The Department for Transport has a consultation out on how we report (and hence record) road collisions, which closes just before midnight on Tuesday (24th). Here I explain the background to the consultation and my response to it. My focus is on road injury data (“Stats19”) as that’s my area of expertise as a researcher. This isn’t the only issue: some organisations representing vulnerable road users and road victims have other concerns, for instance around access to justice.
The consultation is here. In brief, it proposes allowing those involved in a road crash to report online rather than in person. Currently around 25,000 injury collisions are annually reported at a police station, as are around twice as many ‘damage-only’ collisions.
Being cynical, one wonders if the consultation stems from angry drivers complaining to politicians about being forced to attend (increasingly distant) police stations just because they pranged something in their car. The repeated framing of ‘saving drivers time’ suggests it might be.
On the other hand, there are good reasons to support online reporting methods. And whatever the motivation, the implications of changing how road collisions are reported and recorded are substantial, for all of us – particularly the most vulnerable. My focus here is on the implications of the changes for ‘Stats19’ data (the dataset of road injuries reported to police, used in much road injury research). ‘Stats19’ is not mentioned in this consultation, which is worrying. The dataset is due its quinquennial review and one would expect potential changes to reporting methods to also form part of this review.
The reporting of road collisions is governed under the 1988 Road Traffic Act, Section 170. (The RTA refers to ‘accidents’ throughout. Given this is now widely accepted as not best practice to describe road collisions, it would be nice if the opportunity was taken to also revise this language and for the consultation itself not to perpetuate the language of ‘accident’).
The consultation affects a driver’s ‘duty to stop’ after a road crash. Duty to stop and report an incident only includes those collisions involving a ‘mechanically propelled vehicle’, on a road or other public place. ‘Mechanically propelled vehicle’ has been traditionally defined as meaning ‘motor vehicle’. (Incidentally, the consultation suggests that a pedelec might count as a ‘mechanically propelled vehicle’, potentially a major issue.)
‘Duty to stop’ further comes into force where one of the following has happened: (i) personal injury to someone other than the driver, or (ii) damage to another vehicle, to an animal (including dogs, but not cats) outside the driver’s own vehicle, or damage to a property (e.g. building or garden).
Duty to stop is therefore triggered by injuring another person, or someone else’s property, pet, or animal. It’s not triggered by your damaging only your own car, even if you also injure yourself and your dog (if the dog was in your car). It is framed around damage to others and damage to property, on the public highway (or certain other public places) by use of a motor vehicle.
Duty to stop entails the duty to exchange details with others at the scene, including (if injury is caused) an insurance certificate. If this hasn’t happened the driver in question must report the incident to the police within 24 hours.
Case law in the 1980s established that phoning a police station was not enough to count as a ‘report’. However, some police services have started to allow online reporting: some only for non-injury collisions, some also for injury collisions. While online reporting isn’t explicitly sanctioned in the Act, it’s hard to believe any court would condemn a driver for using a reporting channel created for this purpose – as opposed to phoning and hoping.
So it’s odd that the consultation claims ‘At the moment if you need to report a road traffic accident you must do so in person at a police station or to a police constable.’ For some collisions, if it happened in areas including London, Cambridgeshire, Kent, Essex, Thames Valley, and Norfolk, you can report online already.
Online reporting thus seems in principle possible now. The development of a plethora of different forms by different services is another issue entirely. Some may not be fully Stats19 compliant, in terms of producing data comparable to that collected by an officer at the scene or in a station. (I’ve looked at a few of these forms, and they can be a bit confusing). If you’re creating a national online survey (effectively), you really don’t want up to 45 police services (39 in England) all creating their own survey forms in a process of trial and error. It’s not a simple matter, and there’s quite a high risk that one or more services will create an online form that produces invalid data for months, rendering online-reported data from that authority unrecordable and unanalysable.
