Earier I reblogged a post from the House of Commons Library which fact-checked David Cameron’s claims that waiting times at A&E have fallen. It found those claims wanting. About half an hour later, the post was removed and replaced with this message:
“The blog post “Have A&E waiting times fallen?” has been removed by the House of Commons Library as it does not meet our expected standards of impartiality. A revised post will be uploaded as soon as possible.”
It seems probable that this has been done following howls of protest from the Conservative Party. It seems as though stating facts isn’t impartial if those facts are at odds with something the PM says.
Luckily, I was able to find a cached version here. Here’s the original text:
Yesterday at PMQs the Prime Minister stated:
When [Labour was in Government], the average waiting time [in A&E] was 77 minutes; under this Government, it is 30 minutes.
This claim has been made a number of times by Ministers in the House of Commons. But other reports indicate that that A&E performancehas declined, with the 4-hour target having been missed in England’s major departments every week since last July. So where does the PM’s claim come from, and is it correct? Here I’ll suggest that it relies on a simplistic reading of the data, and that the measure he refers to is not the most natural indicator of the “average waiting time” in A&E.
Means and Medians
The PM is referring to the mean “duration to initial assessment” in A&E – i.e. the time between when a patient arrives and when they are first assessed*. (This indicator is not available in the routinely published data, but can be viewed in these tables which were given in response to a parliamentary question). Here is a chart of the mean and median time to initial assessment from 2008 to 2012:
Two things should jump out from this data: first, the large difference between the median and the mean waiting time; and second, the dramatic fall at the beginning of 2011-12.
Whenever there is a large difference between the median and the mean, this usually indicates that the mean is being distorted by some outlying values – in this case, some very long waits for initial assessment. In their publications on A&E waiting times, the Health and Social Care Information Centre warn that these extremes are “particularly sensitive to poor data quality and definitional issues”. This suggests that the mean value here is not a good indicator of time to initial assessment in A&E; so we should rely on the median value to tell us what the typical time to initial assessment in A&E is. But the median has remained more or less unchanged at around 10 minutes to initial assessment, which does not show the trend that the PM refers to.
What about the dramatic fall in the mean in April 2011? This corresponds with the time to initial assessment in A&E being designated as a “care quality indicator” and becoming subject to mandatory reporting (along with several other similar measures). So it’s plausible that the fall in the mean in April 2011 reflects an improvement in data collection, quality and reporting, rather than any genuine change in waiting times. It’s also worth noting that the 2012/13 data has over three times as many data points – “attendances with a valid duration to initial assessment” – than the 2008/09 data, and so is likely to be more reliable than this earlier data.
What is the ‘average waiting time’?
So the data on time to initial assessment does not support the PM’s statement. But there is a further problem here: the time to initial assessment is not a natural indicator of the typical waiting time in A&E. After all, the median time to initial assessment is only 10 minutes: patients will obviously still have to wait in A&E, often for a significant period, after their initial assessment is completed. If, after a trip to A&E, you were asked “How long did you have to wait?”, it is very unlikely that you would reply by reporting the time to your initial assessment. Instead, you’d probably respond by saying how long it took before you were treated, or perhaps giving the total time you spent in A&E.
Thankfully, these measures are both included in the published data on A&E quality. Here is a chart showing how these have changed between 2008 and 2013:
On these measures – which as we saw above, are more natural ways to report “average A&E waiting times” – there has been no reduction in waiting times. Time to treatment is static save for seasonal variation, and total time in A&E has been steadily increasing. A fuller analysis of the published data suggests that total time in A&E is only increasing for patients who require admission – and that total time in A&E for non-admitted patients is also unchanged.
The data does not show that the average time in A&E has fallen since 2008. Rather, the typical total time in A&E has risen (for admitted patients, at least), and the typical time to treatment has remained static.
It is welcome that the rich data on the amount of time patients spend in A&E is becoming part of the wider political debate on the NHS. But in order for it to be useful and informative, it must be discussed in a way which fully respects the data.
* Note that the Health Secretary stated on 9th June that the median waiting time had fallen from 77 minutes. This is false.
Author: Carl Baker