Clinical Whitepaper · Series No. 15

The Real ROI of Executive Mental Health

What the randomized evidence on wellness program ROI actually establishes, and where the return on senior-leader care sits.

24 min read · 5,383 words · 4 figures · 18 references

Executive summary

The best designed studies of workplace wellness programs found close to nothing. Two large randomized trials, covering roughly forty five thousand employees between them, moved some self-reported behavior and left clinical measures, medical spending and employment outcomes statistically unchanged.^{01, 02} That result is real, it is inconvenient, and it is also not an answer to the question a chief financial officer is actually asking, because neither trial tested clinical treatment delivered to the small group of people whose incapacity would show up in results.

For most of two decades the business case for workplace wellbeing rested on one meta-analysis. Reviewing the published literature in 2010, Baicker, Cutler and Song reported that medical costs fell by about $3.27 for every dollar spent on wellness programs, and absenteeism costs by about $2.73.⁰⁵ Those two numbers were quoted into thousands of benefit decks. Then the field ran the experiments. A cluster randomized trial across 160 worksites of a large US warehouse retailer, covering 32,974 employees, found that after 18 months the program had raised two of 29 self-reported health behaviors and had changed nothing else: not one of ten clinical measures, not one of 38 measures of health care spending and utilization, and none of the three employment outcomes, which were absenteeism, job tenure and job performance.⁰¹

The Illinois Workplace Wellness Study, a randomized trial of more than 12,000 university employees, arrived at the same place from a different direction. It found no significant causal effect on medical spending, productivity or self-reported health in the first year, and its confidence intervals were tight enough to rule out 83 percent of previously published estimates on medical spending and absenteeism.⁰² It also found the mechanism behind the old numbers. In the year before the program began, the employees who went on to participate already had lower medical expenditures and healthier behaviors than those who did not.⁰² The observational studies had not measured a treatment effect. They had measured who volunteers.

The trials did not find that spending on mental health fails to pay. They found that spreading a general wellness program evenly across a workforce does not pay. On the randomized evidence in references 01 and 02

Read carefully, the wellness literature is not ambiguous. It is precise about something narrower than the headline suggests. What follows separates four things a budget decision depends on: what the trials actually tested, what they measured, where the return inside a wellness program came from when anyone looked component by component, and how unevenly the underlying strain is distributed across the workforce being averaged.

The four figures point one way. Broad, undifferentiated wellness programs do not produce a measurable financial return, and the earlier claim that they did was largely an artifact of who signs up. But the money inside those same programs was never evenly spread either. When RAND separated the two components, the disease-management component, which reached about 13 percent of participants who already had an identified condition, returned roughly $3.80 per dollar and accounted for about 87 percent of total savings, largely through a near 30 percent reduction in hospital admissions. The lifestyle-management component, which reached about 87 percent of participants, returned roughly $0.50 and reduced absenteeism by slightly more than one hour per employee per year.⁰⁴ RAND's own five-year cost analysis is a caution against over-reading even that: the average annual health care cost difference for a participant was $157, and it was not statistically significant.⁰³ The pattern that survives is about targeting rather than about spending more. Return concentrated where the intervention was clinical and the population was already identified as at risk. It did not appear where the intervention was general and the population was everyone. Most employers still buy the second thing and evaluate it with numbers borrowed from the first, which is also why benefits designed for the whole headcount so reliably fail to reach the people at the top of it.

One column in that table does the work. Every row that reports a return is a row about a population that was already identified, and every row that reports nothing is a row about a population that was everybody. The difference is not effort or budget. It is the denominator, and a denominator this large also hides the loss that never appears in an attendance record, which is the mechanism behind burnout's effect on productivity in senior roles.

A named model earns its place here for one reason. The disagreement about wellbeing return is not really a disagreement about numbers; it is a disagreement about which population the numbers describe. The four stages below hold apart four questions that benefit reporting usually collapses into one. The model describes where financial return concentrates. It does not claim how large that return is, and the fourth stage is the one that decides everything, which is why clinical treatment for executive burnout is not interchangeable with a step-count challenge offered to the same person.

