Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Mental Health and Episodic Disability Claim Classification

Abstract. Mental-health-related disability claims pose a distinctive classification problem because clinical diagnosis, functional limitation, occupational capacity, insurance coverage, and continuity of benefit eligibility are different measurements of a changing condition. Episodes of deterioration and recovery may recur; treatment may improve visible symptoms without restoring sustainable work performance; and the relevant occupation, contractual definition, waiting period, benefit limit, and review date may change during the claim. This monograph examines the classification of mental health and episodic incapacity in private disability income insurance and its interfaces with statutory disability and sickness protection. Its method combines documentary comparison of clinical and functional standards, primary legal texts, supervisory provisions, disability-insurance experience studies, and systematic reviews of occupational interventions. The study constructs an explicit separation between diagnostic evidence, longitudinal function, occupational demands, benefit triggers, and claim administration; derives the implications of recurring episodes for incidence, termination, reserving, and interpretation; and compares the United States, United Kingdom, European Union, and international rights instruments with their respective jurisdictional limits. Evidence from WHO, the U.S. Social Security Administration, the Society of Actuaries and LIMRA, the U.S. Department of Labor, the International Labour Organization, and clinical-occupational research establishes why a single diagnosis, a single examination, or a single return-to-work event cannot by itself characterize the full duration or the financial effects of episodic incapacity. A transparent set of analytic examples illustrates exposure definitions, competing claim exits, recurrent events, and censoring without pretending to estimate population parameters. The central conclusion is that classification quality depends upon temporal consistency, functional specificity, explicit contract interpretation, distinguishable administrative outcomes, and independently reviewable evidence. The monograph presents an original synthesis of existing scholarship and institutions; it institutes no new theory or doctrine.

Keywords: Disability income insurance; mental health; episodic disability; recurrent incapacity; occupational functioning; mental disorder classification; insurance claims; long-term disability; claims adjudication; disability duration; claims termination; incidence; actuarial experience; discrimination; reasonable accommodation; social protection; longitudinal evidence; return to work.

Introduction

Classification is the hinge between a person's changing health, the work that person is expected to perform, and a payment promise made under specified rules. Insurers classify a claim when they decide what condition and limitation the evidence establishes, whether the relevant contract supplies a benefit, how long qualifying incapacity lasts, and which event permits payment to continue, change, or stop. The difficulty becomes particularly visible in mental disorders and episodic conditions. A claimant may meet diagnostic criteria during one interval, recover enough to perform certain tasks, and subsequently experience deterioration in the same or another condition. A transition to part-time work may indicate substantial improvement while leaving an earnings loss. A normal clinical encounter can coexist with difficulty sustaining concentration, attendance, interpersonal interaction, or pace across an ordinary workweek.

The World Health Organization's International Classification of Functioning (WHO, 2001) locates disability and functioning within a relationship among health conditions, activities, participation, and environment. The ICD-11 clinical diagnostic requirements (WHO, 2024) serve a different purpose: identifying disorders for clinical recognition and care. Insurance arrangements introduce further predicates: coverage in force, an insured occupation, an elimination period, required proof, maximum benefit terms, and coordinated earnings or other benefits. The analytical object of this dossier is the relationship among these predicates over time.

The international research record gives good reasons to study the relationship, while also imposing limits on inference. WHO's Guidelines on Mental Health at Work (2022) address the participation of people with mental conditions in employment. The OECD's Sickness, Disability and Work synthesis (2010) examines institutions that shape transitions into and out of disability protection. The Society of Actuaries and LIMRA publish group long-term-disability incidence and termination experience, but their aggregate insurance observations do not constitute individual clinical adjudications. At the other end of the chain, legal rights instruments govern access to work and nondiscrimination and do not automatically determine the terms of every privately insured benefit.

The study is a documentary and analytic synthesis, not a reanalysis of confidential claims microdata. Primary legal and institutional texts anchor the statements about law, regulation, and insurance experience; peer-reviewed systematic reviews and longitudinal research anchor statements about observed work outcomes. Jurisdictional conclusions are explicitly localized. Quantitative examples below are hypothetical and identified as such. No prevalence estimate is treated as a disability-claim incidence rate. No clinical diagnostic category is presumed to entail a particular contractually insured benefit. The substantive chapters follow the approved research design: foundations, episodic measurement, diagnostic and occupational evidence, coverage and duration, validation, and synthesis.

Conceptual and Historical Foundations

The objects classified

A clinical diagnosis identifies a condition according to a diagnostic system. A functional observation records the ability to perform an activity, often in a particular environment. Occupational capacity concerns the sustained performance of the material duties of specified work, with its actual physical, cognitive, interpersonal, temporal, and safety demands. Contractual disability is a legal and economic status arising when evidence satisfies the definition in an applicable insurance contract. Statutory disability is the status created by a particular public law and administrative test. An approved claim means the adjudicator accepted eligibility for some period; a payment is an additional cash-flow event subject to offsets, administration, and payment timing.

These objects require separate variables even when the same case file contains them all. Let D(t) represent diagnostic evidence, F(t) functional capacity, O(t) the occupational requirements at date t, C the insurance contract, and E(t) the adjudicative evidence then available. A claim's contractual eligibility can be written schematically as I(t; C, D, F, O, E), a determination governed by the contract and evidence, rather than a clinical score or universally applicable algorithm. This is a descriptive notation for established legal and actuarial distinctions, not a proposed new theoretical model. For an individual with a depressive condition, a diagnosis can persist during recovery, functional deficits can change faster than diagnostic coding, and insured eligibility can terminate at a contractual cutoff while symptoms and inability to work persist. Each occurrence has a different meaning.

The WHO Disability Assessment Schedule (WHODAS 2.0) measures functioning across domains including cognition, mobility, self-care, interpersonal relationships, life activities, and participation. Its multidomain structure helps separate functioning from diagnostic classification, but generic domain scores alone do not specify a policy's duties-of-occupation test, proof requirements, or threshold for benefit payment. Conversely, insurance documentation can reasonably require the work-related translation of functional limitations: what exact duties are affected, under what conditions, at what frequency, and for how long?

Historical emergence of the modern question

Disability income protection developed through a combination of wage-replacement products, employer-sponsored group cover, public sickness and disability programs, and employment protections. These arrangements evolved along different institutional tracks. Their coexistence explains why an employee may simultaneously possess legal protection against employment discrimination, employer sick pay, a disability-income policy, and potential statutory disability entitlement, yet satisfy different conditions for each.

