consultants on weekends did not affect death rates. (The thing that did affect death rates was the level of trained nursing cover on weekends).

And, as many many people who have actually worked in hospitals have pointed out since Mr Hunt’s speech, no-one can deliver healthcare on their own. The best, most experienced and most dedicated consultant would be unable to deliver any useful care without a team of professionals. Nurses, nursing assitants, pharmacists, physiotherapists, radiographers, junior doctors… the list goes on.

Statement 4: “This will not increase the number of hours worked by any individual doctor”

Mr Hunt seems to believe that he can change the NHS to a “24/7” system by changing our rota around a bit.

How can you extend a system from 5 days to 7 without doctors working more hours? Either you need more doctors (which the government won’t provide), or you need to shift doctors from weekdays to weekends, making an already over-stretched system even worse. Even a politician couldn’t argue with that logic, could they?

So are you saying there is no problem?

There probably is a problem. But the data that exists is not black and white. And it would be a good idea to understand the problem before trying to change things.

One interesting thing to do would be to compare the NHS to other health systems, to see whether we lag behind other countries such as America in weekend care. Luckily, there’s a shiny brand new paper on this topic too…


If you don’t want to read it, I can summarise. People who come into hospital on weekends do worse in other countries too (Holland, USA, Australia). No-one is claiming that the NHS is a perfect system, or that the weekend care is flawless, but it certainly isn’t lagging behind. There is no scandal of poor weekend care here.

Of course, we don’t go to work aiming to be no worse than anyone else. We want the care we deliver to be the best. And if anyone can suggest real ways we can improve, we will listen.


This blog now has a twitter account: @abrainia

Please follow if you want to hear about future posts.


UPDATE (28/7/15):

This article from the New Statesman is an excellent summary of the situation as a whole, including a more detailed look at the economics of 7 day working.


UPDATE (13/5/17):

I’ve written a short update here: “Another Date With Death