Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Dependency-Free Life Expectancy and Long-Term Care Exposure

Abstract. Dependency-free life expectancy measures expected years outside an explicitly defined state of functional dependency. Interpreting that measure as long-term care insurance exposure requires separate information on functional severity, incidence, recovery, mortality, claim triggers, insurer selection, policy persistence, benefit duration, care utilization and prices. This international scientific monograph reconstructs Sullivan's life-table method, distinguishes WHO severity-weighted healthy life expectancy from Eurostat activity-limitation healthy life years and national self-rated-health indicators, and evaluates the historical hypotheses of morbidity compression, expansion and dynamic equilibrium. It formalizes the relationship between health-state occupancy and multistate transition intensities, demonstrating with a hypothetical calculation that identical 10% dependency prevalence can coexist with fourfold differences in episode incidence and markedly different mean spell duration. Source-grounded analysis of OECD, WHO, Eurostat, longitudinal US surveys, Society of Actuaries experience data and national care benefit rules establishes the necessary conversion from population metrics to insured liability. Original analytical synthesis identifies the information requirements and validity limits of long-horizon care projections; no new theory or original actuarial rate is claimed.

Keywords: Dependency-free life expectancy; disability-free life expectancy; healthy life expectancy; morbidity compression; morbidity expansion; Sullivan method; multistate life tables; activity limitation; long-term care insurance; ADLs; cognitive impairment; claim incidence; claim termination; dependent-state mortality; insurer selection; care utilization; long-horizon reserves; international comparative measurement.

Introduction

Life expectancy is a measure of survival under a defined mortality schedule. Dependency-free life expectancy (DFLE) partitions those expected years into states defined by functional limitations. Long-term care (LTC) insurance concerns a further transformation: some people transition into care-dependent conditions, some satisfy contractual eligibility, some remain insured, and some receive payments for covered care. The causal and accounting relationship among these quantities is the subject of this monograph.

Sullivan's mortality–morbidity index first formalized an integrated summary of mortality and observed morbidity in 1971. Since then, health expectancy has become an established part of population-health comparison. The WHO healthy-life-expectancy series, Eurostat healthy life years and ONS methodological framework embody related but distinct definitions. A statistic derived from broad self-reported limitations cannot be substituted directly for an insurer's probability of a benefit-triggering episode of severe dependency. An illness can be present without dependence; dependency may require supervision rather than residential care; a care need may be met informally; and even a severely impaired person may hold no private LTC contract.

Three research questions structure the inquiry. What exactly does a prevalence-based health expectancy measure, and under what statistical conditions can it be interpreted? How do changes in onset, recovery, severity and mortality produce apparent compression or expansion of dependency? What data and contractual transformations convert health-state years into a long-horizon insured exposure and present value of benefits? Each question requires a different denominator, state space and reference period.

The methodology combines primary-document historiography, health-expectancy mathematics, established multistate survival models, actuarial cash-flow representation, international comparison and critical evaluation of insurer-experience evidence. Cross-country numbers retain their original populations, definitions and years. The numerical demonstration is explicitly hypothetical. The authorial contribution is an original critical synthesis and formal illustration of identification limits, without proposing a new theory or claiming original longitudinal estimates. The study belongs jointly to Long-Term Care and Disability Insurance, Actuarial Science and Risk Theory.

Conceptual Foundations and Historical Development

Defining the object of dependency

The unit 'year without dependency' has meaning only after operational dependency has been defined. Limitations in activities of daily living (ADLs), instrumental activities of daily living (IADLs), cognition, mobility, communication, self-care and residential support are different observational categories. A measure of severe dependency may count an inability to complete two or three specific ADLs; a health survey may count any long-standing limitation in normal activities. A threshold for cognitive supervision may admit someone without multiple physical ADL impairments. Published rates must identify the exact criteria and the assessment method.

The WHO functioning classification distinguishes the impairment of body functions, restrictions in activities, participation barriers and environmental conditions. The World Report on Ageing and Health places functioning in the interaction of individual capacities and environments. Consequently, housing adaptation, assistive equipment and accessible transport may change a person's dependence on assistance without eliminating an underlying diagnosis. Caregiver availability can also determine how severe a limitation becomes in daily life. These features matter both to the ethics of measuring dependency and to the economics of formal care.