The OECD's cross-country Mental Health and Work research series documents the economic connections between mental illness, sickness absence, employment and disability-benefit entry. Arends and colleagues (2014), in their integrated policy analysis, show why employment, health care, and social protection institutions have to be examined together when studying work outcomes. Their contribution to this monograph is institutional: a claimant's eventual work status is produced under employment opportunities and administrative rules as well as medical recovery.

In U.S. group long-term-disability plans, mental-health-related duration limits became a particularly important feature of contractual classification. The U.S. ERISA Advisory Council's 2023 report on disability parity describes widespread provisions limiting certain mental-health or substance-use-disorder benefits to approximately 24 months, contrasting them with potentially longer benefits for some other disabling conditions. This is a report of the Advisory Council, including testimony and policy recommendations, rather than a position adopted as federal law by the Department of Labor. The report states that the Mental Health Parity and Addiction Equity Act for health coverage does not, by itself, extend parity obligations to long-term disability income plans. That categorical distinction matters when studying claims: an analyst who transfers medical-plan parity rules automatically into a disability-income contract will misdescribe the legal system.

The Council also discusses disagreement over causal classification where conditions and symptoms overlap. Mental disorders can coexist with chronic pain, neurologic illness, fatigue, sleep disruption, or physical consequences of treatment; some conditions can have psychiatric symptoms and physical causes. Whether a mental-disorder limitation applies depends on operative wording, applicable legal interpretation, and the specific proof of causation. Classification codes can inform a claim without resolving the contractual question. Such disputes are inherently sensitive to policy wording and jurisdiction, so this dossier does not assign a universal outcome to mixed-cause cases.

Temporality as an essential feature

A static account imagines one condition, one assessment, and one decision. An episodic account must observe onset, waiting periods, periods of partial recovery, relapse, claim submission, delayed evidence, review, return to work, and possible re-entry. The same person may pass through several episodes during a calendar year or a multiyear policy term. Because public programs and private insurers may recognize these changes at different thresholds, a period of medical improvement cannot simply be equated with the end of the insured loss.

Temporal questions include the unit of observation, whether brief returns to work interrupt an elimination period, whether recurrent disability provisions connect episodes into a single claim, whether a benefit limitation applies cumulatively or per episode, and whether a later claim must satisfy new evidence and eligibility tests. Each answer is document-specific. The definition of continuous disability under one contract cannot be inferred from the U.S. Social Security Act's disability definition, which includes an inability to engage in substantial gainful activity by reason of a qualifying impairment meeting a statutory duration criterion. Public entitlement is not a ready-made proxy for private policy disability.

Dimensions of measurement and the problem of classification accuracy

Classification accuracy cannot be summarized solely as diagnostic concordance. The diagnostic criterion may be reliable and the insurance decision still wrong if work duties, sustained functional evidence, policy terms, or the relevant time interval were inaccurately evaluated. An equally important distinction separates a claim adjudicated correctly under restrictive wording from a contract design whose coverage leaves a substantial loss uninsured. The first is a question of decision correctness under existing rules; the second concerns risk allocation, distribution, and fairness. Both are legitimate scientific questions, and evidence answering one does not settle the other.

The UN Convention's employment provisions (2006, Article 27) establish an international rights framework for employment access, reasonable accommodation, job retention, and return to work. Its social protection provisions (Article 28) address adequate living standards and social protection. These articles supply evaluative principles for institutional design; treaty implementation and private-law effects depend on national law. The ensuing chapters therefore analyze legal status and benefit design separately before examining their consequences together.

Episodic Impairment Measurement

Diagnostic episodes and functional episodes

An episodic mental condition is one in which symptoms or functional effects vary over time, including recurrence after intervals of improvement. For insurance research, the analyst must define precisely whether an episode refers to symptoms, a clinician-recorded disorder episode, a medically certified work absence, the start of an insured waiting period, or a newly admitted benefit claim. These events can occur in different orders and at different dates. A worker can experience symptoms before any absence, can work while symptomatic, or can cease work before a specialist records a definitive diagnosis. The relation among events is empirical and administrative, not definitional.

A scientific episode table should preserve, at minimum, time of first symptom report where documented, first clinical encounter, diagnostic confirmation, first reduced duties or absence, notification date, elimination-period start, claim admission, benefit commencement, periodic review, and exit. Where dates are unknown, the record must carry the uncertainty rather than silently substitute the next available timestamp. Date of claim payment is particularly unsuitable as a proxy for onset because administrative delay varies across cases and systems.

The WHO's ICD-11 diagnostic guidance offers standardized clinical terminology, including a lifespan and culturally responsive perspective. For longitudinal measurement, an ICD category provides a starting point, while repeated functional assessments and occupational records provide the time structure. Diagnostic coding can change between revisions, providers, countries and claim years; historical coding requires preserving the version employed at the time and recording later harmonization rules. A retrospective recoding exercise can create apparent changes in frequency that originate from taxonomy rather than claimant incidence.

Observation frequency and the hidden trajectory

The observation process is unequal across claimants. A worker experiencing worsening symptoms may consult a clinician more frequently; another may face geographic, financial, language, or administrative barriers to care. Visits are therefore partly triggered by health status, access, and claims administration. If claim severity is reconstructed only from appointment records, both observed episode frequency and apparent recovery may be biased. The research question is whether a missing record represents no impairment, no contact, or no documentation. These possibilities must remain separately coded.

The WHO/ILO mental health policy brief (2022) places work conditions and access to support in the same policy field. This matters to disability classification because work organization, employer adjustment, and treatment availability alter observed functioning without necessarily altering diagnostic labels. At an individual level, evidence of sustained performance with flexible hours is not identical to evidence of sustained performance under the insured occupation's ordinary requirements. A remote or sheltered role may show particular capabilities, while still requiring examination of whether that role represents suitable alternative work under the relevant policy.

Repeated measurement is therefore preferable when a claim turns on sustainability. A defensible longitudinal account can combine clinician observations, treatment effects and side effects, work schedules, attendance, reliable accounts of task performance, and the claimant's own descriptions. Each source measures something different and can contain error. Self-report is important evidence of subjective symptoms and function but is not interchangeable with employment records; employer records can document attendance but rarely establish psychiatric diagnosis; standardized questionnaires can facilitate comparison but require attention to scoring, administration, language, and validation.

