Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Long-Term Care Transitions Across Home Community and Institutional Settings

Abstract. Long-term care is organized through sequences of assistance, assessment, admission, discharge, and financing rather than a single permanent placement. People may receive unpaid assistance in a private household, publicly financed home and community-based services, privately purchased home care, assisted living, or residential nursing care. Hospitalization, temporary rehabilitation, waiting lists, and death interrupt or terminate these sequences. Each movement changes the observed population at risk, the organization of labor, the distribution of expenditures, and the contractual circumstances in which an insurance benefit may become payable. This monograph examines transitions across home, community, and institutional settings as a joint problem in actuarial science, long-term care insurance, health services research, public finance, and the legal organization of care. It establishes a disciplined distinction among functional need, care setting, formal service use, benefit eligibility, claim status, and cash flow; evaluates continuous-time multistate and semi-Markov estimands; explains competing risks, informative observation, and family-provider substitution; compares the United States, Germany, the Netherlands, and Singapore; and derives practical validation requirements for insured benefit projections. A worked numerical example demonstrates how setting-specific payments depend on transition timing and benefit rules. The synthesis finds that transitions can be understood only when risk sets, duration, care intensity, service capacity, household contributions, and payment entitlements are observed together. Published administrative and clinical evidence supports parts of this framework, while proprietary insurance and unpaid-care data remain material limitations. The contribution is an original analytical integration of established methods and verified institutions, without a claim to an independently discovered theory or causal estimate.

Keywords: Long-term care; care transitions; home and community-based services; institutional care; nursing homes; informal caregiving; multistate models; semi-Markov processes; competing risks; insurance claims; functional dependency; benefit eligibility; actuarial valuation; Medicaid; long-term services and supports; care coordination; model risk.

Introduction

Long-term care (LTC) comprises assistance and support required when functional, cognitive, or health-related limitations constrain ordinary living over an extended period. The World Health Organization's International Classification of Functioning situates functioning within the interaction of health conditions, activities, participation, and environmental factors. For insurance research, this framework is indispensable because an observed nursing-home admission cannot by itself reveal the severity of disability, the accessibility of housing, or the amount of support supplied by relatives. An identical functional limitation can result in distinct care arrangements when housing, staffing, and entitlements differ.

Institutional care may provide continuous supervision and intensive services; home care may preserve established community relationships; community programs can replace selected hours of household assistance; and hospital stays may temporarily interrupt long-term care. These arrangements often overlap. The National Association of Insurance Commissioners' long-term care insurance overview describes coverage that may reach home health services, adult day care, assisted living, and nursing facilities. Accordingly, a scientific account of transitions needs a time-indexed record of where the person lives, who supplies assistance, which service is received, and who pays. A classification of location alone is insufficient.

The United States offers a visible illustration. The Centers for Medicare & Medicaid Services (CMS) reports that the Money Follows the Person program supports eligible people moving from institutions into community living. In a September 2026 update, CMS reported 144,651 cumulative transitions under that demonstration from 2008 through the end of 2025. This administrative count documents program-supported transitions; it does not estimate a universal probability that an institutional resident can or will return home. The distinction between a counted transition, a risk-set denominator, and a causal effect underlies the present investigation.

The subject is also financially material. The OECD's 2024 LTC affordability analysis shows that assessed care costs can absorb very large fractions of older people's incomes, while public support varies substantially by country, functional severity, and care arrangement. Private insurers, public programs, households, and providers therefore experience the same episode through different accounting systems. Premiums or public contributions are collected in one period, eligibility may be assessed later, and benefit payments follow an often discontinuous trajectory through multiple settings.

The research question is precise: how should changes among household support, community services, and institutional care be defined, measured, explained, and translated into benefit outgo while preserving differences between clinical needs and legally payable obligations? The method is a documentary and analytical synthesis of official rules, administrative reporting, actuarial experience-study documentation, and peer-reviewed empirical studies available as of November 14, 2026. Each empirical example retains its own population, place, period, and observational design. The formal calculations are derived illustrations and have not been fitted to policyholder microdata.

Conceptual foundations and historical development

The object of a transition

A transition is a change from a defined origin state to a defined destination state over a specified observation interval. A household receiving unpaid help is one observable arrangement; an individual living at home with paid personal assistance is another; an assisted living residence, nursing facility, hospital, and death are additional possibilities. The same person can experience a functional deterioration without moving residence, or move residence without an immediate change in dependency. Consequently, the state needed to answer a clinical question may differ from the state needed to price an insurance product.

For analysis, let the person's underlying condition be represented by functional state F(t), residence R(t), formal service package S(t), informal-care input U(t), legal eligibility E(t), claim status C(t), and payer P(t). These are distinct variables. Their joint observation is the scientific target. An insured claim state is generally a function of qualifying evidence, contract conditions, proof, waiting periods, coordination provisions, and administration; it should not be substituted for F(t). Likewise, residence may be a home, assisted living residence, or institution, while payment may be cash, reimbursement, or direct provision and may cross residential categories.

