The punch was the easy part

I was punched in the face at work once.

Years ago I was supporting a young man with severe autism. I got too close, too far into his space, and he got me right on the nose. It hurt. It was also, honestly, fine. And if we'd done the paperwork, the paperwork would have been easy. Time, place, what happened, who was involved. A sore nose as evidence.

The hardest thing that's happened to me at work has no paperwork. It was a year under a CEO I couldn't trust not to undermine me, in private and in public. The punch stung for a day. That year followed me home. It disrupted my sleep and had me doubting myself for months.

I wrote about this on LinkedIn a few weeks back and asked a question: why do we find it so much harder to capture and report psychosocial risk than the physical kind? I said I had a hypothesis, and that I'd rather hear other people's first. Fair enough. Here's mine, worked through properly this time.

The short version: an incident report is built to catch events, and most psychosocial harm isn't an event.

Something happens, at a moment, to a person, and it gets written down. That works for hazards that behave like that. A fall, a cut, a punch. But chronic overload, role confusion, a manager who slowly grinds someone down, the long tail of a badly run restructure? These aren't events. They're exposures. They accumulate. There's no moment to log, so a system that waits for moments never records them.

You don't have to take that on faith, because we can start from something already well evidenced. Minor physical injuries are under-reported, and not slightly. Across studies, somewhere between a fifth and nine-tenths of workplace injuries never make it into the record. Whether an injury makes it into the record isn't random. It turns on a handful of properties. How severe it is. How visible it is. Whether it happened at a clear moment. Whether it can be traced to a cause. Whether it carries legal weight. The more of those an injury has, the more likely it is to be reported.

Now hold psychosocial hazards up against that list. Invisible to a bystander. Slow to arrive. Hard to pin to a cause. Rarely attached to a rule anyone can point at. They don't sit outside the reporting problem we already accept for minor injuries. They sit at the far end of it. If a cut finger slips through the net, a corrosive year was never going to be caught by it.

Three things do most of the damage.

Delay. When the harm shows up a year after the exposure that caused it, people don't connect the two. Everyone watched the restructure happen. Almost nobody links it to the resignations and the stress leave that follow.

The harm often comes from the person you'd report it to. If the hazard is your manager, the reporting line runs straight through the problem. That was my situation exactly. Who was I meant to hand the form to? Most people in that position do what I did, which is stay quiet and manage it privately.

Habituation. Do demanding, under-supported work for long enough without a dramatic blow-up and it stops registering as a hazard at all. It just becomes the job.

So what do you do with the near-empty psychosocial section in your quarterly health and safety report? I think you read it backwards. A low count is not evidence the risk is low. It's evidence your instrument can't see it. If you measure psychosocial risk by counting incidents, your quietest areas aren't your safest. They're your least visible.

The fix isn't complicated. Stop waiting for forms and go looking for the exposure. Incident forms work for physical hazards because those behave like events. Psychosocial risk needs a different instrument. The recognised tools are built for exactly this. They measure exposure by asking the workforce, through surveys and structured conversations, precisely because there's no event stream to count. A proper hazard assessment, not a bigger incident register.

I think about the gap between those two injuries a lot. One was noticed the moment it happened. The other, the one that did the real damage, never appeared in any system, anywhere. Nobody decided to ignore it. The instruments just weren't built to see it.

Yours probably aren't either. Worth knowing that before the empty report reassures you.