Recent vitalisation of HSE provisioning requirements for mental health and the case for on-site massage interventions in disclosure-averse populations

Why personal stress is the workplace’s real stress, why staff tend not to use EAP lines, and why recent HSE enforcement has made EAP-only provision unsafe to rely on

In two minutes

Stress costs UK employers £51bn a year. In finance and insurance the bill runs at £5,179 per employee, 7% of earnings, the highest of any UK sector.¹ In 2024/25 the HSE recorded 22.1 million working days lost to work-related stress, anxiety or depression.²

The intent to route and optimise work processes around stress is natural, well intentioned, and useless since 2008.³ AI uptake hasn’t changed the minute-by-minute reality of humans sat at their desks either; if anything it has added cognitive load for staff.

PAM Group, themselves an EAP provider, found just 22% of UK workers have access to workplace counselling through their employer, and 11% had at least one contact with the service in the past year.⁴ That 11% bundles actual help-seeking with information-only calls, manager queries and family-member contacts.⁴ Most workers don’t use the counselling on offer because disclosing a mental-health problem at work puts a career at risk, and self-esteem with it.⁵ Confidence in the service is justifiably low: in 2024 the BBC reported that the UK’s largest EAP provider had allowed strangers to listen in on confidential helpline calls; the ICO took no public enforcement action.⁶

In December 2025, the HSE found the University of Birmingham’s lower-than-average sick days and EAP utilisation insufficient evidence of managing workplace stress, and issued a notice of contravention,⁷ the highest-profile instance to date following similar action against the East of England Ambulance Service; the HSE’s own Working Minds campaign signals further enforcement to come.⁸⁹

Workplace massage, historically seen as indulgent and unorthodox, offers no-nonsense and affordable assistance for employees, without crystals, chakras, bells or gongs. It effectively bypasses the disclosure obstacle. The evidence is strong: independent meta-analysis puts adult effect sizes for stress, anxiety and depression in the medium-to-large band.¹⁰ Massage doesn’t compete with CBT for the CBT-engaged population. Its case is reaching the population the EAP doesn’t.

A 12-week pilot costs £3,000 to £4,800 and reaches 30 to 50 staff (pricing detail). A single resignation can cost an employer £40,000–£100,000 for specialist and senior roles, and up to £132,000 for a mid-level manager.¹¹

Disengagement and presenteeism are larger costs already being absorbed, and the same intervention addresses them. Deloitte’s sector ladder puts the per-employee cost of poor mental health at £2,357 in the professional services sector, £2,571 in information and communications, and £5,179 at the top in finance and insurance. Presenteeism alone accounts for £24bn of the £51bn national total.¹

Presenteeism and absenteeism are the daily-occurring cost categories, not tail events; the evidence base above targets their modal drivers (stress, anxiety, depression, pain) at medium-to-large effect sizes. Modest reductions deliver recurring savings against a fixed pilot cost, and interrupt the slow drift from worsening attendance to HR escalation before it triggers a capability process. Resignation prevention sits behind those as a lower-frequency outcome, not the recurring case. The compliance argument runs in parallel: even at low take-up, the firm still holds the documentary trail HSE’s deliberative-quality test requires. The regulatory case below overdetermines a decision the operating economics had already settled.

Hesitation to disclose is a plainly foreseeable barrier to access.⁷¹²¹³ Effective leadership will demonstrate due consideration towards interventions that engage with the issue.

Short on time? The one-page brief · Save as PDF · Share via email

By the numbers

Stress, depression and anxiety cause more than half of work-related ill health. HSE’s own framing.⁹

HSE counted 964,000 UK workers with work-related stress, depression or anxiety in 2024/25, and 22.1 million working days lost to it — significantly higher than the previous year.² The bill lands on employers. Deloitte put it at £51bn, weighted toward presenteeism rather than absence, and concentrated in the sectors that pay the most for talent: finance and insurance loses 7% of earnings per employee to it — the highest of any UK sector.¹

Figure I

The £51bn stress is already costing UK employers

Deloitte 2024 decomposition + sector ladder. A monthly half-day on site costs 1–3% of what stress is already taking out per employee.

Under 3% of workers identify a job-caused clinical-adjacent condition in the Labour Force Survey, which depends on workers reporting it as such themselves. One in four report that their job is harming their mental health.¹⁴

UK GPs issue approximately three million fit notes for mental-health conditions annually, the single largest category of medical sign-off.¹⁵ Around 8.7 million people in England received an antidepressant prescription in 2023/24.¹⁶

The figures above are floors, not measures. The high-functioning professional’s default is “I’d rather not talk about it.”⁵¹² Plainly, the people on the waiting lists are yet to be included in the data sets. EAP use inherits the same trait.

Physiological stress is processed on biological timescales, not procedural ones. A mugging victim does not “come out of shock” if the assailant is arrested ten seconds later. Stress integrates: the human body handles a dying parent and a difficult Q3 close on the same physiology.¹⁷ Whether or not the employer was the main contributor of the stress, they are the ones paying the documented cost or taking the production hit through HR time, resignation costs (hiring, training, integrating), staff disengagement, sickness absence and enforcement risk.

While the mental-health tributaries that flow into the downstream cost for businesses are rarely labelled, the workplace has become the de facto flood defence against them. As and when the banks burst, the business cannot attribute the provenance of individual costs; it just inherits the total.

Whether or not it should is immaterial.

Charlotte

Charlotte, 35. One of the firm’s better lawyers, definitely the firm’s best party organiser. A go-getter who got into law via alternate paths. Or was.

Her father has been in palliative care since February, her siblings aren’t helping, her ten-year-old is being bullied at school and doesn’t want to talk about it. Her eldest, at university in Bristol, met a girl in Lisbon and wants to stay there. The fairly amicable divorce process has turned into a hostile family court process. She wonders if, like her mother, she is beginning to suffer from early peri-menopause. She pushes away thoughts of what Christmas will look like this year.

Charlotte is not going to ring a stranger in a contact centre about menopause, or the resentment she feels towards her siblings. It was never even a question. People talk. Everyone says mental health is the big-button topic; everyone is accepting of it in theory. The quiet remarks in the office, the looks that flick away, suggest otherwise. Besides, this isn’t a mental health issue, she just has a lot on her plate. Everyone’s busy, it’s normal.

After ten months she has the kind of fortnight that lasts three years.

HR acts with consummate professionalism and alacrity, as soon as they become aware of the problem, doing everything they can. Despite this, the HR conversation starts at the beginning of the end.

Charlotte is invited for a chat she can’t refuse, to answer questions not ask them, in meeting room 3. “I have a lot on, I’d rather not talk about it” will not close this conversation. Assessed to the facts, the process is normalised coercion towards a personal examination.¹⁸ The helping hands sit in opposing seats, across the table, with or without the tape recorder. It’s getting written down in triplicate. What choice does she have? Defeated, ashamed, identity permanently damaged, she agrees to an action plan that won’t stop her father dying, but will be parcelled to the seniors as “struggling, needs lighter duties”. The company has acted by the book, identified Charlotte’s personal plights, and over the course of a month, somehow everyone else does too. Later that month the company submits itself to an employer awards competition, emails everyone about it, and tags colleagues on Facebook. Charlotte leaves the business 18 months later.

It’s possible that a sympathetic director, told directly, could have taken the kind of swift action needed: a week off, a month’s guest pass at the Marriott. HR cannot pass on the details safely without breaching GDPR; it would need Charlotte’s consent, which she will not give.¹⁹ Serious firms train HR to maintain professional separation, and they are right to.²⁰

To protect privacy, compliance ensures that leadership updates are vague. Not knowing exactly what is wrong tilts the impression more negative, not less. “She might benefit from a lighter caseload for the moment.” “Her son is being bullied and her dad’s in a hospice at the moment” is an infinitely more relatable fact set than whatever leadership fills the gap with from imagination. The specifics were never shared onwards from HR. They will shape every decision about Charlotte’s career for the next eighteen months, which in the end was worse for her than the truth would have been.

