Mental Health Economics

How economics helps improve mental health decisions

Mental health economics examines how scarce resources can improve mental health, functioning, recovery, and wellbeing across prevention, treatment, and long-term support. It evaluates consequences that extend beyond specialist healthcare into primary care, social care, education, employment, housing, families, and the justice system. This page explains why mental health requires particular attention to outcomes, costs, time, equity, and implementation while using the core methods of health economics.

The field covers more than medicines

Economic questions arise across the full pathway from population prevention to crisis care and recovery support. The relevant intervention can be clinical, social, organisational, digital, educational, or regulatory. Comparators should reflect realistic alternatives, including waiting, fragmented care, or no active programme when those represent current practice.

Area	Example interventions	Important economic question
Prevention	School programmes, workplace interventions, suicide prevention, and public-health policy	Do early benefits and cross-sector savings justify current investment?
Early identification	Screening, outreach, and primary-care case finding	Does additional detection lead to effective care rather than unmet demand?
Psychological treatment	Individual, group, family, and digital therapies	How do therapist time, engagement, fidelity, and waiting affect value?
Medicines	Acute, maintenance, augmentation, and deprescribing strategies	How do response, adverse effects, adherence, relapse, and monitoring interact?
Service delivery	Collaborative care, crisis teams, inpatient care, and community services	Which configuration improves outcomes and continuity within capacity constraints?
Recovery support	Supported employment, housing, peer support, and social care	Are health, participation, autonomy, and cross-sector effects captured?

The decision perspective changes what counts

A healthcare perspective records costs and outcomes falling within the health service, while a public-sector or societal perspective can include education, employment, social care, housing, criminal justice, informal care, and patient time. Mental health interventions often shift costs between sectors, so a narrow perspective can make a beneficial programme appear unattractive to the budget that must initially pay. Results should separate sectors rather than hiding transfers inside one total.

Relevant perspectives include:

• The health and mental healthcare system.
• Health and social care combined.
• The public sector, including education, housing, welfare, and justice.
• Employers and workplaces.
• Patients, families, and unpaid carers.
• Society, including productivity and broader wellbeing when appropriate.

Mental health outcomes are multidimensional

Symptoms matter, but people may also value functioning, relationships, autonomy, safety, participation, cognition, hope, and recovery. A symptom score can improve without restoring everyday life, while meaningful recovery can occur despite residual symptoms. Economic evaluation should identify which outcomes matter to people with lived experience and avoid assuming that one clinical scale captures them all.

Outcome domains can include:

• Symptom severity and remission.
• Relapse, recurrence, crisis, hospitalisation, and self-harm.
• Functioning in daily life, education, work, and relationships.
• Health-related quality of life and wellbeing.
• Personal recovery, capability, autonomy, and social participation.
• Treatment burden, adverse effects, and experience of care.
• Outcomes for family members and unpaid carers.

Quality-adjusted life-years can miss important change

Quality-adjusted life-years provide a common measure across health conditions, but generic preference-based instruments may be insensitive to some mental health changes. Problems can arise when instruments emphasise physical functioning, omit recovery or capability, or do not reflect severe fluctuations. This does not make QALYs unusable, but it strengthens the case for reporting condition-specific and broader outcomes alongside them.

QALYs over periods (t) are calculated as:

$$ QALY = \sum_t u_t\Delta t $$

where (u_t) is the health-state utility and (\Delta t) is time in that state. Analysts should state the instrument, respondent, valuation set, timing, missing-data method, and whether proxy responses were used.

Capability and wellbeing can complement health utility

Capability measures ask what people are able to be and do, such as maintaining relationships, having control, participating, and pursuing valued activities. Wellbeing measures can capture broader effects that are not fully represented in health utility. These outcomes may be particularly relevant to social care, recovery, prevention, and cross-sector programmes.

The choice of measure should follow the decision question and avoid double counting. A cost-utility analysis, capability analysis, cost-consequence analysis, and wellbeing valuation answer different questions and should not be merged without a clear framework.

Fluctuation, relapse, and recovery shape time

Many mental health conditions fluctuate, with acute episodes, partial response, remission, relapse, and persistent symptoms. Short trials can miss recurrence, treatment switching, delayed benefit, withdrawal, or long-term adverse effects. A model should represent clinically meaningful changes over a horizon long enough to capture future episodes and recovery.