An operational taxonomy should specify at least: (a) private residence with unpaid care only; (b) private residence with formal home care; (c) community service participation such as day care or respite alongside a home residence; (d) assisted living or another residential care arrangement; (e) nursing facility or another high-intensity institution; (f) acute hospital care; and (g) death. Community programs can be simultaneous with home residence and therefore need service flags or factorial states rather than mutually exclusive labels. For some datasets a practical five-state representation—home, hospital, waiting for long-term care placement at home, institutional LTC, death—can be more identifiable than a highly detailed state space.

Dufour and colleagues (2026) implemented precisely such a five-state transition model for 3,424 older home-care recipients in Quebec, Canada. The study distinguishes waiting for LTC placement from living at home without that designation and identifies differences in service intensity. Its model supplies a concrete example of why waiting-list status can alter the meaning of apparently similar home-based experience. Its estimates are local to a Quebec administrative cohort rather than private-insurance transition probabilities.

Functioning, need, service, and payment

Functioning describes ability and participation; assessed need translates functional limitations into a care requirement under a specified instrument; service use records actual assistance; and covered expense identifies a payment basis. They can diverge for four reasons. Assessment thresholds may differ; services can be unavailable; care can be provided without a financial transaction; and the contract may limit eligible settings or benefits. The WHO ICF classification provides a common language for functioning but does not establish an insurer's particular benefit trigger.

This distinction has a legal dimension. In the United States, Internal Revenue Code section 7702B specifies conditions for qualified LTC insurance and qualified LTC services. The definition includes the inability to perform at least two activities of daily living for at least 90 days, subject to required certification, or substantial supervision due to severe cognitive impairment. The specified activities include eating, toileting, transferring, bathing, dressing, and continence. The statutory definition should not be read as the universal entry threshold for public programs or every nonqualified private contract. A benefit can be triggered by eligible dependency even when care is delivered at home; a care setting transition may affect the payment amount while leaving trigger status unchanged.

In public financing, a different boundary appears. Medicare's long-term care guidance explains that nonmedical custodial LTC is generally outside Medicare coverage. Medicaid can finance institutional care and specified home and community-based services subject to applicable eligibility and state-program rules. The distinction matters when a patient leaves a hospital: a medical discharge, the end of covered post-acute skilled services, and the commencement of prolonged custodial assistance are separate events with separate payers.

Historical institutional differentiation

Modern long-term care systems have developed different combinations of social insurance, municipal support, needs-tested public assistance, and voluntary private insurance. Germany introduced mandatory long-term care insurance as an independent social insurance branch in 1995; the federal health ministry's official description of care insurance emphasizes care grades, choices between professional assistance and cash support for informal arrangements, and the partial-benefit character of statutory coverage. This institutional architecture means that two people with equivalent functional needs may have different cash entitlements depending on care grade and type of service.

The Netherlands distinguishes municipal social support, health-insurance-financed district nursing and personal care at home, and intensive care under the Long-Term Care Act (Wlz). The government's long-term care at home explanation assigns responsibilities to municipalities, health insurers, and central government. The Wlz eligibility guidance describes assessment through the Care Needs Assessment Centre and allows eligible individuals to remain at home when care there is practicable and cost-efficient. Thus the legal entitlement and the residence state need not coincide.

Singapore's CareShield Life offers a further organizational contrast. The Central Provident Fund Board describes cash benefits after severe disability using an inability to perform three of six specified activities of daily living and a monthly payout that does not depend directly on reimbursing a nursing-home invoice. It follows that the transition from home to an institution need not produce an equivalent step change in the insurer's cash payment. Across these countries, measurement requires recording benefit design before assigning meaning to a care setting change.

Rights and institutional choice

The United Nations Convention on the Rights of Persons with Disabilities, Article 19 on community living, recognizes choices of residence and access to support. The U.S. Medicaid home and community-based settings rule similarly places emphasis on integration, privacy, autonomy, and individual choice in relevant HCBS programs. A transition study concerned with care quality therefore must measure choice and suitability of placement alongside the recorded destination. A lower institutionalization rate can correspond to better supported community living; it can also reflect delayed placement when appropriate home services are unavailable. The observed rate alone cannot distinguish those mechanisms.

Care setting transition intensity

Risk sets, clocks, and estimands

For a person in state i at time t, a transition intensity to state j can be written

qij(t∣X(t))=lim⁡Δt↓0Pr⁡{Z(t+Δt)=j∣Z(t)=i,X(t)}Δt,i≠j.q_{ij}(t\mid X(t))=\lim_{\Delta t\downarrow 0}\frac{\Pr\{Z(t+\Delta t)=j\mid Z(t)=i,X(t)\}}{\Delta t},\qquad i\ne j.