The sequence from morbidity to insured liability comprises at least eight distinguishable objects: disease, functional limitation, care need, care utilization, policy eligibility, claim submission, accepted claim and payment. A functional limitation may not produce a claim; a fixed-cash policy may generate a payment without proving an identical amount of formal expenditure. An international analysis must preserve all these transitions.

Historical hypotheses of morbidity

Fries (1980) articulated the compression-of-morbidity hypothesis: disability and illness may be postponed nearer to death, reducing the period of late-life impairment. Gruenberg's failures-of-success argument described a different mechanism, in which preventing deaths from chronic disease permits more people to survive with continuing conditions. Manton's dynamic-equilibrium account emphasized that mild chronic morbidity can become more prevalent while severe functional impairment is delayed or reduced. These are explanatory hypotheses and analytical perspectives with distinct empirical predictions.

A statement about absolute compression means fewer expected years in the specified dependent state at the same reference age. Relative compression means a lower proportion of remaining expected life spent in that state. Those results may diverge when longevity rises. Moreover, severe-ADL dependency may compress even while diagnoses of chronic disease expand. An undifferentiated assertion that 'morbidity is compressing' cannot support a numerical premium forecast.

Crimmins and Beltrán-Sánchez (2011) examined American patterns over their historical study window and found insufficient support for compression in major disease prevalence and mobility loss. Cai and Lubitz (2007) used a multistate perspective for older Americans during 1992–2003 and documented patterns of active life and disability that depended on definition and age. The studies differ in outcome, time period, and design; their juxtaposition demonstrates why a global one-direction trend is unwarranted.

Demographic versus contractual expectancy

The population life table expresses expected years for a hypothetical group under given mortality schedules. A policyholder's lifetime payment horizon begins at issue or valuation, conditional on age, insured health and in-force coverage. The insurer's cash flows also reflect underwriting selection, premium pattern, lapse, benefits, care prices and interest rates. The fact that a social indicator is expressed in 'years' does not make it an actuarial duration table. This distinction grounds the subsequent formal analysis.

Health State Life Expectancy: Measurement and Estimation

Sullivan construction

Let l_x be survivors to age x in a conventional life table, and L_a the life-table person-years lived in age interval a. Let π_a be the prevalence of an explicitly defined dependency condition for the corresponding population, age and period. The Sullivan estimator of remaining dependency-free life expectancy is

DFLEx=∑a≥xLa(1−πa)lx.DFLE_x=\frac{\sum_{a\geq x}L_a(1-\pi_a)}{l_x}.

Its complement under the same binary partition is

DLEx=∑a≥xLaπalx,ex=DFLEx+DLEx.DLE_x=\frac{\sum_{a\geq x}L_a\pi_a}{l_x},\quad e_x=DFLE_x+DLE_x.

The sum must be constructed consistently for the terminal age interval, sample weights and life-table conventions. The quantities represent expected person-years over the stipulated mortality and prevalence schedule, including years that may occur in nonconsecutive spells. They do not locate a person's future first onset of dependency and do not reveal the number or duration of episodes.

The original Sullivan study established this combined mortality–morbidity indicator. Imai and Soneji (2007) examined its formal statistical properties, showing the significance of stationarity assumptions and providing an extension for cohorts using suitable longitudinal or repeated cross-sectional disability data. Their research makes it essential to distinguish period standardization from actual cohort experience. A synthetic period expectancy may summarize the current social health state without being an unbiased forecast of the next thirty years of an insured person's life.

The prevalence π_a is a stock measure. In a stationary two-state system, the share currently dependent depends on how frequently new cases arrive and how long episodes last before recovery or death. It is therefore possible for observed prevalence to remain stable even as entry hazards fall and survival with dependency increases. Dividing an observed stock prevalence by one year cannot produce an annual LTC claim probability.

