Nailing Hunt's lies
Here is the soundbite which Hunt has been repeating ad nauseam: "And at the moment we have an NHS where if you have a stroke at the weekends, you're 20% more likely to die. That can't be acceptable."
Sounds quite impressive? I tweeted him to ask for evidence and got no reply so I decided to look for my own. It wasn't hard to find. I'll deal with the best and most recent example. The headline finding from this paper is that patients admitted at weekends with a stroke diagnosis have an increased 7-day mortality of 19%, apparently just as Hunt claims: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4487251/
The 7-day mortality for week-end admissions is 12.9% and for week-day admissions is 11.1% and if you factor in comorbidities etc. this gives you an adjusted mortality odds ratio of 1.19. Just as Hunt says.
But let's look more closely. At some numbers buried in the text of the paper and not even given a p value, even though they are probably the most statistically significant thing in the paper: "The overall admission rate for stroke was higher on normal week days (111 per 100 000 population; 95% CI = 110–113) than on weekends (88; 86–90) and public holidays (95; 88–102)."
So the admission rate for stroke is substantially lower at week-ends. 88 versus 111 per 100,000. See those confidence intervals? They imply that the standard error of the mean is around 1. Which means that a comparison of the two means would give a t statistic of around 13, which would give a p value so tiny that most statistics software would not even be able to calculate it.
Given that there is no earthly reason for the incidence of stroke to fluctuate wildly by the day of the week, what is obviously happening is that fewer patients with stroke are getting admitted at week-ends and that the ones who are admitted then are more severe, and hence have a higher mortality.
In fact, we can go further and quantify things. We'll make a couple of assumptions. One is that the incidence of severe stroke, i.e. one with a chance of killing you, is constant for each day of the week. Unlike, say, RTAs, alcohol poisoning or whatever. The second assumption is that somebody with a severe stroke is going to present and get admitted the same day. Strokes happen suddenly, unlike say pneumonia, appendicitis, whatever.
If we accept this, then the day of admission is the same as the day the patient has their stroke. At week-ends, 88 per 100, 000 people have a stroke and get admitted to hospital, of whom 12.9% die within 7 days, which is 11.4 people. On weekdays, it's 111 per 100,000 people having a stroke and coming in, of whom 11.1% die within 7 days, which is 12.3. So out of 100,000 people, every week-end day 11.4 get a stroke which will kill them in 7 days compared with 12.3 for every week-day.
So Hunt is lying. If you have a stroke at the week-end you're not 20% more likely to die. In fact, according to the best evidence, you're less likely to die. Out of people admitted to hospital more die, but that can only reflect their increased severity.
The main point about all this is not just that he is misleading people on this particular issue. It is that this is the best he can come up with. The DoH is packed with people smarter than us whose only job is to find stuff to support the minister. If this is the best they can come up with it doesn't reflect their incompetence. It means that there really is no evidence at all for a "week-end effect". It means there is no problem that needs fixing. It means that Hunt is deliberately picking a fight with the medical profession over a complete non-issue, presumably to further his own political ends.
Sounds quite impressive? I tweeted him to ask for evidence and got no reply so I decided to look for my own. It wasn't hard to find. I'll deal with the best and most recent example. The headline finding from this paper is that patients admitted at weekends with a stroke diagnosis have an increased 7-day mortality of 19%, apparently just as Hunt claims: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4487251/
The 7-day mortality for week-end admissions is 12.9% and for week-day admissions is 11.1% and if you factor in comorbidities etc. this gives you an adjusted mortality odds ratio of 1.19. Just as Hunt says.
But let's look more closely. At some numbers buried in the text of the paper and not even given a p value, even though they are probably the most statistically significant thing in the paper: "The overall admission rate for stroke was higher on normal week days (111 per 100 000 population; 95% CI = 110–113) than on weekends (88; 86–90) and public holidays (95; 88–102)."
So the admission rate for stroke is substantially lower at week-ends. 88 versus 111 per 100,000. See those confidence intervals? They imply that the standard error of the mean is around 1. Which means that a comparison of the two means would give a t statistic of around 13, which would give a p value so tiny that most statistics software would not even be able to calculate it.
Given that there is no earthly reason for the incidence of stroke to fluctuate wildly by the day of the week, what is obviously happening is that fewer patients with stroke are getting admitted at week-ends and that the ones who are admitted then are more severe, and hence have a higher mortality.
In fact, we can go further and quantify things. We'll make a couple of assumptions. One is that the incidence of severe stroke, i.e. one with a chance of killing you, is constant for each day of the week. Unlike, say, RTAs, alcohol poisoning or whatever. The second assumption is that somebody with a severe stroke is going to present and get admitted the same day. Strokes happen suddenly, unlike say pneumonia, appendicitis, whatever.