But the consultation doesn’t address potential impacts on data quality and comparability. It simply focuses on providing formal legal sanction for online reporting (and telephone reporting – although I found it very difficult to see how that might work given the need for locational specificity: all incidents in Stats19 have an Easting and Northing). Effectively then, the consultation formalises something that is happening anyway. To help us provide informed answers to such a consultation, it would have been helpful to be told how the existing reporting portals are doing, so we can see any early indications of how effective (and variable) these are and how good the data is that they generate. But there isn’t an acknowledgement that these systems exist, nor a promise to review them as part of this year’s Stats19 review.
I hope very much that formal evaluation is under way or planned, despite a lack of reference to it. We need to be protecting, and improving, Stats19 data. Our road injury data is some of the best in the world. That’s not to say that it’s flawless or unbiased.
Of course not. Like other police-based injury recording systems we only capture a small minority of ‘slight’ injuries. And detailed incident recording by non-specialist officers has its own problems. I’m generally sceptical of the ‘contributory factors’ fields in Stats19, which rely so much on the officer’s view of what caused the incident, for instance. But we shouldn’t underestimate the tremendous value of nationally and historically comparable, geocoded incident data, with much of the dataset freely available.
Do we need more national organisation? Should DfT be overseeing a process of selection to create a national online road injury reporting system? I think they should. And that we need to be thinking now about how police systems and road user behaviour might adapt to the growing use of online reporting.
The consultation promises substantial savings of both ‘driver’ and police time. I suspect that there are trade-offs, and online reporting will generate more work for police in particular. It’s hard to design a meaningful and complex form to be filled in by members of the public, which precludes invalid or incomplete responses without excluding many people from filling in the form at all. The forms I’ve seen say it’ll take 15-20 minutes to fill them in, a length with which many people may struggle. For comparison, the consultation currently says it takes 20 minutes at the police counter for staff to take down a report.
Information provided at the station has the considerable virtue of being checked by the officer taking the report (e.g. if the person reporting is unclear about how to answer a question, this can be immediately clarified). Where information provided online is unclear or incomplete, or appears inconsistent/implausible upon checking, will (as the consultation suggests) police attempt to contact the respondent, to gather the remaining data needed to complete the form correctly? The consultation suggests only ‘basic’ data would initially be gathered, but the online forms I’ve seen are all quite extensive, and quite complicated.
Post-processing may further increase given that one can imagine a growing number of cases where multiple involved people report the same incident in slightly (or very) different terms. Where data reported by different participants is inconsistent, how will police services proceed? Will there be incentives for people involved in incidents to misreport injuries, if it is believed that these will not be independently checked (e.g. describing injuries to someone else as ‘slight’ may mean a driver avoids investigation, or conversely injuries might be claimed for insurance purposes)?
All this means that although individual road users may well save time (by no longer having to get to a station and wait in line) there may not be much or any saving in police time, per incident. This is of even more concern given that while drivers have duties to report, police don’t have corresponding formal duties to record. If there are sudden and unexpected increases in workload, then there may be unintended consequences in terms of what gets recorded and how well.
As Professor Richard Allsop wrote in a 2006 Review of Road Accident Statistics conducted for the Office for National Statistics (p.5):
‘[It] is by a voluntary arrangement that police forces undertake to record, for the purposes of the national road accident statistics, details of injury accidents which they attend or which are reported to them. This means that the quality and coverage of these data depend on motivating the officers concerned and facilitating the recording task in the face of many competing demands on their time.’
This is not to say online incident reporting is a bad thing. In principle, I think it’s a good thing. But we need to be realistic about how difficult and time-consuming collecting, cleaning, and managing accurate and valid online data can be – and to plan to protect both the data and road victims (who may be traumatised, rather than merely cross at the prospect of having to stand in line).
All this is before considering the impact of a rise in reporting levels. Road injuries are perhaps five times more numerous than Stats19 would suggest, with under-reporting highest for people cycling. Enabling online reporting for all modes is likely to redress this balance, as currently cycling in the UK is skewed towards demographics likely to be at ease with online reporting. However, this will affect time series comparisons. It will affect comparisons between areas, if only some police services implement online reporting. It’s reasonable to assume that in some areas we might get a doubling of reported cycling injuries, when the facility becomes widely known. Increases will happen for other modes, and types of under-reporting bias may also shift within modes (e.g. by age).