Easy to see when assessed from the armchair.

The wellness hustle

It became self-evident and self-supporting in the way all social inertias do, that attacking workplace stress meant attacking work place stress. It seemed obvious. Hundreds, then thousands, of hours went toward workflow improvements, the latest Microsoft Teams package, the smart board, integrated calendars. New magnetic door locks. Look how easy it makes life for you! AI is the latest of these, changing nothing of the minute-by-minute reality of humans sat at their desks; if anything it has added cognitive load for staff.

Both employer and employee feel they are working harder for less. They are. The ONS calls the post-2008 stagnation the “productivity puzzle”; the pandemic forced every office worker onto the SaaS tools the productivity industry had spent a decade selling, and the line never bent. Real wages did not return to their 2008 peak until 2024; median household income in 2024-25 was no higher than in 2019-20.³

Figure II

UK output per hour, 2001–2025

Pre-2008 growth of 1.8% a year collapsed to 0.3% across the puzzle decade. The line never re-meets the trend it left behind. Markers note the era’s signature productivity tools.

The intervening years brought Slack, Zoom, Notion, Okta, DocuSign, and the rest. Yet we still do it. Well intentioned, irrelevant.

When the smarter businesses began to notice this, the wellness industry wasted no time selling solutions to it. The most corporate-looking wellness outfits were reliably taken as the most effective, the conflation accepted uncritically and at chronic cost. The vendor ROI claims, when tested, often evaporate.²¹

Figure III

Three large studies, three null findings

Eighty-four thousand workers across two US randomised trials and one UK matched-pair study. The lifestyle-and-incentive wellness model does not survive controlled testing.

The deliverable is a certificate of wellness, a yoga mat in a cupboard, a trial for a mindfulness app, a leaflet titled “Get more exercise, it’s good for you!”, a cold slice of vegan pizza [tree planting scheme printed on the box].

We might consider the eye-roll and sigh as well deserved.

Holding the bag

1 in 4 employees say work is harming their mental health¹⁴; only 45% believe their employer treats it as a high priority.²² Leadership doesn’t hate its staff. It authorised the spend, the consult and the programmes. But welfare programmes by and large demonstrate provision, not real engagement. For two decades, this point could be ignored, if one were inclined. The legal corpus an HR function rested on was Sutherland v Hatton,²³ which stated, among other things:

“An employer who offers a confidential advice service, with referral to appropriate counselling or treatment services, is unlikely to be found in breach of duty.”

The Court of Appeal qualified that reading early on. In Daw v Intel (2007) Pill LJ rejected counselling-as-panacea.²⁴ Soon after, in Dickins v O2 (2008) Smith LJ warned that “it is dangerous to apply guidance given by the court as though it were a statutory provision.”²⁵ Both judgments are now seventeen and eighteen years old.

The EAP trade body’s own current narrative confirms the displacement. Holding It Together 2023, in the Introduction (p.3) by Eugene Farrell — identified in the by-line as EAPA UK’s immediate past chair and as mental health lead at AXA Health, an EAP provider — still cites the 2002 Hatton reading as the foundation for the market’s growth from £32m in 2003 to £118m in 2023: “The case ruled that employers had a ‘duty of care’ in terms of employee mental health, and that employers who offered workplace counselling services with access to treatment were unlikely to be found in breach of this duty.”²⁶ As of writing, no EAPA UK published report mentions either Daw v Intel or Dickins v O2.²⁶ That is the legal-sufficiency claim in the Introduction to Holding It Together 2023 — made by a party not holding out as legal adviser, with direct commercial interest in the conclusion, and without disclosure of either in the document.

Figure IV

Workplace-stress case law and HSE enforcement, 2002–2025

The counselling-with-referral defence was qualified by 2008. The EAP industry kept selling it for fourteen years. HSE’s enforcement turn arrives in 2025.

The industry-cited EAP utilisation figure of 10–11%⁴ conveniently bundles information-only calls, manager inquiries about staff, and family-member calls with actual help-seeking. EAPA UK, the trade body, admits in its own methodology that “there is no standard method” for separating them. Identifying which inbound calls were actual help-seeking would be straightforward on the call-routing system. Nobody in the industry is rushing to invent the methodology. The apparent reasonable-steps defence thousands of UK firms still rely on has two holes. The utilisation figure is unmeasured. The case law was distinguished within five years of Hatton itself.

HSE’s ten-year strategy Protecting people and places, published May 2022, sets five strategic objectives. The first, verbatim: “Reduce work-related ill health, with a specific focus on mental health and stress.” Supporting text: “the most commonly reported causes [of work-related ill health] in Great Britain are now stress, depression, or anxiety.”²⁷ Working Minds is the operational arm. The 2025 enforcement actions are delivery against a published commitment, not ad hoc action on an inferred trend.

When the HSE acted in 2025, first against EEAST in April and then against Birmingham in December, the tension between this defensive posture and the reality on the ground snapped.

April 2025: First HSE enforcement of its kind

East of England Ambulance Service Trust served Notice of Contravention for failure to manage work-related stress.

Following an HSE inspection in September 2024, EEAST was cited for breaches of Regulations 3, 4 and 5 of the Management of Health and Safety at Work Regulations 1999. Trust Board minutes record the HSE explicitly using EEAST as a “trail blazer” for the ambulance sector.⁸

11 December 2025: Highest-profile HSE enforcement to date

University of Birmingham served Notice of Contravention on materially identical grounds.

The University pointed to sickness absence below the UK average, OH referrals, and EAP usage data. The HSE held those factors had been considered “in isolation” of other significant sources including reports of stress-related incidents, and that the institution “cannot demonstrate sufficient understanding of the risk to your staff from work-related stress.”

Read the Notice · PDF, 5pp

Birmingham mounted the expected defence: it presented sickness absence below the UK average, OH referrals on file, and EAP usage data as evidence stress was being managed. HSE did not dispute the figures, and issued the breach finding anyway.⁷

EEAST’s enforcement six months earlier was on broader regulatory-breach grounds (failure to assess and manage stress risks under MHSWR Regs 3, 4 and 5). Birmingham is the case that tested the deliberative-quality framework against a defensive position that was prima facie strong.

The Birmingham finding was not about outcome metrics. The breach was that Birmingham could not, in the HSE’s words, “demonstrate sufficient understanding of the risk to your staff from work-related stress.” That is a deliberative-quality test, not an outcome test. A defensible position is a documented one: foreseeable barriers identified, additional options provisioned, policy and aggregate uptake recorded. That document exists at 5% uptake or 50%.

Reach is a separate question from utilisation. Utilisation tells you how many employees made contact; reach tells you how many got substantive help. EAPA UK’s 2023 flagship reports 434,250 people receiving counselling across 2022, delivering 1.3 million sessions in total. The arithmetic on those two numbers gives three sessions per person per year as the average dose. NHS/IAPT CBT protocols and NICE clinical guidance for depression specify six to twenty sessions; the dominant UK EAP commercial product (Health Assured) caps at six or eight as standard. An arithmetic mean of three sits below that floor. Under any plausible distribution, at least 60% of EAPA’s counselling recipients receive less than a clinically-meaningful course.

The reach figure works the same way. The UK workforce affected by existing mental health conditions runs at approximately 15%, around 4.5 million people. EAPA’s 434,250 represents about 10% of that population at any level of engagement. Removing the 60% or more who receive less than a clinical course bounds clinical-dose reach at no more than one in twenty-five, around 4%.³² Farrell’s literal phrasing is “significant part” of UK mental-health provision, paired with “13,000 workers per week.” That reads as broad coverage. The arithmetic delivers something else:

Figure V

EAP reach against the 4.5M at-risk workforce

From Farrell’s “13,000 per week” framing down to a clinical-dose reach EAPA’s own numbers cap at one in twenty-five. The bound is the argument.

