Using personal injury collision data properly

Collision data in a planning application is usually presented badly. A table of eleven incidents over five years, a plan with red dots on it, and a sentence saying there is no evidence of an existing highway safety problem. That is a data dump, and a highway officer reading it learns nothing.

The purpose of the analysis is to answer two questions. Is there an existing safety problem at this location, and would the proposed development make it worse. Everything in the section should serve one of those.

Where the records come from and what they contain

Injury collisions attended by or reported to the police are recorded on the national reporting form, and the records pass to the local highway authority, which holds them for its own network. The Department for Transport publishes national statistics derived from the same source.

Each record carries a location, a date and time, the severity, the vehicles and casualties involved and, in most authority extracts, a short description and contributory factors recorded by the attending officer.

Request the data from the highway authority for a defined area and period. Some authorities charge, some supply it free to agents working on applications in their area, and a few publish a public viewer. Ask for the full record rather than a summary, because the contributory factors are where the useful information sits.

Two things the dataset does not contain. Damage-only collisions, which are the majority of all collisions. And near misses, which is where most of the real evidence about a bad junction lives and which no dataset captures.

Under-reporting is documented and it is uneven. Slight injury collisions are reported less consistently than serious ones, and collisions involving cyclists with no other vehicle involved are substantially under-recorded. Say so in the report.

Setting the study area and the period

Both should be agreed rather than chosen. Getting them scoped with the highway authority at the outset avoids the accusation that the study area was drawn to exclude an inconvenient junction, which is the criticism this section attracts most often.

The study area normally covers the site access and the junctions within the area of influence, which is a function of the trip distribution rather than of a fixed radius. Where the development traffic will route through a particular junction, that junction is in the study area even if it is a kilometre away.

Five years is the usual period. It is long enough to show a pattern and short enough to reflect the current layout.

Where the highway has changed during the period, split the analysis. A junction rebuilt three years ago has two datasets: one describing a layout that no longer exists, and one describing the current arrangement. Presenting them as a single five year total is misleading and reviewers spot it.

Analysis, not enumeration

A pattern is something the records have in common. Look for:

• Repeated manoeuvre types, such as several right turns into a side road, or several rear-end collisions on the approach to a signal.
• Clustering at one point rather than an even spread along a link.
• Common conditions: wet surfaces, darkness, a particular direction of travel, a particular time of day.
• Common road user groups, such as pedestrians crossing at a particular gap in parked cars.
• Contributory factors that recur, such as failure to look properly or travelling too fast for conditions.

Then say what it means. If four of the eleven collisions were right turn manoeuvres into the same side road and your development adds right turning traffic to that arm, the analysis has found something and the design has to answer it. If the eleven collisions are unrelated in type, location and conditions, that is a finding too and it should be stated as one.

The temptation to resist is comparing the raw number against some notional average without a defensible basis. Rates depend on flows, and unless you have reliable flow data for the same period, a rate is a made-up number.

Linking the data to the proposal, or ruling it out

The link runs through mechanism, not through volume alone.

Where the proposal adds turning movements at a location with a turning-related pattern, the connection is direct and the answer is a design change: improved visibility, a right turn lane, relocating the access, or a package of measures secured under a highway agreement.

Where the proposal adds pedestrian trips at a location with a pedestrian casualty record, the answer is usually crossing provision rather than a change to the access.

Where the pattern has no relationship to what the development changes, say that explicitly and explain why. A rear-end pattern at a signal 400 metres away is not affected by the addition of eight dwellings, and stating the reasoning is more persuasive than presenting the table and leaving the officer to reach the conclusion.

Visibility is where the analysis and the design most often meet, so cross-refer to the geometry: the visibility splay standards applied at the access should be consistent with the recorded speeds used in the collision analysis. Using an 85th percentile speed for the splay and the posted limit for the collision discussion is an inconsistency a reviewer will pick up.

Common errors that undermine the section

Four recur often enough to be worth naming.

Using a published national summary instead of the authority's own records. Summaries are aggregated and cannot be plotted, and a reviewer who holds the underlying data will see the difference immediately.

Quoting a collision rate with no flow data behind it. If you have not counted the traffic, you cannot express collisions per million vehicle kilometres, and inventing one is worse than omitting it.

Treating contributory factors as findings of fault. They are the attending officer's judgement recorded at the scene, not a determination of liability, and they should be described as such.

Ignoring the years with nothing in them. A location with collisions in two years out of five and nothing in the other three is telling you something about how the site is used, and a total figure hides it.

Presenting it in the report

Keep the structure simple and predictable.

State the source of the data, the area it covers, the period and the date it was obtained. Provide the plan with collisions plotted by severity and year. Provide the table. Then two or three paragraphs of analysis, followed by a conclusion that addresses the policy test directly.

That last point matters. National policy is clear that development should only be prevented or refused on highways grounds where there would be an unacceptable impact on highway safety, or where the residual cumulative impacts on the road network would be severe. The conclusion should engage with that wording rather than asserting that the proposal is safe.

For a scheme large enough to need a Transport Assessment the collision analysis is a chapter. For a smaller proposal it belongs in the Transport Statement and can be two pages. Either way, it should be written by someone who has read the individual records rather than the summary counts.

One honest caveat worth including. Recorded data describes what happened at a location under its current use, and a development changes that use. The analysis supports a judgement; it does not make one.

This is general guidance rather than legal advice. Check the position with your local planning authority or a qualified professional before relying on it. To see which transport documents a site needs, start with the postcode.

PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval.