Why struck-by incidents keep happening around heavy equipment

Every control we use to keep people away from moving equipment has been standard for thirty years or more. Backup alarms, mirrors, high-visibility clothing, spotters, toolbox talks. Workers are still being struck. That is not because sites are careless. It is because almost all of those controls share one design assumption, and the assumption is wrong.

What the national numbers do and do not tell you

The Association of Workers' Compensation Boards of Canada recorded 1,024 accepted work-related death claims in 2024, of which 381 were traumatic injury cases and the remainder were attributed to occupational disease. There were 267,199 accepted lost-time injury claims nationally in the same year. Alberta recorded the highest number of injury-related deaths of any province, at 91.

Two honest caveats. These are all-industry figures, not construction figures, and they are not struck-by figures. And they count accepted claims, which is a narrower thing than incidents: near misses, unreported contacts and the many cases where a worker stepped clear at the last second never appear anywhere. The number that would actually tell you about risk on your site is the one nobody collects.

That gap is worth naming, because it shapes how these decisions get made. Most fleets have no idea how often a person comes within a metre of a moving machine, so the risk is invisible until the day it is not.

The shared assumption

Look at the standard controls and ask a single question of each one: who has to act for this to work?

• Backup alarm. The worker has to hear it, identify which machine it belongs to, and move.
• Mirrors and cameras. The operator has to be looking at the right one at the right moment.
• High-visibility clothing. Someone has to be looking in that direction for it to do anything at all.
• Spotter. A second person has to be present, positioned correctly, watching continuously, and able to signal in time.
• Toolbox talks and procedures. Everyone has to remember, under time pressure, weeks later.

Every one of them is a control that depends on a human reacting correctly in the two or three seconds available. That is the shared assumption. Each control is reasonable on its own, and stacking five of them feels like defence in depth, but they are not independent layers. They fail in the same way, at the same moment, for the same reason.

Where each one actually breaks

Alarms become background noise

Habituation is not carelessness, it is how attention works. On a site where reversing alarms sound continuously for eight hours, the signal carries no information. Add hearing protection and the noise of an active pad, and a control that is legally satisfied can be practically inaudible.

Spotters are expensive, and therefore intermittent

A dedicated spotter is genuinely effective when the arrangement is what the regulations describe: continuous view of the operator, complete view of the area the machine is moving into. The difficulty is that this is a full-time role per machine. Under schedule pressure it becomes a person with other duties who is watching most of the time, which is a different control wearing the same name.

Cameras move the problem rather than solving it

A rear-view camera gives the operator more information, which is a real improvement. But it adds a screen to a cab that already has several, and it still requires the operator to be looking at it in the moment that matters. Detection without a decision is just more data.

The hierarchy of controls, and why this matters

Standard safety practice ranks controls by how much they depend on human behaviour, from most reliable to least: eliminate the hazard, substitute something safer, engineer a control into the work, administrate with procedure and training, and finally protect the individual with PPE.

Sort the list above into that hierarchy and something becomes obvious. Separating pedestrians from equipment routes is elimination and is the strongest thing you can do. Everything else on the list is administrative or PPE, sitting at the bottom two rungs. For most fleets, the entire middle of the hierarchy is empty.

An engineering control is one built into the work itself, so it operates whether or not anyone is paying attention. In this context that means a system that acts on the machine rather than on the person: it detects, it warns, and if nobody responds it stops the equipment. The value is not that it detects better than a human. It is that its response does not depend on a human at all.

What to ask a supplier

1. Does it measure distance, or only motion? A system that cannot tell how far away something is will either miss real hazards or cry wolf until the crew disables it.
2. What is the false alarm rate on a site like ours? Nuisance alerts are how safety systems get switched off, and a disabled system protects nobody.
3. Does it stop the machine, or only warn? These are different controls at different levels of the hierarchy.
4. Does it work without a network connection? Most worksites have unreliable coverage, and a control that degrades with signal is not an engineering control.
5. How do we know it is working today? Self-checks and a visible record matter more than a specification sheet.
6. What are its limits? Any supplier claiming a system sees everything in every condition is telling you something that is not true of any sensor, and you should ask what happens in the gaps.

The short version

Struck-by incidents persist because the controls we have relied on all ask a person to notice something and react in a couple of seconds. They are worth keeping. They are not sufficient on their own, and adding a sixth control of the same kind will not change the outcome. The gap in most fleets is a control that acts on the machine.

In Alberta this is not only good practice, it is written into the Code: s.267 names an automatic stopping system as one of three compliance options when an operator's view is obstructed. You can also see how detection and the stop work on a real machine.