Function domains and occupational granularity

WHODAS 2.0's six domains of functioning permit structured discussion of limitations in cognition, mobility, self-care, interpersonal relations, life activities, and participation. A claims assessor can supplement these domains with job-specific evidence: concentration and persistence in a high-interruption environment; response to emergencies; safe handling of equipment; predictable attendance; effective communication under conflict; and the ability to sustain a full schedule. An occupational demand is not an intrinsic trait of the claimant. It is a requirement of a particular job or qualifying alternative occupation, which may change with employer practices and adjustments.

A good assessment separates capacity, performance, and sustainability. Capacity concerns what the claimant can accomplish under specified conditions; performance concerns what actually occurs in the environment; sustainability concerns whether the performance can be maintained at the necessary frequency and duration. A single successful task can establish capability for that task under the observed conditions. It offers limited evidence regarding the ability to perform it repeatedly across a full workweek. Likewise, a short-term absence is evidence of interruption but does not by itself establish a long-term entitlement.

Recurrent events and thresholds

Episode counts require an explicit gap rule. For illustration, consider two depression-related work absences separated by 20 days of ordinary work. A study with a 14-day clean-gap criterion would count two episodes; one requiring 30 consecutive days without symptoms or work disability would count a single recurrent episode. Neither convention is intrinsically correct outside its scientific purpose. Researchers should publish their gap rule and examine sensitivity to alternatives. They should also report whether a recurrence must arise from the same diagnosis, whether a new cause creates a new event, and whether the insurance contract treats the intervening recovery as continuous disability.

An interval may contain recovery in symptoms with residual inability to work. An empirical recurrence study should specify the endpoint used to certify recovery. Ervasti and colleagues (2014) examined recurrence of depression-related work disability after return to work and observed recurrence in 35% of qualifying prior episodes in that cohort (2014). The finding is valuable precisely because its design identifies a selected population of Finnish public-sector workers with recorded episodes and returns to work. It does not imply a 35% relapse probability for every person with depression, every insured portfolio, or every elapsed duration; follow-up, selection, employment and administrative definitions matter.

Why longitudinal classification matters

Insured claims may contain more than one transition: from working to absent, absent to partially working, partially working to fully working, and working back to absent. Analysts should separate an episode end marked by restored paid work from an episode end marked by exhaustion of a benefit limit. The latter is an administrative termination with potentially continuing incapacity. The 2026 SOA/LIMRA long-term disability termination study studies recoveries and deaths in insured group experience; different administrative exits need explicit treatment to avoid treating all payment cessations as clinical recoveries.

The general interpretive rule is that an event must be classified according to the process that generated it. Diagnosis is generated by clinical practice, work absence by health and employment decisions, entitlement by law or contract, and payment by claim administration. These processes can be linked without being merged. This distinction makes longitudinal data auditable and prevents false conclusions about recovery, risk, or coverage.

Diagnostic and Occupational Evidence

Clinical identification and the evidentiary record

A disability claim involving a mental condition begins with a need to establish what the available clinical information shows. The WHO's 2024 ICD-11 clinical descriptions address identification of mental, behavioral, and neurodevelopmental disorders; they are clinical descriptions and diagnostic requirements, not disability insurance contracts. The translation to insured work incapacity requires evidence about the severity, persistence, occupational implications, and applicable dates of the condition. In a well-documented claim, the information may include structured clinical interviews, medical records, medication history, observed symptoms, psychiatric or psychological assessment, occupational or functional evaluations, collateral descriptions, and the relevant claimant and employer records. Each item should be weighed for its purpose and limitations.

The U.S. Social Security Administration's adult mental-disorder listings provide an unusually explicit public example of separating diagnostic and functional evidence. For several categories the listings require both disorder-specific documentation and functional criteria. The four listed areas are understanding, remembering or applying information; interacting with others; concentrating, persisting or maintaining pace; and adapting or managing oneself. The listings also address serious and persistent disorders through alternative criteria involving ongoing treatment or support and marginal adjustment in specified categories. This is an SSA statutory eligibility framework, not an industry-wide standard for private disability claims.

SSA emphasizes functional performance on a sustained basis and explicitly discusses the limits of assessments conducted in unfamiliar or highly supportive situations. A clinical evaluation can be accurate at the time of assessment while providing only a partial picture of a claimant's typical work functioning. A record may therefore need longitudinal observations and evidence from multiple contexts. The distinction is particularly important where a condition produces severe but intermittent episodes, where treatment changes symptom presentation, or where structured surroundings materially support functioning.

Treatment and functional limitation can point in different directions. An appropriate medication may improve mood or anxiety symptoms yet also cause fatigue or cognitive effects. Clinical improvement may permit partial work before full recovery, while the removal of workplace support may cause a renewed deterioration. The adjudicative question is what the claimant can sustain in the relevant work environment on the relevant dates, after considering medical evidence and contract language. The inference must be documented rather than imported from a diagnostic label or a generalized expectation about the condition.

Occupational evidence and sustainable work

Occupational evidence begins with a precise description of material duties. For a classroom teacher, work might require sustained interpersonal contact, preparation, attention to multiple students, and regular attendance; for a safety-critical operator, even intermittent impairment may have different consequences. The actual job and the policy-defined occupation are not necessarily identical. Some contracts reference the claimant's own occupation; others move after a defined period to a test involving any suitable occupation, sometimes with conditions related to qualifications, training, or earnings. The text of the contract determines which comparison is legally relevant.

Job-demand evidence can be collected through employer duty statements, schedules, performance and attendance records, occupational assessments, and claimant testimony. An adjudicator should identify whether the evidence describes pre-disability duties, temporary modified duties, alternative work, or work performed with accommodations. Otherwise, a finding that the claimant can perform a modified job may be confused with a finding that the claimant can perform the insured occupation. The OECD's integrated-services research underlines that labor-market outcomes are shaped by coordination among employment services, health care, and social protection. This institutional observation is consistent with the need to document actual workplace opportunities rather than assume that medical improvement automatically supplies suitable employment.

