A5 mapped collision record: 39 deaths, 173 serious injuries and a major legal turning point

A single number cannot tell the story of deaths on the A5. Within the road corridor mapped by PureCar from police injury-collision records for 2013–2025, 31 collisions were classified as fatal. The people recorded as killed in the linked casualty files numbered 39. These are different units of measurement: 31 events with fatal outcomes, and 39 human lives lost. Both figures matter, and neither should be casually replaced by a much broader death toll reported for a different period or stretch of road.

The same mapped corridor contains 122 serious-severity injury collisions and 960 slight-severity collisions. Across all categories that makes 1,113 police-recorded injury incidents and 1,848 associated casualties. Behind the fatal count are another 173 people recorded seriously injured and 1,636 recorded slightly injured. The evidence therefore reaches beyond the fatalities that have rightly dominated public discussion of the route.

Those findings now sit alongside a consequential change in the future of the A5 Western Transport Corridor. On 6 October 2026 the Court of Appeal ruled in favour of the Department for Infrastructure on the grounds before it, clearing a major legal obstacle to progressing the proposed new route. On 26 October, the department reported continuing engagement with more than 300 landowners affected by land vesting. It has not announced a firm date for major construction; its stated undertaking is that major work on affected lands will not begin before the end of December 2026.

The figures tell a history of recorded harm. The court decision concerns a proposed infrastructure response. They belong in the same public debate, but it would be wrong to claim the mapped figures prove how many lives a future scheme will save.

Thirty-one fatal collisions, 39 people killed

The distinction between fatal collision events and fatal casualties is essential on a road where successive deaths have shaped a long-running public campaign. A police collision record is a record of an incident. A casualty record refers to an individual harmed in that incident. One collision can involve multiple vehicles and more than one person killed or injured. Here, the 31 fatal-severity incidents have 39 linked deaths, eight more people killed than fatal collision events.

It would be a mathematical error to call the 31 fatal collisions “31 deaths”. It would also be wrong to treat the difference of eight as proof that exactly eight collisions each claimed two lives. Multiple combinations of casualties per event could yield the same totals. Without examining individual linked records, the aggregate alone does not identify how deaths were distributed among those incidents.

PureCar's 13-year A5 mapped corridor identifies 2,256 associated vehicle records. Those are records of vehicles involved in the mapped incidents, not 2,256 separate crashes. Similarly, the 1,848 casualties are not 1,848 collisions. The underlying dataset is organised this way because road harm involves relationships among an incident, the vehicles and the people affected, and each layer answers a different question.

Most recorded injury collisions in the area were classified slight: 960 out of 1,113, about 86.3%. There were 122 serious-severity incidents, approximately 11.0%, and 31 fatal-severity incidents, around 2.8%. That distribution should not be read as diminishing the serious cases. It shows the relative frequency of incident categories over thirteen years, while the 173 serious-injury casualties indicate that serious harm can affect more people than the number of serious-classified incidents.

The fatal classification refers to the worst injury resulting from an event, rather than to the number of fatalities in it. A collision with a fatality could also include other injured people. Reporting both casualty and collision figures is the only way to avoid hiding that additional harm.

The yearly record complicates any straightforward trend

The mapped A5 corridor recorded 81 injury collisions in 2013, 85 in 2014, 91 in 2015 and 97 in 2016. The annual count stayed close at 96 in 2017 before rising to 119 in 2018, the highest year in this series. It then fell to 93 in 2019 and 73 in 2020, before counts of 76 in 2021, 75 in 2022, 73 in 2023, 79 in 2024 and 75 in 2025.

The latest two years were close: 75 in 2025 compared with 79 in 2024, four fewer collisions, or a decline of roughly 5.1%. That is a modest year-on-year difference in a series which has ranged from 73 to 119. It is not evidence that the A5's underlying risk has been solved, nor proof that the route became more dangerous in the years when counts increased.

To see the broader shift, compare equal six-year spans. From 2013 through 2018 the mapped corridor recorded 569 injury collisions, averaging about 94.8 per year. From 2019 through 2024 it recorded 469, around 78.2 a year. The latter period had 100 fewer incidents, a decline of about 17.6% in the recorded count. The year 2025, considered separately, was slightly below that second period's annual average.