The system reaches the inverse of what its marketing implies. Among the workers it does not reach are, by construction, exactly the workers the disclosure-aversion argument has been about.

NICE Guideline NG212 (Mental wellbeing at work, March 2022) is the UK’s standing reference for workplace mental health provision, issued by NICE (an Executive Non-Departmental Public Body of the Department of Health and Social Care, established under the Health and Social Care Act 2012) and developed in partnership with Public Health England under the GRADE evidence-appraisal methodology. It sets a three-tier architecture for workplace mental-health provision. NICE §1.6.1 is explicit that the tiers are additive, not substitutive: individual-level approaches must not replace organisational ones.

Figure VI

NICE NG212 tier architecture

The ‘EAP-only’ firm leaves NICE’s second tier empty. Birmingham made defending that gap a live question, not a settled one.

NG212 cites HSE’s stress risk-assessment template, Management Standards for work-related stress, and employer-support toolkits across recommendations 1.1.4, 1.3.1 and 1.4, anchoring the operational architecture to the duty HSE enforces under the Management of Health and Safety at Work Regulations 1999. A firm offering the EAP alone has discharged tier 1’s universal-availability requirement and left tier 2 empty. The standard Birmingham applied was deliberative: whether the employer had understood the risk to its staff, by consulting them, identifying the hot-spots, and monitoring, rather than pointing to sickness-absence and EAP-access data read in isolation. The extension is the author’s, not the Notice’s: an employer required to demonstrably understand that risk should be able to show, too, that it has considered who its existing provision reaches and what it offers the population the EAP does not. That showing is operationally modest: a booking log, an aggregate uptake report, a written provision policy. That documentary trail sits in the firm’s reasonable-steps file before any inspection arises, not in response to one.

Unsafe provision

The principal limitation of counselling-based provision is that it requires disclosure, and that the disclosure becomes a record that outlasts the call.

Charlotte’s thinking is not fictional. Peer-reviewed research on EAP under-utilisation has consistently found that the more career a worker has to protect, the less likely they are to engage. Surgeons and senior leaders use EAPs at materially lower rates than the average workforce.⁵ The reason is not simply that they have more to lose, though they do. They are also more sophisticated about how disclosure records work and where they end up. The avoidance is informed calculation. The implication for any firm promoting its EAP harder is uncomfortable: a more visible programme may not increase uptake among the populations who already understand the architecture; it may suppress it further.⁵

The records generated are special-category data under UK GDPR, reachable in litigation through summons, disclosure orders, or where a party has put their own mental health in issue.²⁸ They become exhibit material in family court, in custody disputes, in employment tribunals. A person in the middle of family proceedings cannot have “sought help for stress or depression” on paper without it being used against them. Refusing the EAP hotline is the rational calculus of any professional with a private family situation.

Several million UK workers hold roles where mental-health status may be a notifiable, certificate-affecting matter rather than a private one. DVLA Group 2 drivers, FCA-certified financial services staff, SRA-regulated solicitors, GMC-registered doctors, ENG1-medical seafarers, security-cleared engineers and civil servants, CAA-medical pilots and cabin crew, ORR-medical rail staff, HCPC-regulated health professionals, NMC-registered nurses, MoD reservists.²⁹ Thresholds across these regimes vary from notification at one end to suspension or revocation at the other. For this population the confidentiality question is not “how does this look to my line manager” but “what does this do to my licence.” A simple example is the army reservist, where a mental-health record could end the reserve role they hold, so they don’t touch the EAP.

Figure VII

UK workers in licence-affecting roles, by sector

Approximately 3.45 million UK workers in roles where a mental-health record is a regulatory event, not a private one. De-duplicated across overlapping regimes; soft-licence schemes such as SIA security and teaching are excluded. For this population, EAP refusal is informed calculation.

Figure VIII

One in nine of the workforce

About one in nine UK employees works in a role where a mental-health record is a regulatory event, not a private one. The EAP is not built for them.

Further to this, records leak or are abused. In 2024 BBC File on 4 reported that the UK’s largest EAP provider had allowed parties outside the therapeutic relationship to listen in on confidential helpline calls without callers’ knowledge or consent. The industry’s own trade body subsequently concluded the provider had not met the ethical standards expected of members on confidentiality and informed consent.⁶ The ICO has taken no public enforcement action, which surprised 7 people across the country.

In the US, the Federal Trade Commission did the work the ICO did not. Its 2023 final order against BetterHelp, the largest US telehealth mental-health provider, required a $7.8 million settlement and refunds to approximately 800,000 consumers, and prohibited the firm from sharing consumer health data for advertising. The FTC alleged BetterHelp had shared sensitive health data with third parties including Facebook and Snapchat for ad targeting after promising consumers it would not. BetterHelp’s public position was that the settlement was not an admission of wrongdoing and that the conduct was standard for the industry.³⁰ Its own defence is that the regulator is the outlier; the ICO’s silence on the UK equivalent is a regulator’s choice, not a sign the conduct sits in a grey area.

The picture confirms independently in the digital-wellness channel. Mozilla’s Privacy Not Included audit of 32 mental-health apps in 2022 (Talkspace, BetterHelp, Calm, Headspace and Woebot among them) flagged 28 with privacy warning labels and recorded 25 as failing Mozilla’s Minimum Security Standards, with Mozilla concluding mental-health apps were “worse than any other product category” on privacy and security. The 2023 follow-up found only two of 27 apps audited had improved enough to clear the same standards. Cerebral set the audit’s record for trackers loaded in the first minute of download (799); Talkspace was found to be using psychotherapy notes for marketing.³¹ An informed worker reading this is unlikely to download the app.

Figure IX

Mozilla’s audit of mental-health apps

Privacy Not Included 2022 and 2023 follow-up. The product category Mozilla calls “worse than any other” it audits.

Other employee benefits don’t require this. You can use the workplace eye-test without showing the optician your last prescription, or the on-site gym without telling your colleagues about your back. Trying to use the EAP without leaving a trace is like trying to take a shower without getting wet.

No major UK wellness vendor treats the EAP confidentiality problem as a commercial proposition. The pitch on offer remains that more wellness equals more wellness.

An easy way to understand this: provision is not the same as uptake, and a shower makes the gap obvious.

Figure X

Universal access on paper, four users in practice

Two hundred staff wanted showers. Four used the block the firm called universal.

The shower trade body counts the sixteen who came near the block (high). The company counts the four who actually showered (low). HSE accepts neither as adequate provision: there was no deliberative assessment of an obvious barrier, that most people will not undress in front of strangers of the opposite sex. A plainly foreseeable barrier to access.

It is fair to ask why other individual-level interventions named in NICE NG212 §1.6 (yoga, mindfulness, meditation) do not equally well discharge the same requirement. The spatial constraint disposes of yoga first: a mat per participant plus walk-around space is studio geometry, not meeting-room geometry. A massage table with walk-around space is what a six-person meeting room will accommodate; an instructor plus even one or two students on mats is not.

Mindfulness and meditation do fit the room. Their distinction from massage is in what the participant has to do. Both are active disciplines: the participant signs up, performs the practice, and in performing it is identifying themselves to colleagues as someone reaching for a wellness intervention. Signing into a mindfulness app, or showing up to a class, is itself a form of disclosure. Yoga sits between: also active, but mainstreamed enough as a fitness pursuit that attendance carries no wellness-seeking signal to colleagues. Massage carries neither signal: the recipient lies on a table for an hour while someone works on them, with a booking that looks like most other diary slots. For the sophisticated worker whose career calculation already drives them away from the EAP, that distinction is the operative one.