For mental conditions, the most difficult duties are frequently temporal and interactive: sustained concentration, pace, adaptation to changes, consistency of attendance, decision-making under stress, and predictable completion of work. These features can be documented with contemporaneous evidence. Their importance depends on occupation, employer, hours and safety demands. The same clinical impairment may support different occupational conclusions because the insured work comparison differs, even while the health condition remains the same.

Comorbidity, causation, and diagnostic transitions

A claimant's limitations may arise from several interrelated conditions. Depression may coexist with a musculoskeletal condition, another psychiatric condition, cardiometabolic illness, or a neurologic condition. The Finnish register cohort on recurring depression-related work disability found associations between comorbidity and recurrence among a defined population with prior work-disability episodes. These hazard ratios are observational associations, subject to residual confounding, cohort selection and recording conventions; they are not the individual probability that any specific diagnosis causes renewed disability.

Mixed conditions create at least three distinct questions. First, what disorders and impairments does the clinical record establish? Second, which impairments materially prevent the relevant work during each period? Third, under the precise insurance wording, which cause or combination of causes determines a duration limit, exclusion, or continued entitlement? A claim can be clinically complex without its legal interpretation being indeterminate: the contract and applicable law remain the authoritative decision rules. Conversely, a simplified coding field that selects a single primary diagnosis can obscure other causes materially contributing to work incapacity.

The ERISA Advisory Council's mental health disparity report directly discusses conditions with neurological, psychiatric, or mixed presentations and differing policy language concerning limitations. It records a policy debate and particular legal interpretations, rather than establishing one universally operative causal test. Research comparing insurer decisions should therefore retrieve full policy clauses, not merely a dataset's mental-versus-physical claim category. An apparent disparity may reflect underwriting, contract design, claimant composition, causation rules, or error in adjudication; its explanation requires case-level evidence and careful design.

Procedural fairness and decision integrity

Evidence quality includes how decisions are made. In covered U.S. ERISA benefit plans, the Department of Labor's claims-procedure guidance explains the review requirements for disability benefits, including procedures for full and fair review of adverse determinations. The applicable claims procedure regulation includes provisions addressing impartial adjudication for disability benefit claims, notice requirements and review of adverse decisions. The status and applicability of specific procedural provisions must be assessed by plan type and claim date; the rule is not a substitute for state insurance contract law.

Impartiality requires attention to the relationship between medical reviewers, vocational experts, claim handlers, and claims outcomes. The evidence record should indicate what was reviewed, which contrary information was considered, how internal or external conclusions were reconciled, what governing provision supported the decision, and how the claimant can obtain review. A diagnosis-based conclusion that omits function, or a work-capacity conclusion that omits central duties, is analytically incomplete even when the file is extensive. A lengthy list of documents is not equivalent to a transparent inference from evidence to decision.

Reassessment has its own evidentiary demands. A previous finding of disability establishes the outcome under an earlier set of facts, terms and dates. It does not automatically establish all later periods; equally, improvement documented during a follow-up does not prove that every material restriction has ended. The relevant test is whether the current record, read longitudinally, establishes the presently applicable criteria. Care should be taken when the insurance definition changes during a claim, when medication has changed, when the claimant has attempted work, or when newly discovered records clarify an earlier period.

Work participation, treatment, and return-to-work evidence

Interventions aimed at return to work can affect observed claim outcomes and create opportunities for partial recovery. Their evidence base is varied. A 2020 Cochrane review, Interventions to Improve Return to Work, included 45 studies involving 12,109 people with depression or depressive symptoms, comparing work-directed and clinical interventions. A later review by Brämberg and colleagues (2024) found that work-directed approaches for common mental disorders could improve or accelerate return to work relative to comparator conditions, with low certainty for the central findings and important methodological limitations.

The implication for insurance classification is precise. A treatment or workplace intervention is relevant because it may change the claimant's actual ability to perform particular duties, income, participation, or recurrence risk. It does not provide a universal causal rule guaranteeing recovery for an individual. Programs may differ in the degree of employer engagement, job adjustment, clinical care, income support and selection into treatment. A return-to-work event can mean any employment, partial duties, restored pre-claim earnings, or uninterrupted employment across a minimum period. Studies that use different endpoints cannot be combined as though they measured the same termination event.

Rehabilitation participation must be distinguished from demonstrated work capability. Participation may show motivation or access to supportive services and may generate meaningful evidence on functional response. A program's availability is itself uneven across regions, employers and insurance arrangements. A claimant who has been unable to obtain timely treatment or appropriate accommodations should not be mechanically equated with one who rejected a specific available service; claims decisions must be made under the actual contract, law and evidence. At a population level, differential access to services can distort comparisons of claim duration between groups.

Evidentiary hierarchy and limitations

Different classes of evidence answer different questions. Clinical diagnostic standards establish whether a disorder can be classified. A validated functional instrument describes measured difficulty. Job records document work demands and attendance. An insurance certificate supplies the promised coverage and definitions. Claims correspondence establishes which grounds an administrator relied upon. Statutes and regulations govern specified public entitlements and procedures. Actuarial studies describe observed experience in contributing portfolios. Systematic reviews synthesize interventions in defined research populations. An admissible or persuasive source for one question may be weak for another.

The resulting analytical principle is proportionality between a source and the claim drawn from it. A diagnosis alone is insufficient support for a claim about contractual duration; a long-term disability termination table alone is insufficient support for a claim about psychiatric symptom remission; a court decision on particular language should not be converted into universal benefit legislation. A rigorous dossier checks what the source actually measured and the bounds within which its findings travel. This principle governs both the next chapter's legal comparison and the later actuarial sensitivity analysis.

Coverage Interpretation and Duration

The insurance contract as the unit of legal interpretation

Private disability income cover converts defined losses of work capacity into contingent income replacement. The financial promise may depend on employment status, contribution history, an insuring definition, disability onset, a waiting period, ongoing proof, earnings offsets, maximum benefit duration, and termination events. The policy's own-occupation and any-occupation definitions can create different eligibility results for the same medical record. A mental-health-specific duration limit can create yet another boundary independent of medical recovery. The scientific task is to identify which boundary actually operated in each observed claim.

The unit of analysis should be a specific issued contract or certificate, effective during the relevant claim dates, with endorsements and amendments preserved. Group master policies and participant certificates may require careful reconciliation. Research should record applicable definitions, enumerated conditions, exclusions, limitation clauses, recurrence provisions, partial disability and rehabilitation terms, review requirements, and offsets. An isolated quotation of a benefit limit without its exceptions, cross-references and effective date is not enough to determine entitlement.

