The Leave Problem

Mental health leave was once a rare event that most HR teams managed case by case. It is no longer rare. Leaves have surged 300% since 2017, accelerated through COVID, and have not corrected. In 2026, 61% of HR leaders report an increase in mental health leaves in the past year, with 1 in 6 saying that increase exceeded 25%. Mental health is now the most common reason employees request leave, surpassing illness, injury, and parental leave combined in several industries.

Most organizations are managing this reactively: processing leaves as they arrive, managing accommodation and return-to-work logistics, absorbing the cost. Few are asking the more important question: what were the signals before the leave request, and what would it have cost to address them? This report examines what is driving the surge, what a mental health leave actually costs when you count everything, and what the evidence says about the economics of prevention.

The data on mental health leaves is unambiguous and consistent across sources. ComPsych, which manages leave administration for thousands of US employers, reported a 300% rise in mental health-related leaves between 2017 and 2023. A further 22% increase was recorded in 2024 alone. Mental health now represents 1 in every 10 leaves taken, and in white-collar, high-intensity industries, the share is meaningfully higher.

Spring Health's 2026 Workplace Mental Health Annual Report, based on surveys of 500+ HR and benefits leaders across five countries, found that 61% of HR professionals reported an increase in mental health leaves in the past year. One in six said leaves had increased by 25% or more. These are not marginal changes. They represent a structural shift in how workforce mental health is manifesting, not as quiet presenteeism or gradual disengagement, but as acute, documented, operationally disruptive events.

The drivers are compounding. Global anxiety and depression have risen 25% since COVID and have not corrected. The AI transition is generating a category of workplace anxiety, about job security, about capability gaps, about the pace of change, that has no precedent and no established support infrastructure. Political and economic uncertainty has overtaken job demands as the top driver of employee distress in recent measurement. Employees are not struggling more because they are weaker. The load is genuinely heavier.

Mental health leave was the canary. Most organizations are still treating it as an anomaly. The data says it is the new normal.

The direct cost of a mental health leave, salary continuation, administrative management, short-term disability claims, is the number that appears in budgets. It is not the real cost. The real cost is distributed across a set of categories that most organizations have never aggregated into a single figure.

Mental health claims also last significantly longer than other leave categories, 24% longer than the average claim according to ComPsych data. An employee on a mental health leave is not typically back in two weeks. They are back in six to twelve, if they return at all. Research suggests that a meaningful proportion of employees who take mental health leave do not return to the same role, and a subset do not return to the workforce at all within 12 months.

The WHO estimates that depression and anxiety alone cost the global economy $1 trillion annually in lost productivity. In the US, untreated mental illness is projected to cost approximately $477 billion in 2024. These are not figures that belong in a wellness budget. They belong in a risk and finance conversation.

Most organizations know what a leave costs to administer. Almost none have calculated what it costs in total. The gap between those two numbers is where the prevention case lives.

Mental health leaves do not arrive without warning. They arrive after a period, often months, during which an employee was carrying something that was visible in their behavior, their output, and their engagement, if anyone was looking for it.

Research on burnout trajectories consistently identifies a predictable sequence: early-stage stress and decreased motivation, progressing to emotional exhaustion and withdrawal, progressing to functional impairment and eventual inability to continue. The leave request is not the beginning of the problem. It is the end of a long process of degradation that the organization largely missed.

The employees most at risk of mental health leave share a recognizable profile. They are often high-performing, which means their early decline is masked by residual output. They are unlikely to disclose, because self-disclosure requires acknowledging a vulnerability that conflicts with their professional identity. They sit in the subclinical space, stressed, depleted, running on empty, where traditional EAPs don't reach and managers aren't trained to look.

According to NAMI's 2025 Workplace Mental Health Poll, only 13% of employees whose mental health suffered at work told their manager. 72% of employees say discussing mental health at work is appropriate. They know it's acceptable. They still don't speak.

The leave request is not the start of the problem. It is the point at which the problem became impossible to hide.

There is a structural irony at the center of how most organizations approach mental health leave: the majority of investment goes to managing leaves after they occur, and almost none goes to the conditions that produce them. Leave administration, accommodation management, return-to-work programs, short-term disability insurance: these are all downstream of the moment that mattered. The employee who needed support six months ago, before the breaking point, received nothing because nothing was designed to reach them.