The argument is not that mindfulness or meditation have no place; it is that the population this provision is designed to reach is the one most allergic to performing wellness, and a passive intervention is the form that reaches them.

Universal non-disclosure tier-2 provision is a category, not a product. Saunas, hot tubs, isolation tanks, in-house Pilates studios, tennis courts, kickboxing instruction, tai chi classes all qualify in principle. None fit a lunch hour in meeting room 3. And some staff won’t undress at work or be touched by a stranger at all. That’s a separate barrier needing its own provision alongside whatever covers the disclosure side.

Workplace massage is not the only intervention in the disclosure-bypass category. Anonymous digital platforms (Unmind, Headspace for Work and similar) avoid the line-manager problem but inherit the privacy issues flagged above and import the same self-identify-and-act problem the EAP has: the worker still has to recognise their own need and decide to engage. Peer-support programmes (Wellbeing Champions, Mental Health Allies, buddy systems) move disclosure one rung down rather than removing it. Manager mental-health-first-aid training improves response to disclosures that already happen, where it works; even then, it doesn’t typically reach the worker who never raises a hand. It produces a per-session record without producing a personal record. Other interventions belong in a serious tier-2 stack alongside it; none of them is interchangeable with it.

Figure XI

Tier-2 interventions on five disclosure-related attributes

On-site massage is the only intervention without an outright failure across the five tests that determine reach into the disclosure-averse population.

Workplace massage, historically seen as indulgent and unorthodox, requires no disclosure and comes without crystals, chakras, bells or gongs: just a therapist and a table in meeting room 3 (or any quiet space with screening), delivering physical and mental benefit at once, without the necessity of disclosing personal fears and failings. Nor does seeking a massage indicate anything compromising about the recipient. Booking it is not a confession of any particular need, satisfies provision needs for those who are claiming it, and feels like a perk of the job. Perks were historically the default mechanism for retention and engagement of skilled staff.

Emma’s commercial sessions normally engage 20–50% of staff.

Effect sizes for those who engage with touch interventions sit in Cohen’s medium-to-large range; CBT runs slightly ahead head-to-head on depression but in the same class.¹⁰ The chart below sets that against the alternatives a firm is typically sold. Wellbeing apps³³ and gratitude exercises³⁴ reach only the small band; resilience workshops post a small effect that fades to almost nothing within months³⁵; and mental health first aid, for all that it teaches a colleague to recognise a crisis, shows no measurable effect on the mental health of either the trainee or the person they help.³⁶ A quiet room and the time to use it³⁷ out-performs everything else on the chart except massage and CBT. The motivational keynote and the goody bag, the priciest items on the menu, have never been put to a controlled test.³⁸

Figure XII

Adult effect sizes, with CBT comparator

Hedges’ g; Cohen bands shaded. Massage and CBT sit in the medium-to-large range. The bought-in wellbeing menu (apps, gratitude, resilience training, mental health first aid) clusters small-to-nil, and a free quiet room outscores all of it. Not plotted: the keynote speaker (£5,000 to £50,000 a booking) and the goody bag, neither of which has ever been measured.

Offering massage demonstrates that foreseeable disclosure barriers have been considered, not that any single intervention reaches everyone. A serious tier-2 stack will include several options.

A half-day on site each month, four one-hour sessions at the £250 visit rate, works out at roughly £60 per employee per year across a 50-staff firm (£250 per visit × 12 months, divided by 50). Against Deloitte’s per-employee stress numbers earlier in this page (£2,357 in professional services, £2,571 in information and communications, £5,179 in finance), the intervention costs 1–3% of what stress is already imposing per head. The question is no longer “should we look outside the standard wellness offerings?” It is “why haven’t you?” When the HSE asks, you cannot say the leaflet about eating a salad took up any real strain. You cannot claim the EAP number on the toilet wall was adequate. This will not suffice.

If leadership cannot demonstrate they have taken reasonable steps to reach the people who do not raise their own concerns, the provision is not credible. Sophisticated employees notice this; it informs their read of how seriously the firm takes the duty. The PM-commissioned Stevenson/Farmer review records that “8 in 10 employers report no cases of employees disclosing a mental health condition” and that “only 11% of employees discussed a recent mental health problem with their line manager.”¹² Provision that depends on the worker raising their hand cannot reach the worker who won’t.

Men “bottling things up” and women saying “I’m fine” are reasonably foreseeable challenges to delivering employee assistance. Everyone knows this. Everyone always did know this. The time where companies could pretend otherwise is in the past.

The leadership read

Effective leadership will immediately understand the following:

The state is acting against the economic costs of mental health.
The state is unlikely to accept the bill or the blame.
People do not talk about compromising situations in their lives.
They do not need to, in order to benefit from provision.
Massage reaches employees who will never use counselling.
Even if take-up were poor, the provision demonstrates engagement with tier-2 and understanding of the risk.
Massage creates fewer GDPR headaches than counselling-based provision and runs at a fraction of the mental-health costs the firm is already absorbing.
Massage tackles stress, anxiety and depression directly, the largest single driver of work-related ill health. It might even prevent a resignation.
HSE is not the only route to exposure. Civil claims under occupational stress (Daw v Intel was a civil case), employment tribunals, ICO action on the data-handling side, and individual regulator scrutiny of conduct in regulated sectors (FCA, SRA, GMC) all add their own vectors.
Expectations on firms to demonstrate engagement with mental health are only going to grow.