If we accept this, then the day of admission is the same as the day the patient has their stroke. At week-ends, 88 per 100, 000 people have a stroke and get admitted to hospital, of whom 12.9% die within 7 days, which is 11.4 people. On weekdays, it's 111 per 100,000 people having a stroke and coming in, of whom 11.1% die within 7 days, which is 12.3. So out of 100,000 people, every week-end day 11.4 get a stroke which will kill them in 7 days compared with 12.3 for every week-day.
So Hunt is lying. If you have a stroke at the week-end you're not 20% more likely to die. In fact, according to the best evidence, you're less likely to die. Out of people admitted to hospital more die, but that can only reflect their increased severity.
The main point about all this is not just that he is misleading people on this particular issue. It is that this is the best he can come up with. The DoH is packed with people smarter than us whose only job is to find stuff to support the minister. If this is the best they can come up with it doesn't reflect their incompetence. It means that there really is no evidence at all for a "week-end effect". It means there is no problem that needs fixing. It means that Hunt is deliberately picking a fight with the medical profession over a complete non-issue, presumably to further his own political ends.
Sympathetic though I am to your message, it surely means little that the NUMBER of people who die is lower after weekend admissions. If you are admitted, what matters to you is the probability that you'll die, and that is a bit higher ate weekends.
ReplyDeleteThe real question is why admissions are lower at weekends. Given the smaller number of admissions at weekends, it does indeed seem likely that the admissions are more ill and that alone could easily account for the observed increase in the probability of dying for weekend omissions. I guess that, to be devil's advocate, other explanations are possible. For it could be that the severity is always the same but the more relaxed life-style at weekends makes strokes slightly less likely. There is no way to be sure from the data that we have.
In any case, by quoting relative rather than absolute risks, Hunt has made it sound as alarmist as possible, The difference between 12.9% and 11.1% sounds very small even if it's real, not sampling error. A change in relative risk of 19% sounds much more serious. Most people think it's better to look at absolute risk.
I'm afraid I don't agree. Probability of having a stroke and dying is same every day. Probability of having a (?mild) stroke and being admitted is higher weekdays. Medically, it's very implausible there is any marked difference in incidence of stroke by day of week.
ReplyDeleteWhat on earth is your justification for assuming that because fewer stroke patients are admitted at weekends, they have more severe strokes? There are very good reasons why more people should suffer strokes during the working week than at weekends. I think you should rewrite this piece with more research and less prejudice.
ReplyDeleteYou've misunderstood the point. It is that fewer people are admitted on weekends, thus the ones who are admitted are the ones who are most severe and more likely to die. As for your second point, I would argue that we are more likely to abuse our bodies on the weekend with things like higher alcohol consumption and number of cigarettes smoked. You're the one who comes across as having an agenda or prejudice. Your argument simply doesn't hold up.
DeleteI agree that it's implausible, but plausibility is never a strong argument. I suspect that it simply isn't known.
ReplyDeleteMore to the point, do you disagree that what matters is you probability of dying? If the number that sie is smaller simply because the number of admissions is smaller, that doesn't mean you are safer.
Since leaving my first comment, I've looked at Freemantle (http://www.bmj.com/content/351/bmj.h4596.full.pdf+html ) and that says that weekend admissions are indeed more severe. That reduces the weekend effect, but doesn't entirely eliminate it, sadly.
Thanks for this.
ReplyDeleteUnderstand the analysis but is there not an inate contradiction in the assumption that the day of admission is the day that the person has their stroke and the fact that less people present at the weekend but they have more serious conditions. Is the conclusion therefore that some people with less serious strokes do not present at the weekend but wait until Monday with a likely poorer outcome for them. This would be an argument for a seven day NHS although it is not the one that Jeremy Hunt is making.
It also doesn't require a change in working patterns, only a change in the public's understanding about what weekend care is provided.
DeleteSpecialist stroke units are fully staffed 24/7. If the weekend effect is due to staffing there should not be one for people admitted here. The authors say there was insufficient data to show if the effect was eliminated in a specialist stroke unit. Surely stroke units have done an audit?
