Guidance

Reported road casualty statistics: background quality report

Updated 28 September 2023

This background quality report relates to the publication of statistics on reported road casualties and collisions in the series of publications labelled Reported Road Casualties Great Britain (RRCGB).

These statistics are largely based on the STATS19 collection of collisions reported to the police, but also include data on drink-driving, other sources of casualty data (for example survey estimates) and valuation of collisions and casualties.

The purpose of this document is to provide users of the statistics with information about the quality of the outputs, measured against different dimensions of statistical quality. As a result, this document helps to demonstrate how the department complies with the Code of Practice for Official Statistics principle on quality.

These statistics as of 12 February 2013 were designated as National Statistics. National Statistics are produced to high professional standards as set out in the Code of Practice for Statistics. They undergo regular quality assurance to ensure they meet customer needs.

Code of Practice for Statistics:

The Code of Practice sets out common standards that should be followed by all UK organisations that produce official Code of Practice statistics.

Section 1: Background to the statistics

History of the data collection

The current system for collecting data about road collisions and casualties dates back to 1979. The Department for Transport holds detailed records about road collisions since that year in a database. A number of published data tables include a time series back to 1979. Earlier data are available because the post-second World War data collection system was set up by a Home Office Circular dating from 1948, and before that, figures were collected from the First World War onwards. The post war figures have always been based on police reported data.

Data are collected by the police using a data definition which is known as STATS19, after the name of the data collection form used. Full guidance on completion of STATS19 is outlined in the STATS20 document.

Police forces pass the STATS19 data onto the Department for Transport (for English forces), Welsh Government and Scottish Government. The Police Service for Northern Ireland performs a similar role there.

Coverage of the statistics

The majority of the statistics in the RRCGB reports are from road traffic collisions (see box on terminology below) in which at least one person was injured and which were reported to the police, and then by the police to the department via the STATS19 collection.

The collision must have involved at least one vehicle. This includes pedal cycles and ridden horses. The accident did not have to include any motor vehicles – for instance, an accident involving a pedal cyclist falling off their bicycle on the road would be included, even if no other vehicle or pedestrian was involved. The collision also has to take place on the public highway. Collisions in car parks, on private driveways and off-road are not included.

As long as all drivers exchange details, there is no legal obligation to report a road traffic collision even if someone is injured. This leads to the problem of under-reporting of injury collisions in scope of the collection. Estimates are made of injury collisions which are not reported to police based on data collected in the department’s National Travel Survey.

Collisions which do not result in a personal injury (such as ‘damage-only’ accidents) are not in scope and are not estimated, except for the purpose of estimating cost of collisions.

Terminology: accidents or collisions?

Historically, these statistics have followed the terminology of the Road Traffic Act in referring to road traffic accidents. Following stakeholder feedback, from 2022, where possible this terminology has been changed in the statistics and published data tables to refer to collisions. In this context, these terms are interchangeable and have the meaning as in the Act.

Notes and definitions

A full list of notes and definitions relating to these statistics is available separately, though the following are particularly important:

STATS19 data: The core set of statistical data which the police have agreed to provide about personal-injury road traffic accidents.

Collision (or accident): Involves personal injury occurring on the public highway (including footways) in which at least one road vehicle or a vehicle in collision with a pedestrian is involved and which becomes known to the police within 30 days of its occurrence. Damage-only accidents, with no human casualties or accidents on private roads or car parks are not included.

Collision severity: The severity of a collision is based on the severity of the most severely injured casualty, as outline below

Casualty: A person killed or injured in an accident. Casualties are sub-divided into killed, seriously injured and slightly injured.

Killed: Human casualties who sustained injuries which caused death less than 30 days (before 1954, about two months) after the accident. Confirmed suicides are excluded.

Seriously injured: Injured casualties are classified as seriously or slightly injured based on the type of injury sustained, as recorded by the police on the basis of information available within a short time of the accident. This generally will not reflect the results of a medical examination, but may be influenced according to whether the casualty is hospitalised or not. Injuries classed as serious include: fractures, concussion, internal injuries, crushings, burns (excluding friction burns), severe cuts, severe general shock requiring medical treatment and injuries causing death 30 or more days after the accident

Slightly injured: An injured casualty that is not classified as seriously injured, having an injury of a minor character such as a sprain (including neck whiplash injury), bruise or cut which are not judged to be severe, or slight shock requiring roadside attention. This definition includes injuries not requiring medical treatment.

