On 24 May 2025, three crew were cleaning soot from inside a boiler on the bulk carrier Olivia, alongside at South Port in Bluff. The ship had been in port 32 days. Repeated attempts to restart a burner had failed, so the night shift was using a compressed air lance to clear the soot. The air went into a hot furnace that still held unburned fuel. Two of the three were seriously burned.
The Transport Accident Investigation Commission released its final report on 30 October 2026. The crew thought of boiler cleaning as familiar work, and they were right. They normally did it at sea, after the boiler had cooled for several days. This time the ship was alongside, nobody had objectively confirmed the furnace was cool, and the job went ahead anyway.
What the Commission concluded about ‘why’ is the bit I want to talk about.
They wrote “It is very likely that senior members of the engineering team were fatigued as they sought to complete the generator repairs, which adversely affected their decision-making.” The experienced daytime engineers approved the job. The night shift that carried it out was smaller and had nobody authorised to supervise work at that risk level. One of the two safety issues the Commission identified was “a lack of guidance for managing crew during sustained intensive work periods”.
Here’s the pivotal bit again.
It is very likely that senior members of the engineering team were fatigued as they sought to complete the generator repairs, which adversely affected their decision-making.
I’ve done night shift before. The stakes were lower, but what counted as acceptable risk was materially different from day shift. Ask if that sentence would apply to any of your teams.
Fatigue does not degrade everything at the same rate
I think we often talk about fatigue as though it makes us uniformly less capable; a slightly worse version of our rested self at every part of the job. That’s not how it shows up though. The well-practised, procedural parts of work hold up the longest. What goes early is the work of noticing that circumstances have moved, and then judging what to do about it.
This is work that extends outside the lab, with workplace evidence too. Quan and colleagues followed 60 attending surgeons through 362 procedures, comparing operations performed after a night on call against the same surgeons operating when they were not. The sleep difference was modest, roughly five hours against roughly six and three quarters. Technical performance in theatre held up. What fell were the observed ratings of situational awareness, decision-making, and communication and teamwork. The surgeons also took longer, which looks a lot like people slowing down to keep quality where it was. (Nod in there to the compensatory control model.)
You could think about that as a map of where damage might get done. Their hands were fine, and you’d expect that because they’re practised hands.
But judgement, awareness of the room, and communication? They all degraded. Those last three are the content of a handover.
Why the handover is where it surfaces
A handover has work to do. The incoming shift needs the current picture, and it also needs the reasoning that produced it, the options that were considered and dropped, and the things nobody is sure about yet. That underlying layer is what lets them notice when something shifts.
It’s also the layer that degrades. So the outgoing team, lead, or controller gives an account that is accurate as far as it goes but potentially paper thin underneath. They know something left alone, undone even. Why it was left alone, maybe why it wasn’t done, has faded. They know somebody said an access route was marginal, or a situation was tenupus, without quite holding on to who said it or how much weight it deserved.
The person receiving it can’t audit any of that, because they weren’t there. A fluent (even confident) handover sounds complete, and there’s often little in the process that’s designed to test the difference. Receivers also may not close the loop. In a study of urgent interfacility transfer calls into a paediatric intensive care unit, only 9% of receiving physicians summarised back what they had been told, even though code status came up in 19% of calls.
The standard fix fails
The standard organisational answer is to ask people to speak up when they are likely too tired to hand over properly.
Self-assessment degrades alongside everything else, and the people deepest into an event are the worst placed to rate themselves. Any control that depends on somebody noticing their own decline isn’t an effective control. The same objection applies to anything resting on the outgoing shift remembering to mention the important thing, because remembering to mention it is itself the function under load. Prospective memory, commonly known as remembering to remember, where you aim to remember to do something later, is negatively affected by cognitive load, stress, lack of sleep and so on.
Two changes that don’t rely on self-awareness
Build the handover record while the shift runs rather than at the end of it. A note made at the moment a decision is made can carry the reasoning, because the reasoning is sitting right there with you. The same note written at shift end carries only whatever has survived. This is a change to when the writing happens, not to how much of it there is, which is why it tends to survive contact with a busy operation.
Then make the things that can get dropped into required fields rather than feats of recall. Contingency plans and open uncertainties could be items the handover format demands, so that leaving one out is visible to both parties. There’s evidence this works. When a paediatric ward moved to a structured handover format, the proportion of handovers containing a contingency plan went from 16% to 73%, and action lists from 31% to 81%, with no meaningful increase in how long the handover took.
Those are massive increases for little effort, no? This is especially important in higher stakes, sometimes life-and-death situations.
The Commission has already written the principle down
Finding 1.13 of the Olivia report says procedures “should contain clear task-classification, stop-work, escalation and supervisory triggers so that changed conditions are recognised and reassessed before work continues”. Note that. Triggers, rather than judgement calls. On day 32 of a hard job you simply cannot rely on tired people to spot that the conditions have moved. Consequently, the recognition has to sit in the system instead of in somebody’s head.
Most organisations already accept that logic for physical risk. Nobody asks a crew to decide on the day whether a confined space warrants a gas test. Handover still gets treated as a professional courtesy between competent adults, which is roughly how confined spaces were treated sixty years ago.
Photo by Giulia May on Unsplash
References