Accident statistics and data accuracy

[Woodbridge Pete]

Does anyone know how accurate the published statistics and data are for incidents. 5yrs ago my wife and I were involved in a bad accident (hit from behind @ 70mph, air ambulance, A road closed) but cannot find any details of the incident on the mapping type sites, e.g. www.Crashmap.co.uk
It made the local newspaper, but doesn't seem to have got on the records correctly. I've found what looks like the incident on a gov.uk file, based on date and time on the same road, but it puts it about 5 miles further along the road, so to me that twists the stats on what is a poor junction.

[The utility cyclist]

According to Bez who runs the BeyondtheKerb blog, STATS19 can be quite a bit out in terms of cycling deaths alone so this doesn't surprise me.
Maybe you could contact the government stats people directly and maybe see if you can get the data changed roadacc.stats@dft.gsi.gov.uk

As you rightly point out, when you have missing information or that information relates to a totally different area it can falsify/mislead what is actually happening at a location and potentially why it happened. Some might say why bother to dig these things up but to you it's a massively traumatic event that I would think you want others to be able to not be in same/similar scenario so want to attempt to force local authority to make changes in whatever way you can using data as part of your persuasion.

Personally I think it's all down to motorists, there are no dangerous junctions/roads, only dangerous motorists (and indeed dangerous cyclists and pedestrians) whose actions put others at risk of harm and indeed do harm to the point of death or life changing injuries. it's sickening and makes many more than a little bit angry that those at national and local level can do something but won't, often citing 'well there have being no incidents or few incidents at x so nothing needs to change'.

I hope you can get your incident amended and that helps with your end goal whatever that may be.

[Bez]

Yes, there's a reasonable bit of under-recording. I've not made a statistical estimate as to the extent of it, though I know a handful of people who may have. When it comes to fatalities specifically, I'm not sure that under-reporting is an issue, but there are many reasons for fatalities either appearing as serious injuries or not being recorded in this data set at all. One county a couple of years back had four cycling fatalities but none appeared in Stats19: they were all recorded as medical deaths. (I keep meaning to trawl through the coroners' reports for those; one or two cases raised an eyebrow to say the least, given the events reported in the media.)

I would think you want others to be able to not be in same/similar scenario so want to attempt to force local authority to make changes in whatever way you can using data as part of your persuasion. Personally I think it's all down to motorists, there are no dangerous junctions/roads…

Rather a contradiction there. If there are no dangerous junctions or roads, what point is there in seeking changes to them? In reality, of course, both behaviour and engineering are factors. Or rather, if behaviour is always the problem then it can be greatly mitigated through engineering. The thing about engineering is that you can absolutely change it, and by doing so you can increase a system's resistance to problematic behaviours (especially those which arise from known and universal cognitive failings such as saccades, vision cones, assumptions, etc rather than outright malice or recklessness such as drunk driving, racing or fleeing police). There are quite a lot of nuances here: some incidents may be behavioural in the sense that two people acted and a collision resulted, but if you say the problem is behaviour then you can't solve problems which will inevitably repeat due to either the fact that most people are pretty careless by nature or the fact that our brains operate in specific ways that we have no conscious control over (even if we can, with some effort, train ourselves to minimise the problems those ways cause). Engineering is a catalyst, if you like, and for the most part once you understand the behaviour it's perfectly possible to figure out what the catalyst is and what is necessary to fix it.

So—returning to the original question—if you know of an incident, for which you have a police incident number or other unique and reliable identifier, why not email the stats folk and ask why it's not recorded. Even if it doesn't result in a change to the data set, you may find out something illuminating about how these incidents do (or don't) end up in the records.

[The utility cyclist]

Yes, there's a reasonable bit of under-recording. I've not made a statistical estimate as to the extent of it, though I know a handful of people who may have. When it comes to fatalities specifically, I'm not sure that under-reporting is an issue, but there are many reasons for fatalities either appearing as serious injuries or not being recorded in this data set at all. One county a couple of years back had four cycling fatalities but none appeared in Stats19: they were all recorded as medical deaths. (I keep meaning to trawl through the coroners' reports for those; one or two cases raised an eyebrow to say the least, given the events reported in the media.)

I would think you want others to be able to not be in same/similar scenario so want to attempt to force local authority to make changes in whatever way you can using data as part of your persuasion. Personally I think it's all down to motorists, there are no dangerous junctions/roads…

Rather a contradiction there. If there are no dangerous junctions or roads, what point is there in seeking changes to them? In reality, of course, both behaviour and engineering are factors. Or rather, if behaviour is always the problem then it can be greatly mitigated through engineering. The thing about engineering is that you can absolutely change it, and by doing so you can increase a system's resistance to problematic behaviours (especially those which arise from known and universal cognitive failings such as saccades, vision cones, assumptions, etc rather than outright malice or recklessness such as drunk driving, racing or fleeing police). There are quite a lot of nuances here: some incidents may be behavioural in the sense that two people acted and a collision resulted, but if you say the problem is behaviour then you can't solve problems which will inevitably repeat due to either the fact that most people are pretty careless by nature or the fact that our brains operate in specific ways that we have no conscious control over (even if we can, with some effort, train ourselves to minimise the problems those ways cause). Engineering is a catalyst, if you like, and for the most part once you understand the behaviour it's perfectly possible to figure out what the catalyst is and what is necessary to fix it.

So—returning to the original question—if you know of an incident, for which you have a police incident number or other unique and reliable identifier, why not email the stats folk and ask why it's not recorded. Even if it doesn't result in a change to the data set, you may find out something illuminating about how these incidents do (or don't) end up in the records.