The model explains two uncomfortable facts at once. A program can be well designed, well liked and well attended and still return nothing, because stages two and four worked against it. And a program reaching very few people can carry real financial weight, because stage four is where the leverage lives. None of that is an argument for spending less on population wellbeing, which has its own justification in duty of care and in the working conditions the Surgeon General's framework describes.¹⁷ It is an argument for not asking population wellbeing to produce a return it was never positioned to produce.

The concentration argument only holds if the strain is real in the roles where the financial weight sits. It is. Three groups below have documented exposure, measurable replacement cost, and the structural reasons a standard benefit is least likely to be used by them.

The cost of doing nothing is easy to state at the level of the world economy and hard to state at the level of one company. The World Health Organization estimates that 12 billion working days are lost every year to depression and anxiety, at a cost of about US$1 trillion a year in lost productivity, and that 15 percent of working-age adults have a mental disorder at any given point.⁰⁶ Gallup puts the cost of low engagement at roughly $10 trillion, or 9 percent of global GDP.⁰⁹ Those numbers are useful for direction and useless for a budget. The three below can be attached to identifiable people.

Good care for this population is defined by three properties, none of them exotic. It has to be clinical rather than educational, because the presenting problems are depression, anxiety, insomnia and the erosion of judgment under sustained load, and those have treatments with an evidence base; the National Institute of Mental Health notes that for many therapies, research involving large numbers of patients has shown the treatment reduces symptoms of depression, anxiety and other disorders.¹⁶ It has to be confidential in a way the person can verify rather than merely be assured of. And it has to be available at a depth and a cadence that fits a calendar which is frequently not the client's own.

CEREVITY is built as a nationwide network of independent licensed clinicians, matched to the person rather than assigned by geography, delivered by secure video, and paid for privately so that the work does not generate a shared record. The scheduling model reflects the range of clinical need rather than one default, in three depths: the 50-minute format carries ongoing weekly work, and 90-minute extended work exists for sessions where an hour ends in the middle of something.

The third depth is a 3-hour block. Where a leader cannot hold a weekly rhythm through a quarter close, a filing or a diligence period, single-block intensive work concentrates the same clinical work into one sitting. The reasoning behind those three depths, and behind matching rather than assigning, is set out in how the clinical model is structured.

The practical work here is mostly definitional. Most of the confusion about wellbeing return comes from one budget line being asked to do two jobs, so the first move is to stop asking it to.

1. 01

   Split the budget in two, and give each half its own test

   Population wellbeing is justified by duty of care and by working conditions, along the lines the Surgeon General's workplace framework sets out, and it should be reported on participation, conditions and retention rather than on a return multiple.¹⁷ Targeted clinical care is justified financially and should be tested financially. Merging the two ensures that one of them is measured with the wrong instrument.

2. 02

   Define the covered population by exposure, not by title

   The question is not seniority but concentration: whose incapacity for a quarter would be visible in results, in a filing, in a clinical schedule or in a client relationship. That usually produces a list shorter than the executive team roster, and occasionally one that is longer.

3. 03

   Design for uptake, because uptake is where the return is destroyed

   Every population in section 04 shares one barrier: the belief that using the benefit creates a record. The World Health Organization's 2022 guidelines on mental health at work set out organizational interventions, manager and worker training, individual interventions, return to work and employment support as the intervention set,⁰⁸ and every one of them assumes people will actually engage. Route the benefit outside the internal reporting line, and say plainly and in writing what is and is not shared.

4. 04

   Pre-register the outcomes and the measurement window

   Name the measures before the program starts, choose ones the intervention could plausibly move inside the window, and record baseline cost and turnover for the covered group first. Doing this in advance is the only real defense against the selection problem that produced the original wellness ROI literature.