Dr Foster Intelligence is hosting its first ever live tweet chat on Tuesday 16th October, to discuss the findings of a preliminary data release, due to be published exclusively in the Guardian that day, in advance of its annual Hospital Guide.
The tweet chat is the first ever opportunity for the public, healthcare professionals and politicians to discuss live with Dr Foster, journalists and eminent clinicians why pursuing transparency through releasing such data is important.
The latest findings shed new light on weekend care in the NHS. A range of health experts will be on Twitter to discuss why patients are more likely to die when admitted on a Saturday or Sunday, and what the NHS can do to improve the quality of weekend care. Ahead of the publication of Dr Foster’s annual Hospital Guide in November, our panel will also be fielding questions around how mortality figures are calculated and what this data means for patients.
The chat will be taking place at 12:30-13:30 GMT on Tuesday 16th October and will feature:
- Roger Taylor – Co-Founder and Director of Research, Dr Foster Intelligence (@RTaylorDrFoster)
- Denis Campbell – Health Editor, Guardian (@Denis_Campbell)
- Prof Brian Jarman – Emeritus Professor at Imperial College (@jarmann)
Join the discussion using the Dr Foster Hospital Guide hashtag, #DrFHG, or tweet (@drfosterintel) or email (info@drfoster.co.uk) your question in advance.
Perhaps Dr Foster could address the denominator issue once and for all by comparing death rates calculated using day of admission and those calculated using day of discharge, evaluating both elective and non elective care seperately. Worth discussion and directly related is the etent to which length of stay is a feature in these assessments of death rates.
Many Thanks
Luke Readman
The analyses we publish of weekend mortality are based on day of admission not day of discharge. Analysis by day of discharge does not work, as ‘live discharges’ are heavily influenced by work patterns and change dramatically over the weekend. It might be possible to do an analysis taking as the denominator all the patients in hospital on a particular day and calculating a mortality rate for the day and seeing how that changed at weekends. However, it’s not unlikely that one of the factors that would be important in such an analysis would be length of time since admission so it could end up getting rather circular. The standard academic approach to this topic has been to use day of admission as the point of reference.
To Gavin’s question – The data we published looks at crude rates to demonstrate the absolute change in death rates. There are fewer admissions on weekends than weekdays but a rise in the mortality rate. In the conditions looked at in the published data, the drop in the rate of admissions at weekends is greater (more than 10 per cent) than the increase in the crude mortality rate. Thus, the absolute number of daily deaths at the weekend is lower than the absolute number of daily deaths on a weekday, despite the fact that the mortality rate has increased. There is also an increase in the riskiness of patients admitted at weekends, though this is much smaller than the increase in the mortality rate; after risk adjustment there is still a significant increase in the mortality rate. This picture is consistent with other studies of the topic.
You are right about the need to look in more detail at the data – although I don’t think that is needed in order to establish whether the effect is real – there is plenty enough evidence of that. However, examining this at a local level is needed to better understand what specific issues may exist. The analysis by diagnosis shows the extent to which it is as much about community services as it is about hospital services.
– Roger Taylor
Director of Research, Dr Foster
I agree with Luke Readman that we need more detail. Actual numbers rather than ratios. The case has not been proved that the casemix is the same weekdays and weekends (are fewer patients admitted weekends). Standardised mortality ratios are a pretty crude measure, especially when we have other ways to look at the data.
The question I would like answered by the stats is do more people die after being admitted on Saturday and Sunday than during the week (and not just a bigger ratio)
There are lots of problems with the data which can only be said to generate hypotheses rather than proving anything.
It’s observational – so subject to bias and confounding and so cause and effect cannot be linked. (I wish John Humphrys on the Today programme would go on a basic stats course before he gets so outraged about this sort of thing)
Many people assume that the difference in mortality must be because there are fewer doctors in hospitals at the weekend but this is not necessarily the case.
The first obvious problem is that patients admitted at weekends may “on average” be sicker than those admtted during the week – fewer emergency patients are admitted during weekends than during the week – very sick patients are likely to present to hospital throughout the week but mild or moderately sick patients may find it difficult accessing medical advice at weekends (no GPs) and so hold out until Monday. This could be checked by comparing the Charlson Morbidity index of patients admitted at weekends to those admitted as emergencies during the week – Dr Foster should have this data but have not presented it.
Dr Foster does not give the actual figures (as pointed out above) just relative risk. A 10% increase in mortality on an assumed 3% 30 day mortality for admission, gives a 3.3% mortality (not really a very substantial change). You would have to treat 300 patients as emergencies to see one additional death. Differences on this scale are very hard to prove with such inadaquate data.
Dr Foster has corrected the data for morbidity (very sick patients with lots of problems have a higher than expected risk of dying). The data for adjusting patient risk comes from hospital coding data which may be inaccurate and incomplete (many audits have shown this). A systematic error, such as fewer diagnoses being coded for weekend patients compared to patients admitted during the week would result in an underestimate of the weekend patients’ morbidity. This will mean more patients will appear to have died than expected at weekends compared to weekdays.
Strangely, the data published by Dr Foster last November showed that whilst there appeared to be a higher mortality in patients admitted at weekends, the patients were most likely to actually die on a Wednesday. The lowest mortality rate was actually at the weekend. I suppose this suggests that patients admitted over the weekend are surviving until Wednesday before dying. What happened on Monday and Tuesday? Certainly at the very least this would suggest that the time delay between admission and death should be scrutinised to see if other factors may be important. Dr Foster has this data but has not released or analysed it as far as we know.
There are other problems, too boring to go on about such as the fact that the data on how many “senior” doctors (consultants and registrars) were available on weekends was extracted from survey data that looks very dubious and inconsistent. It was provided by people in trusts who may have had little idea of what the data was going to be used for and little incentive to make sure the data was accurate. An inspection of some of the ratios of senior doctors to hospital beds reveals different figures for the same hospitals for different tables of data. This lack of consistency within Dr Foster’s dataset is not explained.
I think the data are interesting and suggests there may be problems to be addressed but the interpretation put on this data may be rubbish. The trouble is that interested parties (politicians, senior doctors at RCP, Dr Foster, leaderwriters etc) have jumped on a bandwaggon of dubious provenance. Of course everyone wants to see a “senior” doctor ASAP when they go to hospital and, in fact, in many hospitals they do.