It's not a contradiction at all, a road or junction in itself is not dangerous, how can it be, the danger comes from human behaviour, however as you well know in lieu of government/police actually forcing/coercing change in that behaviour of those that cause the harm (or simply removing a user type) we have only other pitiful options like changing road design and speed limits. Even that doesn't really do much, lower speed limits would be great IF they were enforced but they aren't and so has next to little bearing when the vast majority ignore it.

Modern motorvehicles 'straighten' out roads and bends can now be taken at speeds without any braking due to wider wheels, improved suspension, power steering etc. All down to human behaviour and not the design of the highway. Brighter lights allow motorists to drive faster at night, have less reaction time and so more tech is brought in and yet incident numbers are not really improving, again all down to human behaviour.

[Bez]

I think you missed the points I made after the bit you highlighted

Engineering isn't an alternative to enforcement. Enforcement can (given the status quo) be the better way to deal with certain behaviours: ie those which are malign in nature rather than those which are errors of judgement. But engineering can be used to tackle behaviours which we all exhibit by our nature, as well as often preventing harm from outright recklessness. Engineering doesn't make bad guys good, and it doesn't change how the human brain works. What it can do is to prevent those things from resulting in harm.

Moreover, while enforcement is by nature reactive, engineering is preventative. You don't have to wait for someone to commit an offence before attempting to fix the problem, you can minimise the chance of the problem ever occurring in the first place.

And, a point of detail: I disagree with the suggestion that changing a speed limit is engineering. It merely changes the parameters of enforcement.

[Bez]

An analogy:

A rise in laceration injuries is observed outside pubs in summer months. Some are caused by people being glassed in fights, others by people stumbling and falling onto broken glass lying around as a result of accidental drops, and so on.

Do you (a) ban anyone with a conviction for assault and anyone you see accidentally dropping a glass, or (b) use plastic glasses?

[Cyril Haearn]

(b) & (c) make driniking less attractive, more expensive
Stoptober!

[mjr]

I would think you want others to be able to not be in same/similar scenario […]

I never wrote that. Clean yourself up!

[mjr]

Does anyone know how accurate the published statistics and data are for incidents.

We know they're inaccurate, based on comparisons with hospital data, such as that done in https://injuryprevention.bmj.com/content/15/1/13 - what we don't know is how inaccurate they are, especially about things like location. It might be interesting for someone to go through a sample from Bez's casebook from a while ago and see how well the RRCGB incidents on the crashmap or collideoscope match the press descriptions... but even then, which is in error?

[Bez]

I would think you want others to be able to not be in same/similar scenario […]

I never wrote that. Clean yourself up!

LOL, sorry, that's a full-on brain fail for me there. Sorry! I have duly cleaned up

[thirdcrank]

We've discussed the severity of injuries as recorded in the Stats booklets before. Somebody, possibly Road Peace, did a comparison of police stats with hospital records and IIRC, there were signs that the police stats had recorded minor injuries when the hospital had them as severe.

Re fatals, unless something has changed, the Stats booklets have to be finalised within 30 days so if somebody dies after that time, I'm not sure of the mechanism for getting the death included.

AFAIK, The stats results are published annually and include a detailed explanation of how the different categories are allocated.

I'm puzzled about the possibility of the location being wrong as the Stats booklets require a six figure grid reference and certainly in my day, the computer couldn't accept it if it didn't correlate with the street name location.

(If nobody like gaz can remember the earlier discussion, I'll try to find it when I have more time.)

[Bez]

It might be interesting for someone to go through a sample from Bez's casebook from a while ago and see how well the RRCGB incidents on the crashmap or collideoscope match the press descriptions... but even then, which is in error?

I did pick through the 2015 data to identify discrepancies…

https://beyondthekerb.org.uk/casebook/c ... repancies/

Unfortunately by the time the 2016 data came out late last year I'd gone off the boil and wasn't putting much effort into this stuff, so I never got round to doing the same for the following year (or completing a more detailed look at the 2015 Lancashire fatalities—though I believe I do have a half-written analysis sitting behind the scenes).

Press descriptions can often be way off the mark, though. One the one hand, they usually give a lot of information which you obviously don't get from the data set (particularly photos, although even they can be misleading if you're not careful); but on the other there's quite a bit of work involved if you want to start checking the validity of that information, and in most cases you simply can't. They're useful for prompting questions about what really happened, but it's best not to assume that they give the answers.

[thirdcrank]

A bit of thread resurrection here containing a detailed analysis by snibgo although I've not yet found the stuff comparing police and hospital stats

viewtopic.php?p=470167#p470167

[Bez]

Re fatals, unless something has changed, the Stats booklets have to be finalised within 30 days so if somebody dies after that time, I'm not sure of the mechanism for getting the death included.

Yes, it's still 30 days. Normally the result is that a casualty would be recorded as a serious (or sometimes slight) casualty, provided an injury is apparent at the time, and I'm fairly sure there is never a retrospective change to the data if death occurs later.

[mjr]

Re fatals, unless something has changed, the Stats booklets have to be finalised within 30 days so if somebody dies after that time, I'm not sure of the mechanism for getting the death included.

My understanding is that there is no such mechanism. If the victim doesn't die quickly enough, they will only ever be a serious injury in the road casualty data.

I'm puzzled about the possibility of the location being wrong as the Stats booklets require a six figure grid reference and certainly in my day, the computer couldn't accept it if it didn't correlate with the street name location.

I understand that this should improve now with the CRASH (what a bad name) data entry system which I think actually submits the location from GPS on-site. However, it has not been a smooth roll-out of the new IT equipment, as I'm sure you're all very surprised to read(!)