The U.S. ERISA Advisory Council's 2023 disability benefit inquiry provides evidence of the policy debate over mental-health duration limits in employer-sponsored long-term-disability plans. Its observed widespread 24-month provisions and testimony-based estimates refer to particular U.S. group-market conditions; they do not establish a 24-month rule for all U.S. policies, all states, individual disability insurance, or disability schemes outside the United States. Contract wording and applicable law decide the case. The Council recommended further research concerning actuarial consequences of removing differential limits, which confirms that the published industry evidence does not resolve the financial question by itself.

United States: private disability, ERISA claims, and Social Security

Employer-sponsored disability benefits governed by ERISA can trigger claims-process requirements whose purpose is a fair review of benefit determinations. The ERISA claims-procedure regulation governs specified processes rather than dictating the substantive content of every privately insured disability promise. The Department of Labor's explanatory FAQs emphasize that procedural disability rules follow a benefit conditioned on an adjudicator's disability determination, subject to the regulation's scope. State insurance rules and judicial interpretation may independently affect substantive policy terms; self-funded and insured arrangements can differ.

Social Security Disability Insurance follows a separate federal statute. Section 223 of the Social Security Act defines disability by reference to an inability to engage in substantial gainful activity because of a medically determinable impairment satisfying statutory severity and duration criteria. It also refers to ability to perform previous and other work under statutory considerations. The SSA's mental-disorder listings supply structured diagnostic and functional criteria for specified stages of evaluation. A favorable SSA decision may be evidence relevant to a private claim, but it does not automatically replace a different contractual definition, and a private insurer's admission does not itself grant statutory entitlement.

A full analysis should also distinguish claim administration from benefit coordination. A group policy may provide for offsets based on other disability income or require application for public benefits. In a claims dataset, reduced private payment may therefore indicate a public benefit offset rather than a reduction in gross insured disability or restored work capacity. A payment amount, entitlement determination, and functional status must be stored as separate outcomes. The analysis must also avoid importing the parity requirements applicable to health-service reimbursement automatically into disability-income benefit duration, as the Advisory Council report expressly distinguishes them.

United Kingdom: recurring impairment, workplace rights, and claim handling

Section 6 of the Equality Act 2010 defines disability for the statute's protected-characteristic purposes by reference to a physical or mental impairment with a substantial, long-term adverse effect on normal day-to-day activities. Schedule 1's recurring-effects provisions treat an impairment's substantial effect as continuing where it has ceased but is likely to recur, subject to the statutory conditions. The government guidance on fluctuating effects explains the approach to episodic or recurring effects. These provisions show an explicit legal recognition that intermittent impairment can have continuing significance in equality law.

The statutory recognition of recurring disability for equality purposes should not be read as a universal private income-protection contract definition. Income-protection eligibility depends on the insurance policy and applicable regulatory and legal rules. For UK insurers, the Financial Conduct Authority's ICOBS 8.1 claims-handling rules require prompt and fair handling, reasonable guidance to claimants, avoidance of unreasonable rejection, and prompt settlement once terms are agreed. Those obligations concern fair processes and outcomes within the regulated insurance setting; they coexist with separate equality and employment rights. A claims investigator needs all relevant documents rather than assuming one statutory definition determines the others.

Public sick pay also occupies its own legal category. The 2026 commencement regulations brought specified changes concerning statutory sick pay into force on April 6, 2026. Those changes affect the statutory sickness-income interface and the timing of support, but do not by themselves alter preexisting private income-protection definitions. In historical or comparative research, UK sick-pay law must be matched to the exact year being studied, since retrospective application of a later waiting-period rule would materially distort measured income gaps.

European Union and international employment rights

The EU's Employment Equality Directive 2000/78/EC establishes a general framework for equal treatment in employment and occupation and addresses reasonable accommodation in Article 5, subject to the Directive's scope and implementation by member states. It is directly relevant to the actual environment in which work capacity is exercised and to the possibility of employment despite impairment. It is not a general EU-issued long-term disability insurance policy and does not standardize every private contract's claims trigger. EU employment nondiscrimination, national social insurance, and private disability income cover therefore need separate documentation.

The UN Convention on disability and work (Article 27) recognizes equal access to employment, measures to promote retention and return to work, and reasonable accommodation. Its social protection article addresses broader income security and support. These standards provide an international evaluative frame for distinguishing legal opportunity to work, accessible workplaces and the financial adequacy of benefit protection. Country-specific implementation, enforceability and definitions require separate legal verification; a treaty article cannot be cited as a complete insurance entitlement schedule.

The International Labour Organization's World Social Protection Report 2024–26 documents significant cross-country gaps in public disability benefit coverage. Its estimates include 38.9% effective disability-benefit coverage among people with severe disabilities under the report's underlying measures. That figure measures a global social-protection outcome; it is neither the share of workers insured under private group disability plans nor the prevalence of mental-health-related private claims. Comparisons should retain the report's year, geography, target population and methodology.

Duration and the economically relevant endpoint

Duration has several meanings. Clinical duration may run from symptom onset to remission. Work-disability duration may run from first incapacity to restored sustainable duties. Contractually covered duration begins under the policy's benefit conditions and may end at recovery, exhaustion of benefits, eligibility change, death, or other contract events. Payment duration can be shorter or interrupted by administrative delay and offsets. A reserve valuation seeks to estimate future covered cash flows, including applicable termination and payment rules. Conflating these durations creates misclassification in both scientific interpretation and insurer financial reporting.

For mental-health-related claims, the distinction is sharp where a plan imposes a cause-specific cap. A claimant can remain unable to work after benefits end. If a claims system codes both genuine work recovery and exhaustion of the cap simply as terminated, an apparent improvement in claim duration can be produced by contractual restrictions alone. The resulting termination rate cannot be interpreted as a rate of functional recovery. Conversely, a claimant returning to partial employment may retain a reduced insured benefit under a residual disability clause. Such transitions require explicit cause-coded states and payment records.

The 2015–2024 SOA/LIMRA group LTD termination experience study covers more than 43 million life-months and reports approximately 937,000 recoveries and 181,000 deaths in its public overview. Those counts characterize the contributed study's definitions and data, not every exit mechanism in all contracts. The 2015–2022 group LTD incidence experience study reports about 294 million life-years exposed and 1.2 million claims. The studies use different observation units, claim processes and calendar windows. Their aggregate counts cannot be divided against one another to derive a valid duration estimate, recurrence frequency, mental-health share, or claimant-level hazard.

