Collision data sections and how to read them

The personal injury collision data section of a transport assessment exists to answer one question: is there an existing road safety problem that the development would make worse, or that the development could help to fix. A list of incidents with a total at the bottom does not answer it. A pattern does.

The data comes from the police record of personal injury collisions reported under the national road safety statistics regime, obtained from the local highway authority or from the published national dataset. Damage-only incidents are not in it, near misses are not in it, and collisions nobody reported are not in it. Say so in the report, because the gap matters when an objector cites a junction everyone locally knows is dangerous and the record shows nothing.

The word to use is collision, not accident. The change in terminology reflects a view that these events have causes, and officers notice when a report uses the older word.

Where the data comes from and what it covers

Each record carries the date and time, the location to a grid reference, the severity, the number and class of casualties, the vehicle types involved, and contributory factors recorded by the attending officer. Contributory factors are the most useful and the least reliable field, because they represent the officer's judgement at the scene rather than a finding after investigation.

Two consequences follow. Use contributory factors to identify candidate patterns, then test the pattern against the geometry on site rather than treating the field as proof. And never characterise an individual collision as the fault of a named party, because the record does not establish that and the report may be read by people connected to it.

Obtain the data rather than relying on a summary. A tabulated extract supplied by the authority sometimes omits the contributory factors and the casualty classes, which are the fields that let you see what is actually happening.

Study area and period conventions

Both are matters of judgement that should be stated and justified rather than assumed.

The study area is the set of links and junctions where development traffic would form a material part of the flow. That normally means the site access, the junctions immediately adjacent, and any junction on the main distribution route where the development adds a meaningful percentage to a turning movement. Extending the area to capture a cluster that development traffic will never reach is padding. Excluding a junction the traffic will use because its record is poor is worse.

The period is conventionally the most recent five full calendar years, with three years sometimes accepted for a small scheme. Use full years, state the start and end dates, and note the date the data was obtained, because the most recent year is often provisional.

Identifying patterns rather than totals

A total is almost meaningless without exposure. Twelve collisions on a busy urban distributor over five years may be unremarkable, while three on a quiet rural lane at the same bend is a pattern.

What to look for, in order:

1. Spatial clustering, meaning several collisions at the same junction or on the same short length.
2. A repeated manoeuvre, such as right turn against oncoming traffic, or emerging from a minor arm.
3. A repeated casualty class, particularly pedestrians or cyclists at the same crossing point.
4. Time, lighting and surface conditions, such as a cluster in darkness or on wet surfaces.
5. A repeated contributory factor across several records.

Then ask whether the pattern has a physical explanation you can see on a site visit. A right turn cluster at an access with substandard visibility to the right is a pattern with an explanation. A scatter of unrelated single incidents across five junctions is background.

And resist the temptation to find a pattern because the scheme needs one. Reports that discover a convenient pattern justifying works the developer wanted anyway are read sceptically, and reports that discover none are often the honest answer.

Severity and what it changes

Severity drives attention more than numbers do, and there is a trap in the figures.

The method by which police forces record injury severity changed as forces adopted injury-based reporting systems, and the transition happened at different times in different force areas. The effect was to reclassify as serious a proportion of injuries previously recorded as slight. A five-year dataset that straddles the transition in that force area will show an apparent rise in serious collisions that reflects recording rather than road safety.

Say this in the report. An analysis that presents a rising serious-injury trend across the transition, without noting it, is wrong in a way that a highway officer will spot immediately and that an objector's consultant will use.

Fatal collisions are different again. They are rare, individually significant, and should be described factually with no speculation about cause. Where a fatality occurred within the study area, expect it to be raised at committee.

Linking a pattern to the proposal

The link has to be specific. For each pattern identified, the analysis should say which movement the development adds traffic to, how much it adds as a proportion of the existing movement, and whether that increase makes the conflict more likely or more severe.

Where development traffic does not use the movement involved in the pattern, say so plainly. A cluster of right turn collisions at a junction a kilometre from the site, on a route the development will not use, is context rather than impact.

Where it does, the proportional increase is what matters. A development adding two per cent to a turning movement with an existing pattern is in a different position from one adding 40 per cent. Set the reasoning out in the Transport Assessment, or in a shorter Transport Statement where the scheme is small enough that a full assessment is not required.

Where the data supports mitigation

The useful outcome of a collision analysis is a proportionate, evidence-led intervention rather than a general offer of improvement.

Interventions that follow from particular patterns include improving visibility at an access where emerging collisions cluster, a right turn lane or ghost island where right turn conflicts recur, a controlled or uncontrolled crossing where pedestrian casualties cluster at a desire line, signing and lining changes where a bend or junction is poorly perceived, speed reduction measures where speeds and collision severity correlate, and lighting where a darkness cluster exists.

Each of those is delivered through a highway agreement with the authority rather than through the planning permission itself, so the assessment should identify the mechanism as well as the measure. Behavioural measures belong in a Travel Plan, which is the right place for mode shift targets and monitoring but is not a substitute for a physical fix where the data shows one is needed.

One honest limitation to acknowledge: five years of records at a single junction is a small sample, and small samples fluctuate. An assessment that treats a two-collision difference as a trend is overreading the data, and a reviewer who knows the dataset will say so.

This is general guidance, not legal advice. Data sources, severity reporting conventions and local analysis requirements change, so confirm the study area and period with the highway authority before the work is done.

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