ReplyDeleteMany questions emerge from this article including the ambulance service and care of the elderly at weekends but this is an interesting speculation.
https://petition.parliament.uk/petitions/112404
ReplyDeleteProfessor Curtis, thank you for your note. I too looked for the paper and played about with a spreadsheet. The conclusion you reached can be arrived at more straightforwardly. The foot of the first data column of Table 2 gives the no. of admissions for each day of the week. The next column gives the 7 day mortality, so the no. of deaths for each day can be calculated. This shows clearly that admissions at the w/e are significantly lower; it is this (the denominator) that causes the higher mortality rate not the higher number of deaths (the numerator). In fact the no of deaths at the w/e is also lower than those during the week (!) but not as much lower than the level of admissions.
ReplyDeleteIt is not medically improbable! People do different activities at weekend: some drink (alcohol) more, others drink less(water, as carers are away). Stressors increase at weekends. For more information on publication, kindly search AHA journals. Stroke is not the only condition with weekend effects on incidence and severity.
ReplyDeleteSurely this is an argument for focussing resources on weekday strokes rather than weekend strokes :)
DeleteMany patients with mild strokes may present to their GP after a weekend having not appreciate the seriousness of their symptoms. They then get admitted on a weekday, hence making the weekday figures look better.
ReplyDelete"Medically, it's very implausible there is any marked difference in incidence of stroke by day of week."
ReplyDeleteBit it's very plausible that people suffering 'unusual symptoms' (of which they do not know the significance of) delay or avoid making contact with primary or emergency care services at the weekend because they don't want to cause a fuss - especially with the constantly reinforced messages about A&E delays and lack of resources, thus the difference is rooted in behavioural responses rather than purely medical. There could be a variety of outcomes here, in that hospitals are more likely to see 'serious' strokes at the weekend, balanced against the fact that during the week the strokes may be less serious, but with a poor clinical outcome due to the delay in treatment. A complex interaction of symptom, reaction and treatment that I think you cannot dismiss the significance of easily.
I'm mostly saying what you are, which is that less severe strokes will present during the week. But if delay was producing worse outcome then this would increase the proportion of people dying after admission on Monday. Which is not what is observed. I maintain that for "severe" strokes people generally get admitted on the day it happens. Delaying may indeed cause problems for less severe strokes but is not reflected in an increase in 7-day mortality.
DeleteIf somebody can show me a proper peer-reviewed paper which says different I'll be willing to change my mind. I'm a perfectly competent medical practitioner, albeit specialised in psychiatry, and am, for example, a statistical reviewer for the Lancet. I am not aware of any literature or any physiology which makes me think it likely that incidence of severe stroke will have strong associations with day of the week. And don't forget, the vast majority of strokes occur in people who are not of working age. Occam's razor says the incidence is the same every day. The mortality is the same every day. Therefore, there is no effect on mortality from having what is currently provided at NHS hospitals at week-ends. In Wales.
ReplyDeleteI could add that my father died of a stroke and that his father died of a stroke and I suppose it's pretty likely that I will die of a stroke. I have no personal interest in stroke services being anything but optimal. So far as I can see, in terms of impacting short term mortality, there isn't much room for improvement.
Fantastic blog Dave.
ReplyDeleteEven if you are wrong in your analysis (which I don't think you are) you clearly demonstrate that what Hunt is claiming is untrue. The strongest thing the Rt Hon Gentleman could say would be something like "The evidence seems to suggest that there MIGHT be a higher rate of death for people who have strokes at weekends. However when you look at the data properly there is a very big difference in the number of patients admitted and thus the data is without doubt SKEWED hence this is most likely to be a statistical anomaly. What this paper clearly shows is that we need to look at this more closely."
Now, I could forgive the Health Secretary for getting this wrong once (although he has a whole department who's job it is to help him get it right) but after it has been repeatedly pointed out to him that his statistical claims are not just flawed but wrong (by the editor of the BMJ no less) I find myself unable to give him any benefit of the non-existing doubt.
Hence, Mr Secretary, this is what I think: www.tinyurl.com/jeremyhunt
Please share this image widely
AFZ
Like the previous note from AFZ I find it incredibly frustrating to hear the same Govt line 'all we are trying to do is prevent worse clinical outcomes at the weekend'. This is a convenient line, appears to be at best unproven and is clearly Govt policy to keep repeating it, regardless of whether or not it is true, in an effort to convert more people to the view they are trying to improve the NHS, rather than take this first step to removing all unsociable hours payments for all NHS employees.This nonsense was repeated by Michael Portillo last week on the Andrew Neil programme 'This week' and well done to the young Doctor who calmly refuted the claim. As many people have said many times - I would believe Junior Doctors every time versus Jeremy Hunt, Michael Portillo, David Cameron this morning on Radio 4. He backtracked when challenged by the lady interviewer on the 20% claim showing again that the figure is not correct.
ReplyDelete