References

Deloitte, Mental Health and Employers: The case for investment (2024), Figure 16, page 18 (cost per employee by SIC sector). Deloitte figures are reported as-published in 2024 prices.
Health and Safety Executive, Work-related stress, depression or anxiety statistics in Great Britain, 2025.
Office for National Statistics, “Productivity flash estimate and overview, UK: July to September 2025 and April to June 2025”, ONS statistical bulletin, released 13 November 2025. Verbatim on the lost decade: “the 2009 to 2019 trend was historically weak and is commonly described as the ‘productivity puzzle’”. The 0.3% per year figure used in the supporting clause is the ONS labour productivity series (output per hour, whole economy) the Royal Statistical Society named UK Statistic of the Decade in December 2019. By Q3 2025 output per hour stood 3.1% above 2019 average levels, with the ONS noting growth “has slowed and begun to stabilise around the Quarter 1 (Jan to Mar) 2009 to Quarter 4 (Sept to Dec) 2019 weak trend growth”. Corroborated by Martin, “The UK Productivity Slowdown”, Productivity Institute IPM 48 (July 2025): 2001–2007 average 1.8%/yr, 2010–2019 average 0.4%/yr. On real wages and household income: Corlett, The Living Standards Outlook 2025, Resolution Foundation (June 2025), records median non-pensioner household income in 2024-25 “no higher than in 2019-20” and the 2019–2024 Parliament “the worst on record for income growth”; CPI-adjusted real wages returned to their 2008 level in 2024.
EAPA UK, the trade body, admits in its ROI Report 2023 (p.4; mirror hosted on this site, SHA-256: e561bf8f9fc888abe7aa4a0a7bf834383032515cb409c5d3c66cf2546d6bf50c; original) that the headline utilisation figure has no methodological floor at all: “Employers also enter an EAP service ‘usage’ figure. There is no standard method for this, but the great majority of calculator users enter a figure that represents the number of individual cases raised with an EAP.” The trade body further admits in the same report that its 10–12% headline comes from a self-selected sample: “This is because the users are a self-selected group: HR professionals and managers engaged with the importance of organisational wellbeing and an active interest in the performance of their EAP.” Independent surveys put industry-wide utilisation considerably lower: the Work Foundation’s 2016 EAPA-commissioned survey found 5%; the 2017 Barnett Waddingham Workplace Wellbeing Index (145 UK employers, 250+ employees) found 3%. EAPA’s own once-every-two-years methodology (Holding It Together 2023; SHA-256: 2fd362d848398e12d0fed1e296c9338610ce2d6d461365a1e40a92d8e7fefc05; original) reveals that of 640,250 people who “used an EAP service” in 2022, 434,250 (68%) were offered counselling; the remaining 32% bundled into the headline figure are information requests, manager-initiated referrals (mandated by EAPA UK Standards 2023 §21 to be counted as service use), and family-member calls. PAM Group, Health at Work Report 2025, finding 57% of UK employees want workplace counselling, 22% report access, and 11% used it in the past year. PAM Group is itself an EAP provider, making the gap harder to dismiss as advocacy. Structured counselling under the dominant UK EAP commercial offerings is in any case capped at six to eight sessions per issue per year (Health Assured, supporting around 13m UK and Ireland employees, offers tiers of 6, 8, 10 or 12 sessions; 6 or 8 is standard); a structural ceiling on usefulness even for the population who do disclose.
Long, T., “Why are employee assistance programmes under-utilised and marginalised and how to address it? A critical review and a labour process analysis,” Human Resource Management Journal 34(4), 1134–1153 (2024). Open-access critical review of 24 peer-reviewed articles plus industry sources, applying labour process theory to explain why EAPs are systematically under-used despite near-universal employer provision (88% of UK employers offer them). The career-aspirations finding (“employees with substantial career aspirations are less likely to utilise EAP services,” p. 1141) draws on Hu, Y. Y., et al., “Physicians’ needs in coping with emotional stressors: The case for peer support,” Archives of Surgery 147(3), 212–217 (2012), a study of surgeons. The upper-echelon finding (“individuals occupying the roles in the upper echelons are less willing to access EAPs” despite knowing more about them, p. 1139) draws on Quinane, V. R., Bardoel, E. A., & Pervan, S., “CEOs, leaders and managing mental health: A tension-centered approach,” International Journal of Human Resource Management 32(15), 3157–3189 (2021). Long frames the avoidance not as employee failure but as “the benign coping strategy” (p. 1138) against an employer-sponsored programme workers perceive as an extension of managerial surveillance. The body’s implication that more visible EAP promotion may further suppress uptake among these populations follows from Long’s labour-process framing rather than from a measured finding in any of the cited papers; it is presented as inference, not as a published outcome.
BBC News, “Strangers ‘eavesdropped’ on confidential helpline,” File on 4, 12 July 2024, following an earlier March 2024 investigation by the same programme into the provider’s referral and call-handling practices. The UK Employee Assistance Professionals Association launched an investigation immediately after the broadcast: EAPA UK statement, 12 July 2024. EAPA UK’s investigation concluded in 2026 with a finding that the provider had not met the ethical standards expected of members on confidentiality and informed consent, reported by Health & Protection, “EAPA finds Health Assured did not meet expected ethical standards” (March 2026). The British Association for Counselling and Psychotherapy separately suspended the provider’s counselling service accreditation pending its own investigation. As of writing, the ICO’s public enforcement register records no action against the provider in connection with these allegations. The ICO’s 2024/25 Annual Report indicates 70% of data-protection complaints received no response within the 90-day target, and enforcement throughput has fallen materially against prior years.
Health and Safety Executive, Notice of Contravention to the University of Birmingham, 11 December 2025 (reference 4826898). Full Notice (PDF, 5 pages, 310 KB) hosted on this site for archival durability. SHA-256: b74a5d64730ae686dd43778367950c3391f825cbfd070422ed964df1b251c2f2. Originally obtained from Birmingham UCU, who filed the original April 2024 complaint that triggered the HSE inspection (HSE does not publish Notices of Contravention directly). HSE wording cited or paraphrased in the body, traceable to the primary text: “clear and significant inconsistencies in your management arrangements and control measures for work-related stress”; the favourable indicators Birmingham presented were considered “in isolation” of other significant sources including reports of stress-related incidents; “the lack of consultation and information gathering means you cannot demonstrate sufficient understanding of the risk to your staff from work-related stress”; “where it has been identified that workload is a factor, there is no assessment of what else can be done to reduce workloads”; the University “[has] not involved your employees in the development of the organisational risk assessment or in determining how to address the failures in current performance.” An HSE spokesperson described the action as “proportionate enforcement action.” The Notice details six required actions Birmingham must complete by 30 September 2026, with an action plan due to HSE by 28 January 2026. Birmingham’s defence (sickness absence below the UK average, EAP usage data, occupational health referrals on file) and HSE’s rejection of that defence are reported by Times Higher Education, “University of Birmingham reprimanded over management of work-related stress,” 18 December 2025. Background on the union complaint that triggered the inspection: Birmingham UCU, 13 December 2025. For the wider regulatory trajectory, see Weightmans, “From guidance to enforcement – HSE steps up enforcement action on workplace stress” (March 2026), and Clyde & Co, “HSE increase their focus on work-related stress” (January 2026), both legal commentary.
East of England Ambulance Service Trust, Health and Safety Report, July 2025; CQC and HSE Quality Improvement Plan Progress Report, November 2025. HSE Notice of Contravention issued April 2025 following inspection in September 2024; breaches recorded under Regulations 3, 4 and 5 of the Management of Health and Safety at Work Regulations 1999. Trust Board minutes record HSE describing EEAST as a “trail blazer” for the ambulance sector. The Notice itself is not publicly available (HSE does not publish Notices of Contravention, and EEAST has not released its copy). An FOI request to HSE is pending; the document will be hosted on this site once received, to match the Birmingham Notice in [fn7].
Health and Safety Executive, Working Minds campaign. Campaign-page headline: “Stress, depression and anxiety cause more than half of work-related ill health.”