For a simple state-transition model with remission, symptomatic illness, relapse, and death:

$$ \mathbf{s}_{t+1}=\mathbf{s}_t\mathbf{P}_t $$

Transition probabilities may depend on treatment history, adherence, time since remission, comorbidity, and prior episodes. A memoryless Markov assumption should be tested when recurrence risk changes with history.

Suicide and premature mortality require careful modelling

Some mental health conditions are associated with suicide, accidental death, and excess mortality from physical illness. These outcomes are rare in many trials but highly consequential. Estimates should use credible population and clinical evidence, avoid double counting within all-cause mortality, and distinguish association from causal treatment effects.

Modelling should consider:

• Baseline mortality by age, sex, condition, severity, and setting where justified.
• Time-varying risk during crisis, discharge, treatment initiation, or discontinuation.
• Uncertainty in rare-event rates and treatment effects.
• Competing risks and physical-health comorbidity.
• The ethical and communication implications of valuing mortality outcomes.

Physical and mental health are interdependent

Mental illness can affect physical-health risk, self-management, healthcare access, and treatment adherence. Physical illness, pain, disability, and medication effects can also worsen mental health. Evaluations that isolate one condition may miss costs and outcomes produced through the other.

A complete pathway may include physical-health monitoring, metabolic or cardiovascular adverse effects, substance use, chronic pain, multimorbidity, and diagnostic overshadowing. Interventions that improve coordination can generate value across several services even when their direct mental health effect is modest.

Treatment engagement is part of effectiveness

An intervention cannot produce its expected benefit if people cannot start, attend, use, or continue it. Waiting time, therapeutic alliance, acceptability, digital access, side effects, stigma, transport, language, and competing responsibilities affect engagement. Trial efficacy should therefore be connected to realistic uptake, adherence, fidelity, and persistence.

Expected population benefit can be represented conceptually as:

$$ Population\ benefit = Eligible\ population \times Uptake \times Completion \times Effect\ among\ participants $$

This identity highlights the implementation pathway but does not imply that the factors are independent. Poor capacity can reduce uptake and completion at the same time.

Psychological therapies depend on workforce capacity

The cost and effect of psychotherapy depend on therapist grade, training, supervision, session length, group size, modality, fidelity, and non-attendance. Scaling a programme can strain workforce supply or reduce fidelity, so trial delivery costs may not predict system-wide implementation. Capacity constraints should be represented explicitly when they affect access or outcomes.

A simplified therapist capacity calculation is:

$$ Annual\ completed\ courses = \frac{Available\ clinical\ hours}{Sessions\ per\ course \times Hours\ per\ session}\times Completion\ rate $$

The model should also include assessment, documentation, supervision, cancellations, leave, and case complexity. Treating every paid hour as direct clinical contact will overstate capacity.

Digital mental health has distinct economic questions

Digital interventions can reduce marginal delivery cost and expand reach, but they still require development, maintenance, clinical support, safeguarding, data governance, and escalation pathways. Engagement often declines after initial use, and access can differ by device, connectivity, language, disability, and digital literacy. Low unit cost does not guarantee effective or equitable care.

Evaluation should report acquisition, onboarding, support intensity, active use, completion, clinical outcomes, adverse events, replacement of other care, and who is excluded. Human-supported and fully automated interventions should not be assumed equivalent.

Medicines create benefits, harms, and service needs

Pharmacoeconomic analysis of mental health medicines should include response, remission, relapse prevention, adherence, discontinuation, switching, adverse effects, monitoring, and withdrawal. Weight change, metabolic effects, sexual dysfunction, sedation, movement disorders, cognition, and other harms can affect quality of life and physical health. Acquisition price is often only one part of the economic consequence.

A treatment sequence model may be needed when patients move through acute, maintenance, augmentation, and alternative treatments. Line-specific evidence and reasons for discontinuation should be retained when they change later outcomes.

Informal care can be substantial

Family members and friends may provide supervision, emotional support, transport, crisis response, advocacy, and help with daily activities. Caregiving can affect employment, health, sleep, relationships, and wellbeing. Excluding these effects can understate both illness burden and intervention benefit.

Informal care time can be valued using replacement cost, opportunity cost, or another justified method:

$$ Informal\ care\ cost = Hours\ of\ care \times Value\ per\ hour $$

Time and health effects should be reported separately where possible to avoid double counting. Carer outcomes can also be included directly when the decision perspective permits.