Here Z(t) is a properly specified mutually exclusive operational state, X(t) contains measured covariates, and an intensity is a conditional rate per unit of person-time. If a service flag can be present alongside multiple residences, the model must expand the state vector or treat the service as a time-varying covariate. For a continuous-time Markov chain with generator Q, diagonal entries are minus the sum of outgoing intensities and the transition probability matrix over elapsed time s is exp(Qs) when Q is constant. The diagonal identity is an accounting requirement, not an empirical finding.

The origin risk set is essential. The institutional-to-home rate uses time spent in the relevant institution by individuals eligible and observable for that destination. A hospital-discharge count cannot be divided by community-resident person-years and interpreted as an institutional-discharge intensity. Similarly, annual program transition counts cannot be translated to a probability without knowledge of eligible exposure, the qualifying waiting period, and competing exits.

A conditional cumulative transition probability is obtained through the full multistate process. In a simple two-exit state i, with constant cause-specific intensities a for transition to j and d for death, the probability of entering j before time T is

Pij(T)=aa+d(1−e−(a+d)T).P_{ij}(T)=\frac{a}{a+d}\left(1-e^{-(a+d)T}\right).

This expression demonstrates that a cause-specific transition rate and an observed cumulative proportion are different quantities. Death removes people from the risk set; an apparent reduction in institutional admissions can arise because competing mortality increased. A valid analysis reports both the cause-specific intensity and the competing-risk cumulative incidence, with time origins and follow-up explicitly defined.

Illustrative competing-risk calculation

Consider a purely illustrative cohort of 1,000 insured people already receiving formal home care. Assume the next transition is either institutional admission at annual intensity 0.12 or death at annual intensity 0.08, with constant intensities for one year and no other exits. The one-year probability of institutional admission is 0.12/0.20 times (1 − exp(−0.20)), approximately 0.10876. The death probability is approximately 0.07251; remaining in the home-care state has probability exp(−0.20), approximately 0.81873. The expected counts are approximately 109 institutional admissions, 73 deaths, and 819 people remaining, subject to rounding. Dividing 120 theoretical admissions by 1,000 as if everyone were exposed for a full year would conflate an intensity with a realized one-year incidence. These values illustrate arithmetic only and are not empirical experience rates.

If mortality rises to 0.16 while admission intensity stays 0.12, the one-year institutional-admission probability becomes approximately 0.10466. A smaller incidence proportion under those assumptions does not mean lower institutional-admission propensity among survivors. Changes in transition counts must therefore be decomposed into exposure, baseline severity, mortality, provider access, and the definition of the destination.

Estimation from longitudinal records

For a transition i→j, the observed counting process N_ij(t) records qualifying transitions, and Y_i(t) indicates whether a person belongs to the observed origin risk set. A cause-specific intensity model takes E[dN_ij(t) | history] ≈ Y_i(t) q_ij(t|X(t))dt under the model's observation assumptions. Estimators can use piecewise constant exposure cells by attained age, duration, severity, calendar period, and care setting; competing-risk survival methods; or multistate regression. In practice, analysts must document whether dates refer to clinical need onset, first service, signed assessment, facility occupancy, claim approval, or first payment. Each date produces a distinct observable process.

An important actuarial source is the Society of Actuaries' 2000–2016 LTC aggregated database documentation, which describes separate claim-incidence and claim-termination files and their data definitions. These files can inform experience analysis of insured claims, but a claims study does not automatically capture every change of care residence, unpaid support arrangement, or unsuccessful eligibility application. The SOA's 2018 incidence trend research identifies observational and actual-to-model approaches to calendar-year trend analysis. Trend findings depend on the studied contract cohorts, underwriting, observation windows, and revisions to claims practices.

The SOA and LIMRA also announced a 2000–2023 LTC experience data request to update industry evidence on stand-alone policy incidence, termination, and utilization. The request itself is a data-collection initiative, not a finalized 2000–2023 public result. It explicitly excludes hybrid and combination products from that collection scope. This illustrates why actuarial evidence must be labeled by the exact population and stage of publication.

Duration dependence and incomplete observation

A homogeneous Markov model assumes that the next transition depends on the current modeled state and covariates, rather than unrecorded time spent in that state. This can be restrictive. Admission intensity may rise after several months of high-intensity home care, while readmission risk after institutional discharge may be unusually high early in the community. A semi-Markov approach permits q_ij(t,u), where u is duration since entering the current state. A tunnel-state representation can discretize duration if a given software system supports only Markov transitions. The choice should follow empirical duration diagnostics and the intended valuation use.

