Road casualty statistics: up a cul-de-sac

How many people are seriously injured on Britain’s roads every year? It sounds a simple question, but it certainly doesn’t have a simple answer.

Yesterday the Select Committee on Transport of the House of Commons quizzed Richard Aldritt, Head of Assessment at the UK Statistics Authority, about the discrepancies between figures collected in different ways.
 
First, there are the figures of the numbers “killed or seriously injured in accidents reported to the police” which are used to see if the Department for Transport is meeting its targets. These are going down quite nicely, and the target (a 40 per cent reduction by 2010 against a baseline which is the average for 1994-98) is in line to be met.
 
The actual figures, collected in the data series called STATS19, are shown below. They have come down from just under 40,000 in 1995-96 to 26,000 in 2007-08. But note that hospital admissions for road traffic casualties, collected from Hospital Episode Statistics, show no such decline. As Helen Joyce reported earlier on Straight Statistics, the discrepancy between these two sets of data needs explaining.
 
 
  
 
In addition, in the STATS19 series, injuries have fallen faster than deaths – by 41 per cent compared to 29 per cent. Deaths are reliably recorded, with very few missed. So either the two trends are diverging, with the ratio of deaths to serious injuries changing, or some serious injuries are not being recorded as such.
 
A second set of figures comes from the National Travel Survey, which since 2007 has asked a sample of people if they have suffered an injury in a road accident. This suggests that in 2007-08, 1.8 per cent of all adults suffered an injury of some sort. However, the STATS19 figures indicate that only 0.4 per cent did. So if the NTS is right, STATS19 is picking up only about a quarter of all injuries.
 
For serious injuries, STATS19 produces a figure of 26,000 for 2007-08, but the NTS a figure of 220,000 – eight and a half times greater. “Is this not a very dramatic difference?” asked the committee chairman, Louise Ellman MP. Mr Aldritt agreed. 
 
Faced with this huge discrepancy, Matthew Tranter, a DfT statistician, makes a heroic guess at the actual numbers (below) and comes up with a “best approximation” for the annual number of road casualties for the UK of 800,000, to the nearest 10,000, and of serious injuries of 80,000 - but it could be as few as 40,000 or as many as 120,000. That’s a pretty wide range, for something that it ought to be possible to count with reasonable precision. (His detailed analysis of the figures, in section 5 of the report, is well worth reading.)
 
             
 
It’s also some way from the STATS19 figure of 26,034 serious injuries against which success in the target is measured. But Mr Aldritt refused to be drawn on whether this meant the department hadn’t met its target. Because the target was clearly defined as “those killed or injured in accidents reported to the police”, then only STATS19 data applies. (Whether accidents reported to the police are identical to accidents recorded by the police is not entirely clear but if there is a difference – and Mr Aldritt conceded there may be – nobody knows how to measure it.)
 
The figures are the responsibility of the department, not of the Statistics Authority, he made clear. The authority has already made plain it expects improvements. “So has the Government met its target?” Mrs Ellman insisted. “Yes, as crafted” said Mr Aldritt. “Because it is couched in those terms, it is the target.”
 
Among the changes the authority sought were changing the name from “reported casualties” to “police-recorded casualties” to make it clear what the source actually is. The department prefers “reported road casualties” which Mr Aldritt said he could live with.
 
“There is no evidence that the department is doing anything wrong” he said. “STATS19 data is a very powerful source, and a good job is being done, within its limitations.” He added that the National Travel Survey had been a very important step, but questioned whether the sample size was sufficient to meet a wide range of needs.
 
The suspicion remains that STATS19, the police-reported data, has been the victim of targetitis, and changing policing methods. With fewer police on the roads, some accidents may go unrecorded, and with targets to meet, some injuries may be under-reported. An electronic system for recording, called CRASH, might help, Mr Aldritt suggested.
 
Instead of licking their pencils and scribbling in their notebooks, police will enter details of accidents on a hand-held computer. CRASH should be trialled in the middle of next year, and available to all forced by the end of 2010, says the National Policing Improvement Agency. Better allow a bit of lee-way – it’s an IT system.
 
Meanwhile, we have only the haziest idea of how many people are actually seriously injured on the roads. It could be 26,000, 80,000, or 220,000. Amazing but true.