You already know this site.
Output is up eleven per cent on last year. Headcount is flat, because the business case for two more operators didn't survive the budget round. Three experienced shift supervisors have retired inside eighteen months and their replacements are capable people who have been in post for four months. The permit backlog is growing. The maintenance planner has started ringing people on their days off. Nobody has been hurt, the SHE stats look respectable, and the last audit closed out cleanly.
Then absence starts to move. Short-term, self-certified, scattered across the shifts rather than concentrated in one team. Two grievances land in the same month. A team leader you would have described as unshakeable goes off sick and doesn't come back for eleven weeks.
And the organisation responds by procuring a resilience workshop.
To be clear, this is not an argument against resilience training. It is an argument about where resilience training belongs. Because in every other part of this profession there is already a precise term for a control that sits at the very end of the chain, protects one person at a time, depends entirely on correct fit and correct use, and does nothing whatsoever to the hazard itself.
We call it PPE.
An old idea, with better evidence than most people realise
The intellectual foundations here are older and considerably more robust than the current wellbeing conversation suggests.
In 1979, Robert Karasek published the demand-control model, arguing that strain arises not from demand alone but from the combination of high demand and low decision latitude. It was a claim about the design of jobs, not the constitution of the people doing them.
The most striking test of it came out of Whitehall. Bosma, Marmot and colleagues followed 10,308 London civil servants for a mean of 5.3 years and reported the results in the BMJ in 1997. Participants with sustained low job control had an odds ratio of 1.93 for newly reported coronary heart disease, and the association held after adjustment for employment grade, personality and conventional cardiac risk factors. The finding that ought to stop a safety professional in their tracks is the one that gets quoted least often: in that cohort, job demands showed no significant association with subsequent coronary events. Control did.
Read that again, because it inverts the folk model. The hazard was not that people were working hard. The hazard was that they had very little say in how.
HSE built on exactly this evidence base. The 1999 discussion document Managing Stress at Work established that stress was a health and safety matter to be handled under existing law. Tackling Work-Related Stress (HSG218) followed in 2001 with seven stressor areas. By the time the Management Standards were consulted on and issued in 2004, culture had been folded into the others as the thing underpinning all of them, leaving the six we use now: Demands, Control, Support, Relationships, Role and Change.
Every one of those six is a property of the work. Not one of them is a property of the worker. That was a deliberate design decision, taken twenty years ago, on the basis of occupational epidemiology. Somewhere between then and now, a great many organisations quietly reversed it.
The current numbers
In 2024/25, HSE estimated 964,000 workers in Great Britain suffering from work-related stress, depression or anxiety - a rate of 2,770 per 100,000 workers, statistically significantly up on the previous year and a record high. Those cases accounted for 22.1 million working days lost, out of 40.1 million lost to work-related ill health and injury combined. Roughly one working day in two lost to work in this country is now lost to this.
In April 2026 the ILO published The psychosocial working environment: global developments and pathways for action. Its estimate is that more than 840,000 deaths a year are attributable to health conditions linked to psychosocial risks - principally cardiovascular disease, mental disorders and suicide - alongside close to 45 million disability-adjusted life years and economic losses in the order of 1.37 per cent of global GDP. The named exposures are long working hours, job insecurity, high demand with low control, harassment and effort-reward imbalance.
EU-OSHA's Healthy Workplaces Campaign for 2026-2028, Together for Mental Health at Work, is now running on the same theme.
If a physical agent were producing numbers of that order, we would not be commissioning webinars about it.
The HartfellCo Psychosocial Hierarchy of Control
This is the framing HartfellCo uses, and the one worth defending in front of a board. It is not a new hierarchy. It is the hierarchy every safety professional already knows, applied without flinching to the design of work.
Tier 1 - Eliminate. Remove the hazard from the work. Kill the duplicated approval step. Stop producing the report nobody reads that generates the Thursday-night deadline. Withdraw from the contract whose terms make safe resourcing impossible. Elimination is rarer here than in physical safety, but it is not as rare as people claim, and it is almost never attempted before the alternatives.
Tier 2 - Substitute or redesign. Change how the work is done rather than how much of it there is. Different shift pattern. Different sequencing so two peaks stop colliding. Decision rights pushed down to the people with the information. This tier is where Karasek's control variable actually lives, and it is the most under-used tier in practice.
Tier 3 - Engineer at source. Build the resource into the system: staffing levels, span of control, handover structure, tooling that removes rework, systems that don't require heroics at month-end. Structural, protects everyone exposed, and does not depend on any individual behaving well on a bad day.
Tier 4 - Administrative and organisational controls. Role clarity. Escalation routes that work. Workload caps that are honoured. A communication rhythm during change. Competent, resourced line management. Consultation that has teeth. These are real controls - and like a permit-to-work system, they depend on people following them, which means they degrade precisely when pressure is highest.