Comparative implication

The comparative finding is that the same person's episode can carry different statuses in clinical reporting, equality law, public disability entitlement, private insurance and employment practice. This is a predictable result of distinct purposes and definitions. Scientific classification becomes robust when researchers keep those purposes explicit and identify the document producing each decision. It becomes unreliable when labels are moved from one institution to another without translation.

Evidence Validation and Sensitivity

The observation population and event denominator

An incidence rate needs a population at risk and an observation period. In employer group long-term-disability insurance, the denominator may consist of insured employee life-years exposed, potentially stratified by age, occupation, industry, salary, gender, geographic area, waiting period and plan features. The numerator may count claims incurred, reported, approved, payable after an elimination period, or persisting beyond a chosen duration. Each numerator defines a different empirical question. Claims admitted after a waiting period are a selected subset of all episodes of lost work capacity, which are themselves a selected subset of all clinically recognized mental-health conditions.

The SOA/LIMRA 2015–2022 incidence study reports around 294 million exposed life-years and 1.2 million claims from contributing U.S. group long-term-disability business. Dividing the rounded totals yields about 4.1 claims per thousand exposed life-years for the study's aggregate observed claims. That quotient is a rounded descriptive calculation only: it is not age-adjusted, diagnosis-specific, policy-standardized, representative of all markets, or an estimate of mental-health disability incidence. The published landing page does not provide a public claim-level dataset sufficient for the more detailed stratified comparisons required by this topic. A scientific analysis must avoid inventing such rates.

The same principle governs prevalence and service-use statistics. An estimated proportion of working-age adults living with a mental disorder in WHO's mental-health-at-work guidelines concerns population health and work, not insured incidence. The ILO's disability benefit coverage statistics concern social protection under their own denominators. Neither statistic can be inserted into the SOA incidence denominator to derive insured claim probabilities.

Time origin, left truncation and selection

The appropriate time origin depends on the estimand. Incidence may use the first day of coverage as the start of exposure. A disability-episode study may use onset of qualifying work incapacity. Benefit duration can begin on the first payable day after an elimination period. A recurrence analysis may begin at the first verified restoration of work. Survival analyses that use a later administrative date for an earlier clinical event are vulnerable to delayed-entry and immortal-time problems. The researcher must preserve the actual sequence and use delayed-entry methods where a person only enters observation after already surviving a prior period.

A long-term-disability portfolio selects claims that met plan conditions and survived the waiting period. It may disproportionately exclude brief episodes and people without stable employment, eligibility or timely medical documentation. A study of claims approved after a long elimination period says relatively little about short sickness absences unless the selection mechanism is measured. Comparison across different elimination periods should therefore condition on exposure and survival to claim entry, and should describe what portions of morbidity remain outside the insured portfolio.

Selective observation arises later as well. Insurers may cease observing a claimant after a benefit limit, an occupational change or plan transfer, while public health records continue. That administrative exit is informative for medical and occupational outcomes: the end of the insurer's record is linked to policy conditions and sometimes to severity, not necessarily to recovery. Missing post-termination evidence is an evidentiary gap, rather than proof of renewed work capacity.

Recovery, death and other competing exits

In survival analysis, mutually exclusive first exits from a paying claim may include recorded recovery, death and benefit termination for a contractual reason. The cause-specific hazard for exit type k at duration t is the instantaneous rate of that event among claims still observed and at risk, conditional on the covariate definition and observation regime. An illustrative cause-specific cumulative incidence can be expressed as F_k(t) = integral from 0 to t of S(u−) times lambda_k(u) du, where S is the probability of remaining in the state of interest through u. This is standard competing-risk notation used here to clarify the question, not an estimation performed on proprietary data.

A recovery-specific quantity answers a different question from the proportion of all claims that cease payment. If claims exit due to a mental-health-specific 24-month limit, counting those exits as recovered creates an upward bias in apparent recoveries precisely at the limit. If a researcher instead censors them as though they were independent random withdrawals, the resulting estimate of recovery can also be biased, because the limit is determined by contract and diagnosis and may be related to the process under study. A full analysis should identify such exits as a competing administrative event and, where appropriate, model underlying functional recovery using independent follow-up data.

The SOA/LIMRA 2015–2024 termination study supplies an industry benchmark for specified group-claim terminations and includes recovery and mortality counts in its public overview. It cannot by itself identify how an unobserved claimant's symptoms, residual earnings, workplace accommodations or unpaid needs evolved after contractual benefits ended. That limitation affects causal comparisons of mental-health and physical-diagnosis claims, particularly if their policy benefits differ.

Recurrent events and person-level dependence

For an episodic claim, one person can generate several observations. Treating these as independent cases underestimates the uncertainty in a simple model and may overrepresent individuals who repeatedly experience illness or administrative recertification. Standard recurrent-event analysis can use person-clustered variance or appropriate individual-level frailty, depending on the research question and available data. It must also identify whether a second claim is genuinely a new episode, a reopened case, a change of diagnosis, or an administrative duplicate.

An individual-level longitudinal analysis could distinguish transitions among: insured and working; temporarily absent; eligible but within elimination period; fully payable; partially payable and working; restored work; benefit-exhausted but still functionally limited; and deceased. This list is an analytic classification of observed states for data design, not an original theory or claim that every contract recognizes the same states. State definitions must be mapped to existing policy and public-program categories before estimation. Some transitions may not be observable from insurance records alone and require linked employment or clinical data.

The Finnish recurrent work-disability cohort illustrates why person-level recurrence and comorbidity matter. Its reported 35% recurrence among qualifying prior episodes is not transferable to a private U.S. insurance book without accounting for employment selection, public-sector membership, national certification rules and follow-up period. Conversely, its register structure provides a useful example of what administrative linkage can reveal that a snapshot of current open claims cannot.

Confounding, differential measurement and causal claims

Comparing average durations for mental-health and other claims requires accounting for differences in age, sex, occupation, earnings, job security, severity, comorbidity, plan design, waiting periods, care access, calendar period and data completeness. Adjustment does not automatically solve the problem. Insurer coding may classify diagnoses differently over time; employers may select different plans; workers with severe or recurrent conditions may leave insured employment; and recorded recovery may partly depend on a claimant's ability to find accommodated work. These mechanisms affect both claim selection and outcomes.