Section 2: Data collection and processing overview

Data collection method

Every personal injury road traffic collision that is reported to the police is recorded in an administrative system that includes an element of statistical reporting. This statistical reporting is done in a consistent way across Great Britain using the STATS19 variables and STATS20 guidance. However, police forces are free to use whatever form and technology they choose and have no obligation to use a standard approach.

Historically, the STATS19 variables have been captured in the standard STATS19 form (sometimes known as MG NSRF), though reporting police forces have been free to use their own locally defined version, provided that the core STATS19 variables are captured by the police officer attending the scene of a collision, or when recording details of a collision reported by a member of the public at a police station (around a third of collisions, mostly less serious ones). Forces are able to collect any additional data they choose, based on local requirements.

In the past, STATS19 forms were usually a series of paper forms which are completed by the police officer and then keyed in by back office staff in the police force. Though a small number of forces use paper forms, STATS19 data is now more commonly completed directly onto a digital version of the form or increasingly captured directly through hand-held mobile digital devices.

In recent years, there have been two notable developments in the form of data collection:

Introduction of injury-based reporting, including through the Collision Reporting and Sharing (CRASH) system. In these systems, casualty injury severity is coded from a list of injuries, rather than being based on the judgement of the reporting officer. In some forces, officers complete the information required on a mobile device

Introduction of online reporting. While collisions reported by members of the public over the counter at police stations have long been included in the statistics, in recent years several police forces have introduced means of online reporting such as via the ‘Single Online Home’ platform which has made this self-reporting easier.

Both the introduction of injury-based reporting and online reporting have implications for the quality of the statistics, considered further below.

The STATS19 collection is part of a larger administrative form that is completed by the police officer for each collision. This larger form contains additional fields about the collision that are relevant to the decision whether or not the police will prosecute any of the people involved. It also contains the police officer’s description of the circumstances of the collision, details of any witnesses and additional variables which are not shared for statistical purposes.

Variables collected

The purpose of the STATS19 variables are to provide detailed information about the collision itself, about the attendant circumstances, the vehicles involved and the resulting casualties.

The STATS19 form is really a series of forms:

• An overall collision record setting out collision circumstances, including date and time, location (grid reference), road details; light, weather and road surface conditions
• A vehicle record for each vehicle involved, including vehicle and driver details, movements during the collision and whether or not it was a hit and run collision.
• A casualty form for each of the casualties involved, including casualty type and severity of injuries and demographics (age, sex, home postcode)
• A contributory factors form which allows a reporting police officer to record up to six facts they consider may have contributed to the collision

STATS19 collects a large amount of data. Each collision record contains 26 data fields, each vehicle record has 22 and each casualty record has 14. Additionally, up to 6 contributory factors can be recorded for each collision.

Some of the fields on the form are completed with a numerical code (for example ‘road type’ is 1=roundabout, 2=one-way street and so on). Other fields are filled with numerical data, such as ages of drivers and casualties, the number of vehicles involved and so on; or alphanumeric data, for example like the vehicle registration marks (number plates) of the vehicles involved and the postcodes of drivers and casualties.

In order to match up the different parts of the STATS19 form, all the parts of each forms have a single collision reference number; each vehicle and casualty involved also has a reference number, so that casualties that are drivers of, or passengers in, a vehicle can be linked to their vehicle, and so that the contributory factors section can refer to the relevant vehicles or pedestrians.

Data entry

While data are collected to a consistent specification, practices vary in how these data are inputted onto computer systems, validated and transmitted to the department.

For forces using the CRASH system, data are validated on entry via a series of error or warning messages. Once a collision record has been entered, the overall number of records can be seen by members of the DfT road safety statistics team. Once considered sufficiently complete, records are marked as signed off for statistical purposes by the reporting officer and at this point, the STATS19 variables can be viewed and exported for statistical purposes.

Some police forces not using CRASH key the information into their own databases as part of back-office procedures and then carry out validation themselves. Other forces provide the forms to the local highway authority which enters and validates the data.