Productivity effects require transparent methods

Mental health can affect absence, presenteeism, job retention, educational attainment, and unpaid work. Interventions may generate substantial gains, but estimates are sensitive to whether time is valued using wages, replacement costs, the human-capital approach, or the friction-cost approach. Employment effects may also be confounded by economic conditions and workplace selection.

Reports should state whose productivity is counted, the valuation method, duration, taxes or transfers, and whether improved employment is treated as a cost offset, outcome, or both. Productivity should not crowd out health and distributional consequences simply because it is monetisable.

Cross-sector costs should remain visible

Mental health programmes can change use of emergency care, inpatient beds, primary care, social services, supported housing, education, police, courts, and welfare. Aggregating these effects can inform societal value, but individual budgets still need to know who pays and who benefits. A cost-consequence table can show sector-specific impacts without forcing them into one ratio.

For sector (j), incremental cost is:

$$ \Delta C_j=C_{intervention,j}-C_{comparator,j} $$

The total across sectors is meaningful only for the stated perspective. Transfer payments should be distinguished from real resource use.

Cost-of-illness and economic evaluation answer different questions

A cost-of-illness study estimates the economic burden associated with a condition. It can demonstrate scale and identify cost components, but it does not show which intervention should be funded. Economic evaluation compares alternative actions and links incremental costs to incremental outcomes.

Large burden does not guarantee that every intervention is cost-effective. Conversely, a low-cost or uncommon condition can still have a highly cost-effective treatment.

Prevention can involve long time horizons

Preventive interventions may act before symptoms, diagnosis, or service contact and can affect education, employment, relationships, substance use, physical health, and future treatment. Benefits may emerge years after the programme while costs occur immediately. Long-term models require cautious causal assumptions and clear separation of observed from extrapolated effects.

Prevention analysis should consider:

• Universal, selective, and indicated target populations.
• Uptake and reach beyond people already connected to services.
• Spillovers to peers, families, schools, and workplaces.
• Possible labelling, anxiety, or unnecessary treatment from screening.
• Duration of effect and need for booster activity.
• Distribution of benefit and risk across population groups.

Early intervention can shift costs forward

Early intervention often increases near-term identification and treatment, so health-system spending may rise before later benefits or offsets appear. A budget holder may therefore face an affordability problem even when the intervention is cost-effective over a longer horizon. Economic evaluation and budget impact analysis should be presented together but kept conceptually distinct.

The evaluation should test whether early detection leads to timely, effective capacity. Screening without treatment availability can increase waiting, distress, and inequity without delivering the expected outcome.

Mental health services are constrained systems

Adding demand to one service can lengthen waits, shift thresholds, or displace other patients. Staff shortages, bed occupancy, referral criteria, crisis pathways, and geographic capacity can alter real-world cost-effectiveness. Static models may miss these system effects.

Queueing models, discrete-event simulation, or constrained scenarios may be useful when timing and capacity drive outcomes. At minimum, analyses should test uptake and throughput against plausible workforce and service limits.

Equity is central to mental health economics

Mental health need and access are socially patterned, and discrimination or stigma can affect exposure, diagnosis, treatment, and outcomes. Average cost-effectiveness can conceal lower uptake or benefit among groups facing greater barriers. Equity analysis should examine both the distribution of mental health and the distribution of intervention access and burden.

Relevant dimensions can include income, deprivation, ethnicity, sex, gender, age, disability, neurodivergence, housing, migration status, rurality, justice involvement, and severe mental illness. Categories and mechanisms should be locally justified and co-interpreted with affected communities.

Co-production improves relevance and validity

People with lived experience can identify outcomes, burdens, pathways, and implementation barriers that routine datasets or clinical experts miss. Co-production should influence the decision question, model structure, outcome selection, assumptions, interpretation, and communication rather than appearing only as consultation after analysis. Participation requires accessible materials, support, compensation, and transparent influence on decisions.

Lived experience does not replace empirical evidence, and technical analysis does not replace lived knowledge. Credibility comes from integrating both and documenting areas of agreement and disagreement.

Modelling should represent meaningful pathways

The model structure should fit the condition, intervention, and decision. Decision trees may suit short acute pathways, while state-transition or patient-level models can represent recurrence, treatment sequencing, comorbidity, and long-term recovery. Structural detail should be added only when it affects the decision and can be supported.