That result is important because it prevents selective attention to the single high year of 2018. But it has a serious limitation. Both six-year windows count injury collisions, not traffic exposure; the comparison does not tell us how many people used the road, how far vehicles travelled or how traffic moved between adjacent routes. Nor does it correct for the exceptional travel conditions of 2020, which lies in the later window.

A rolling average offers a different reading. PureCar's published three-year average for the corridor was 104.0 collisions per year for the period ending in 2018. The equivalent figure ending in 2025 was approximately 75.7. The moving average helps prevent a single year from overwhelming a discussion, but it does not turn a collision count into a measure of danger for an individual driver.

A smaller recorded collision total can coexist with a very serious history of fatal and life-changing incidents. Conversely, a rising count of slight-injury events does not automatically imply a matching rise in deaths. The casualty breakdown must remain visible throughout the comparison.

Which A5 is being counted?

The geographic definition is perhaps the most consequential qualification in this investigation. PureCar's road page matches historical PSNI collision coordinates to 60 mapped sections of the DfI road network and includes incidents within 100 metres of the selected alignment. A collision at an adjoining junction, local road or connecting section can fall inside the search area even if it did not take place on the A5 carriageway itself.

That makes the 1,113 incidents a count for the mapped A5 corridor, not a certified whole-road collision tally. The deaths and serious injuries refer to people linked to those same selected incidents. The definition offers a reproducible way to explore a complex route and its immediate surroundings, but it cannot identify every location as a confirmed A5 carriageway incident without case-level checks.

A separate official statement concerning the A5 Western Transport Corridor says more than 50 people have died on the existing A5 since 2006. PureCar also displays a historic reported figure of 58 deaths on its road-safety hub. Those broader public statements are not directly interchangeable with 39 deaths in the 2013–2025 proximity dataset: they use a different starting year and may cover different sections and rules for assigning a collision to the route.

Rather than declaring one tally correct and another wrong, an authoritative comparison would establish the precise time frame, sections, collision references and geographical inclusion criteria used in each. In a public debate shaped by grief and urgent questions about investment, misleading precision can be as damaging as an obvious error.

The corridor model also includes 1,636 slightly injured people. Those records matter for assessing repeated lower-severity harm around junctions and local access points, yet they should not be used to infer the effect of a proposed new dual carriageway. The source data records what happened in historical locations, not a controlled comparison of old and new road designs.

A5 plans enter a new phase after the October court ruling

The Department for Infrastructure's A5 proposal is a major transport project, not a minor alteration to the existing road. Its October 2026 update describes a scheme of approximately 85 kilometres of new road, around 82 kilometres of it dual carriageway, linking New Buildings towards the border south of Aughnacloy through the north-west. The department describes safety as central to the proposal alongside connectivity and economic objectives.

In June 2025 the High Court ruled against the earlier decision allowing the authorised section to proceed. The department appealed. Following proceedings that included consideration of obligations under Northern Ireland's Climate Change Act, the Court of Appeal ruled in the department's favour on 6 October 2026. The department has since said further legal, commercial, financial and governance requirements must be completed before construction arrangements can be finalised.

That matters for how the road-safety record is reported today. A court judgment permitting officials to advance the scheme is not the same thing as construction beginning immediately. The department's current public update says negotiations with contractors, completion of environmental requirements, approval of the full business case and mobilisation planning remain part of the process. A further attempt to appeal could affect the timetable. Any article that treats the court result as the opening of the new road would be inaccurate.

The department's 26 October statement says it has continued to engage with more than 300 landowners affected by vesting arrangements. It acknowledges uncertainty surrounding land use and compensation after the earlier ruling and says claims arising from evidenced financial impacts will be assessed under the relevant arrangements. The minister has committed to no major construction on affected land until after 31 December 2026. These are live consequences for rural residents and businesses along the proposed alignment, not footnotes to a collision graph.

The latest departmental questions-and-answers document places accumulated project spending from inception to 2025/26 at £192.5 million, including £38.8 million in 2025/26, with the latter figure not yet audited. It says the full scheme's 2024 emerging cost estimate was around £2.1 billion and that estimates are under review. The department also records the Irish Government's commitment of €600 million. Project expenditure, road-safety benefits and casualty counts are separate subjects; comparing them as if each death carries a simple price tag would misrepresent both public finance and human loss.