The new biases are not necessarily worse or better than the old ones, but we need to think about how we minimise bias, communicate to the public and researchers how and why the data is changing, and where comparability is affected do what we can to enable comparisons (e.g. by allowing analysts to exclude cases that would not have been recorded under earlier conditions, where we have increased the scope of reporting).
We should also take the opportunity to re-consider what we record (online and in-person) to improve future analysis and hence prevention of road injuries. This year’s review of Stats19 should discuss issues raised by changing reporting systems and how to use them to best effect (informed by data about use of the first wave of online tools) alongside possible changes to data collection and recording, including those facilitated by changes to reporting.
In order of increasing ambition, I’d suggest:
(i) Recording start location for trip or trip stage for all casualties – this will enable a whole world of research on injury risk, by allowing the modelling of journeys and selection of ‘control sites’ to be compared with ‘injury sites’, at virtually no cost (it could replace the not very useful ‘journey purpose’ question). Like home postcode data, start locations would not be included within the public dataset. They could be held as safeguarded data for sharing with accredited researchers for safe use in approved research projects (e.g. within the UK Data Service Secure Data Service).
(ii) Also recording start location for ‘vehicles’ (i.e. anyone involved in a crash whether injured or not) – this could allow similar and ground-breaking research to be conducted on where drivers pose the greatest risks to others.
(iii) Recording the involvement of uninjured pedestrians in the ‘vehicles’ dataset. At present, we only learn about pedestrian involvement in a crash if the pedestrian is also injured and hence appears in ‘casualties’. This means our data is skewed if for instance we seek to compare impact of pedestrian-cyclist collisions on pedestrians and cyclists.
(iv) Also including injuries off the public highway, to also cover for instance pedestrians injured by cyclists or moped riders on canal towpaths.
(v) Also including pedestrian falls. This has the potential to generate a LOT more data but increasingly is seen as important by road safety researchers. We currently know very little about locational risk factors associated with falls, as we don’t have routine geographical data on them. There are issues here with online reporting: many victims are older women unlikely to report their fall online. I would suggest a pilot project involving hospitals in one police authority area (which could therefore only cover more major injuries), where subject to patient consent, research nurses could facilitate recording such falls into an (or the) online reporting system. This would of course not save money (it would need to be funded). But it might save lives.
There are variables I think could be removed (such as the ‘journey purpose’ question which is not useful in my view and has substantial missing data). As mentioned above I’m sceptical about the usefulness of ‘contributory factor’ data for research (although others may disagree, and CF data may be needed for purposes other than research).
Even without changes (iv) and (v) suggested above, online reporting may lead to many more incidents being reported. This will have implications for police resources, especially if there is not the hoped-for per-incident time saving (which I think is unlikely, for the reasons suggested above).
Will we see incidents with missing data simply deleted?
Or even (I hope not) will we see moves to remove slight injuries from Stats19 entirely, removing responsibility for data cleaning? This would remove many incidents currently available for analysis – a paper I published recently wouldn’t have had enough statistical power for some of the analysis without including slight injuries. But this could in theory happen.
As a half-way house, police services may start to expect (or even mandate?) slight injury collisions to be exclusively reported online. That would mean reporting of these collisions would skew towards those able competently to use online reporting systems. Some currently used forms are only available to drivers, not cyclists, pedestrians, and horse riders. This is a matter of urgency and makes it even more worrying that this consultation focuses entirely on drivers.
Another question: will online reporting be available in multiple community languages? (The online reporting systems I looked at seemed only available in English, but translating surveys is relatively straightforward). Will it be accessible to disabled people? Will a helpline be available to talk users through the system? I worry that our data will become skewed in dramatically different ways to the present, and we are unprepared for this.
