Workplace Mental Health and Burnout — Costs and Consequences

Poor workplace mental health is often described with very large, very confident dollar figures. Most of them do not survive a look at where they came from. This page does that look. Its through-line is provenance: for each well-known statistic on absence, presenteeism, return on investment, excess health costs, burnout, and turnover, it traces the number to its origin, states the strength of the evidence behind it, and says only what the figure can honestly support.

The result is a set of distinctions worth holding onto. Some claims are measured (Statistics Canada absence days). Some describe a real phenomenon with untrustworthy magnitudes (presenteeism). Some are peer-reviewed models of something other than employer ROI (the WHO “fourfold return”). Some are a single modelled estimate from another country’s health system (excess deaths and health costs). And some are replicated, predictive associations that are not the same as a dollar cost (burnout’s link to health, performance, and turnover). The restraint is the point: a business case built on inflated numbers is fragile, and a small employer does not need fabricated billions to take exhaustion seriously.

What does stress- and mental-health-related absence actually cost a business?

The honest split is measured days versus modelled dollars. Absence days are well measured: in 2024, private-sector full-time employees in Canada lost roughly 9.3 total days per worker and public-sector employees about 15.7. But the headline figure of about $51 billion a year is commissioned simulation modelling of all mental illness across all of Canada — not a measurement of what any employer actually pays.

The measured side. Statistics Canada’s Labour Force Survey tracks work absence well for days lost — not “cost.” In the most recent release of Table 14-10-0196-01 (covering reference years through 2025), private-sector full-time employees lost roughly 9.3 total days per worker in 2024 and public-sector employees about 15.7. The long-run structure is visible in the agency’s analytical work: in 2011 the average full-time employee lost 9.3 days for personal reasons — 7.7 for own illness or disability, 1.6 for personal or family responsibilities. Crucially, Statistics Canada cautions that the survey does not ask whether an illness is work-related and does not directly measure stress, so these days cannot be attributed to “neglected well-being.” They are a ceiling that includes all illness.

The modelled side. The famous ”~$51 billion a year” Canadian figure traces to Smetanin and colleagues (2011), prepared by the private firm RiskAnalytica on its proprietary simulation platform for the Mental Health Commission of Canada. The base report estimated about $42.3 billion in direct plus $6.3 billion in indirect costs; the ~$51 billion version adds broader categories (including dementia care) in follow-on documents. This is commissioned simulation modelling, not a measurement of what employers pay — a population-level estimate for all of Canada and all mental illness, not a per-employer number. The same RiskAnalytica model is the lineage behind the “$50 billion annual cost” framing that appears in consultancy ROI material (see the ROI section below).

For a 20-to-200-person Kitchener-Waterloo employer, the defensible statement is that illness and disability absence is real and measurable — single-digit days per worker per year in the private sector — but any dollar figure derived from the $51-billion national model is a modelled aggregate and should never be presented as “what poor mental health costs our business.” Smaller workplaces historically record fewer days lost than large ones, partly reflecting lower paid-sick-leave coverage rather than healthier workforces.

Source: Statistics Canada, Table 14-10-0196-01, Work absence of full-time employees by public and private sector, statcan.gc.ca; structure from the Work Absence Rates analytical report (71-211-X). The national modelled figure: Smetanin et al. (RiskAnalytica) for the Mental Health Commission of Canada, The Life and Economic Impact of Major Mental Illnesses in Canada 2011-2041.

Confidence: the StatsCan absence days are verified (a national statistical measurement). The $51-billion figure is a modelled aggregate, not a measured employer cost.

What is presenteeism, and are the productivity-loss figures trustworthy?

Presenteeism — reduced productivity while attending work unwell — is a real, peer-reviewed phenomenon, but essentially all of its dollar magnitudes are vendor estimates with weak measurement and should not be asserted as fact. The most-quoted claim, that it costs US companies over $150 billion a year, comes from a 2004 Harvard Business Review essay that itself attributes the number to unnamed “some estimates.”

This is the softest-number topic in the set, and the candour is the point. The concept is well-grounded: Gary Johns’s 2010 peer-reviewed review in the Journal of Organizational Behavior traces presenteeism and is itself notably skeptical — it stresses that the effect is hard to measure, that productivity-loss estimates lean on self-report and arbitrary assumptions, and that it is confounded with absenteeism. So the phenomenon is industry-consensus; the magnitudes are not.

The most-recycled dollar claim originates in Paul Hemp’s 2004 Harvard Business Review article — a practitioner essay, not a research study. Hemp’s own wording is hedged: “By some estimates, the phenomenon costs US companies over $150 billion a year.” The “$150 billion” is explicitly attributed to unnamed “some estimates.” Other circulating numbers — a “$1.5 trillion” figure, “costs 2-3 times direct medical care,” “61% of total cost” — come from wellness vendors, employer surveys, and consultancy syntheses re-citing one another. None rests on a transparent, replicated measurement.

What can be said responsibly is that research consistently finds lost productive time from health conditions to be substantial, that much of it happens on the job rather than through absence, and that working while ill is associated with later health deterioration and burnout. But the specific dollar magnitudes are directional at best.