Most rigorous single citation: Packheiser et al., “A systematic review and multivariate meta-analysis of the physical and mental health benefits of touch interventions,” Nature Human Behaviour 8, 1088–1107 (2024). 137 randomised trials, n=12,966 across adults and newborns. Adult effect sizes: pain g=0.69, depression g=0.59, state anxiety g=0.64, trait anxiety g=0.59. Clinical cohorts benefit more than healthy populations (g=0.63 vs g=0.37). Packheiser’s bundle spans therapeutic massage, ICU and neonatal contact, parental skin-to-skin and other clinical and quasi-clinical touch interventions; the authors describe the dominant adult intervention as massage sessions, with kangaroo care being newborn-specific and not contributing to the adult effect sizes. Workplace massage delivered by a qualified MT to self-selected staff sits in the adult-massage population. The clinical-cohort vs healthy-population split (g=0.63 vs g=0.37) is a real uncertainty: a self-selecting workplace recipient is not a clinical-trial inclusion case but is also not the general healthy population, having self-identified as wanting the intervention. Effect size at this population is directionally positive within the band the meta-analysis reports; quantitative point-estimate transfer is not claimed. The authors’ meta-regression found session count positively correlated with effect size while session duration did not, so the comparison to corporate-format massage does not depend on session-length equivalence between trial and workplace. The same finding implies the 12-week pilot’s low per-recipient session count (one to two sessions) sits at the lower end of the dose-response curve tested in the included studies; effect-size transfer at this dose is therefore directionally preserved but quantitatively uncertain. For comparison: Cuijpers et al., “Cognitive behavior therapy vs. control conditions, other psychotherapies, pharmacotherapies and combined treatment for depression: a comprehensive meta-analysis including 409 trials with 52,702 patients,” World Psychiatry 22(1), 105–115 (2023). CBT g=0.79 vs control conditions (95% CI 0.70–0.89). The comparison in the body to CBT is class-vs-class, both arms reported at their respective trial doses. The Cuijpers comparator (g=0.79) is averaged over trials at clinically-meaningful CBT courses, typically 6–20 sessions in line with standard NHS/IAPT protocols. Any reader inclined to dock the massage arm against a one-to-two-session pilot dose must apply the same reduction to the CBT arm to preserve the comparison; at matched low dose, both arms sit at the lower end of their own dose-response curves and the symmetry of the class comparison is preserved. The asymmetric reading (supposing a standard CBT course on one side of the comparison while reducing the massage side to pilot dose) is not available.
PayFit, Cost of Employee Turnover (UK payroll software; last updated 17 July 2025). Verbatim: “the average cost of replacing an employee in the UK is around £25,000 per worker, which accounts for rehiring and lost productivity”; “for specialist and senior roles, this number can range higher, think £40,000 to £100,000 per head”. NatWest Mentor, How much does it cost to employ someone? (19 August 2025): first-year employment cost “often exceeds the base salary by 50–110%” (worked example £62,890 for a £27,600 role); “the total cost of replacing a mid-level manager may reach £132,000”. A separate NatWest piece, Staff turnover rates by industry — UK 2026 benchmarks (15 April 2026), adds: “total replacement costs can reach 150% to 200% of annual salary” for senior positions. Historically anchored by Oxford Economics, The Cost of Brain Drain: Understanding the Financial Impact of Staff Turnover (2014, commissioned by Unum), UK average £30,614 per employee replaced, CPI-uplifted to approximately £43,000 in current prices (ONS CPI rose roughly 40% from 2014 to 2026); consistent with ACAS rule-of-thumb guidance putting replacement cost at six months’ loaded salary minimum. No UK-specific empirical study of equivalent scope to Oxford’s 2014 work has been published since; PayFit and NatWest Mentor (both 2025) are the most current published references for this article’s target audience of professional-services firms.
Stevenson, D. & Farmer, P., Thriving at Work: The Independent Review of Mental Health and Employers (October 2017), PM-commissioned, jointly sponsored by DWP and DHSC. Records (pp.29–30, citing Business in the Community’s Mental Health at Work Report, October 2017) that “only 11% of employees discussed a recent mental health problem with their line manager, and half of employees say they would not discuss mental health with their line manager”, and that “8 in 10 employers report no cases of employees disclosing a mental health condition.” The Government’s response, Improving Lives: The Future of Work, Health and Disability (DWP/DoH, 30 November 2017), accepted the review’s recommendations.
National Institute for Health and Care Excellence, Mental wellbeing at work, NICE Guideline NG212 (2 March 2022). Recommendation 1.1.1 prescribes a tiered architecture: “organisational-level approaches as the foundation for good mental wellbeing (the first [bottom] tier), followed by individual approaches (the second [middle] tier) and targeted approaches (the third [top] tier).” NICE places the EAP at tier 1: Recommendation 1.4.6, within §1.4 “Organisation-wide approaches,” reads “Consider giving all employees free access to an employee assistance programme and occupational health services.” Tier 2 is §1.6 “Individual-level approaches”: Recommendation 1.6.4 names mindfulness, yoga and meditation as examples; on-site massage is in the same individual-level category. Recommendation 1.6.1 makes the architecture additive: “Do not use individual-level approaches to replace organisational strategies for reducing work stressors.” Recommendation 1.9.2 identifies stigma as a barrier “affect[ing] [workers’] ability to discuss any difficulties or engage with certain forms of support.”
CIPD, Good Work Index 2025 (June 2025), survey of 5,000 working people. Headline finding: 25% of UK workers (an estimated 8.5 million people) report their job has harmed their mental health.
NHS England Digital, Fit Notes Issued by GP Practices, England, quarterly publication, most recent December 2025. Mental and behavioural disorders are consistently the single largest diagnostic category, accounting for around a third of fit notes with a known diagnosis. Total fit-note volume across all diagnoses ran at approximately 11 million in 2024/25 (Q1–Q4 quarterly releases summed); the mental and behavioural subset is therefore approximately three million annually. Fit-note counts are not unique-individual counts: a person with an ongoing condition will receive multiple sequential notes within a year.
NHS Business Services Authority, Medicines used in mental health, England, 2015/16 to 2024/25. Over 8.7 million identified patients received antidepressants on prescription during the 2023/24 financial year, up from approximately 7.6 million in 2018/19.
McEwen, B.S., “Stress, adaptation, and disease: Allostasis and allostatic load,” Annals of the New York Academy of Sciences 840(1), 33–44 (1998); McEwen, B.S. & Wingfield, J.C., “The concept of allostasis in biology and biomedicine,” Hormones and Behavior 43(1), 2–15 (2003). The allostatic-load framework establishes that the physiological cost of adaptation accumulates across all sources of demand (psychosocial, physical, occupational) into a single integrated load that determines health outcomes. Multi-domain stress integration in human populations is confirmed in subsequent work: Juster, R.-P., McEwen, B.S. & Lupien, S.J., “Allostatic load biomarkers of chronic stress and impact on health and cognition,” Neuroscience & Biobehavioral Reviews 35(1), 2–16 (2010).
Information Commissioner’s Office, on consent in an employment context. On the power imbalance: “Consent must be freely given, which is unlikely to be possible in an employer/employee relationship given the imbalance of power.” On detriment: “Consent should not be regarded as freely given if the data subject has no genuine or free choice or is unable to refuse or withdraw consent without detriment.” An employee who discloses a mental-health crisis to access a wellbeing service does so knowing the disclosure will be filed, recalled, and weighed against them in any future decision the employer makes. The detail of what is wrong is not, in any case, required to offer a remedy for it. The employer does not demand an optical prescription before granting access to the workplace eye test. Sources: ICO, When is consent appropriate? and What is valid consent?
UK GDPR Article 9 provides bases for processing special-category data beyond explicit consent: employment-law obligations (Art. 9(2)(b)), preventive or occupational medicine (Art. 9(2)(h)), vital interests (Art. 9(2)(c)). Each requires a statutory hook, a clinical-context boundary under appropriate confidentiality, or an emergency where the data subject cannot consent. A sympathetic-director chat about a colleague’s family situation in ordinary working conditions engages none of these. Consent under Art. 9(2)(a) remains the only applicable basis, and on the ICO’s settled position (see footnote 45) consent in an employment context is not freely given. The conversation is unlawful in substance, not on a technicality.