Related indicators and their limits

WHO's HALE indicator expresses equivalent years in full health after adjusting for the severity and duration of health losses. Since it weights less-than-full health, it is not simply the expected number of calendar years free of any disability or policy-defined dependency. Eurostat's healthy-life-years measure uses activity limitation information with a life-table basis; it concerns limitations rather than the severe-care trigger of a given insurer. ONS quality and methodology notes show the national distinction between self-rated good health and disability-free life expectancy. Accordingly, the phrase 'healthy life expectancy' must always be paired with its exact underlying definition.

Eurostat's 2023 EU results placed healthy life years at birth at an average 63.1 years, with 63.3 for women and 62.8 for men, using its activity-limitation indicator. These figures describe the European Union in the reference year 2023, published in August 2025. They are not 2026 LTC insurance claim durations. Their correct use is to document an international health-state measure and illustrate that total life expectancy and limitation-free expectancy can exhibit different sex gaps.

Sampling and classification

Surveys of private households may miss persons living in nursing facilities, changing measured severe disability prevalence. Hospitalization, assisted living and the presence of informal helpers may be recorded inconsistently. Survey response modes, proxy interviews, reporting of long-standing limitation, changes to question phrasing, mortality-linked attrition, and undercoverage of particular ages can bias comparisons. Insurance datasets, by contrast, may record only lives who purchased products and survived underwriting. The appropriate statistical response begins by disclosing the target estimand and designing corrections or uncertainty bounds rather than assuming exchangeability.

Different age structures also change crude prevalence, even when state-specific hazards remain constant. Rates should be age-sex specific before aggregation; cohort or period life tables and international standardizations must be labeled. Where respondent-level transitions are available, model-based estimates permit additional inference about incidence, progression and recovery. Where only independent cross-sections exist, a prevalence-based estimate should retain a clear limitation that duration and transitions are only partly identified.

Interpretation of an expectancy number

The reader needs seven descriptors: the health or dependency definition, mortality basis, population and sampling frame, age at which expectancy begins, calendar reference, statistical uncertainty and period/cohort character. A further descriptor is the relation, if any, to the contractual eligibility threshold. The scientific meaning of DFLE is established by these descriptors and the mathematics; a headline number by itself does not support insurer-level conclusions.

Morbidity Compression, Multistate Dynamics, and Competing Exits

Markov and duration-sensitive states

A useful actuarial representation distinguishes I (independent), M (moderate functional limitation), C (severe care dependency) and D (death). People can move from I to M, M to C, recover from M to I, improve from C to M, enter C directly from I, or die from any living state. In a time-inhomogeneous Markov model, let Q(a,t) denote an age- and calendar-time-dependent transition generator. For the row vector of state occupancy probabilities p(a), the Kolmogorov equation is

dp(a)da=p(a)Q(a).\frac{dp(a)}{da}=p(a)Q(a).

Off-diagonal Q entries are nonnegative transition intensities, and each diagonal entry is minus the row sum of outgoing intensities. A living-state occupancy integrated over future years gives expected time in that state. For an individual beginning in state i at age x,

Ei,x[TC]=∫0∞PiC(x,x+s) ds.E_{i,x}[T_C]=\int_0^{\infty}P_{iC}(x,x+s)\,ds.

The probability P_{iC} is conditional on the starting state, age, calendar conditions and transition assumptions, with death absorbing. This expectancy depends on the full transition structure, unlike a period Sullivan estimate that combines prevalence and an external mortality life table.

The Markov assumption may be implausible where duration since illness onset affects recovery or mortality. Care episodes following fractures, stroke, advanced dementia and progressive chronic disease have different dynamics. A semi-Markov specification or a model with duration-since-entry and cognitive status as states can represent such effects, subject to sufficient data. States should correspond to observable measurements and contractual triggers. Excess granularity relative to the sample generates unstable estimates; insufficient granularity conceals material differences in severity and payment.

A mathematical demonstration of non-identification

Consider two hypothetical stationary models with only independent state I and dependent state C. For transparent algebra, exclude death from this illustrative subsystem and use exponentially distributed times. Let the I-to-C entry intensity be λ and the C-to-I recovery intensity be μ. Its stationary dependent prevalence is

πC=λλ+μ.\pi_C=\frac{\lambda}{\lambda+\mu}.