What the evidence says about present safety

The 31 fatal-severity collisions were not distributed evenly by definition through an unchanging transport system. The published annual corridor table tracks all injury collisions, but the road-level summary does not show annual fatal collisions, annual serious injuries or casualty ages in the same time series. Therefore this article cannot assert that the fatal toll accelerated or slowed in the same way as the overall number of recorded injury incidents.

The PSNI's national 2025 data add essential context: 56 people were killed and 1,066 seriously injured across Northern Ireland, with 5,015 injury collisions in total. Of those national KSI casualties, 115 were associated with inattention or diverted attention as the principal factor; 95 with wrong course or position; and 86 with excessive speed for the conditions. But that national breakdown cannot be lifted out and ascribed to individual A5 crashes. Doing so would misstate what was investigated and where responsibility lies.

An examination of the corridor's 2,256 linked vehicle records might provide clues about collision configurations if analysed alongside locations and casualties, but the total alone does not tell us which vehicle types dominate fatal incidents. A researcher could examine whether severe collisions were concentrated around particular junction approaches, overtaking opportunities or access roads. The current published aggregates do not establish those patterns, and this report does not claim they have been found.

The wider Northern Ireland Road Safety Strategy to 2030 gives another caution. In 2025 it recorded 43 fatalities on rural roads, representing most road deaths, and 1,066 serious injuries, 42% above the strategy's baseline figure of 751. The strategy is concerned with trends, road users and exposure-adjusted rates. It does not certify the risk per mile on this particular A5 corridor. Describing the A5 through its own mapped collisions and the wider rural problem are complementary, not equivalent, exercises.

Many of the considerations that attract attention in the political argument—the need for a safer route, environmental obligations, landowner protections, public spending and the pace of approvals—have independent evidence and competing claims. A collision database should inform that debate without being stretched into a verdict on every policy question.

Even when a collision is mapped near the A5, the injury severity is recorded at the event level and not an assessment of the design of that road. The Police Service of Northern Ireland's source files cannot by themselves distinguish a road-layout contribution from a series of independent driver actions. For the route's public inquiry and policy arguments, these limits are substantial: they mean that any prediction of benefits from the new alignment must be based on a separate appraisal, not merely the 31 fatal-severity events displayed in a proximity search.

What would demonstrate an improvement?

When a new road or substantial junction changes eventually become operational, testing road-safety outcomes will require a baseline that matches the locations actually affected by the works. PureCar's 100-metre matched corridor is one source of historical evidence, but a formal before-and-after evaluation would need to define old and new sections precisely, account for altered traffic routes and distinguish traffic growth from changes in collision risk.

That evaluation should not be satisfied merely by fewer incidents on one surviving piece of the existing road. A bypass may divert vehicles elsewhere; the harm across a relevant network needs to be assessed. Nor should a reduction in slight-injury collisions obscure any movement in severe outcomes. Fatal and serious casualties should be reported separately, alongside counts adjusted for how much travel has occurred.

For now, the records establish that 39 people were killed and 173 seriously injured in police-recorded injury collisions selected by PureCar's A5 mapped corridor between 2013 and 2025. They show an overall injury-collision count lower in recent years than around the 2018 peak, within a series with significant annual variation. The official project documents establish that the A5 scheme has a new legal footing after the October appeal result but still faces necessary approvals and practical steps before building.

There is no contradiction between those findings. Fewer recorded injury collisions than in the late 2010s do not erase decades of bereavement; the severity of past harm does not by itself prove which engineering design will deliver the largest benefit. The proper public-interest test is whether future decisions are grounded in consistent, independently checkable measures of injury and loss, with the human and legal consequences disclosed as carefully as the headline totals.

Sources and definition

PureCar's A5 collision investigation supplies the 2013–2025 100-metre corridor counts, annual series and linked casualty data. The underlying records derive from PSNI injury-collision datasets and DfI road-network geometry. The PSNI's 2025 Key Statistics Report supplies the national comparisons. The Department for Infrastructure's A5 overview and its 26 October 2026 landowner update establish the project's current procedural position. None of the corridor's proximity-matched totals should be described as a confirmed whole-road death count.