The imbalance persists because of how the two kinds of spending behave. Reaction requires no decision. It arrives as a disability claim, a replacement hire, an overtime line, a manager's lost hours: each one small, individually defensible, and never assembled into a mental health total. Prevention requires a decision: a budget line, an owner, a justification presented in advance against competing priorities. Reaction spend is diffuse enough to escape scrutiny. Prevention spend is visible enough to attract it. That asymmetry, more than any disagreement about the evidence, is why the gap survives every planning cycle.

The economics of investing in care are not actually in question. Published research on Spring Health, the company behind Thoughtful, offers the clearest picture available: a peer-reviewed study in JAMA Network Open found that employers saved $1.90 in health plan costs for every $1 invested in the benefit, and a Validation Institute–certified analysis found lower total healthcare costs among participants than among matched non-participants. An earlier Validation Institute–certified analysis, from 2023, found $2.20 in health-plan savings per $1 invested, rising to nearly $4.10 when reduced absenteeism and turnover were included. That broader figure includes modeled workplace savings as well as healthcare savings, so it should not be read as a healthcare-cost return alone. These findings answer the question most budgets ask, whether support pays for itself, and answer it in the affirmative.

What makes prevention difficult is not economics. It is reach. The employee most likely to go on leave might be the employee least likely to self-identify, least likely to use an EAP, and least likely to raise their hand. Prevention requires a model of care that reaches people before they've decided they have a problem, which is a fundamentally different design requirement from the one that produced the EAP category. This is the problem Thoughtful is built for: support available at any hour, at a threshold low enough that reaching it doesn't require deciding you have a problem first. What reaching people earlier looks like in practice is the subject of the next section.

The most expensive mental health spending in most organizations is never labeled as mental health spending.

Effective prevention of mental health leave is not a matter of awareness campaigns or manager training in isolation. It requires infrastructure that can reach the subclinical population: the 80% of employees who are stressed, depleted, or quietly struggling but who don't meet the threshold for clinical care and won't seek it out proactively.

The evidence for early, low-friction intervention is consistent. Thoughtful's randomized controlled trial found statistically significant improvements across five performance-relevant outcomes within two weeks, including meaningful increases in behavioral activation and help-seeking behavior, two factors closely tied to whether people reach support early. Critically, the intervention worked not by identifying people who were struggling and routing them to care, but by being available before that identification step was required.

Thoughtful's own customer data points in the same direction. In a pilot with a Fortune 100 insurance carrier, 48% of users said Thoughtful's AI helped them address something they otherwise would have ignored or handled alone, and 52% of participants came to Thoughtful because they needed immediate support or were feeling stressed and overwhelmed. Together, the two figures describe the population prevention is supposed to reach: people taking a first step toward support, some because they didn't think their situation warranted a traditional benefit, others because they needed help in the moment rather than an appointment.

Early intervention at scale requires three things the EAP model was not built for: availability outside business hours, an access threshold low enough that an employee doesn't need to have already decided they have a problem, and continuity: the ability to follow someone over time and notice when patterns are shifting. The organizations getting ahead of the leave surge are not doing something radically different. They are doing the ordinary work of mental health support, redesigned around the behavioral reality of when and how employees actually seek it.

Prevention doesn't require knowing which employee will go on leave. It requires building something that reaches all of them before any of them have to.

The mental health leave surge is not going to reverse without deliberate intervention. For HR and benefits leaders trying to understand where their organization stands, the following questions are a starting point.

What is your current leave trajectory? Do you have year-over-year data on mental health leave frequency, duration, and return-to-work rates? Most organizations don't. If you don't have the baseline, you can't measure the intervention.

What is the true all-in cost per leave? Most leave cost figures count direct salary continuation. Few count replacement, productivity loss on return, management time, and downstream turnover. Until you have the real number, the prevention ROI case is invisible to finance.

What are you offering the pre-leave employee? The employee who goes on leave in Q3 was likely struggling in Q1. What did they have access to? What was the threshold to access it? Did it reach them at 6am when the weight became acute, or only during business hours after they'd already decided they needed help?

Where does your EAP reach and where does it miss? Utilization data tells you who engaged. It tells you nothing about the much larger population who needed something and found nothing in the moment they looked.

What happens after a return? Most return-to-work programs focus on legal accommodation and gradual hours. Few address the psychological conditions that led to the leave, which means recurrence rates are high and organizations are cycling through the same employees repeatedly.

To calculate your own cost-of-leave baseline, use the financial case guide for workplace mental health investment. To examine where your current support reaches employees, and where it does not, use the benefit reach guide. For the manager-level signals that may appear before a leave request, read The Stress Your Team Can't Name.

The organizations that solve the leave problem won't do it by managing leaves better. They'll do it by building something that gets to employees six months earlier.