Set System A to λ=0.04 and μ=0.36 per year, and System B to λ=0.01 and μ=0.09 per year. Both have exactly 10% dependent prevalence. When combined with an identical life-table survival schedule, both therefore produce the same Sullivan dependency-free proportion and expected dependent years. Yet the stationary rate of entry into C per total person-year equals (1−π_C)λ: 0.036 in A and 0.009 in B. A has four times as many episodes entering dependency. Mean uninterrupted time in C under the exponential assumption is 1/μ: approximately 2.78 years in A and 11.11 in B.

The values are purely illustrative; they are not estimates of real recovery from severe dementia, nursing-home entry or any LTC insurance population. The example deliberately excludes death and lapse. It proves an identification point: equal health-state prevalence and equal prevalence-derived DFLE do not determine claim incidence, claim duration, recurrence or the effect of a contractual elimination period. A benefit that commences only after substantial continuous dependency may respond very differently to the two hypothetical systems. Long-term care valuation therefore needs flow and duration information in addition to occupancy.

Death, recovery and lapse as competing processes

In real populations, dependence can end through improved function or death. At the policy level, lapse, benefit exhaustion and administrative closure may also terminate benefits. An observed termination dataset therefore needs cause-specific categories. The probability of recovery before death differs from a recovery intensity under competing risks. Censoring and withdrawal from a study require special care where poor health predicts missingness. A claim closure date may not be the last day of medical dependency.

Improvements in mortality before care onset create more surviving persons who remain exposed to later onset. Improvements in mortality after entering C may extend benefit duration. Postponing care onset reduces one component of liability, while longevity following onset can increase another. These channels should be measured jointly and decomposed in pricing and reserve stress tests.

Absolute and relative compression

Define e_x^C as expected remaining time dependent and e_x total remaining life expectancy at a specified reference age. Absolute compression corresponds to a reduction in e_x^C; relative compression corresponds to a reduction in e_x^C/e_x. A population can experience relative compression while absolute dependent years rise more slowly than total lifespan. Severe disability can compress while mild chronic illness expands. The source-specific results of Crimmins and Beltrán-Sánchez and Cai and Lubitz illustrate the need for explicit time windows and severity constructs. An insurer must ask which of these metrics affects its actual claim threshold, rather than convert a generic narrative about healthy ageing into a premium trend.

Long-Horizon Insurance Exposure, Benefits, Pricing, and Capital

From health state to eligible insurance payment

The financial exposure of a policy is generated by four conditions: the insured is alive and covered, enters or remains in a contract-defined eligible dependency state, satisfies any waiting and claim-assessment terms, and generates a cash entitlement or reimbursable service expenditure. Each condition has distinct evidence. The NAIC long-term care guidance summarizes the variety of private LTC contract provisions in the United States. Singapore's CareShield Life scheme uses severe disability measured by inability to perform three of six activities of daily living, under its specified assessment and payment rules. These functional triggers cannot be replaced by Eurostat's measure of any long-standing usual-activity limitation.

For a policy issued at age x, a general expected-present-value representation is

EPVx=∫0ω−xv(s)∑j∈Spjinsured(x,s)gj(x,s)bj(x,s) ds,EPV_x=\int_0^{\omega-x}v(s)\sum_{j\in\mathcal{S}}p_j^{\mathrm{insured}}(x,s)g_j(x,s)b_j(x,s)\,ds,

where v is a discount factor, p_j^{insured} is the modeled probability of being alive, in force and in state j at duration s, g_j specifies contract eligibility including its timing, and b_j represents the state-dependent payable benefit or eligible expense rate. The sum is over modeled living states. The formula is a standard expectation identity, not a newly invented pricing system. If benefit payment is path-dependent—because of elimination periods, recurrent episode rules, limits, inflation protection or remaining benefit pools—these path histories must be included in an augmented state space or simulated explicitly; the simple occupancy expression alone is then inadequate.