Data processing

Following data entry, the STATS19 data are validated through various processes which take place within police forces, local highway authorities and national governments.

In some cases (including where a police force is using the CRASH system) data are supplied directly from the police force to the DfT, which then validates and checks the data, raising any data quality problems directly with the relevant police forces. In these cases the police force also sends data to the relevant local highway authority (or national government, in the case of Scotland) which also carry out validation.

In other cases, police forces send the data directly to the local highway authorities who carry out the majority of the validation tasks and corrections required. Once this has been completed, the authority provides the data to their national government (DfT for English forces, and the Welsh Government (WG) for the 4 Welsh forces; in the latter case, the data are sent to DfT by the WG). These processes take place on different cycles, ranging from weekly through to quarterly. This process is currently followed for London and in Wales, among others.

On receiving the data from a data provider, the road safety statistics team at DfT carry out a range of further checks, querying with data providers where appropriate.

Data validation and verification

STATS19 is an administrative system, which, in principle, cover the entire set of road collisions within scope of its definitions. Therefore the resulting dataset does not contain statistical errors. Instead errors arise from:

• Individual police officers failing to complete STATS19 forms
• Individual police officers misreporting the details they enter on to the form,
• Other errors made by police forces and local authorities in processing the data, including deleting or otherwise failing to report collision records

There are a number of layers in the validation process aimed to minimise these errors. Police forces have internal checking processes designed to ensure that police officers complete this form. They also can use in-house checking of the quality of the data on the forms. For those forces using CRASH, a series of validation checks and warning messages are built into the system. Many police forces use local authorities to check details of accidents, particularly the information about the location of accidents.

The validation and verification procedures carried out by the DfT road safety statistics team include ensuring that:

• police and local authorities send figures, on time.
• the forms are fully completed.
• data provided is within scope of the STATS19 definition (for example, collisions away from the public highway are excluded)
• each collision is allocated to the correct local authority by checking that the grid reference falls within the boundaries of the authority concerned
• duplicate records have not been sent through.
• the department holds the same number of records as the data providers for any given time period
• the dataset is consistent, for example by eliminating orphan records (such as casualty records without an associated collision record)

In addition, a standard set of validation checks are run on each record. If a record fails validation, then a report is generated which is sent back to the data provider so that they can correct the record. Where data errors are identified, they are classified as hard, severe or other errors:

• Hard errors are those which prevent data being loaded into the database, and are always resolved with data providers
• Severe errors, prioritised for follow up, include for example missing data for fatal collisions and implausible values for more important variables like driver age.
• For other errors, including missing data for less serious collisions or less important variables, resolution is sought, but if not received inconsistent values are typically set as ‘unknown’

A series of additional validation checks are carried out, for example checking that the collision location is consistent with the road network, and within the area of the reporting police force.

While some validation is carried out throughout the year, the majority of validation is done following receipt of a full year of data. Following completion of validation checks and correction of any errors or inconsistencies, the STATS19 database is closed for use in production of the casualty statistics.

Review of STATS19

The information collected as part of STATS19 is revised approximately every 5 to 10 years. The current version of the form was introduced from the beginning of 2011. It was introduced following a review carried out in 2009.

The latest review of STATS19 was carried out in 2018, with the recommendations agreed and due to be implemented from the start of 2024. Further details are available in the published review reports.

Reviews are steered by the Standing Committee for Road Accident Statistics (SCRAS), first set up in 1977 to oversee the new STATS19 process for road accident data collection. The membership of SCRAS is drawn from a wide range of bodies and includes representatives of central government, local government and the police. The committee is chaired by DfT.

Section 3: Quality assessment

In this section, the quality of the statistics is considered in relation to the different dimensions of quality as stated in the European Statistical System (ESS) quality framework.

European Statistical System (ESS) quality framework:

The European Statistical System (ESS) handbook for quality reports provides comprehensive guidance on measuring the quality of statistical processes.

1. Relevance

Relevance is the degree to which a statistical product meets user needs in terms of content and coverage.

Statistical products for road casualties and collisions

STATS data are administrative rather than statistical data, so that all figures that are collected are (with the few exceptions for reasons of disclosure) available to be released after the quality checking process described above.