Common states or events include:

• No or subthreshold symptoms.
• Acute episode at different severity levels.
• Response, remission, recovery, relapse, and recurrence.
• Treatment initiation, continuation, discontinuation, and switching.
• Crisis, self-harm, hospitalisation, and supported care.
• Physical-health complications and adverse effects.
• Death from suicide, other condition-related causes, and background causes.

Economic evaluation compares alternatives incrementally

Costs and outcomes should be compared with the relevant alternative rather than reported in isolation. If QALYs are used, incremental cost-effectiveness and net monetary benefit can support a decision at a stated threshold. Broader outcomes may require cost-consequence, cost-benefit, capability, or multi-criteria approaches.

Incremental net monetary benefit is:

$$ INMB=\lambda\Delta QALY-\Delta C $$

The interpretation depends on perspective, time horizon, discounting, evidence, and whether the QALY captures the outcomes that matter. A positive value does not remove the need to examine affordability, implementation, and equity.

Uncertainty is both evidential and structural

Mental health evidence can be limited by short follow-up, attrition, unblinded outcomes, heterogeneous diagnoses, changing usual care, and exclusion of people with comorbidity or severe illness. Structural assumptions about relapse, recovery, suicide, productivity, and service capacity can dominate long-term results. These uncertainties should be tested separately.

Useful analyses include:

• Probabilistic analysis of parameter uncertainty.
• Alternative relapse, recurrence, and treatment-duration assumptions.
• Scenarios for uptake, waiting, fidelity, adherence, and workforce capacity.
• Alternative utility, capability, and wellbeing measures.
• Healthcare, public-sector, employer, and societal perspectives.
• Subgroup and distributional analyses.
• Value-of-information analysis for consequential evidence gaps.

Worked cross-sector example

Suppose a collaborative-care programme costs the health service an additional $500 per participant and saves $150 in other healthcare. It also reduces social-service use by $100 and improves productivity by $400 under the chosen valuation method. The incremental cost is $350 from the health-system perspective and negative $150 from the stated societal perspective before valuing health outcomes.

$$ \Delta C_{health}=500-150=350 $$

$$ \Delta C_{societal}=500-150-100-400=-150 $$

The programme is cost-saving only under the broader stated perspective and assumptions. The analysis must still report the health outcome, distribution of effects, uncertainty, and whether savings fall to the organisation funding implementation.

Common mistakes

Mental health economic studies can mislead when they use narrow measures, optimistic implementation, or unsupported long-term effects. The following errors are especially important because they can systematically undervalue or overstate interventions. Each should be checked before results inform policy.

• Treating symptom change as the only meaningful outcome.
• Assuming a generic utility instrument captures recovery, capability, and social participation fully.
• Ignoring suicide, physical-health comorbidity, adverse effects, or treatment withdrawal when material.
• Applying trial engagement and fidelity unchanged at scale.
• Counting cross-sector savings without showing which sector receives them.
• Treating welfare transfers as real resource savings without explanation.
• Assuming digital delivery has no ongoing support, safeguarding, or exclusion cost.
• Valuing productivity without reporting the method and distributional implications.
• Using burden-of-illness estimates as proof that an intervention is cost-effective.
• Reporting average value without examining access, stigma, and equity.

Reporting a mental health economic evaluation

Transparent reporting should connect the decision problem, lived experience, clinical pathway, outcomes, costs, and sectors affected. It should show where evidence is direct and where long-term or cross-sector consequences are modelled. Results should be understandable to patients, practitioners, commissioners, and analysts without concealing technical assumptions.

• Define the condition, severity, population, setting, intervention, and comparator.
• State the perspective, time horizon, discounting, price year, and sectors included.
• Report symptoms, functioning, quality of life, recovery, harms, and broader outcomes as relevant.
• Describe uptake, waiting, adherence, fidelity, persistence, and implementation capacity.
• Separate costs and savings by sector and distinguish transfers from resources.
• Explain utility, capability, wellbeing, carer, and productivity methods.
• Report uncertainty, subgroup effects, equity, validation, and limitations.
• Describe lived-experience involvement and how it changed the analysis.

The decision standard

Mental health economics should identify how limited resources can produce meaningful, sustainable, and fairly distributed improvements in mental health and life. A credible analysis follows people across fluctuating illness and recovery, captures consequences beyond the clinic, and tests whether services can deliver the modelled benefit in practice. The final decision should consider value, affordability, capacity, equity, and the perspectives of people whose lives are represented in the evidence.