Insurance selection and persistence

A national dependency rate describes a general population. Insured purchasers reflect medical underwriting, price, wealth, education, insurance preferences, risk expectations and possible cognitive or family circumstances. Finkelstein and McGarry (2006) show that multiple dimensions of private information complicate simple adverse-selection tests in US LTC insurance. A low observed claim rate in insured lives cannot automatically be attributed to a healthier national population; underwriting and lapse may remove higher-risk or lower-risk lives from exposure in different ways.

The in-force exposure denominator is critical. A claim rate per policy-year at risk differs from a rate per all people aged 65 or more. At older durations, policies that survive lapse and mortality may be systematically selected. Insurers should model issue age, attained age, duration, benefit design, underwriting era and cohort effects. Pool experience can change even when underlying health hazards remain constant.

The SOA intercompany LTC database contains separate aggregate claim-incidence and claim-termination files for experience in 2000–2016; the supporting report describes definitions and confidentiality constraints. SOA basic-table research addresses long-term care incidence and termination experience by actuarially relevant segmentation. The SOA–LIMRA 2000–2023 request defines a later study's requested experience and product scope. A solicitation for updated data should not be represented as a completed, published set of validated new tables.

Care setting, benefit form and inflation

A person may need help at home, in a community program, in supported living or in a nursing institution. Eligible services, staffing requirements and allowed providers differ by contract and country. A fixed cash benefit might be paid when severe disability criteria are met, while indemnity and reimbursement contracts depend on service use, eligible cost, deductibles and daily or lifetime maxima. Family members may provide care unpaid, reducing formal utilization even when disability is severe. If informal caregivers become unavailable, paid demand can rise with no change in clinical disability prevalence.

The OECD care-affordability study documents the high potential household cost of sustained support and how severity affects affordability. Under a long-duration contract, care-specific wage inflation, scarce labor, service transformation and utilization intensity can materially alter payout even if dependent years are unchanged. Conversely, home modification can delay institutional care while maintaining substantial functional support. Financial projection requires separate modeling of health transition, service choice, payer eligibility and price.

Public insurance and payer boundaries

The Medicare long-term-care coverage guidance explains that US Medicare generally does not cover most custodial long-term assistance; certain skilled services have different rules. Medicare coverage, Medicaid means-tested long-term services, employer plans, private indemnity and family resources therefore occupy different portions of US care financing. Another country may cover a much wider range through compulsory social insurance. International liability estimates must respect these institutional financing borders and avoid counting a public benefit both as a private insured payment and as an independent social cost.

Reserving, interest rates and capital

A reserve depends on the conditional discounted value of future contractual payments less relevant premium income on the prescribed basis. Mortality, policy lapse, first incidence, recurrent disability, claim termination, escalation and investment discounting affect projected cash-flow timing and amount. A stressed scenario should vary onset and recovery, care-dependent mortality and inflation separately, and should test dependencies among them. A decline in first incidence combined with longer survival after entry need not reduce reserves. Similarly, higher care prices combined with limited provider supply can change actual utilization differently from the nominal benefit guarantee.

A national forecast of total care expenditure and the actuarial liability of an individual LTC insurer are different accounting objects. Both use health and ageing evidence but are conditional on different payer obligations, coverage selections, inflation assumptions and time horizons. This is the principal financial consequence of treating DFLE as a contextual input rather than a claim table.

Comparative Evidence, Data Validation, and Measurement Risks

International official evidence

OECD's 2024 affordability assessment found that close to one in four older people across the OECD countries studied had LTC needs under its defined assistance threshold and projected growing demand under demographic assumptions. Need prevalence is a stock of older people requiring assistance, not a new-claim probability. It includes people served informally, people with no insurance and persons below private benefit triggers.

OECD's 2025 access chapter reports that around 12% of older people aged 65 and above received some LTC across 31 reporting OECD countries in 2023. Certain country series cover restricted portions of care delivery; receipt cannot simply be subtracted from need to compute a universal gap. The denominator, assessment method, service types and coverage of institutions must be harmonized before any comparison.

The OECD workforce analysis documents an average of about five long-term care workers per hundred people aged 65 and above in its reporting sample for 2023, with extensive country variation. These statistics concern people employed in care relative to older residents, not individual insured claim entitlements. Nevertheless they explain why an insurance benefit's practical purchasing power depends on availability of services. Demand measured in health-state years may exceed deliverable care hours even if a public budget or private policy nominally provides coverage.