It is impossible to disseminate more than a small fraction of the available data in an aggregated format, particularly as it also has a geographical element. The approach taken is to provide a range of outputs to meet the needs of a range of users. The statistical outputs presented within road casualty statistics include:

• statistical publications and factsheets presenting high-level figures and commentary on road collisions and casualties overall, and for selected groups of high-risk or vulnerable road user
• an interactive casualty dashboard allowing users to explore the key trends visually
• a series of aggregate statistical data tables, covering the key variables and typically providing data for a 10 year period
• a data download tool covering many of the casualty and collision variables, allowing users to download bespoke aggregations of the data including for selected geographical areas
• raw data extracts published on data.gov.uk

The road safety statistics team also produce ad-hoc tables and statistics on demand, and respond to several hundred requests every year, including in response to Freedom of Information (FOI) requests and Parliamentary Questions (PQs).

Known users and uses of the statistics

STATS19 data and the department’s statistical products have a wide range of known users, summarised separately on our user engagement page.

How well the statistics meet user needs

It is considered that, on the whole, the range of statistical outputs meet most needs for data from both official and non-official users of data.

This assessment is based on regular user engagement, including direct feedback about from people and organisations involved in road safety policy and delivery in Great Britain (for example road safety policy teams within DfT road safety groups such as the Parliamentary Advisory Council for Transport Safety (PACTS), local authorities, researchers and academics). The road safety statistics team also requests feedback, particularly in response to proposed changes to outputs. Website analytics are also used to assess the level of interest in different outputs.

In addition to the standard high-level publication, topical factsheets aim to focus on areas of topical or user interest. For example, in 2021 a focus on road user risk was published to address feedback from PACTS on the presentation of casualty rates within the annual report.

Additionally, the DfT provides an on-demand service for answering requests for data that are not covered by the published statistics (for example a request for details about accidents along a particular stretch of road) or signposting to products produced externally using the STATS19 data which may address their need (for example, online casualty mapping).

Further details of the extent to which the statistics meet the needs of users, changes in response to specific user feedback, and currently unmet needs, are available in our separate document on user engagement.

2. Accuracy and reliability

Accuracy refers to how close the estimated value in the output is to the true result.

The concept of statistical accuracy can be broken down into sampling and non-sampling error. Non-sampling error includes areas such as coverage error, non-response error, measurement error and processing error.

The collisions and casualties data are based on personal injury accidents that are reported to the police. The STATS19 figures, in principle, have complete coverage of these collisions so there is by definition, no sampling error in the figures.

There are issues of accuracy, or rather incompleteness, to the degree that:

• personal injury collisions are not reported to the police; or
• collisions and casualties are reported to the police but are not recorded on the Stats19 form; or are
• reported to the police but the details of the accident and collisions are not recorded accurately

Provisional figures (see below) are based on data for most, but typically not all, police forces. This is because there is a trade off between timeliness and accuracy and some forces, for various reasons, are unable to supply data in time for publication. Imputation is used to estimate for missing data, and provisional statistics are published only at higher levels of aggregation. The final figures for the year always include full returns from all police forces.

In a dataset of the size of STATS19, completed by many different data providers, it is inevitable that there will be inaccuracies. The data validation work, as outlined above, aims to mitigate these particularly for the more serious collisions, though the final dataset will still contain some errors and missing data. On the whole, these are not considered to impact on the patterns and trends shown.

However, there are two notable issues related to accuracy of the STATS19 data – under-reporting, and the recording of injury severity. Each is considered in further detail below.

Under-reporting of collisions

Under-reporting of collisions covers both cases where collisions are not reported to the police, or where collisions are reported but are not recorded. Of these, it is believed that the former occurs to a greater extent.

This is linked to the issue about the definition and recording of personal injury; in other words what is, or should be, the circumstances when someone involved in a traffic collision has suffered a ‘personal injury’. For example the STATS19 definition of a collision will include single-vehicle pedal cycle collisions, if the rider (or a pedestrian in collision with the cycle) is injured. In practice, only a small proportion of these types of collision are reported to the police.