For an Ontario small or mid-sized business, the usable takeaway is behavioural, not financial: pressuring sick or struggling employees to “show up” tends to trade visible absence for invisible underperformance and longer-term health costs. The honest move is to cite the phenomenon and the behaviour — not a per-employee presenteeism dollar figure — in any business case.

Source: Johns, Presenteeism in the workplace: A review and research agenda, Journal of Organizational Behavior 31(4), pp. 519-542; the recycled dollar claim: Hemp, Presenteeism: At Work — But Out of It, Harvard Business Review 82(10).

Confidence: industry-consensus on the phenomenon; the dollar magnitudes are directional at best and should not be asserted as fact.

Are the business-case and ROI-of-workplace-mental-health figures true?

Not in the form they are usually quoted. The WHO “one dollar invested returns four” line is a peer-reviewed model of scaling up clinical treatment of depression and anxiety across 36 countries, not the ROI of an employer’s wellness program. The Deloitte Canada figures of $1.62 to $2.18 returned per dollar come from a 2019 consultancy press release based on self-reported data from just seven large firms — not peer-reviewed research.

This is the trace-to-origin section. The WHO “$1 to $4” headline comes from Chisholm and colleagues (2016, The Lancet Psychiatry) — a genuine peer-reviewed study, but one that models scaling up clinical treatment of depression and anxiety across 36 countries from 2016 to 2030, not the ROI of an employer’s wellness program. The paper’s actual bottom line: “benefit to cost ratios amount to 2.3-3.0 to 1 when economic benefits only are considered, and 3.3-5.7 to 1 when the value of health returns is also included.” The “$4” is a rounded midpoint, which the WHO later re-stated as “$5.” All of it is peer-reviewed modelling with stated assumptions — including a prevalence-reduction assumption that other researchers contested in published correspondence — and none of it measures what a single employer gets back.

The Deloitte Canada figures come from a 2019 press release: a median annual ROI of “$1.62 for every dollar invested” after one year, rising to “$2.18” for programs running three or more years. Deloitte drew these from self-reported historical data from just seven Canadian companies, complemented by interviews at ten large firms. This is consultancy work, not peer-reviewed research: a tiny, self-selected, large-employer sample with a proprietary method. Its “$50 billion annual cost” framing traces back to the same RiskAnalytica model as the absence-cost section above.

For a Kitchener-Waterloo small or mid-sized business, the honest position is that there is peer-reviewed evidence that treating depression and anxiety yields positive societal economic returns (Chisholm), and vendor evidence suggesting employer programs can pay back over years (Deloitte) — but the specific ROI multiples should be cited as modelled or consultancy estimates from large firms, not guarantees for a 50-person company. Whether the programs themselves work is a separate question, taken up in the practice note on EAPs and mental-health supports.

Source: Chisholm et al., Scaling-up treatment of depression and anxiety: a global return on investment analysis, The Lancet Psychiatry 3(5), pp. 415-424; the popularized framing: WHO news release; the Canadian employer figures: Deloitte Canada press release (vendor).

Confidence: directional. The Chisholm model is peer-reviewed but measures societal returns, not employer ROI; the Deloitte figures are consultancy estimates from a tiny, self-selected sample.

Does a stressful, unsupportive workplace drive real health and cost burdens?

The strongest single academic estimate links more than 120,000 excess US deaths a year and roughly 5 to 8 percent of US health-care costs to how companies manage their workforces — but it is one US-based modelled estimate that does not transfer cleanly to a small Ontario employer.

Goh, Pfeffer and Zenios (2016, Management Science) built a model estimating the aggregate health impact of ten workplace stressors: unemployment, lack of health insurance, shift work, long hours, job insecurity, work-family conflict, low job control, high job demands, low social support, and low organizational justice. Drawing relative risks from the epidemiological literature, US health spending, and exposure data, the paper concludes that “more than 120,000 deaths per year and approximately 5%-8% of annual healthcare costs are associated with and may be attributable to how US companies manage their work forces” — roughly US$125 to US$190 billion a year. The largest single modelled contributors are lack of health insurance (about 49,000 deaths) and unemployment (about 34,000). Pfeffer’s trade book Dying for a Paycheck (2018) popularized this, framing the workplace as effectively a leading cause of death.

The caveats are substantial, and the authors are transparent about them. This is a model, not a measurement: it chains relative risks, exposure estimates, and cost data — each uncertain — and explicitly assumes the epidemiological associations are causal. It is entirely US-based and built on the US health-financing system. Note that its single largest modelled mortality driver, “lack of health insurance,” is largely irrelevant in Ontario’s single-payer context. The mortality and dollar magnitudes therefore do not transfer to a Canadian small or mid-sized business.

What does transfer is the directional logic. The job conditions in the model — low control, high demands, long hours, insecurity, weak support — are the same psychosocial hazards independently linked to ill health, and they are largely within an employer’s control. For a 20-to-200-person Ontario firm, the lesson is qualitative: these stressors are plausibly health-damaging. The imported death and dollar counts are not.