Measurement comparison

Source construct Measured object Potential insurance misinterpretation
WHO HALE Equivalent full-health years with severity weights Converting burden-adjusted years into severe-ADL claim years
Eurostat HLY Life-table years free of long-standing reported activity limits Interpreting broad limitations as contractual eligibility
ONS healthy expectancy Self-rated health-based expected years Treating a self-rating as a clinical dependency assessment
Longitudinal disability panels State sequences measured at survey visits Ignoring transitions between visits and survival attrition
Insurer claim incidence First or recurrent qualifying claims per insured exposure Treating a selected policy population as all older people
Insurer claim termination End of an insurer-defined claim episode Confusing administrative closure, recovery and death
LTC recipient counts People receiving defined support services Treating formal service receipt as all underlying need
Prospective LTC reserve Discounted liability under a contract and valuation basis Replacing cash-flow projections with a demographic indicator

This table is an analytical mapping based on WHO indicator documentation, Eurostat's series metadata, ONS measurement methodology and SOA's experience definitions. It is an interpretive aid, not an assertion that these sources use interchangeable categories.

Longitudinal evidence and causal caution

The US Health and Retirement Study demonstrates how a panel can observe recurring limitations and changes in health, wealth, family support and cognitive function. Its observation schedule is not a continuous monitoring system; onset time between two interview waves may be interval-censored, and death and institutional movement can cause selective attrition. Statistical correction requires explicit assumptions. An insurer dataset may contain claim dates and payments but only sporadic clinical observations; linking the two types of evidence requires a privacy-compliant linkage protocol and a documented common state definition.

A credible prevalence trend can reflect changes in incidence, survival, recovery, diagnosis, medical care, assistive technology or reporting behavior. Each mechanism has a different actuarial effect. Aggregate claims can also change because policy provisions, premiums or assessment practice changed. A regression of claim counts on national healthy-life expectancy without controlling such factors does not identify a causal effect of DFLE on insured costs.

Practical validation sequence

Measurement audit should verify the survey wording, dependency threshold, sample frame, residential facility coverage, age bands, mortality input and calendar period. Transition audit should compare model-implied state occupancies, onset entries, recovery counts, death by state and spell durations against observed data, not only match overall prevalence. Insurance audit should reconcile in-force exposure, underwriting, benefit triggers, waiting periods, claims closure causes, lapse and benefit form. Financial audit should test claim timing, care inflation, reimbursement conditions, discount assumptions and stress dependency.

Sensitivity analyses should separately vary changes in onset, mortality among dependent people, recovery, care-service uptake, cost inflation and contract eligibility. Parameter uncertainty and model uncertainty must be reported as such. When claim data are proprietary or sparse, give ranges or qualitative conditional scenarios rather than unsupported numerical tariff assertions. Historical public results should be used to scrutinize plausibility, while country-specific pricing must rely on legal text and relevant insured experience.

Bibliographic and geographic integrity

Each source should retain its year, population and publication type. The OECD cross-country average is not a global estimate. Eurostat's 2023 healthy-life-years table is not a 2026 insurer experience study. Society of Actuaries datasets represent reporting insurers and their specified product years, not all people living with disability. Claims on public programs are bound to the effective jurisdictional rules identified. In comparative research, retaining these identities is part of scholarly accuracy.

Integrated Findings and Research Agenda

Synthesis of established evidence

The first result is definitional: DFLE is expected time in a specified health state, not an expected number of paid claims. It gains scientific value when used with remaining total life expectancy, disability severity, uncertainty and a stated cohort or period convention. The distinction persists whether the data are national or insured.

The second result is mathematical: under a prevalence-based Sullivan construction, two populations can have exactly the same dependency fraction while exhibiting different entry frequencies and durations. The illustrative two-state calculation shows a fourfold difference in episode entry, with identical 10% prevalence. A long-term care insurer that prices only from the prevalence number cannot identify elimination-period losses, recurrent claims or mean payment duration. A validated multistate or duration-sensitive model supplies the missing structure.