This is a serious issue concerning the quality of the road casualty data. If under-reporting remains unrecognised, then the true magnitude of any road safety problems cannot be known, or could be underestimated. This could in turn lead to incorrect prioritising of policy measures to improve road safety, or could lead to less efficient or inappropriate countermeasures. These issues also affect the ‘coherence’ strand of quality.

Extensive work has been done by the department and others to understand the extent of under-reporting of road casualties. A summary is given in the annex to this document below.

Severity recording

It is known to sometimes be a challenge for police officers in classifying casualties correctly into the ’slight’ and ‘serious’ categories at the roadside. Whilst it is clear which category many injuries should be classified (a fatality is very simple to classify, for instance), there are a wide number of injuries which might require hospitalisation yet this fact is not obvious at the roadside. Although officers rely on experience and, where possible, input from paramedics, it is inevitable that they will get this wrong sometimes.

In recent years, some police forces have moved to injury-based reporting systems (notably CRASH) where the reporting police officer selects the most severe injury and the casualty severity is coded from this. While this still requires judgement of injury, it is considered to improve the consistency of recording of severity across officers and forces. The introduction of injury-based reporting has typically been associated with an increase in the number of casualties classed as seriously injuries. This has implications for the comparability of the data over time, considered further below.

3. Timeliness and Punctuality

Timeliness describes the time between the date of publication and the date to which the data refers, and punctuality describes the time between the actual publication and the planned publication of a statistic.

Timeliness

Road casualty statistics are published as soon as is practical after the end of the time period to which they relate.

Currently, statistics are published as follows:

• Provisional mid-year estimates covering January to June are published in November (a gap of around 5 months between the final data period and publication)
• Provisional annual estimates are published in May (a gap of around 5 months)
• Final annual estimates are published in September (a gap of 9 months)

Ideally this time would be shorter, but it is dictated by the large number of police forces and local authorities in Great Britain, plus a small amount of validation, processing and production time within DfT. As an illustration, for the final annual statistics in 2021, complete data for the final police force was received around end July, validation was complete by end August and the statistics were published at the end of September.

We are aware of user needs for more timely statistics, and in recent years have improved the timeliness of the provisional figures, which do not rely on returns for all police forces. In 2022 and 2021, provisional annual statistics were published in May, rather than June as in 2020 (and July as in 2019).

Drink-drive statistics are produced to a slower timetable as more time is required to obtain data from coroners which is needed for these statistics. In particular, in many cases coroners provide data following completion of an inquest, which can be a year or so after the date of the collision in some cases. Further work is then required to match coroner returns to STATS19 data.

From 2022, drink-drive statistics have been produced annually in July for the last but one complete year, that is around 19 months after the end of the period to which they relate.

Punctuality

Punctuality is usually a minor issue as all outputs adhere to the Code of Practice by pre-announcing the date of publication on the gov.uk website.

In recent years, there have been some cases where data supply issues have resulted in statistics being published later than the usual timetable. For example, provisional mid-year estimates for 2020 were pushed back from November to the following January to allow time to improve the completeness of the dataset. In this case, changes to the planned publication date were announced well in advance, and there have been no occasions where pre-announced statistics have not been published as planned.

4. Accessibility and Clarity

Accessibility is the ease with which users are able to access the data, also reflecting the format in which the data are available and the availability of supporting information. Clarity refers to the quality and sufficiency of the metadata, illustrations and accompanying advice.

Accessibility

The outputs are published on the GOV.UK DfT statistics page in accessible formats.

Statistical releases are available as html files, which are accessible for those who use assistive technologies.

Data tables are available in ODS file format which can be accessed by using freely available software. The spreadsheets have been designed to be used easily by assistive technologies. The published tables include information on collisions, casualties and vehicles involved. More detailed breakdowns are available via the data download tool or open dataset.

Clarity

The statistical releases use plain language, in which technical terms, acronyms and definitions are defined where appropriate. The main findings are presented using a series of text, charts and maps, with maps used to show the location of the major ports. Full details of the strengths and weaknesses of the data are provided at the end of each statistical report for users who are interested in this.

In addition to the statistical releases, a notes and definitions document has been published which outlines the coverage of the statistics as well as the key definitions.

5. Coherence and Comparability

Comparability is the degree to which data can be compared over time and domain. Coherence is the degree to which data that are derived from different sources or methods, but refer to the same topic, are similar.

