Darllenwch y dudalen hon yn Gymraeg.

In our previous blog on understanding road collision and casualty statistics, we described changes to how injury severity is recorded by police forces across Great Britain. We also committed to reviewing how these changes affect our statistics.

This blog provides further information on the changes to injury severity reporting, the expected impacts and adjustments needed, and our plans for communicating these changes to users.

Change in police recording injury severity

Casualties in road collisions are recorded as either killed, seriously injured or slightly injured. Historically, this was based on the judgement of reporting police officers. However, many police forces across Great Britain have adopted an injury-based reporting system (IBRS), where severity is determined using a predefined list of injuries. This approach reduces subjectivity and is intended to improve both accuracy and consistency across forces.

By 2016, around a half of English police forces had adopted IBRS and the Department for Transport (DfT) noticed that the change in system led to a change in the reported severity of road casualties, with a higher proportion of injuries categorised as serious after IBRS was introduced.

Dyfed-Powys police force became the first Welsh force to move to IBRS in May 2023 followed by South Wales police force in November 2025. Gwent and North Wales police forces have not yet moved to IBRS.

Severity reporting under the old and new systems are not consistent, and so we need to adjust historical road casualties data by severity to allow us to compare the data consistently over time and between police forces. The methodology is explained in the next section.

In our next publication, Quarter 1 (January to March) 2026 data due to be published on 29 July 2026 we will be publishing adjusted figures for the first time whilst continuing publishing unadjusted figures. The new adjusted casualty severity data will be published as ‘official statistics in development (Office for Statistics Regulation)’. In addition to adjusted casualty data, we also plan to introduce adjusted collisions data from the Quarter 2 (April to June) 2026 publication onwards.

The move to an injury-based reporting system does not affect the total number of collisions and casualties, or the total number of fatalities.

How the adjustment works

In 2019, the Office for National Statistics (ONS) and DfT developed a methodology designed to adjust the impact of IBRS on reported injury severity (DfT). This methodology uses a statistical model to estimate how severe each case would be if it had been recorded using IBRS at the time rather than based on the police officer’s judgement. The model then adjusts the figures to reflect what it estimates the severity would have been if IBRS had been used. This means that the trends in severity reflect what really happened rather than reflecting the change in reporting method.

DfT implemented the severity adjustment methodology for the first time in their 2018 statistical release (published in 2019) and published both adjusted and unadjusted road casualty figures. DfT continue to publish both adjusted and unadjusted figures in their data tables for all the police forces in Great Britain.

Testing the adjusted method for Wales

Due to 2 of Wales’s 4 police forces moving to IBRS, now is an appropriate time for us to move towards implementing adjustments.

We tested 2 models which both used ONS’s model with different input data:

1. Model 1: uses the statistical model on all available GB data (2004 to 2024) and the latest available data for Wales (2025).
2. Model 2: uses the same model on all available data for Wales only.

We tested these 2 methods because model 1 aligns more closely with the input data used by DfT, but we also wanted to assess the impact of a Wales-only model to understand whether that would be more suitable for our statistics.

The results for model 1 are very similar those reported by DfT. This is to be expected as incorporating 2025 Wales data represents only a minor adjustment to DfT’s original model.

Under model 2, the adjusted estimates of serious casualties were consistently higher than the model 1 and DfT results. This could be due to the limited amount of IBRS data that was used in the model (primarily Dyfed-Powys police force data over a period of 2 years). Collisions occurring in Dyfed-Powys police force area may not be representative of the rest of Wales because the methods used to evaluate the severity of injury can vary between police forces. For example, Dyfed-Powys have larger proportion of rural roads compared to other parts of Wales.

Future plans

From the July 2026 publication, we will use model 1 to produce adjusted casualty severity estimates for all non-IBRS data from 2004 onwards. This model draws from a large GB wide data set, so we have more confidence in the severity adjustment than if we used the alternative method where the sample size of Welsh data using IBRS is too small at the moment. Method 1 is also consistent with the approach taken by DfT.

These adjustments affect the breakdown between slight and severe injury. This means that most historical data on road casualty severity will be revised to enable us to compare consistently across time and between police forces. We will advise users to use the adjusted figures when comparing severity data over time or between forces, but for transparency we will also publish the unadjusted data.

The adjustments do not affect the total number of collisions, casualties or fatalities. They will not affect Dyfed-Powys Police data from May 2023 onwards or South Wales Police data on road casualty severity from November 2025 onwards when they started using IBRS.

We aim to publish adjusted collision data in the Quarter 2 2026 publication.

We will run the model annually from our 2026 annual release in summer 2027. This means that the historical data will again be revised as the model will include more data.

We will monitor the scale of annual adjustments. As other police forces move to IBRS and we have more IBRS data, we will continue to work with our partners (including DfT, ONS, and independent academics) to ensure we are using the most appropriate methodology for Wales.

Over time, the scale of revisions will reduce. Once there is enough Welsh IBRS data and the annual revisions are minimal, we will finalise the adjustments and stop revising the historical data.

We will continue to work with our users as the remaining Welsh police forces move to IBRS to ensure we have the most accurate police recorded road collisions data for Wales.