The third result concerns morbidity compression. The historical mechanisms of Fries's compression hypothesis, Gruenberg's survival-expansion hypothesis and Manton's severity equilibrium generate different outcomes across disability thresholds. Consequently, falling severe-ADL dependency can coincide with longer survival and a growing burden of milder diagnosed conditions. The cited empirical literature records conditional evidence for the populations and periods actually studied, rather than one inevitable universal trend.

The fourth result is institutional: care need becomes funded support through families, markets and public programs, and becomes an insured payout only through contract eligibility and available benefits. Singapore's benefit threshold, Medicare's payer scope, NAIC policy descriptions and OECD affordability comparisons show the heterogeneity of legal and economic conversion steps. Health indicators must be translated through these steps before any cash-flow interpretation.

The fifth result concerns selection. The LTC private-information evidence cautions against transferring national health experience to privately insured populations without underwriting, preference and lapse adjustment. Insurer intercompany datasets, exemplified by SOA LTC experience reports, provide separate claim incidence and termination measurements and require their own exposure denominators and policy definitions.

The sixth result links financing to care supply. A benefit may pay cash with little formal service use, or formal care need may remain unmet despite nominal support because of constrained caregiving labor. The OECD workforce comparison makes that real-resource boundary visible. Expected financial liability and expected socially useful assistance are associated but do not coincide automatically.

A rigorous agenda for future empirical work

A useful next-stage research program would link age-specific population functional states, state-dependent death and recovery, family assistance, private LTC insurance entry and termination histories, and public benefit assessment. It should document informed consent, privacy, missingness, institutional sampling and the exact version of national eligibility criteria. Cohort projections should be contrasted with period health expectancies rather than silently substituted for them. Severity thresholds should be defined before interpreting compression.

Stratified validation should check older age groups, sex, social disadvantage, ethnic and geographic variation where lawfully collected and relevant, and institutional versus community care. Caregiver loss, dementia, recovery episodes, cognitive supervision and frailty create important heterogeneity. Models should disclose where restricted data prevent a reliable quantitative estimate and where comparative evidence only supports conditional scenario analysis.

Authorship and limits

The authorial synthesis specifies the transformations necessary to move from demographic healthy-life expectancy to insured liability and provides an internally reproducible mathematical counterexample to the inference that state occupancy determines insurance episodes. It also organizes measurement, transition, institutional, contract and cash-flow validation as an analytical sequence. All components draw on established demography, epidemiology, insurance economics and actuarial mathematics. No new doctrine, model family, empirical claim, or priority over earlier mathematical results is asserted.

The sources reviewed consist of research publications, official national and international health indicators, actuarial experience-study documentation, government benefits pages and bibliographic archives. The study does not include original patient-level datasets, insurer microdata, audited 2026 reserve estimates, or product-specific actuarial calibration. The mathematical rates are explicit examples only. These boundaries preserve scientific usefulness without inventing unsupported generality.

Conclusion

Dependency-free life expectancy partitions expected remaining population person-time by an operational functional threshold. It must be interpreted within its population, mortality schedule, disability definition, reference period and survey coverage. HALE, healthy life years, self-rated healthy life expectancy and insured benefit eligibility convey different scientific meanings and retain their own definitions.

The principal finding is that state occupancy alone does not determine claim onset, spell length, policy survival or contingent benefit payments. The hypothetical two-state example demonstrates mathematically how equal measured dependency prevalence can accompany different claim frequencies and durations. Mortality improvements, changes in disease severity, caregiver availability, underwriting, private insurance design and public payer eligibility influence that mapping independently. Appropriate insurance analysis therefore brings demographic evidence into a multistate, contract-specific, duration-sensitive and financially validated projection.

Historical hypotheses of morbidity compression, expansion and dynamic equilibrium remain empirically conditional. Their use in actuarial science requires a stated disability severity and reference period. The correct international research program combines verified primary statistics, longitudinal functional transitions, insured experience, legal benefit triggers and uncertainty-aware cash-flow projections. This monograph supplies an original integrated interpretation of those established requirements, while leaving product-specific numerical rates to research using the pertinent data and legal instruments.

References