Comparability

Comparison between UK countries

The STATS19 variables and definitions are commonly used across England, Scotland, Wales and Northern Ireland (though some of the variables are slightly different or omitted entirely in Northern Ireland). All of these data are collected from the same source and are, therefore aligned in terms of the definitions and coverage of the data. In addition the data are given a similar set of validation checks to ensure the internal consistency of the data provided in various sections of the STATS19 form.

There is one difference between the statistics DfT produce and the statistics produced by the Welsh and Scottish Governments. Whilst we close down our database for the year at the point of producing the final annual statistics, Scotland and Wales continue to accept changes. This means that although the figures may agree at first, over time the figures for Scotland and Wales in the publications produced by the devolved governments diverge from DfT’s figures.

The reason for this difference is in the volume of data. The number of collisions in Scotland and Wales are much smaller than the whole of Great Britain (or even England), therefore whilst it is possible for the Scottish and Welsh Governments to continue to update their databases, this is impractical for DfT.

International comparisons

Some comparisons can be made internationally, as most countries record road traffic accidents, usually based on police collision reports (as in Great Britain). Figures for fatalities in different countries are broadly comparable; and this is the basis of the usual international comparisons. There are considerable differences in the way different countries treat road traffic-related injuries. So while figures are available for injuries internationally as well, these are not used for comparison purposes.

Comparison over time

The changes to the STATS19 form take place approximately every five to ten years and are controlled by the Standing Committee of Road Accident Statistics (SCRAS). One of the parameters in deciding on changes to the form is the need to retain long-term consistence in the main statistical series. DfT holds a database with comparable data going back to 1979. The main categories, such as total casualties or total collisions are comparable with data from before that date.

Coherence

Other sources of data relevant to road traffic accidents and casualties for Great Britain and England include:

• death registrations data
• Hospital Episode Statistics (hospital admissions), and HES data linked to STATS19
• National Travel Survey questions on road collisions and casualties
• Crime Survey data on road collisions and casualties.

The STATS19 data remain the single most useful source of data on road collisions and resulting casualties in Great Britain. In particular, it is the only source to provide detailed information on collision circumstances, vehicles involved and resulting casualties.

However, as has long been known, STATS19 is not a complete record of all injury accidents and resulting casualties, and this can lead to discrepancies with the other sources of data listed above.

Although STATS19 does not provide complete coverage of road collisions and casualties, this does not in itself make it unsuitable for monitoring changes over time, assuming that levels of reporting to police have not changed. There is no clear or conclusive evidence of a systematic change in levels of reporting at national level. In addition, most, if not all, road accident fatalities are included in the police data.

Other datasets can be useful both as a check on the quality and completeness of Stats19 and in providing information which is not collected by the police, for example relating to more detailed medical consequences of road accidents.

Further details of these other sources, and comparisons with STATS19 data, are available in our note on other sources of road casualty information.

6. Trade-offs between Output Quality Components

Trade offs between output quality components describes the extent to which different aspects of quality are balanced against each other.

The main trade offs involved in producing road casualty statistics are between accuracy and timeliness. Producing more timely statistics would likely involve basing them on less complete data, or with less validation.

We attempt to address this by publishing provisional statistics in advance of the final annual report, which are based on data for the majority of police forces and involve less validation, but are published around 4 months earlier.

7. Assessment of User Needs and Perceptions

Assessment of user needs and perception covers the processes for finding out about users and uses, and their views on the statistical products.

DfT regularly engages with users by social media, email and face to face methods when possible. This includes requesting feedback on each publication, with contact details provided in every statistical release. Each publication is promoted via X (formerly Twitter). Further details are in our separate page on user engagement.

DfT also regularly analyses web page usage, ad-hoc requests and social media analytics to monitor activity over time. In line with the Code of Practice for Statistics, users will be informed about any changes or revisions to the data series. Less frequently, the department reviews the STATS19 data collection (last done in 2019) and presents key results at seminars, for example those arranged by the Transport Statistics User Group, the annual Joining The Dots conference (last presentation 2023) or the Parliamentary Advisory Council for Transport Safety (PACTS) technical user conference (last presentation 2022).