Hospitalization creates another measurement challenge. Some administrative datasets record a move from home to hospital and then hospital to nursing facility; others record the same interval as one uninterrupted LTC claim with a change of service provider. Dufour et al. (2026) report that the Quebec cohort's observed transitions included frequent movement involving hospitals and that people waiting for LTC placement received lower-intensity home services. The results support a richer operational history but do not establish that a waiting list caused a particular subsequent admission.

Left truncation occurs when observation starts after care began; interval censoring occurs when the setting is known only at periodic assessments; informative dropout occurs when records disappear due to coverage termination, disenrollment, or relocation; and administrative censoring occurs at the study endpoint. Cases observed through an insurer may overrepresent claims that satisfy particular thresholds and underrepresent people using care without submitting benefits. Sensitivity analysis should reconstruct alternative plausible dates and missing-service scenarios. Simply treating unseen intervals as unchanged residence artificially inflates state persistence.

Population and calendar comparability

The CMS 2022 LTSS user and expenditure bulletin reports that 86.6% of Medicaid LTSS users received HCBS and 64.6% of LTSS spending was for HCBS in 2022. These are user and spending shares in a specific public program, not transition intensities, and some beneficiaries used both institutional and community services during the year. The 2026 MFP transition report notice, by contrast, counts transitions under a demonstration. Pooling the two types of statistic without a common eligibility denominator would create an incoherent estimate.

A comparable international analysis reports definitions of residence, care duration, severity, service supply, denominator age, payment system, and assessment method. Calendar changes in public eligibility, private claims administration, or coding can mimic health trends. Calendar-year coefficients have interpretive value only after those sources of compositional change are examined.

Provider and family substitution

The production of care

Care is jointly produced by formal workers, household caregivers, adapted housing, technologies, and public services. A person's effective support at home depends on the timing and quality of each contribution. An hour of unsupervised household presence is not necessarily equivalent to an hour of skilled nursing or structured respite; a paid service may substitute for a task, complement existing household support, or make a previously impossible home placement feasible. Family caregivers also undertake scheduling, monitoring, transportation, advocacy, and continuity work that simple service-hour datasets often omit.

Let H_f(t) denote hours of formal personal assistance, H_u(t) reported unpaid task-specific hours, and V(t) an indicator for specialized nursing or other professional care. They are exposure descriptors, not automatically additive units of homogeneous care. A useful empirical comparison fixes a defined care task—such as assistance with bathing—then compares changes in formal and unpaid provision within a cohort with similar baseline needs. Without a fixed task and severity measure, a fall in unpaid hours following formal service expansion may reflect substitution, deterioration, shorter observation, or a shift to institutional care.

The OECD's care affordability study documents substantial differences between compensation for informal and formal care in the social protection systems it examines. A household with strong unpaid support can face lower out-of-pocket expenditures while bearing substantial time costs and forgone employment. The recorded insurer or government claim is one component of the full resource burden. Valuation of unpaid labor should distinguish replacement cost, opportunity cost, and reported hours; different valuation choices answer different questions.

Three substitution mechanisms

The first mechanism is task substitution. Paid home aides may provide direct ADL assistance previously supplied by household members. Its observable signature is a task-specific decline in unpaid assistance after formal provision rises, conditional on comparable need and observation. Its counterfactual challenge is that formal hours are often assigned precisely when need increases, generating a positive raw association even when the causal effect of formal assistance on family hours is negative.

The second mechanism is complementary coverage. Professional nursing can stabilize a complex care plan while relatives continue supervision and coordination. The empirical signature can be simultaneous increases in formal and informal hours, improved adherence, or reduced avoidable acute episodes. Such correlations cannot establish clinical effectiveness without credible confounding control. A substitution coefficient should be estimated separately for personal care, skilled care, transportation, supervision, and administration rather than imposing one exchange rate between total formal and informal hours.

The third mechanism is institutional substitution. A stable home arrangement may replace some residential-care days if the person's needs, housing, caregiver capacity, and formal services allow it. Conversely, an institutional admission can displace household care while still requiring family visits and advocacy. Li, Cai, Kim, and Kim (2024), using Health and Retirement Study observations, examined informal care inside U.S. nursing homes. The study demonstrates empirically that informal assistance can continue within institutions, an important reason to reject a binary coding of institutional residence as 'no family care.' Its observational design is not a causal estimate of the effect of Medicaid status or residence on family effort.

Selection, preferences, and causal identification

Households select care arrangements under resource, health, and social constraints. Dependence severity can increase both formal service demand and family assistance. Available children, household composition, housing accessibility, labor market attachments, local provider supply, and preferences affect the setting reached. Residential status is therefore an outcome of selection as well as a predictor of spending.