These issues shouldn’t just be the responsibility of individual police services: DfT need to take a lead on this and ensure we don’t get into a mess and lose data comparability across the country. At a minimum we need to be working towards a national online road incident reporting service, not leaving police services to ‘do their own thing’ (or not, as they prefer). We should be considering how this fits with the growth of ‘near miss’ / dangerous driving non-injury reporting. Can we start standardising these reporting systems and hence datasets too, and making edited versions publically available as we do with Stats19?
We should be involving researchers in developing these new services and datasets. Many of us, while appreciating Stats19 and wanting to protect it, also have come up against its limits and have ideas for improving data collection and dissemination. (We might also have ideas on developing online surveys!) The Standing Committee for Road Accident Statistics (SCRAS) should have academic/researcher members, who could represent the research community and would take on responsibility to liaise with them on potential improvements, changes, and threats to the data.
In response specifically to the questions posed by the consultation:
What currently available channels of communication do you think need to be considered for reporting road accidents at this time?
Online collision reporting is already starting to happen and in principle a good idea. However, its introduction needs to be closely monitored, evaluated, and improved, given (i) the potential for introducing new biases and (ii) the likely increase in police workloads due to data processing requirements and a potential increase in overall incident reporting (see Q6). This consultation should have provided information about existing online reporting systems and the likely impacts of further roll-out on Stats19 data.
What reporting methods do you believe might exist in the future?
Online reporting is likely to increase, particularly as even without the proposed RTA amendments police services view it as permissible. I doubt telephone reporting (also mentioned in the consultation) has much promise. There is a great need for standardised reporting systems, which would for instance include helpline for those having difficulty, options to use community languages, and accessibility for disabled respondents.
What are your views about giving each separate police force the power to decide which communication channels they offer for reporting road accidents?
I do not support this. There are already well-known issues with data comparability across forces, and the current use by different services of three different electronic reporting systems (CRASH, COPA, and NICHE) is a concern. Changes to reporting systems should be trialled and evaluated but following this the aim should be a national system which provides a high-quality online reporting system for all road users. This will also make it easier to communicate to road users about how and when to use the new system and avoid the risk of some services ending up with invalid data due to problems with local online reporting systems.
What is your opinion of allowing the use of automated processes for the reporting of road accidents?
As above, in principle I see online reporting as potentially positive. However, the impact of these systems needs to be properly evaluated and consideration needs to be given to unintended consequences (including through increasing police workloads, due to an increase in reporting, for example). Reporting systems must be inclusive of all modes and users and must not create incentives to under-report injury severity (for instance). The need for post-processing, checking, and contacting respondents must not be under-estimated and this and other issues raised here must be considered as part of this year’s quinquennial Stats19 review. For instance, what procedures will be in place for incidents reported multiple times by different participants (from identifying that reporting indeed relates to the same incident, to establishing a version of record?)
Do you agree that using new methods of reporting road accidents would deliver benefits to drivers? Why?
I would challenge the primary focus on benefits to drivers here. It might even be argued that where drivers cause damage or even injury, having to spend time travelling to a police station may represent some small deterrent. The system needs to work well for all users, particularly those who experience injury or who are particularly vulnerable (e.g. pedestrians and cyclists, older and disabled people). Hence the consultation and the upcoming Stats19 review must consider the impact on these users if injury collisions are increasingly reported online. As a researcher who makes substantial user of Stats19 I would also highlight the need to consider benefits and disbenefits to research users (which will affect the production of evidence to support injury prevention), especially given potential impacts on bias, data completeness, comparability etc.
Do you agree that using new methods of reporting road accidents would deliver benefits to the police? Why?
Potentially – but please do not under-estimate the likely additional time needed to deal with incidents reported online. Compared to incidents recorded at a station, data is likely to need more (perhaps much more) checking and cleaning, and there may be greater requirements for post-processing e.g. combining incident reports. As there is currently a substantial under-reporting of slight injuries, the numbers of incidents reported is likely to increase. Increased completeness in reporting is a positive thing in principle, but the extra demands placed on police time may lead to unintended and adverse consequences for Stats19 data (and potentially for road victims).