As part of the provisional publication of 2021 data issued in May 2022, the department published a note to explain proposed changes including to the data tables that accompany the publication. This included informing known users about the planned changes and requesting feedback. Following analysis of feedback received, changes were made alongside the final statistics in September 2022 and for all subsequent statistical publications.

A similar exercise was carried out for the drink-drive statistics, with feedback on proposed changes sought in February 2022 and a response to this published alongside the final 2020 statistics in July 2022.

8. Performance, Cost and Respondent Burden

Performance, cost and respondent burden describes the effectiveness, efficiency and economy of the statistical output.

The STATS19 form is part of a wider administrative process that is carried out by police forces in respect of personal injury road collisions that come to their attention. It is therefore difficult to establish the response burden that relates solely to the completion of the STATS19 form.

Previous estimates, now somewhat out of date, have been that the response burden amounted to around £2.6 million for Great Britain as a whole. This covered the time individual police officers spend completing the form, the back office work in compiling the data, work across local authorities in checking the location of accidents, and finally the checking of records by police and local authorities of collisions and casualties which fail the various validation checks carried out.

9. Confidentiality, Transparency and Security

Confidentiality, transparency and security refers to the procedures and policy used to ensure sound confidentiality, security and transparent practices.

The report of the details of a traffic collision and the resulting casualties (and the nature of these casualties) is a report of a public event. For example it may be reported in the press, or on radio or television. So there are considered to be no confidentiality issues posed by reporting about an individual collision and its attendant circumstances, together with details of the vehicles and casualties involved. This means that there is not need to impose any disclosure control to the publication of most of the statistical information about the collision.

DfT aims to publish as much data as is possible whilst ensuring that confidentiality is maintained, and raw data containing non-sensitive fields are released as part of the road casualties open dataset.

Confidentiality protection is relevant to a few potentially personally disclosive data items on the STATS19 form. These items cover information about people involved in the accident which are considered to be confidential and relate to:

• postcodes of the addresses of drivers
• postcodes of addresses of casualties
• the registration numbers of vehicles involved in accidents
• the contributory factors for the accident (as they can imply blame and be sub judice to any investigation or prosecution)
• breathalyser tests and results

Within the statistics team at DfT, the above sensitive items are processed only for specific purposes. For example, postcode data is used for statistical analysis to investigate the links between collision and casualties and deprived areas. DfT links the vehicle registration data with DVLA records in order to analyse collisions and casualties by the make and model of the vehicles involved. Breath test information is required to produce statistics on drinking and driving.

The contributory factor data offer a particular challenge with regards to the balance between confidentiality and access which was investigated in more detailed by the Methodology Advisory Unit at the Office for National Statistics with a published report. In practice, the main change this has brought about is allowing us to release a table containing the total number of contributory factors by local authority, which was published for the first time in 2013.

These pieces of sensitive information from the STATS19 form are not released outside DfT by the road safety statistics team, except where an application for use is made and approved. Researchers working within universities and suitable research institutions can apply to get access to the sensitive variables under an end user licence, provided they set out a clear purpose and provide details of how they will protect the sensitive data.

All STATS19 data are held within DfT in a secure database, only accessible to members of the road safety statistics team. All data is stored, accessed and analysed using DfT secure IT systems. Data protection regulations are adhered to throughout the road casualty statistics production process, and any sensitive data provided to DfT by police forces is kept securely where access to data is controlled in accordance with departmental policy.

DfT adheres to the principles and protocols laid out in the Code of Practice for Statistics and comply with pre-release access arrangements. The pre-release access lists are available on the DfT website.

Section 4: Summary and conclusions

This background quality report presents information for users of the road casualty statistics covering different aspects of their quality. The department concludes, on the basis of the assessment outline above, that the statistics are of a quality which is considered fit for the purposes for which the statistics are being used.

Comments and feedback on this report, or any other aspect of these statistics are welcome, and can be provided by email to the road casualty statistics team.

Further information about these statistics is available, including:

• road casualty statistics: data tables

• road casualty statistics: guidance

Annex: Under-reporting of road casualties

Issue

The collisions and casualties data are based on personal-injury collisions that are reported to the police. These data are incomplete as some personal-injury collisions are not reported to the police; or are reported to the police but the details of the collision and casualties are not recorded accurately, or indeed not recorded at all.