A causal substitution study must identify the source of variation in formal services or benefit generosity. Potential approaches include a policy eligibility discontinuity, rollout difference, or instrument that affects care supply without independently affecting the outcome; each requires a credible design-specific identification argument. For example, geographic care-worker availability may correlate with housing costs, urbanicity, and family labor markets, so it cannot simply be assumed exogenous. Difference-in-differences requires evidence about pretreatment trends and concurrent reforms. Regression adjustment on functional severity alone cannot eliminate unobserved household preferences or crisis-driven selection.

Long-term care insurance purchasing itself exhibits selection. Finkelstein and McGarry's (2006) study of private insurance information shows that private expectations of risk and preference for insurance can exert different influences on observed insurance purchase and subsequent outcomes. Brown and Finkelstein's (2007) analysis of the small LTC market identifies pricing and benefit limitations alongside demand-side explanations. These papers concern insurance selection and market design, not direct measurements of household-care substitution; their relevance here is methodological. An insured cohort cannot be treated as a random sample of all people needing care.

Capacity and care continuity

Provider capacity constrains which transitions can happen, how quickly, and with what consequences. In Health at a Glance 2025, the OECD reports that the number of long-term care workers averaged about five per 100 people aged 65 and over across 31 reporting countries in 2023. A benefit promise to reimburse home care may therefore yield different actual utilization in a well-served district and one with shortages. For insurers, lower realized home-care claims can indicate unmet supply rather than improved dependency.

The CMS MFP transition program description identifies housing coordination, home modifications, equipment, and transition support among practical measures to enable community living. A return-home transition often requires a one-time investment and a stable subsequent service package. Studies limited to recurring daily care costs may omit the expenditure that made the transition feasible. Conversely, an initially successful discharge is not sufficient evidence of durable community residence when readmissions and emergency use are unobserved.

Distributional consequences

A setting change redistributes time, payment, and risk. A move from nursing care to a home arrangement may lower a public or insurer's direct facility payment while increasing rent, utility, home modifications, paid transport, or household caregiver time. The same change can expand the recipient's autonomy and preserve relationships. Analytical evaluation should identify whose perspective is being used: insurer claims, total public expenditure, household consumption, societal resource use, care quality, or the recipient's preferred living arrangement. These objectives are not measured in interchangeable monetary units.

Gender, age, socioeconomic position, disability severity, and geography influence care opportunities and household labor supply. The empirical task is to report subgroup exposure and outcomes and to explain the limitations of group comparability. A socially attractive average shift toward community care may conceal burdens on households with little capacity to substitute unpaid care. The ethical and legal relevance of personal choice, recognized in Article 19 of the CRPD, should be incorporated into the outcome definition rather than treated as a narrative appendix.

State-dependent benefit outgo

The bridge from care histories to cash flows

Actuarial benefit outgo is a contractual mapping from time-indexed events and claims characteristics into payment. Let b_s(t,d,k) denote the payable benefit per unit time in service or claim state s at time t, conditional on episode duration d and policy features k. Let p_s(t|x) be the probability that a covered insured of profile x occupies that payable state at t. Ignoring for the moment claim expenses, taxes, and lapses, an illustrative expected present value is

EPV(x)=∫0Tv(t)∑s∈Spayps(t∣x)bs(t,d(t),k) dt,EPV(x)=\int_0^T v(t)\sum_{s\in\mathcal{S}_{pay}}p_s(t\mid x)b_s(t,d(t),k)\,dt,

where v(t) is a discount factor and the occupation probabilities reflect eligibility, transition timing, competing death, and duration. When benefits depend on complete history—for example, a cumulative benefit pool or elimination-period day count—state s must be expanded to carry the relevant history, or the calculation must simulate paths. A mere residence-state occupation probability is insufficient for history-dependent contracts.

Distinguish four quantities: gross resources consumed by care; services billed by providers; allowed expenses under a public or private scheme; and cash actually paid after limits, coordination, deductibles, and waiting rules. These may differ substantially, especially where a caregiver is unpaid or an insurer pays a fixed cash indemnity. Cost-of-care studies cannot be inserted directly as claim benefits without this mapping.

The NAIC's LTC insurance consumer guidance identifies the relevance of coverage categories, benefit amounts, and contractual features. The U.S. tax-qualified LTC framework regulates specified eligibility conditions, but tax qualification alone does not dictate every policy's daily benefit, reimbursement percentage, or inflation feature. Contract wording, issue cohort, jurisdiction, and amendment history therefore belong in the valuation record.

Payment designs and setting transitions

A reimbursement policy may pay actual eligible covered expenses up to setting-specific limits, perhaps following a stated elimination period. A fixed cash-benefit policy may pay a specified amount when eligibility is satisfied, with little immediate dependence on care residence. A pool-of-money design imposes a cumulative payment ceiling, making an early institutional spell relevant to later home-care benefits. A partnership or public coordination mechanism may change the payer sequence as eligibility evolves. In Germany, the partial-benefit design and care-related choice between cash and professional support create further benefit categories; in Singapore, CareShield Life cash payouts illustrate the separation between care setting and a cash entitlement.