A causal study of removing a mental-health benefit limit would need to specify the counterfactual coverage, the affected insured population, behavioral responses, and the full distribution of benefits and labor outcomes. Observing claims censored at 24 months does not reveal what those claims would have cost under unlimited coverage. The Department of Labor Advisory Council expressly called for research into actuarial effects of eliminating limits. This is a substantive identification problem: the available covered-duration data truncate the part of the distribution most relevant to the proposed change.

Evidence for work-directed intervention likewise varies by design. The 2020 Cochrane review pools or interprets randomized studies with distinct interventions and populations; Brämberg and colleagues' 2024 review finds low-certainty improvement in some return-to-work outcomes. It would be inappropriate to treat either as a calibration of claims termination hazards across all insurance types. A valid actuarial application would require external validation of case mix, support conditions, outcome definition and duration.

Worked sensitivity examples: explicitly hypothetical

Consider an artificial cohort of 1,000 newly opened claims, of which 200 are recorded under a mental-health-related category. Suppose that over a two-year follow-up, 80 of the 200 are recorded as recovered, 10 die, 60 reach a diagnosis-specific benefit cap and 50 remain payable. If a system labels all three exit causes as closed, it reports 150 closures, or 75% of the category. If closure is interpreted as recovery, recovery is overstated by 70 cases. With these stipulated counts, the recorded recovery count is only 80 of 200, or 40%; the crude fraction does not by itself estimate a duration-specific cause-specific cumulative incidence or account for the different times at risk. These figures are arithmetic illustrations, not observed insured experience.

Now change the administrative coding. Suppose 20 of the 60 capped claims have later evidence of restored employment, while 40 remain unobserved after the cap. A dataset limited to payment files still cannot conclude whether all 60 recovered at cap, whether all 40 unobserved cases remained disabled, or what eventual private benefit payments would have occurred in the absence of a limit. Its evidentiary scope ends at observed administrative status. A linkage to consented employment data could reveal some later labor outcomes, subject to measurement and access bias, but even observed return to work would not necessarily resolve all clinical or counterfactual benefit questions.

For a recurrence example, imagine 100 workers each with one observed absence in a first year. Twenty return and experience a second absence after a gap of 21 days. Under a 14-day clean-gap rule, the dataset contains 120 counted episodes; under a 30-day rule, the count remains 100 episodes. The denominator remains 100 persons. Neither episodes per person nor proportion experiencing recurrence should be reported without the gap convention and person-level follow-up. The sensitivity interval from 100 to 120 records is generated entirely by a classification decision; it is not evidence of any biological change.

For a benefit-offset example, suppose a private plan promises a hypothetical gross monthly benefit of $3,000 when eligible and reduces the amount by a qualifying public benefit of $1,000. A payment of $2,000 need not signify one-third recovery of work capacity. It reflects the contractual offset under the assumed terms. A register that only stores payments would make this cash-flow change observationally similar to a reduction in partial-disability benefit, unless gross eligibility, offsets and earnings are separately recorded. This small example demonstrates the same problem for monetary outcomes that diagnostic coding presents for clinical outcomes.

Reliability, reproducibility and missing data

A replicable study requires a codebook covering source system, jurisdiction, law date, policy version, observation unit, diagnosis-code system and revision, functional instrument, occupational duty definition, absence intervals, claim states, termination causes, appeal outcomes and missingness codes. Each data transformation should preserve its input provenance and alteration date. Manual recoding decisions should be traceable and reviewed, ideally by separate assessors blinded to irrelevant outcomes. Where classification involves judgment, the study should report agreement measures and the adjudication protocol, while recognizing that agreement alone does not establish truth.

Source reliability is partly domain-specific. WHO clinical and functioning materials support diagnostic and functional descriptions; SSA and legislative texts establish public-law definitions in their jurisdictions; regulator rules establish conduct obligations; actuarial study landing pages support the published scope and rounded aggregate counts; and systematic-review abstracts support bounded conclusions about available intervention evidence. This dossier uses no proprietary insurer claim rows, no unpublished occupational dataset, and no person-level medical records. The distinction between material retrieved publicly and inaccessible paid study packages is preserved rather than converted into invented numerical precision.

Scientific Synthesis and Research Agenda

Integrated findings from the evidence

The research yields six connected findings. First, clinical disorders, functional impairment, occupational incapacity, statutory disability, contractual disability and payment are different classifications, each created under a particular purpose. Second, episodic conditions make time origin, observation frequency, relapse and administrative state changes central to sound interpretation. Third, job demands and support conditions influence the observed relationship between impairment and work capability. Fourth, contract language determines when mental-health-specific limits and recurrence clauses affect payment duration. Fifth, published industry termination and incidence aggregates are valuable context but cannot reconstruct missing post-limit functional outcomes or the counterfactual cost of extending benefits. Sixth, defensible comparisons require legal specificity, explicit source-to-claim reasoning, and a reproducible longitudinal record.

These findings are an analytical synthesis of the verified source classes. They do not assert that clinical or functional measurement can be reduced to one score, that an actuarial table can settle individual legal entitlement, or that one jurisdiction's policy principle universally governs private disability contracts. The primary research value lies in showing exactly where measures join and where evidence is lost when they are treated as interchangeable.

Implications for claims operations

A claims file should contain an identifiable path from evidence to conclusion. The diagnosis and relevant medical period should be established; the functional limitations and their temporal variability should be characterized; the materially relevant occupational demands should be recorded; the exact policy clauses and effective dates should be identified; and the reasons for approval, partial benefit, further information, or adverse decision should be reviewable. If a case involves mixed causes, the record should distinguish clinical comorbidity from the contractual causal criterion. If benefits cease, a termination reason should distinguish recovery, death, policy exhaustion, loss of eligibility, noncompliance under applicable rules, and other administrative causes.

The DOL disability claims guidance and the FCA's claims-handling requirements illustrate how reviewability and fairness have specific regulatory expression in their respective jurisdictions. Implementation should be matched to the governing legal framework, with clinical and occupational expertise engaged where needed. No automated risk score can substitute for the evidentiary and procedural duties required by the applicable law.

