On 20 November 2025 the Health and Safety Executive published its annual figures for work-related stress, depression or anxiety in Great Britain: 964,000 workers affected in 2024/25 and 22.1 million working days lost. The same report also says what causes it. That part rests on survey years 2009/10 to 2011/12, and the report says so, in the source line under the chart. Both facts are worth holding at once, because the first one gets quoted and the second one gets turned into programmes.
Two dates in one report
The headline numbers come from the Labour Force Survey, which the Office for National Statistics runs across around 31,000 households a quarter and in which HSE commissions annual questions. For 2024/25 the report gives 964,000 workers suffering from work-related stress, depression or anxiety, a prevalence rate of 2,770 per 100,000 workers, which HSE describes as statistically significantly higher than the previous year. New cases were 409,000. Working days lost were 22.1 million, an average of 22.9 days per case. Stress, depression or anxiety accounted for 52% of all work-related ill health and 62% of the working days lost to it.
Then comes the section headed by cause. The main causes, according to self-reports in the LFS, were workload, particularly tight deadlines, too much work or too much pressure and responsibility, followed by a lack of managerial support, violence and bullying, organisational change and role uncertainty. The source line under that statement reads "LFS, average estimate over 2009/10-2011/12". The report's second cause source, the GP reporting network THOR-GP, averages 2013 to 2015; HSE's funding for it ended in 2016, so 2015 is the last year of its data available to HSE.
None of this is hidden and none of it is a scandal. HSE labels every chart with its source years, and the second sentence of the introduction reads "By its very nature, stress is difficult to measure." The point is narrower. The number most people quote and the breakdown most people quote come from different decades, and the breakdown is the part that gets acted on.
Why the cause breakdown cannot be a diagnosis
The survey years behind the breakdown predate the pandemic, hybrid working and the Worker Protection (Amendment of Equality Act 2010) Act 2023. That alone should make anyone pause before writing "workload and tight deadlines" into a board paper as the reason for this year's programme.
The deeper limit is not age but type. The LFS figures are, in HSE's own definition, based on individuals' perceptions: people who have conditions which they think have been caused or made worse by their current or past work. That is the right way to count how many people are suffering. It is a weak way to establish what to change, because the attribution is the respondent's, made after the fact, and HSE itself notes that its two cause sources may reflect different perceptions of work-related attribution. A national average of self-attributed causes, gathered more than a decade ago, tells you nothing about which team in your organisation is under what pressure this quarter.
We see the same gap between perception and behaviour at a much smaller scale in our own research. In the randomised 2020 UCL study by co-founder Ben Laumann (n = 57), participants learned the same feedback model by rehearsal with a role-play coach, by slideshow or by video. Self-rated confidence rose in every condition with no significant difference between them, while delivery of the skill, rated blind by two independent assessors, was significantly better after rehearsal than after either passive format (Cohen's d = 0.75 against slides, 1.08 against video). The study found no relationship between what participants thought of their own skill and what the assessors saw. Self-report measures perception. Perception is worth knowing. It is not the thing you are trying to change.
What we collect instead, and what it cannot show
Our engagements start with a diagnostic phase, not with a national statistic. Using COM-B (Michie, van Stralen and West, 2011) we work out with the organisation whether the behaviour that is missing is missing for lack of capability, opportunity or motivation. If the answer is opportunity, because there is no forum for raising workload and the last person who tried was punished, no rehearsal will supply it, and we would rather say so in scoping than stage a production into it.
Then we define one observable behaviour before anything is designed, count it before the work begins, and count it again three to six weeks after delivery, with a comparison group or a phased rollout where the organisation can support one. Self-report is a supplement, never the spine. For a manager population under workload pressure the indicator might be the quality of a manager's first response when a direct report says they are not coping, scored against a defined rubric, or how often the manager asks an open question before offering a solution. The full method is in how to measure behaviour change.