A transition from home to residential care may therefore increase the payable limit in one product, leave it unchanged in another, or terminate a particular home-care benefit while introducing a facility benefit. A transition back home may restart an elimination period under some wordings, or carry forward an existing episode under others. No general algebraic sign can be assigned to the effect of institutional admission on net benefit outgo without a specified contract and history.

Worked benefit-outgo example

Consider a hypothetical 12-month episode under a reimbursement contract. Assume an eligible member receives formal home care for four months and covered residential care for eight months. For simplicity, months are exactly 30 days, the contract has no deductible or elimination period, claims are accepted, and there are no discounts or coordination offsets. Home care's covered billed amount is $120 per day, with a benefit maximum of $100 per day; residential care's billed amount is $240 per day, with a benefit maximum of $180 per day. The total payment is (4 × 30 × $100) + (8 × 30 × $180) = $55,200. The provider invoices total (4 × 30 × $120) + (8 × 30 × $240) = $72,000, leaving a $16,800 gap before any other resources or public payments.

If the same member instead remains in home care for eight months and spends four months in residential care, with otherwise identical assumptions, contractual payments are (8 × 30 × $100) + (4 × 30 × $180) = $45,600. The difference is $9,600 in this contrived example. It reflects the assumed per-day setting limits and days occupied; it is not an estimate of economic savings, clinical quality, or the likelihood of either pathway. Additional home-care hours, travel, supervision, housing adjustment, or unpaid work could change the total social cost even if the insurer pays less. If a single $140 daily cash benefit applied to every eligible day, the twelve-month payment would be $50,400 under either setting sequence, subject to the assumptions.

A second illustrative intervention pays $2,400 once for home adaptations and transition coordination. If it replaces 30 residential days at an insurer payment of $180 with 30 home-care days paid at $100, the difference in recurring benefits is $2,400 and exactly offsets that one-time payment in the first 30 days, before discounting. A longer stable community spell could alter the lifetime balance, but only a model including readmission, mortality, unmet care, and claims eligibility can assess it. This calculation illustrates a break-even identity and supplies no claim of causal program effectiveness.

Severity, incidence, persistency, and reserves

A realistic projection needs a distribution over care needs, claim incidence, claim termination, mortality, transitions, and policy persistence. The SOA's LTC intercompany database guide supplies institutional evidence on claim-level data definitions, while its 2018 trend methods address temporal incidence analysis. Neither automatically provides an identified transition-intensity matrix across every physical care setting and every unpaid-care arrangement. An insurer must establish whether its observed claim data can support residence-specific utilization assumptions or whether external information is needed.

The capital and reserve consequences depend on what changes. Greater home-care duration can reduce average daily benefits in a reimbursement design while extending claim duration. Expanded coverage can increase incidence by making benefits useful earlier, even if a fraction of residential days is avoided. Delayed institutionalization can shift costly claims to later attained ages or extend the period during which payment obligations remain active. Mortality assumptions interact with all of these effects. Duration-specific adequacy should be tested using present values, distributional tails, and stress scenarios, not solely an undiscounted average annual cost.

For a cohort projection, model reserves at valuation time t from expected future payable cash flows under the contract and applicable valuation basis, allowing for future premiums where relevant. The exact statutory or accounting reserve rule is jurisdiction- and product-specific; the present study addresses the economic building blocks rather than representing this generic formula as an authoritative regulatory calculation.

Public costs and private coverage in comparative context

The OECD's 2025 LTC spending comparison reports that the estimated cost of severe-needs home care is, on average in its specified sample, markedly greater than the median income of older people; institutional care also exceeds that income in the country observations considered. These modeled standardized packages are not insurer-specific claims or national average transaction prices. The 2024 OECD affordability study uses comparable hypothetical need profiles to analyze public support and out-of-pocket burden. The studies make a persuasive case for reporting both cost and financing but cannot determine an individual policy's reimbursement amount.

In the United States, a public expenditure rebalancing ratio is often used to describe the proportion of LTSS financing directed to HCBS. The CMS 2022 Medicaid ratios combine different service populations and payment mechanisms; they are not a causal measure of service substitution, nor a private-policy reserve basis. In the Netherlands, the Wlz coverage framework combines intensive care eligibility with residence options. Cross-country financial comparison is therefore meaningful only after defining whether the metric is payer expenditure, household outlay, modeled need cost, or total resource consumption.

