Methodology note for reported road casualties in Great Britain involving illegal alcohol levels
Published 31 July 2025
Applies to England, Scotland and Wales
In 2025, changes were made to the methodology. More information is in the changes to methodology section.
Introduction
This methodology note describes the estimation of casualties arising from reported road accidents involving at least one motor vehicle driver or rider over the legal alcohol limit for driving, in Great Britain. Figures are derived from the STATS19 forms completed by the police plus toxicology data for road fatalities from coroners and procurators fiscal.
Drink-drive legal limits and definitions
For the purpose of these drink-drive statistics, a drink-drive collision is defined as being a reported incident on a public road in which someone is killed or injured, and where at least one of the motor vehicle drivers or riders involved met one of these criteria:
- refused to give a breath test specimen when requested by the police (other than when incapable of doing so for medical reasons)
- failed a roadside breath test by registering over 35 micrograms of alcohol (reduced to over 22 micrograms in Scotland only from 5 December 2014) per 100 millilitres of breath
- died and was subsequently found to have more than 80 milligrams of alcohol (reduced to 50 milligrams in Scotland only from 5 December 2014) per 100 millilitres of blood
Drink-drive casualties are defined as all road users killed or injured in drink-drive collisions.
Data sources
Two sources of data are used to estimate the number of drink-drive collisions in Great Britain. These are:
- Coroners’ data: Information about the level of alcohol in the blood of road collision fatalities aged 16 or over is provided by coroners in England and Wales and by procurators fiscal in Scotland. Only samples taken within 12 hours of the collision are considered to be suitable to be included
- STATS19 breath test data: The personal injury road collision reporting system (STATS19) completed by police provides data on injury collisions in which the driver or rider survived and was also breath tested at the roadside. If the driver or rider refused to provide a breath test specimen, then they are considered to have failed the test unless they are deemed unable to take the test for medical reasons
The number of fatal collisions where a driver or rider died with an illegal alcohol level is estimated from the coroners’ and procurators’ fiscal data. The number of collisions where a surviving driver or rider had an illegal alcohol level is an estimate based on a calculation of the proportion of these alcohol-related collisions which can be identified from the STATS19 breath test data.
Completeness of data and reliability of drink drive estimates
Both sources of data from the police and coroners on drink-drive collisions are incomplete.
- In the case of coroners and procurators fiscal data, suitable toxicology data are not available for all killed drivers or riders recorded in STATS19 and are typically available for around 60 to 70% of relevant cases
- In the case of the STATS19 breath test data, some drivers and riders are not breath tested since it is not always possible to administer a test to all drivers involved. Some drivers and riders not tested might have failed if a test could have been administered. Others may have departed the scene of the collision before the police arrive
Adjustments to the reported data are required to estimate the actual number of drink-drive collisions and their related casualties.
Two adjustments are applied to the data: one to account for the incompleteness of coroners and procurators fiscal data, and one to account for hit and run collisions (where a breath test cannot be administered).
Adjustment for coroners’ and procurators’ fiscal data
For many drivers or riders killed in road collisions, a suitable blood alcohol level is not available, either because the sample was taken more than 12 hours after the collision, or because no test was carried out, or because some of the data are not reported to the Department by coroners and procurators fiscal.
To take account of this, the proportion of reports received indicating that the deceased driver or rider was over the drink-drive limit is applied to reports that are not received. The same coroners adjustment factor is applied for all breakdowns of collisions and casualties relating to killed drivers over the limit, and therefore does not take into account differences in response rate between subgroups.
The coroners’ data adjustment is therefore:
- DE is the estimate of driver fatalities above the legal limit
- CF is the number of driver or rider records returned from the coroners where the alcohol level recorded is above the legal limit
- CR is the number of all records returned from the coroners
- FD is the number of all drivers fatalities in reported road collisions
Adjustment for hit and run
The breath test failures (including refusal to provide a sample) need to be adjusted to take account of drivers who have departed the scene of the collision before a police officer arrived (called ‘hit and run’ in this section).
This involves an estimate of drink-drive collisions from recorded hit and run drivers where no other driver has failed a roadside breath test or refused to provide one.
This adjustment is based on the 1993 TRL article entitled “The actual number of non-fatal drink-drive collisions”.
There are four component parts which are as follows:
- N1 is the total number of collisions where at least one driver was recorded in STATS19 as a hit and run
- N2 is the number of hit and run collisions (N1) where at least one driver was not contacted at the time
- N3 is the number of hit and run collisions (N1) where at least one driver was not contacted at the time (N2) and at least one other driver failed a breath test or refused to provide a sample. We already know that alcohol was involved in the collision, so we do not need to estimate the proportion of those not tested who would have failed
- N4 is the number of hit and run collisions (N1) where at least one driver failed a breath test or refused to provide a sample
These component parts combine into two types of hit and run collisions, H1 and H2, as shown in the diagram below:
-
H1 are hit and run collisions where either all drivers are contacted, or at least one driver was not contacted but one other driver failed a breath test or refused to provide a sample. The total number of H1 hit and run collisions is N1 – (N2 – N3)
-
H2 are hit and run collisions where at least one driver was not contacted and there are no failed breath tests. The total number of H2 hit and run collisions is N2 – N3
The number of drink-drive collisions within type H1 is known and is N4. The number of drink-drive collisions within type H2 is unknown (X) and needs to be estimated.
It may reasonably be assumed that the police test hit and run drivers whenever practicable, so all drink-drive collisions within type H1 can be identified from the STATS19 data. However, there is no direct evidence to identify drink-drive collisions within type H2.
