More than a billion people live with mental health conditions, yet mental health remains the least standardised benefit in medical insurance. Why coverage lagged, what good design now looks like, and what is still unsolved.
The most consequential gap in health insurance today is not a missing hospital or an uncovered drug. It is a category: mental health, where the mismatch between need and coverage is wider than anywhere else in the benefit table. The World Health Organization reported in 2025 that more than a billion people live with mental health conditions — anxiety and depressive disorders the most common — while governments still devote only around 2% of health budgets to mental health, and treatment reaches fewer than one in ten affected people in low-income countries against roughly half in wealthier ones. Depression and anxiety alone are estimated to cost the global economy about a trillion dollars a year, mostly through lost productivity.
Insurance did not create that gap, but for decades it mirrored and reinforced it. Mental health was the benefit most likely to be excluded outright, capped hardest when included, and least standardised across otherwise comparable plans. That is now changing — unevenly, and in ways that make mental health the most interesting design problem in medical insurance.
Why insurers stayed away so long
The historical exclusions were not pure prejudice; they were prejudice plus three actuarial anxieties, and it is worth naming them because they still shape wording today.
- Definitional softness. A fracture is visible on an X-ray; a depressive episode is diagnosed through clinical judgement. Insurers feared open-ended claims against conditions they could not objectively verify.
- Duration risk. Therapy has no natural endpoint the way a surgical episode does. Without design controls, a covered course of treatment could run indefinitely.
- Adverse selection. People aware of their own mental health needs were expected to seek out generous cover, concentrating cost in the insurers that offered it.
Each anxiety had a kernel of truth and an oversized consequence. Blanket exclusion did not remove the cost — untreated mental illness surfaced in insurers' books anyway, as somatic claims, medication, longer recoveries and disability — it just removed the ability to manage it. Stigma did the rest, keeping utilisation low enough that the exclusion rarely got challenged.
What changed
Three forces pushed mental health from exclusion toward the centre of benefit design. Regulation moved first in several markets, with parity rules requiring mental health benefits comparable to physical ones. Employers moved second: in tight labour markets, mental health support became a visible differentiator in benefits packages, and corporate buyers began asking insurers directly what a plan offers. And delivery economics moved third — teletherapy collapsed the cost and friction of providing care, turning a benefit that once meant scarce psychiatrist visits into one that can start with an app-based session this week.
The result is that the question has shifted from whether to cover mental health to how — and the how is genuinely unsettled, which is what makes this a frontier rather than a settled feature.
The design space taking shape
Look across modern group medical products and a recognisable architecture is emerging, typically in layers. A first layer of low-friction access: counselling sessions or an employee assistance line, often unlimited or generously allowed, no diagnosis required — designed to catch problems early and cheaply. A second layer of outpatient clinical care: psychiatry and psychology consultations and medication, managed like other outpatient benefits with visit limits or sub-limits. A third layer of inpatient psychiatric cover for the acute minority, usually with pre-authorisation. The design controls that replaced the old exclusions are utilisation-shaped rather than absolute: session counts, stepped-care pathways that begin with digital or group formats, and case management for complex conditions.
Honest design also means honest limits. Session caps are a blunt instrument for conditions with very different treatment lengths. Networks are thin — in many markets the binding constraint is not benefit wording but the supply of clinicians, and a generous benefit against an empty panel is marketing. And measurement is immature: unlike surgical outcomes, therapy outcomes are rarely captured in claims data, so insurers are mostly managing cost, not yet quality.
The old model treated mental health as a risk to exclude. The new model treats it as utilisation to design for. The unfinished work is treating it as outcomes to measure.
What distribution changes
For platforms embedding medical cover, mental health has quietly become a comparison point buyers actually use. Employers scanning otherwise similar plans notice the line that says counselling included; younger workforces in particular weight it. That argues for surfacing mental health benefits explicitly in plan-choice journeys — sessions, access channel, whether a diagnosis is needed — rather than leaving them inside the general outpatient wording where no one will find them.
There is also a privacy obligation that platforms must take more seriously here than anywhere else in the benefit table. Mental health utilisation is among the most sensitive data a health product touches, and the trust cost of mishandling it — especially in employer-sponsored schemes, where members fear visibility to their employer — is unrecoverable. Aggregate reporting to employers, strict separation of individual utilisation, and conservative data minimisation are not compliance niceties; they are conditions for the benefit being used at all.
Mental health coverage will not be finished for years — parity of wording is not parity of access, and access is not yet outcomes. But the direction is set, and the plans being designed now are deciding what the standard benefit table of the next decade looks like. Frontier is the right word: unmapped in places, moving fast, and where the interesting work is.