What this cannot show matters just as much. It does not measure stress. A count of one behaviour in one group says nothing about prevalence, about whether anyone in the organisation is less anxious, or about whether workload was the cause. It cannot be set against HSE's rate per 100,000. It shows whether a defined behaviour became more frequent or better executed in the window we looked at, and it says nothing about month seven. Our own study carries the same kind of limit: one skill, one setting (its rehearsal condition was a 20-minute online coaching session, so the study tests the principle rather than today's programme), no control condition, no pre-measurement of ability, no retention test. Stating the limit is part of citing the evidence honestly, and it applies to ours as much as to HSE's.
A note on source lines, including our own
Reading the source line is a habit, and we have needed it ourselves. In September we removed a per-employee training-spend figure from our own pages, because the report it was attributed to could not be found and the number traced back only to our own site. The figure we use now comes from the Department for Education's Employer Skills Survey 2024: total UK training expenditure of £53.0 billion, which equated to a spend of £1,700 per employee, down from £1,960 in 2022.
Around the same time, we decided to publish at most one post a day, and only where it carries something of our own: a design decision, an observation from delivery, a measurement practice or a limit. This one carries a measurement practice and what it cannot show; the source line under HSE's cause chart is the occasion, not the reason.
What to do with the HSE report
Use the headline numbers for what they are: the best available count of how many people say work is making them ill, with confidence intervals, from an accredited official statistic. Use the cause section as a prompt for questions, not as an answer. If a board paper turns workload and tight deadlines into a programme, ask when that finding was gathered and whose attribution it rests on. Then find out what is actually happening in your own teams, with a method that would survive the same question.
If you want to talk through what a baseline would look like for your managers, book a free 30-minute call. Bring the conversation your managers avoid, and we will say what we could count, and what we could not.
The stress figure is new, but the cause data is not: The Takeaways
HSE's November 2025 statistics put work-related stress, depression or anxiety at 964,000 workers and 22.1 million lost working days in 2024/25. The cause breakdown in the same report is a Labour Force Survey average over 2009/10 to 2011/12, and the GP-reported causes stop at 2015. The post sets out why a decade-old national average of self-attributed causes cannot serve as a diagnosis for one organisation, what Sidestream measures instead (one observable behaviour against a baseline, re-measured after three to six weeks), and what that measurement cannot show, including that it does not measure stress at all.
- HSE, published 20 November 2025: 964,000 workers with work-related stress, depression or anxiety in 2024/25, 22.1 million working days lost, 22.9 days per case.
- The cause breakdown in the same report is an LFS average over 2009/10 to 2011/12; the GP-reported causes (THOR-GP) average 2013 to 2015, and 2015 is the last year of GP data available to HSE.
- Both cause sources are attributions made after the fact by respondents or GPs; HSE notes they may reflect different perceptions of work-related attribution.
- A Sidestream measurement plan would count one observable behaviour before delivery as a baseline and again three to six weeks afterwards. It does not measure stress and cannot be compared with HSE's rates.
- Read the source line. This month we struck an untraceable training-spend figure from our own pages and replaced it with the DfE's £1,700 per employee.
Frequently Asked Questions
How current are HSE's work-related stress statistics?
The headline figures are for 2024/25 and were published on 20 November 2025: 964,000 workers affected and 22.1 million working days lost. The breakdown by cause in the same report is an average of Labour Force Survey years 2009/10 to 2011/12, and the GP-reported causes average 2013 to 2015. HSE states the source years under each chart.
What does HSE say causes work-related stress?
According to self-reports in the Labour Force Survey, the main causes were workload, particularly tight deadlines, too much work or too much pressure and responsibility. Other factors were a lack of managerial support, violence and bullying, organisational change and role uncertainty. This is a national average of respondents' own attributions over 2009/10 to 2011/12, not a diagnosis of any one organisation.
Does Sidestream measure whether training reduces stress?
No. A typical measurement design would define one observable behaviour with the organisation, count it before delivery and again afterwards, and isolate the effect with a comparison group or phased rollout where possible. That would show whether the behaviour changed. It would not show prevalence of stress, whether anyone feels less anxious, or what caused the pressure in the first place.
Sources: HSE, Work-related stress, depression or anxiety statistics in Great Britain, 2025 · Department for Education, Employer Skills Survey 2024 · Sidestream, The Study Behind the Method · Sidestream, How to Measure Behaviour Change · Sidestream, What Is Immersive Theatre Training