Nothing in this section criticises HR. HR’s institutional role is risk containment and documentation, and the section depicts that role executed correctly and promptly. The Charlotte outcome is what happens when a function correctly designed for institutional risk containment is asked to resolve a private crisis it was not designed for. The case for disclosure-free tier-2 provision is the response to that structural constraint, not to HR underperformance.
Song & Baicker, “Effect of a Workplace Wellness Program on Employee Health and Economic Outcomes: A Randomized Clinical Trial,” JAMA 321(15), 1491–1501 (2019). An 18-month RCT in which 20 of 160 BJ’s Wholesale Club worksites were randomised to treatment, covering 32,974 employees. The programme produced positive effects on self-reported health behaviours, but no significant effects on clinical health measures, healthcare spending, absenteeism, tenure, or job performance. Companion finding: Jones, Molitor & Reif, “What Do Workplace Wellness Programs Do? Evidence from the Illinois Workplace Wellness Study,” Quarterly Journal of Economics 134(4), 1747–1791 (2019). Individual-level RCT of nearly 5,000 employees at the University of Illinois. Financial incentives increased participation; no significant effects on healthcare costs, health behaviours, or productivity. Both trials tested the dominant “lifestyle education plus financial incentives” model: coaching, screenings, app modules, nudges, prizes. Neither included any hands-on remedial care delivered by a therapist. The null findings indict the specific model that constitutes the majority of corporate wellness spending, not interventional alternatives outside it. EAPA UK’s own ROI methodology (self-reported absence reductions and productivity gains entered into the EAPA ROI calculator by HR professionals from a self-selected sample, with CIPD defaults filling missing data, no control group, no randomisation) is structurally identical to the construction Song & Baicker and Jones/Molitor/Reif tested against RCT controls and found to systematically overstate effects. EAPA’s £10.85-per-£1 figure (ROI Report 2023, p.6) is a textbook example of the methodology that does not survive controlled testing. Fleming, W.J., “Employee well-being outcomes from individual-level mental health interventions: cross-sectional evidence from the United Kingdom,” Industrial Relations Journal (2024), DOI 10.1111/irj.12418, extends the same finding to the UK individual-level mental-wellbeing intervention market: cross-sectional analysis of 46,336 UK workers across 233 organisations, matched-pair design within each organisation so workplace-fixed effects are controlled out. Participants in resilience training, mindfulness, well-being apps, time-management training, relaxation classes and stress management showed no improvement on multiple subjective well-being measures over non-participating co-workers in the same workplace. Of 90 wellness offerings tested, only one produced any well-being boost: getting employees involved in charity work or volunteering. The two US RCTs and the UK matched-pair study together cover both intervention models (lifestyle/incentive and individual-level mental-wellbeing) in both jurisdictions; the case against the wellness vendor catalogue rests on RCT-grade evidence in the US and matched-control evidence at scale in the UK.
Benenden Health, Mental Health in the Workplace Report 2024, research with 2,000 UK employees and 500 UK employers. 75% of UK employers offer workplace mental health support; only 45% of employees believe their mental health is a high priority for their employer; 87% of employees affected by stress, anxiety, burnout or depression.
Sutherland v Hatton [2002] EWCA Civ 76, per Hale LJ, practical proposition 11.
Daw v Intel Corporation (UK) Ltd [2007] EWCA Civ 70. Pill LJ: “The reference to counselling services in Hatton does not make such services a panacea by which employers can discharge their duty of care in all cases.” Mrs Daw had access to the employer’s counselling service. Intel still lost.
Dickins v O2 Plc [2008] EWCA Civ 1144. Smith LJ, on relying on Hale LJ’s proposition 11 as if it were a controlling rule: “serves to demonstrate how dangerous it is to apply guidance given by the court as though it were a statutory provision.”
EAPA UK, Holding It Together (2023; mirror hosted on this site, SHA-256: 2fd362d848398e12d0fed1e296c9338610ce2d6d461365a1e40a92d8e7fefc05; original). EAPA UK’s biennial summary of the UK EAP industry. Body quotes on the 2002 Hatton-foundation framing and the market-size figures (£32m in 2003, £118m in 2023) are from Holding It Together 2023, Introduction, p.3 (mirror SHA-256: 2fd362d848398e12d0fed1e296c9338610ce2d6d461365a1e40a92d8e7fefc05), by Eugene Farrell — the by-line identifies him as “Immediate past chair, EAPA UK, and mental health lead at AXA Health.” AXA Health markets an Employee Assistance Programme as part of its corporate offering (see AXA Health, “Employee Assistance Programme”); the conclusion Farrell is asserting in that Introduction supports the commercial product his employer sells. No “not legal advice” disclaimer appears in the document. The Introduction also inverts the case name (it is Sutherland v Hatton [2002] EWCA Civ 76, not Hatton v Sutherland) — independent corroboration on the document’s own face that the legal-sufficiency claim was not legally reviewed. EAPA describes the change as “368% market growth”; on its own £32m → £118m figures the conventional reading is a 269% increase, with the 2023 market approximately 3.7× its 2003 size. The body’s absence claim covers EAPA UK’s complete published-documents catalogue. On 2026-05-27, EAPA UK’s resources page (HTML snapshot mirrored at /docs/eapa-resources-snapshot-2026-05-27.html, SHA-256 21288653780c437004c90dc4922a241ca0b3e2c2dc0a45ff47cddc9adcd3f4b0) listed 27 items; 26 are EAPA UK PDFs (the 27th, an EAPA Black Lives Matter statement, is hosted at the US EAPA site eapassn.org and is out of scope). All 26 PDFs are mirrored on this site alongside a corpus manifest (SHA-256 c4282e462b1f1bec582484f6a63b35569e5d9a291c225d4e6900226c5353dce7) that lists every mirror filename, SHA-256, and EAPA original URL. The four most recent items by publication date are: (1) Focusing On The Job (October 2025); (2) UK EAPA MHFA Statement (July 2024); (3) UK EAPA Securing the Future of the EAP Report (July 2024); (4) the present report, Holding It Together (March 2023; mirror and SHA-256 above). Verification: pdftotext extraction across all 26 PDFs (~467,000 characters of text in total) was run on 2026-05-27 with case-insensitive regex Daw|Dickins|Intel|O2 Plc; zero true matches across the entire catalogue. One substring artefact: the regex matched “Intel” inside “Intelligence” in the 2024 Securing the Future report (the phrase “Artificial Intelligence” in a paragraph on digital service evolution); not a reference to Daw v Intel.
Health and Safety Executive, Protecting people and places: HSE strategy 2022 to 2032 (PDF, 14 pages, 755 KB; May 2022). Hosted on this site for archival durability; SHA-256: 1bcbcc01a33a2e07b9f62098065e35a264d4eaa1b1f27579562deef856615c4e. Original at hse.gov.uk/aboutus/the-hse-strategy.htm. Five strategic objectives are listed; the first (page 5), verbatim, is “Reduce work-related ill health, with a specific focus on mental health and stress.” Signed by Sarah Albon (Chief Executive) and Sarah Newton (HSE Board Chair); foreword by Chloe Smith MP, Minister for Disabled People, Health and Work.
EAPA UK, EAPA UK Standards (January 2023; mirror hosted on this site, SHA-256: 945c112afddfe3a2621ff02e6ad7d360e334d3df48ad3e39afa422649eccf9f0; original). §6 (Information Control: Record Keeping): “An EAP must record all services delivered to the purchaser. The purchaser shall be informed that a record will be kept of any contact with the service. Employees shall be informed that a record will be kept of any contact with the service. The content of records will be consistent with the scope of the service and detailed enough to provide management reporting… Individual Service users have a right to review their own records on request in accordance with Subject Access Request legislation (Article 15 of UK GDPR).” The trade body’s own minimum standard mandates record creation, employer-reportable detail, and Article 15 SAR access. Charlotte’s calculus is correct by the industry’s own rules.
