Sixty percent of them were not at a junction.
The mental model of cycling danger is the intersection. The 2024 fatality data puts most deaths somewhere else entirely.
Ask most people where cyclists get killed and they will describe a junction. A turning lorry, a driver who did not look, a collision at a place where two streams of traffic were always going to meet. It is the scenario that infrastructure design spends most of its attention on, and reasonably so.
In 2024, 30 percent of American cyclist fatalities happened at an intersection. Sixty percent happened somewhere that was not an intersection. The remaining 9 percent were in bike lanes, on shoulders, at driveways, on sidewalks and medians.
What “not at an intersection” does and does not tell you
It is worth being precise here, because this statistic is easy to over-read and we would rather not.
The FARS field being reported is a location. It says where on the road network the rider was struck. It does not say what the vehicle was doing. “Not at an intersection” includes being hit from behind while riding in a lane, being hit while a driver overtook and cut back in, being struck head-on by a vehicle on the wrong side, and being hit by a car that left the road entirely. Those are different failures with different remedies, and this number does not separate them.
So we will not claim that 60 percent of cyclists killed were rear-ended. That would be reading a manoeuvre out of a location field, and it is not there.
What the number does establish is narrower and still significant: most people who died were killed in open road away from a junction, which is exactly the stretch where the conventional model has least to say. Junction design is a mature discipline. Mid-block, the standard interventions are thinner, and the dominant variable is how a driver behaves in the seconds while passing.
The other two figures in the same chart
Two more from the same release complicate the received picture further. Eighty-one percent of cyclist fatalities happened in urban areas, not rural ones. And 52 percent happened in the dark, against 44 percent in daylight, despite far fewer people riding at night.
The dark figure is the one that should be uncomfortable. It is usually reached for as an argument about lights and conspicuity, and lights obviously matter. But the same data shows the single worst window is 6 to 9 pm, which in much of the country is not dark for a good part of the year, and which is also when the roads are fullest and drivers are most tired. Attributing the night-time share to visibility alone is a guess, and it happens to be the guess that places the responsibility on the person who was killed.
Where this leaves an instrumented bike
We build a system that measures the lateral distance of every vehicle that overtakes you, on every ride, and records the ones that come close. It is worth being clear about what that can and cannot contribute to the picture above.
It cannot prevent a collision. There is no intervention in the loop: nothing warns the driver, nothing steers the bike. A camera measuring a pass is a witness, not a guardrail.
What it can do is produce the measurement that the fatality data structurally lacks. If most deaths happen away from junctions, then the behaviour that matters most is overtaking behaviour on ordinary road, and nobody currently measures it at scale. We know how many people died. We do not know, for any given road, how close the average pass is, whether it is worse at 7 pm than at 7 am, or whether it improved after the council painted a line.
Those are answerable questions. They need a lot of rides and a consistent method rather than a new theory. That is the part we can actually help with, and we would rather describe it in those terms than claim the technology saves lives, which we have no evidence for and no way to demonstrate.
Source: NHTSA, Traffic Safety Facts: Bicyclists and Other Cyclists, 2024 Data (DOT HS 813 822, July 2026).