Sources of underreporting and under-recording of road collisions and casualties

Sources of under-reporting and under-recording include:

• Not all road collisions are ‘reportable’: for example, if no injury occurs or the collision takes place on private land away from the public highway
• There is no legal obligation for drivers to report road collisions to the police, provided the parties concerned exchange personal details at the scene ^{[footnote 1]}
• Some collisions that should be reported by drivers to the police are not reported, because the driver is ignorant of the legal requirements or is reluctant to do so, for example, if they have been drinking
• The police do not record all accidents reported to them
• It is often difficult for a police officer to judge whether a casualty should be classified as having a serious or slight injury. ^{[footnote 2]}

Potential implications

This is a potentially serious issue concerning the quality of the road casualty data. If this underreporting is unrecognised, then the true magnitude of road safety problems are unknown or could be underestimated. This could in turn lead to incorrect prioritising of policy measures to improve road safety, or could lead to less efficient or inappropriate countermeasures.

The impact could be:

• That under reporting is greater for certain types of accident and casualty; and this would lead to a lack of priority for dealing with the relatively under-reported types. ^{[footnote 3]}
• If the degree of under-reporting changes over time, then the STATS19 data will not be a representative measure of progress towards monitoring progress in road casualty reduction

The first point to note is that this is a long-standing issue which affects the road casualty figures throughout Great Britain. It is also a problem internationally as most countries’ road casualty data are based on police reports and so are affected by these issues. this is particularly important as globally, road traffic fatalities are consistently amongst the top three causes of death for people aged between 5 and 44 years old.

In Great Britain, this issue has been raised in a number of policy and statistical contexts, and these discussions are summarised in the UK Statistics Authority Assessment Report of road casualty statistics. Examples of occasions where this issue has been raised include the 2006 National Statistics Quality Review of Road Accident Statistics, or the House of Commons Transport Committee October 2008 report that expressed concern about the quality of the STATS19 data.

DfT response

The Department for Transport has carried out a considerable amount of research into this topic over a number of years.

This included work to address the requirements set out in the UK Statistics Authority assessment of road casualty statistics, which included:

Requirement 1: Develop a best approximation of the numbers of casualties based on research into the undercounting associated with the Stats19 form. These estimates should then be included in the published counts to inform the use of the scale of the problem.

Requirement 3: Bring together as much relevant data as possible – including sources that are not currently exploited – at the time the statistics are released in order to help explain the weaknesses in the STATS19 data, and the implications of these.

In response to these requirements, DfT introduced questions in the National Travel Survey (NTS) asking about whether respondents have been involved in a road traffic collision in which someone was injured in the last three years and last 12 months.

These responses are used to estimate the total number of personal-injury casualties each year including those that were not reported to the police. This estimate was published for the first time in the 2008 annual report and has been updated many times since then. The latest estimates are included as part of the data tables published with the annual casualty statistics, in table RAS4201.

Additional information about unreported road casualties can be found in the tables on hospital admissions for road traffic casualties, which are available in table RAS4101.

Instructions for printing and saving

Depending on which browser you use and the type of device you use (such as a mobile or laptop) these instructions may vary.

You will find your print and save options in your browser’s menu. You may also have other options available on your device. Tablets and mobile device instructions will be specific to the make and model of the device.

How to search

Select Ctrl and F on a Windows laptop or Command and F on a Mac

This will open a search box in the top right-hand corner of the page. Type the word you are looking for in the search bar and press enter.

Your browser will highlight the word, usually in yellow, wherever it appears on the page. Press enter to move to the next place it appears.

1. The requirements to stop, provide information and report a road traffic accident are set out in the Road Traffic Act 1988 (section 170), as amended by the Road Traffic Act 1991 (Schedule 4) ↩

2. For example, the full severity of the injury may not be apparent until some time after the collision when the police officer is no longer present. Research has found that the police tend to underestimate the severity of the injury. ↩

3. For example comparison of STATS19 data with hospital admissions data (for England) shows that the STATS19 relatively under represents serious injury casualties from cyclist accidents, particularly if no other vehicle is involved ↩