Statutory and regulatory fitness regimes across UK occupational sectors that treat mental-health disclosure as material to continued certification: Driver and Vehicle Licensing Agency, Assessing Fitness to Drive: a guide for medical professionals, Group 2 (HGV/PCV) standards with explicit psychiatric notification requirements; Financial Conduct Authority, FIT Sourcebook, the Fit and Proper Test for Approved Persons and SMCR Certified Persons, with PRA equivalents at SS28/15; Solicitors Regulation Authority, Standards and Regulations including the Suitability Test; General Medical Council, Good Medical Practice 2024 and fitness-to-practise procedures; Maritime and Coastguard Agency, Seafarer Medical Certificate (ENG1) standards per MSN 1815; UK Security Vetting, National Security Vetting (AC, CTC, SC, DV); Security Industry Authority, SIA licence conditions: mental health; Civil Aviation Authority, Medical Certification Class 1 and Class 2; Office of Rail and Road / Railway Safety and Standards Board medical fitness standards for safety-critical rail roles; Ministry of Defence, Joint Service Publication 950 (medical policy, applying to regulars and reservists). Thresholds range from notification requirement to refusal, suspension or revocation. Working population summed across the listed regimes is approximately 3.5 million: DVLA Group 2 vehicle licence holders (~420k), FCA SMCR Senior Managers and Certified Persons (~200k), SRA-regulated solicitors (~210k), GMC-registered doctors (~330k), MCA ENG1 seafarers, UKSV-cleared personnel (~500k+), SIA-licensed security workers (~440k), CAA Class 1 pilots and cabin crew (~50k), ORR-medical rail safety-critical roles (~150k+), HCPC-regulated health and care professionals (~290k), MoD JSP 950 regulars and reservists (~180k), NMC-registered nurses (~770k). Excludes the wider SMCR Conduct Rules tier and teaching/care-work regulatory regimes; including those would add materially.
Federal Trade Commission, “FTC Gives Final Approval to Order Banning BetterHelp from Sharing Sensitive Health Data for Advertising” (July 2023). Final order required a $7.8 million settlement, refunds to approximately 800,000 consumers, and prohibited the firm from sharing consumer health data for advertising purposes. The FTC alleged BetterHelp had shared sensitive health data with third parties including Facebook and Snapchat for ad targeting after promising consumers it would not. Consent order at the FTC case file 2023169. BetterHelp’s public position: the settlement was not an admission of wrongdoing and the conduct was standard for the industry. The defendant’s framing of the regulator as the outlier is itself the operative datum on what the industry considers normal practice.
Mozilla Foundation, *Privacy Not Included guide: Mental Health Apps, 2022 audit. 32 mental-health apps reviewed, including Talkspace, BetterHelp, Calm, Headspace and Woebot. 28 flagged with privacy warning labels; 25 failed Mozilla’s Minimum Security Standards. Mozilla’s framing: mental-health apps were “worse than any other product category” on privacy and security. 2023 follow-up: “Shady Mental Health Apps Inch Toward Privacy and Security Improvements But Many Still Siphon Personal Data”. Of the 27 apps in the 2023 round, only two had improved enough to meet Mozilla’s minimum security standards. Specific findings: Cerebral set the audit’s record for trackers loaded in the first minute of download (799); Talkspace was found to be using psychotherapy notes for marketing.
Holding It Together (2023), Eugene Farrell Introduction, p.3, verbatim: “EAPs are supporting 13,000 workers per week in the UK overall and more than 8,300 with counselling… 1.3 million actual sessions.” The same paragraph cites 434,250 people receiving counselling across the year. Sessions divided by people gives an arithmetic mean of approximately 3.0 sessions per recipient per year. The relevant comparators are the NHS/IAPT CBT protocols, which run six to twenty sessions, and NICE CG90 (depression), which specifies the same range. The dominant UK EAP commercial offering, Health Assured, covering approximately 13 million UK and Ireland employees, sells tiers of six, eight, ten or twelve sessions, with six or eight standard. The EAPA-reported mean of three sessions sits at half the lowest commercial tier and well below the clinical floor. A mean of three sessions cannot reflect uniform delivery. Five sessions to everyone would produce a mean of five, not three. For the mean to land at three against a six-session clinical floor, roughly 60% of users would need to receive one session and 40% the full six. Under any plausible distribution, at least 60% of recipients fall below clinically-meaningful dose. The reach calculation uses a UK non-self-employed workforce of 30.3 million (ONS payrolled employees, PAYE RTI, March 2026) at the Stevenson and Farmer prevalence figure of 15% (fn12), giving a mental-health-affected workforce of approximately 4.5 million. Workforce figures here use the most recent ONS release available, chosen for currency and accuracy. EAPA’s reported annual counselling reach of 434,250 is therefore approximately 10% of that population at any engagement level. That same 60/40 split is the distribution most favourable to dose: at most 40% reach the six-session floor, so clinical-dose reach is bounded at no more than approximately 4%, one in twenty-five. EAPA does not publish the session distribution, so this is a ceiling, not a measured point. A residual bundling sits within the 434,250 figure itself: EAPA UK Standards 2023 §21 mandates that manager-initiated counselling referrals be counted as service use, and family-member counselling sessions feed into the same headline. This is a separate bundling from the 32% pre-counselling bundling addressed in fn4 (information requests, manager enquiries about staff, and family-member calls that did not result in counselling). The reach figures here are not double-counted with fn4; the residual bundling within the 434,250 would, if quantified, lower the reach figure further. EAPA may respond that the EAP is triage and brief intervention with onward referral, not a clinical CBT course, and that comparing dose to IAPT is category-confused. The defence concedes the substantive point. Triage and referral are routes to treatment, not treatment; under the deliberative-quality framing the HSE applied at Birmingham, that is not tier-1 sufficiency for the duty of care.
Gál et al., “The efficacy of mindfulness meditation apps in enhancing users’ well-being and mental health related outcomes: a meta-analysis of randomized controlled trials,” Journal of Affective Disorders 279, 131–142 (2021). 34 RCTs, n=7,566. Depression Hedges’ g=0.33 (95% CI 0.24–0.43), anxiety g=0.28, perceived stress g=0.46; the broadest measures were not significant (general well-being g=0.14, distress g=0.10). Small effects, predominantly against waitlist or no-treatment controls.
Cregg, D.R. & Cheavens, J.S., “Gratitude Interventions: Effective Self-help? A Meta-analysis of the Impact on Symptoms of Depression and Anxiety,” Journal of Happiness Studies 22, 413–445 (2021). 27 studies, n=3,675. Depression: a symptom reduction of Hedges’ g=0.29 at post-test (reported as −0.29, SE 0.06) and g=0.23 at follow-up; the figure plots the post-test estimate with a 95% confidence interval of 0.17–0.41 derived from the reported standard error. The authors found effects larger against waitlist than against active controls, i.e. smaller still when measured against another writing task.
Vanhove, A.J. et al., “Can resilience be developed at work? A meta-analytic review of resilience-building programme effectiveness,” Journal of Occupational and Organizational Psychology 89(2), 278–307 (2016). 37 studies. Overall Cohen’s d=0.21, a small effect that decays with time: proximal (immediately post-training) d=0.26, distal (later follow-up) d=0.07. The figure’s bar spans that decay. One-to-one coaching outperformed classroom, train-the-trainer and computer-delivered formats.
Morgan, A.J., Ross, A. & Reavley, N.J., “Systematic review and meta-analysis of Mental Health First Aid training: Effects on knowledge, stigma, and helping behaviour,” PLoS ONE 13(5): e0197102 (2018). 18 trials, n=5,936. MHFA reliably improves what it is designed to improve: first-aid knowledge (Cohen’s d=0.72 immediately, 0.54 at ≤6 months, 0.31 beyond), recognition of disorders, and a small reduction in stigma. Its effect on mental health is a separate question, and the answer is none of note: on trainees’ own mental health d=−0.04 at post-test, 0.16 at ≤6 months and 0.02 beyond, all non-significant; on the mental health of the people they help, d=0.14 and −0.09, also non-significant. The authors rate the evidence on these mental-health outcomes ‘low’ and call the effects “less convincing.” The bar plots that null; the knowledge gain is real but is not a mental-health outcome.
Manzoni, G.M., Pagnini, F., Castelnuovo, G. & Molinari, E., “Relaxation training for anxiety: a ten-years systematic review with meta-analysis,” BMC Psychiatry 8, 41 (2008). 27 studies. Relaxation training (progressive and applied relaxation and related quiet-rest methods) reduced anxiety with a between-group Cohen’s d=0.51 (95% CI 0.46–0.63), a medium effect that exceeds the bought-in apps. ‘A quiet room’ in the figure is shorthand for providing the time and space such methods need; the benefit still requires the person to use it, which the figure’s reach argument treats separately.
Speaker-fee ranges from Kruger Cowne, “Understanding Speaker Fees in the UK”: UK motivational and keynote speakers command roughly £5,000 to £50,000 or more per booking, with wellness keynotes from about £1,000 to £5,000. No randomised trial measures the effect of a one-off motivational address, a fruit basket or a branded goody bag on any mental-health outcome; there is nothing to plot, which is the point.