Integrated findings and research agenda

Synthesis of the evidence

The documentary and methodological analysis supports six conclusions. First, setting transitions are jointly determined by care needs and the institutions that translate needs into services. Residence, clinical condition, service use, and payer liability require separate observation. Second, multistate transition intensities provide a coherent time-at-risk language, but cumulative movement also reflects competing death, censoring, waiting periods, and observation design. Third, informal care continues across settings and can substitute for some tasks while complementing others. Fourth, insurance payment depends on contract design and path history, which explains why the same physical transition can have different financial effects. Fifth, comparative social protection systems define different eligibility and payment categories, making direct international rate comparisons conditional on explicit crosswalks. Sixth, reported rebalancing toward community care has to be interpreted with provider capacity, household burden, supported choice, and unmet needs.

These findings rest on distinguishable evidence. The WHO functioning classification supports the conceptual separation of functioning from environment; the Quebec multistate cohort supplies empirical transition evidence; the CMS LTSS and MFP materials establish program concepts and observed transitions; and the SOA experience database documentation supplies insurance-specific observation rules. The original contribution here is the explanatory connection among these evidence types and their consequences for actuarial measurement. No new formal theory or independently estimated parameter is claimed.

An actionable research design

An empirically stronger next study would link longitudinal assessments, residence episodes, hospitalizations, formal service visits, unpaid-care diaries, insurer claims, public payments, and death registration for a well-defined inception cohort. Its first output should be an audited state and event dictionary and a flow chart of inclusions, losses to follow-up, and exposure. Analysts would then estimate transition intensities and cumulative probabilities by severity and duration, separately examine paid and unpaid task substitution, and project policy-specific cash flows using contract wording. A prospective participant component could document the person's preferred living arrangement and whether formal alternatives were actually available.

Such a design should report total institutional and home-care episodes, transitions in both directions, readmission after home discharge, unpaid hours, hospital days, quality-of-life measures, and the entire payment distribution. Care setting should be treated as a trajectory subject to shocks, not as an absorbing administrative label unless the observed data warrant that assumption. Time-varying confounders need specific causal strategies when an intervention effect is sought. If linkage is unavailable, smaller single-domain studies can provide credible estimates within their stated boundaries rather than an artificially complete unified result.

Actuarial research priorities

For pricing and reserving, the most valuable missing estimates are joint distributions of functional severity, place of care, waiting periods, claim eligibility, paid benefit level, and duration under contemporary contract designs. Data should be stratified by issue cohort, attained age, calendar period, underwriting regime, and benefit options, with clear treatment of lapses and deaths. The SOA's current experience-development program indicates continuing work on claims incidence, termination, and utilization. A separate residence-transition supplement would require valid links between service events and insurer claims rather than inference from claim files alone.

Formal sensitivity should explore at least the following competing trajectories: early moderate home care that prevents or postpones institutional admission; extended severe home care with escalating professional input; sudden hospitalization followed by facility admission; a successful facility-to-community return with one-time adaptation costs; an attempted return followed by readmission; and long episodes ending in death. Each scenario needs its own clearly labeled hypothetical assumptions and the same benefit contract for comparisons. For public programs, the perspective shifts to fiscal expenditure and social objectives; for private insurers, policyholder risk and contracted liabilities remain central.

Implications for scientific and public interpretation

A policy that raises the share of HCBS may expand access and choice, shift spending categories, or change the balance of unpaid labor. Its causal effects and distributional consequences depend on the setting and evidence. The OECD workforce comparisons identify staffing as a relevant constraint. The UN community living provision establishes an equally relevant perspective on supported residential choice. Scientific evaluation gains explanatory power when these dimensions are made operational rather than treated as mutually exclusive policy slogans.

The long-term value of transition research lies in maintaining separate ledgers for need, service, location, family contribution, and financial liability and then explaining their relationships. This produces models that can be tested against actual trajectories and rules. It also makes visible why a setting-specific actuarial benefit projection, a public rebalancing statistic, and a household's account of care may all describe the same person accurately while yielding different numbers.

Conclusion

Long-term care transitions are observable changes in living arrangements, services, and institutional responsibility that unfold within a continuing trajectory of functional need. Their actuarial significance follows from the exact risk sets, transition intensities, competing exits, duration dependence, benefit eligibility, and contract-specific payments that govern each episode. Their social significance follows from care quality, accessible support, family time, affordability, and the recipient's supported choice. The empirical record confirms the importance of distinguishing these dimensions and demonstrates the feasibility of multistate analysis in defined administrative cohorts. It also exposes the limitations of translating public expenditure shares or insurance claims into universal statements about movement between home and institutions.

The decisive scientific requirement is a validated link between observed care histories and the relevant financial or social outcome. Where that link is available, transition modeling can improve the interpretation of benefit outgo, public financing, and service-system change. Where it is absent, the correct result is a bounded estimate with an explicit measurement domain. This monograph provides an integrated methodological and comparative account based on established research and verified original sources, with a documented agenda for the linked data and causal designs necessary for further inference.