In order to estimate their number (X), some assumption must be made about the proportion which would have yielded a positive test if a test had been possible. The proportions for groups H1 and H2 would be equal if it were a matter of chance whether the police can trace a hit and run driver successfully.
Alternatively, the proportion would be lower for H2 than for H1 if the police were more likely to successfully trace a hit and run driver when they have particular reason to. The lower percentage of hit and run drivers in fatal collisions may suggest this. The article suggests that the true drink-drive rate for hit and run collisions of type H2 is Δ times the known drink-drive rate for hit and run collisions of type H1, and assumes that Δ lies between 0.5 and 1. Our estimates use the central value in this range of 0.75.
Sensitivity analysis was conducted and showed that varying the parameter has very little effect on estimates for the number of fatalities. When testing using 0.5 and 1 as the parameter, the estimates for fatalities only varied by around 1% in the years 2013 to 2022. Changing the value had a greater effect on the lower severities, with serious (adjusted) casualties varying by around 6% and slight (adjusted) casualties varying by around 11%.
With no other evidence to base our choice on and varying the parameter having little effect on estimates for more serious casualties, we decided to continue to use 0.75 as the value for Δ.
Thus,
Therefore, the estimate of the number of H2 type drink-drive collisions (HR) is estimated for each severity level from STATS19 data as follows:
The adjustment factor (separately calculated for each severity level) adds this estimate of hit and run drink-drive collisions proportionally to failed breath tests:
- HAF is the hit and run factor
- FB is the number of failed breath tests
- HR is the estimate of hit and run drink-drivers to add to known cases as calculated above
The hit and run factor (HAF) is used to estimate the number of drink-drive collisions and casualties (by severity) from STATS19 data by multiplying the appropriate factor with the recorded number of collisions and casualties where one of the drivers involved has failed a breath test (or refused to provide a specimen).
Combining these results with the estimated drink-drive fatalities from coroners’ and procurators fiscal’ data described above (DE) gives the estimated total of collisions and casualties believed to have involved at least one driver or rider over the drink-drive limit.
Uncertainty
In order to avoid the impression of false accuracy we round to the nearest 10, but this should not be taken as an indication of that being our assumed level of uncertainty.
We also provide 95% confidence intervals for the estimate of the number of drink-drive fatalities. This is calculated by assuming a Poisson distribution, Po(λ), with λ equal to the central estimate. Due to the central estimates being large (over 150), a Normal estimation of the Poisson distribution is used. Therefore, the distribution used for calculating the confidence intervals is Normal with mean λ and variance λ, N(λ,λ).
It should be noted that an assumption taken with modelling using a Poisson distribution is that all fatalities are independent from each other. However, some fatalities occur in the same collision and therefore are not independent.
On average, 17% of drink-drive fatalities occur in collisions in which other were other fatalities. If we assume that drink-drive collisions are independent and allocate a single fatality to each collision, we can assume that those fatalities are all independent. However, there are still 10% of fatalities left over which therefore aren’t independent from the already-allocated fatalities. This is a known limitation on our methodology and should be kept in mind when interpreting the confidence intervals for the fatality estimates.
Investigating missing coroners data and its potential bias
To investigate the cause of missing and potentially biased toxicology data from coroners, we consulted several coroners to understand the process of obtaining and testing samples.
Admission samples are often taken when a casualty arrives at hospital, but they are sometimes used by hospital staff or discarded before a coroner is able to request them for testing. This can be days later and therefore any other samples taken wouldn’t be an accurate reflection of the alcohol levels at the time of the collision. This accounts for some of the missing data.
Coroners do not prejudge the results and therefore should test samples from all deceased drivers. This means that the data shouldn’t over-represent drink-drivers because no matter the demographic of the driver, there sample should be tested.
However, some jurisdictions do not routinely test passenger and pedestrian samples unless there is reason to. Therefore, there may be some bias towards passengers and pedestrians having alcohol in their system.
Timeliness
Coroners’ and procurators fiscal’ data is typically available for analysis a year later than the main road collision data. This is due to the time required to obtain information on fatalities from police forces and then get records from coroners and procurators fiscal. Typically around half of the data expected to be available for analysis are ultimately available for inclusion in the provisional estimates. The estimates for fatalities depend mainly on coroners’ data and are particularly susceptible to revision between the provisional and final figures.
Changes to methodology
In 2025, the methodology was changed in two ways, which slightly affected the estimates between 2005 and 2022. These revisions to the statistics do not change the major trends in the data.
The first change was to stop double-counting collisions which had deceased drink-drivers and drivers who failed or refused a breath test. Previously, these collisions were being categorised as both a collision with a deceased drink-driver and a collision with a failed or refused breath test, resulting in double-counting. To improve the accuracy or our estimates, we have categorised these collisions only as a collision with a deceased drink-driver.
The second change was relating to the severity breakdown of the hit and run scaling factor. The calculation of the hit and run scaling factor is based on the number of drink-drive collision, split by collision severity but previously, casualties were being scaled based on the casualty severity, not the severity of the collision it was in. Therefore, we have corrected this so that casualties in a collision are scaled using the relevant scaling factor based on the severity of the collision.
These changes result in minor revisions to estimates of the number of fatal collisions and casualties of all severities. We have backdated these changes to 2005 so any estimates for 2004 and earlier are the same as they were in the figures published in 2024.
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.
Tablets and mobile devices normally have the option to “find in text” and “print or save” in their sharing or quick options menu of their browser, but this will vary by device model.
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.
Contact details
Road safety statistics
Email roadacc.stats@dft.gov.uk