Discussion in 'Advocacy and Cycling Safety' started by kingrollo, 29 Dec 2017.
You do realise that the world is not uniform?
With only something like 110 cycling deaths a year nationally, it's unusual to have 3 nearby. Is it an area that's especially busy like London or shoot for cycling like the West Midlands?
Yeah, you got it - West Midlands !
I've a question with the injuries per x number of miles thing. The number of injuries comes from hospital records, where does the mileage come from?
I think the number of casualties usually comes from police records, not hospitals (I think I've written about that above) and I think the mileage most often comes from the National Travel Survey - but basically dig into any time someone mentions cycling casualty rates per mile and they should be able to tell you where they're getting the mileage from.
There are often problems with using the NTS for cycling estimates, both due to the small sample size and classifying journeys where you used something else for a greater distance as "train" or "car" or whatever and ignoring the cycling in some measures BUT those problems are pretty consistent over time, so you can compare the casualties per mile over time and between areas unless told of a method change which means you shouldn't.
For example, the latest https://tfl.gov.uk/corporate/publications-and-reports/travel-in-london-reports (number 10) says on page 57 that their cycling mileage is taken from the London Travel Demand Survey. If only more local government collected such detailed data. Norfolk has been decommissioning their counters recently, so we're going from poor data to less of it!
The injury stats come from NHS records. It’s part of the standard reporting primary and secondary care are required to make. The data is very patchy.
That has to be true. The average copper is not known as a diagnostician, so won't be able to record a serious injury.
They use a very simplistic method which is described in some document on gov.uk but IIRC boils down to serious meaning being taken to hospital or a few other visible injuries and all else being minor or killed.
Reg is correct that there are injury stats from hospitals, but I think they're not as widely used, partly because they're patchier. I've yet to see my local public health department use them at all.
I think the confusion comes from the fact that the main database, STATS19, is administered by the police. However, the data in there comes from a variety of sources,
The police data in STATS19 provides data on things like the types of vehicles involved, what roads they were on and the consequent casualties.
The details around the types of injuries treated and clinical outcomes come from the NHS data which is fed into STATS19.
Just a health warning for any NHS reporting...
That data might be gathered by audit (people reporting particular events, Which is generally of good quality but can be under reported), or it can be gleaned from operational data.
Operational data doesn't exist to give accurate statistics, but merely move people from ward to ward, get them their meds, and operations. Those too are subject to errors, but more endemic than an audit. If you are searching on a term, it might not be labelled correctly or systematically one hospital might use a different terminology. It could contain coded items, free text... any number of inaccuracies.
Given the drive for insights, banks of humans review data to bend it to reality. With a growing backlog, and little clarity, things often don't even get a bit closer.
Then the assumption is, the data is handled, extracted, cropped by appropriate cohort with all the correct kind of joins then reassembled in a national dataset... with extreme professionalism and accuracy at each step?
So, you know...NHS data... It's as good a guess as anyone could make.
I asked the county analyst responsible "is the data only from the police, or does hospital data get into it?" and got this reply: "CRASH data, as was the case with STATS19, comes exclusively from the Police. Hospitals have their own system for recording admissions known as HES (Hospital Episode Statistics) which is not linked to CRASH. Efforts have been made in the past to connect the two data sets, but this has proven to be almost impossible due to NHS data protection policies."
Maybe it happens in some parts of the country, but not in Norfolk and almost certainly not in Suffolk (which shares a roads policing unit).
It's alarming ... and completely wrong. The government statistics given below show 26 deaths per billion cyclist kilometres. That works out as one cyclist death per 24 million miles. You're out by a factor of twelve.
Passenger casualty rates for different modes of travel - GOV.UK
Separate names with a comma.