Tier 5 - Individual protection. EAP, counselling, mental health first aiders, resilience and coping skills. Legitimate. Sometimes essential. Always last. Never a substitute for the four tiers above it.
The diagnostic is uncomfortable and takes about ninety seconds. Pull your psychosocial actions and sort them into those five tiers. If everything lands in tier five, you do not have a risk register. You have a welfare offer.
The strongest objection, and why it doesn't hold
The serious counter-argument runs like this: you cannot eliminate workload the way you eliminate a solvent. The hazard is frequently the work itself, which the organisation exists to do. And psychosocial hazards interact - they don't present one at a time in tidy isolation, so single-hazard control logic breaks down.
All of that is true, and the last point is one HSE makes itself: the six areas overlap, combine and interact. High demand with low control is a materially different risk from either factor alone. Poorly communicated restructuring landing on a team with weak support is not the sum of two problems.
But none of it defeats the hierarchy, for the simple reason that a hierarchy of control was never a promise of elimination. It is a priority order for effort and expenditure. We have never eliminated the hazard of working at height; we still don't lead with harnesses. And this is not a fringe reading. NSW's code of practice on managing psychosocial hazards states it plainly - start with higher-order controls such as flexible working or redesigning how work is completed, before relying on lower-order controls such as training or coping strategies. HSE's own primary, secondary and tertiary intervention distinction says the same thing in different vocabulary. ISO 45003 sets the whole question inside an ISO 45001 management system, which is where it belongs.
The hierarchy is not a trick for making psychosocial risk feel like machinery guarding. It is a discipline for stopping an organisation reaching for tier five first because tier five is the cheapest thing to buy and the easiest thing to announce.
Three failure modes worth naming
The survey is not the risk assessment. It is evidence. HSE is explicit that its Management Standards survey is a starting point, not a diagnosis of cause. A defensible assessment triangulates: indicator scores alongside absence, turnover, overtime, exit interviews, grievances, harassment reports, vacancy duration, incident and near-miss trends, and - non-negotiably - conversations with the people doing the work. Numbers tell you where to look. They rarely tell you why.
Stopping at "what" instead of "why". A finding of excessive workload is where the assessment starts, not where it finishes. Why? Insufficient staffing. Why? Recruitment hasn't kept pace with growth. Why? The plan assumed productivity gains that never materialised. Only at that fourth answer do you reach something a board can actually authorise. We do this instinctively for a loss of containment. We stop at symptom level for this.
The action plan that isn't one. "Improve communication" is an aspiration. An action reads: fortnightly operational briefing for the duration of the restructure, owned by the Operations Director, starting 1 October, effectiveness reviewed at three months against team feedback and absence and turnover data. Owner, date, resource, measure. HSE's own guidance asks that each action be given an order of priority, adequately tackle the issue with a review built in to check that it does, have sufficient resources allocated, be assigned to a named individual or function, and carry an agreed and realistic completion date. Five tests. Take your last psychosocial action plan and score it honestly against them.
And it isn't annual. Every restructure, acquisition, system implementation, rota change, leadership change and new target changes the risk profile - which means the assessment has a shelf life measured against organisational change, not against the calendar.
The question for the board
If we discovered a physical exposure affecting forty per cent of this workforce, would the response be to teach those forty per cent to tolerate it better?
Nobody says yes. That is the whole argument, and it takes one sentence.
We would investigate. Identify the source. Quantify exposure. Control it as far up the hierarchy as reasonably practicable. Assign ownership. Fund it. Verify that the control worked. Then review it.
None of that is a wellbeing programme. It is ordinary competent safety management, applied to a category of harm that happens to arrive through the design of work rather than through an energy source. The subject is different. The discipline is identical - and it is a discipline this profession already owns.
So keep the EAP. Keep the mental health first aiders. Keep the resilience training, and fund it properly.
Just stop calling it a control strategy. It's PPE. Put it where PPE goes, and then go and look at the work.
A note on tools
HSE publishes its Management Standards indicator tool and analysis tool free of charge. Any organisation starting this work should download them and use them. They are competently built, they are the reference implementation, and nothing sold commercially replaces them. Anyone who tells you otherwise is selling you something.
Sources
- HSE, work-related stress, depression or anxiety statistics in Great Britain 2024/25; HSE Management Standards (2004) and Tackling Work-Related Stress HSG218 (2001); Bosma et al., BMJ 1997; Karasek, Administrative Science Quarterly 1979; ILO, The psychosocial working environment, April 2026; EU-OSHA Healthy Workplaces Campaign 2026-2028; SafeWork NSW Code of Practice, Managing psychosocial hazards at work; ISO 45003:2021.
First published on the HartfellCo LinkedIn page, 2 September 2026.
Want this built into your training programme? See our Workplace Stress & Mental Health Training Course.