Source: Goh, Pfeffer & Zenios, The Relationship Between Workplace Stressors and Mortality and Health Costs in the United States, Management Science 62(2), pp. 608-628; popularized in Pfeffer, Dying for a Paycheck (HarperBusiness).

Confidence: single-source. One US-based modelled estimate; the magnitudes do not transfer to a Canadian employer, though the directional logic does.

What are the documented consequences of burnout — health, performance, and turnover?

Burnout is prospectively linked to serious physical and psychological health outcomes and to withdrawal behaviours — absenteeism, turnover, lower performance — in peer-reviewed meta-analyses and systematic reviews. These are correlational and predictive findings, not validated dollar costs.

This is the rigorous consequences evidence. Salvagioni and colleagues (2017, PLOS ONE) systematically reviewed prospective studies — 36 high-quality studies from 993 screened — and found that burnout at baseline significantly predicted later physical outcomes, including type 2 diabetes, coronary heart disease, cardiovascular hospitalization, musculoskeletal pain, prolonged fatigue, and mortality below age 45, plus psychological outcomes (insomnia, depressive symptoms) and occupational ones (job dissatisfaction, absenteeism, intention to leave). Because these are prospective designs, they support a predictive and longitudinal link — stronger than cross-sectional correlation, though not a clean randomized causal proof.

On work outcomes specifically, Swider and Zimmerman (2010) meta-analytically estimated that burnout correlates roughly .23 with absenteeism, .33 with turnover, and .36 with job performance — modest-to-moderate associations, with burnout fully mediating the personality-absenteeism link. Lee and Ashforth (1996), a foundational meta-analysis, established that the three burnout dimensions — especially emotional exhaustion — relate to job demands, turnover intentions, and lower commitment.

The critical boundary is that these findings show association and prediction, not a dollar cost. “Burnout is associated with turnover intention (r is about .33)” is defensible; “burnout costs $X” is not — the dollar translation needs separate, weaker modelling. For a Kitchener-Waterloo small or mid-sized business, the actionable read is that emotional exhaustion is an early-warning indicator of both health deterioration and impending departures. The dollar cost of the resulting turnover is calculated separately, in the cost-of-turnover note (the “Rule of Three”).

Source: Salvagioni et al., Physical, psychological and occupational consequences of job burnout: a systematic review of prospective studies, PLOS ONE 12(10):e0185781; work-outcome correlations: Swider & Zimmerman, Born to burnout: A meta-analytic path model, Journal of Vocational Behavior 76(3); foundational: Lee & Ashforth, A meta-analytic examination of the correlates of the three dimensions of job burnout, Journal of Applied Psychology 81(2).

Confidence: verified — peer-reviewed prospective and meta-analytic evidence. The boundary is that these are predictive associations, not validated dollar costs.

Does workplace stress and poor mental health predict turnover and quitting intentions?

Yes. Burnout, and emotional exhaustion in particular, is a consistent meta-analytic predictor of turnover intentions and actual turnover, with correlations in the .3 range. This establishes a replicated link from neglected well-being to staff loss — though the relationship is associational and predictive, and the dollar cost of each departure has to be calculated separately.

Retention is usually the cost a small employer feels most directly, which is why the turnover link is worth isolating. Swider and Zimmerman’s 2010 meta-analysis estimated burnout’s correlation with turnover at about .33 and with absenteeism at about .23, and found that burnout fully mediated the personality-absenteeism relationship and partially mediated the personality-turnover relationship. Lee and Ashforth (1996) earlier established that emotional exhaustion in particular is robustly tied to turnover intentions and reduced organizational commitment.

The boundary matters: these are associational and predictive findings. They support the statement that “employees experiencing burnout are meaningfully more likely to intend to leave, and to actually leave” — a strong, replicated relationship — but they do not by themselves attach a dollar figure to each departure. They establish the mechanism: neglected well-being leads to exhaustion, which leads to withdrawal, which leads to turnover. The per-departure dollar cost — roughly three times salary — is calculated separately in the cost-of-turnover note (the “Rule of Three”).

For a 20-to-200-person Ontario firm, where a single departure can represent a large share of a team’s capacity and institutional knowledge, this is arguably the most operationally relevant stake: exhaustion is an early, measurable leading indicator of resignations.

Source: Swider & Zimmerman, Born to burnout: A meta-analytic path model of personality, job burnout, and work outcomes, Journal of Vocational Behavior 76(3), pp. 487-506; foundational: Lee & Ashforth, A meta-analytic examination of the correlates of the three dimensions of job burnout, Journal of Applied Psychology 81(2).

Confidence: industry-consensus — a strong, replicated meta-analytic relationship. The link is associational and predictive, and the per-departure dollar cost lives in the turnover-cost note.

General information, not legal advice

This page is general information about workplace mental health and its costs, not legal, medical, or actuarial advice. The figures here are reported with their original strength of evidence on purpose: measured statistics, peer-reviewed models, single-source estimates, replicated associations, and vendor claims are not interchangeable, and none of the national or US aggregate dollar figures should be presented as “what poor mental health costs our business.” Before building a business case or a wellness program on any of these numbers, confirm the current figures at the cited sources and treat modelled or consultancy estimates as directional rather than guaranteed.