References

  • World Health Organization (WHO) (2001; living reference). International Classification of Functioning, Disability and Health (ICF). WHO classification and implementation materials. Source.

  • National Association of Insurance Commissioners (NAIC) (n.d.; accessed 2026). Long-Term Care Insurance. Consumer education and insurance product overview. Source.

  • Centers for Medicare & Medicaid Services (CMS) (n.d.; accessed 2026). Money Follows the Person (MFP). Medicaid demonstration program overview. Source.

  • Centers for Medicare & Medicaid Services (CMS) (2026). Money Follows the Person Demonstration: Report on Transitions. Medicaid bulletin, September 17, 2026, reporting data through December 31, 2025. Source.

  • Organisation for Economic Co-operation and Development (OECD) (2024). Is Care Affordable for Older People? Cost-Sharing and the Current Adequacy of LTC Systems. OECD Publishing, chapter on cost sharing and affordability. Source.

  • Dufour, Isabelle, et al. (2026). A Multi-State Transition Model Among Older Adults Receiving Home Care Services: A Population-Based Cohort Study. BMC Geriatrics 26, article 365. DOI: 10.1186/s12877-026-07150-1. Source.

  • United States Congress (current codification, accessed 2026). 26 U.S.C. § 7702B — Treatment of Qualified Long-Term Care Insurance. United States Code, Internal Revenue Code. Source.

  • Centers for Medicare & Medicaid Services (n.d.; accessed 2026). Long-Term Care. Medicare.gov, official coverage information. Source.

  • German Federal Ministry of Health (Bundesministerium für Gesundheit) (n.d.; accessed 2026). Long-Term Care Insurance. Federal ministry overview, social long-term care insurance. Source.

  • Government of the Netherlands (n.d.; accessed 2026). Long-Term Care at Home. Government.nl, care and support at home. Source.

  • Government of the Netherlands (n.d.; accessed 2026). Long-Term Care Act (Wlz). Government.nl, nursing homes and residential care. Source.

  • Central Provident Fund Board (Singapore) (n.d.; accessed 2026). CareShield Life. Singapore government long-term care insurance program. Source.

  • United Nations (2006; living treaty reference). Convention on the Rights of Persons with Disabilities, Article 19. UN treaty Article 19: Living Independently and Being Included in the Community. Source.

  • United States Department of Health and Human Services (current regulation, accessed 2026). 42 C.F.R. § 441.301 — Contents of Request for a Waiver. Federal Medicaid Home and Community-Based Services regulation, via Cornell LII. Source.

  • Society of Actuaries (SOA) (2020). 2000–2016 Long-Term Care Aggregated Databases. SOA experience-study documentation and linked aggregate tables. Source.

  • Society of Actuaries (SOA) (2018). Researching Long-Term Care Insurance Incidence Rates Over Time. SOA Research Institute, methods report. Source.

  • Society of Actuaries (SOA) and LIMRA (2025). 2000–2023 Long-Term Care Experience Data Request. SOA research opportunity and data request. Source.

  • Centers for Medicare & Medicaid Services (CMS) (2024). CMS Releases New Publications on Medicaid Long-Term Services and Supports (LTSS) Users and Expenditures. CMS Medicaid bulletin, November 21, 2024, summarizing 2022 LTSS statistics. Source.

  • Li, Yue; Cai, Xueya; Kim, Yeunkyung Yoon; Kim, Jihye (2024). Informal Care Provided in US Nursing Homes: Reduced from 2010 to 2021 and Lower for Medicaid Residents. Journal of the American Geriatrics Society 72(6):1741–1749. DOI: 10.1111/jgs.18904. Source.

  • Finkelstein, Amy; McGarry, Kathleen (2006). Multiple Dimensions of Private Information: Evidence from the Long-Term Care Insurance Market. American Economic Review 96(4):938–958. DOI: 10.1257/aer.96.4.938. Source.

  • Brown, Jeffrey R.; Finkelstein, Amy (2007). Why Is the Market for Long-Term Care Insurance So Small?. Journal of Public Economics 91(10):1967–1991. DOI: 10.1016/j.jpubeco.2007.02.010. Source.

  • Organisation for Economic Co-operation and Development (OECD) (2025). Health at a Glance 2025: Long-Term Care Workers. OECD Health at a Glance 2025, long-term care workforce indicator. Source.

  • Organisation for Economic Co-operation and Development (OECD) (2025). Health at a Glance 2025: Long-Term Care Spending and Unit Costs. OECD Health at a Glance 2025, LTC spending and unit-cost indicator. Source.

  • United States Congress (current codification, accessed 2026). 42 U.S.C. § 1396n — Additional Provisions Relating to Medicaid. United States Code, Social Security Act §1915. Source.