A low risk scoreis a decision that needs a date on which it is reopened

VAt about 17:13 on 19 June 2026, East Midlands Railway train 1H46 collided with the rear of train 1B67, which was stationary on the Up Fast line about 2.5 miles south of Bedford. The driver of 1H46 died. Of an estimated 257 people on board the two trains, at least 160 were injured and 102 were taken to hospital¹².

VSignal WH154, the last signal before the stationary train, was correctly showing red. Train 1H46 passed it at about 58 mph and accelerating. Why it passed the signal remains under investigation¹.

VThe Rail Accident Investigation Branch (RAIB) records three facts about the signal. It had no train-stop equipment, which the Railway Safety Regulations 1999 did not require for this signal. Network Rail scored it at 3.844 x 10-7 fatalities and weighted injuries a year on 29 October 2025, against a threshold of 3.000 x 10-5 for detailed assessment. And its warning magnet was found 16 mm below railhead level, outside a tolerance of plus or minus 12 mm¹.

VRAIB's interim findings may change, and it states that its reports are not to be used to assign fault or blame¹. This article attributes no cause to the collision.

VNetwork Rail assesses each signal at least every five years and after specified trigger events. Only the 1,260 signals above the threshold go to the detailed model, where fitting train-stop equipment becomes possible. WH154 scored well below it¹.

VEast Midlands Railway began introducing class 810 trains in December 2025, after the October assessment. Train 1B67 was one, with a control panel that differs from the standard panel its driver knew. It received a warning at a green signal, which went unacknowledged, and witnesses did not recognise the alarm. A second class 810 train received a false warning at the same signal on 9 July 2026¹.

AA score can be correct against its method and silent about a system that has changed. The track magnet, the train and its cab panel sit with different owners, so no single assessment sees them meet.

VRAIB's remaining scope includes the processes used to assess overrun risk at this location and the effectiveness of the control measures intended to address it. It found the magnet's position only after the second false warning¹.

AA low score keeps the asset out of the group that gets scrutiny. Every signal below the threshold depends on the first screen's assumptions staying true, and when traffic, fleet or equipment moves, the screen cannot say so unprompted.

AInterfaces carry the exposure. A tolerance reading can be monitored as asset condition, and a new train meeting old infrastructure can trigger a review. Neither needs a failure first.

AKIS PLUS LTD's view is that a risk score belongs in the decision record. Three controls would test it: a reopen date for every low-scored safety-critical asset, set by what changes around it as well as by the calendar; trackside tolerances monitored as asset condition; and an introduction check when new rolling stock or software meets legacy infrastructure.

IIf the final report finds that the assessment method or the magnet position contributed, owners of other similarly scored assets will want to know when each score was last reopened.

OUR PERSPECTIVE

A risk score records a decision, and a decision needs a date on which someone reopens it.

1. For each safety-critical asset we scored as low risk, when will that score be reopened, and what change would bring the date forward?
2. When did we last measure the tolerances of equipment our low-scored assets depend on, and who owns the reading?
3. When new rolling stock or software arrives, who checks how it behaves with the oldest asset it meets?