Implications for product design and distribution

A mental-health-related duration limit affects risk allocation and can create a difference between insured benefit termination and continuing financial need. Policyholders and employers need comprehensible information about what the clause covers, how it interacts with mixed conditions, whether a longer option exists, and how recurrent episodes are treated. The 2023 ERISA Advisory Council report identifies coverage options, education of plan sponsors, and further actuarial inquiry as relevant policy concerns. Those observations support transparent comparison of design choices while leaving the empirical cost and distributional effects open for rigorous estimation.

Fairness has more than one dimension. The correct application of a clearly written exclusion or limit addresses administrative consistency; the existence and distribution of that limit addresses substantive product design; access to employment accommodation addresses a separate rights and labor-market question. All three deserve explicit examination. The EU employment equality framework and the UN disability employment principles help locate work participation within broader institutional obligations, without treating those rights instruments as universal insurance policy templates.

Implications for actuaries and reserve estimation

A disability portfolio's observed payment duration is shaped by insured work incapacity, contractual eligibility, benefit limits, mortality, claimant behavior and administrative practice. A reserve valuation should therefore use appropriately defined claim termination assumptions and contract features. Where a portfolio combines capped and uncapped mental-health benefits, pooled terminations may confound risk and design. A change in mix or a benefit amendment may create a structural break that ordinary extrapolation cannot identify. Experience periods, claim entry, diagnosis groups, limit status and termination cause must be harmonized before comparative analysis.

The SOA/LIMRA incidence study and termination study are important public starting points for exposure and termination concepts. Their public summaries do not supply the case-level causal evidence necessary to infer the cost of removing a specific mental-health benefit cap. Such projections should disclose assumptions, observed support, scenarios, uncertainty and the limits of generalizability. Numerical results derived from paid or confidential study packages require lawful access and independent validation before appearing as evidence in an approved published work.

Implications for occupational reintegration

Sustainable return to work depends on capacity, occupational demands and the availability of meaningful support. The WHO mental health at work guidelines provide an evidence-based occupational-health perspective, while the 2020 Cochrane review and 2024 work-directed intervention review identify both possibilities and limits of intervention effectiveness. Claims researchers should distinguish any return to work, stable return to work, restored hours, and restored earnings. The duration of follow-up after an initial return matters because an early return may be followed by renewed disability.

A joint medical-occupational record can improve understanding of the actual path without implying that every claimant should return at the same pace. The individual's duties, support needs, treatment response and work context are central measurements. The goal of research is accurate attribution and evaluation of observed outcomes, not the manufacture of an administratively convenient label.

Priority research questions

Several questions remain empirically testable. How frequently do alternative recurrent-episode gap rules change mental-health claim incidence and duration estimates? How do actual policy clauses allocate benefit continuity after short returns to work? What proportion of payment terminations caused by benefit caps coincide with sustained restoration of work, and how often is the later outcome simply unobserved? How do diagnosis coding, occupational information and adjudicator review affect inter-rater agreement and successful appeals? What contractual and labor-market effects follow when employers select plans without mental-health-specific limits? How do workplace accommodations influence the relation between medical improvement, claim termination and earnings recovery?

Each question requires an estimand and a feasible design. The first two can be investigated through full contract-text coding and longitudinal case files. Post-cap outcome questions require linked administrative, employment and clinical data with lawful access and adequate follow-up. Adjudication consistency requires auditable decision records and blinded or appropriately controlled review. Coverage design comparisons may use natural experiments or policy changes where selection and pretrend evidence support them. The effectiveness of workplace support can be examined using randomized trials or carefully constructed quasi-experimental designs. None should be described as answered merely because a statistical association is observed.

Strengths and limits of this dossier

The main strength is its cross-source separation of diagnostic, functional, occupational, legal and actuarial meanings, supported by direct institutional and peer-reviewed evidence. The primary limits are the absence of person-level private claim data, restricted access to detailed commercial actuarial experience packages, incomplete global harmonization of disability insurance contract language, and the jurisdiction-specific character of legal findings. There is also publication and ascertainment bias in clinical and occupational research, and selection into insured employment. These are substantive limitations on the conclusions drawn here, not deferred editorial tasks.

A further source limitation concerns the 2023 ERISA Advisory Council inquiry: it is an advisory report based on evidence and testimony and expressly disclaims representing the Department of Labor's position. It is therefore cited to establish the Council's study, the reported prevalence of duration limits, and the policy debate, rather than to establish binding U.S. law or a fully representative market rate. The SOA/LIMRA landing pages establish the existence, scope and rounded totals of studies; the paid granular datasets were not reviewed. The legal provisions were consulted as online publisher or legislative texts available in November 2026; exact versions and applicability remain essential when reproducing a historical adjudication.

Conclusion

Mental-health and episodic disability classification is a longitudinal insurance problem situated at the intersection of clinical evidence, work performance, contract language, actuarial observation and procedural rights. A diagnosis identifies a condition, a functional assessment describes abilities in context, an occupational assessment concerns the sustainable demands of particular work, and a disability insurance contract specifies the conditions under which income replacement is payable. The relation among these components changes over time and can be affected by treatment, accommodation, relapse, changing policy definitions and administrative decisions.

The evidence supports a concrete scientific conclusion: the quality of classification depends on preserving the distinction between a medical episode, work disability, contractual entitlement, claim termination and benefit payment. The distinction is especially consequential at recurrence boundaries and at mental-health-specific duration limits, where a closed claim can remain associated with continuing incapacity. Epidemiologic prevalence, public disability protection, private claim incidence and administrative termination therefore require their own denominators and causal interpretation.

Comparative legal materials confirm that disability is defined for different institutional purposes. U.S. Social Security and ERISA claims rules, UK equality and insurance conduct provisions, EU employment accommodation, and international disability-rights instruments each make specific contributions without becoming interchangeable. Clinical standards and occupational-intervention research improve the evidentiary context while retaining their limits of application to particular contracts and cases.

The monograph's original contribution is an integrated and source-traceable interpretation of those established distinctions, with explicit implications for claims adjudication, contract analysis, experience studies and future empirical research. Future progress requires policy-level text, longitudinal function and work records, independently classified administrative exits, and transparent sensitivity analysis. These requirements define a rigorous research agenda for studying episodic disability within the existing disciplines of insurance, actuarial science, occupational health and comparative law.

References