RESEARCH DOSSIER
Morbidity Trend and Care Cost Inflation in Long-Term Care Insurance
Abstract. Long-term care insurance liabilities are determined by the interaction of functional dependency, survival, claim incidence and duration, the organization and price of care, and the exact promises made by insurance contracts. This monograph examines how morbidity trends and care cost inflation jointly alter the expected cost and distribution of insured long-term care obligations. Its central proposition is a measurement and valuation conclusion: changes in the prevalence of dependency, the incidence of payable claims, the duration of insured episodes, the intensity of purchased services, and the price of a comparable unit of care must be separately identified before changes in reserves or premiums can be attributed to morbidity or inflation. The analysis reconstructs the evidentiary development of long-term care experience studies, evaluates observation-based and actual-to-expected trend methods, and distinguishes cohort, period, age, selection, eligibility, benefit design, and provider-market effects. It integrates primary actuarial experience programs, longitudinal disability research, labor-market and care-price measures, insurance regulation, international care-system comparisons, and relevant scholarly work on private information and demand. Formal multi-state, cash-flow, and sensitivity illustrations demonstrate why the effects of higher morbidity and care prices can compound, while contractual benefit ceilings can transfer a substantial portion of the increase to households rather than insurers. The findings support more granular experience analysis, explicit consideration of dependent uncertainty, transparent inflation assumptions, and governance that links policyholder protection with reserve adequacy. All numerical projection exercises are labeled illustrative; no proprietary insurer portfolio, causal experiment, or new authorial theory is claimed.
Keywords: Long-term care insurance; morbidity trend; functional dependency; activities of daily living; cognitive impairment; incidence; claim continuance; competing risks; recovery; disabled mortality; care cost inflation; labor cost; care intensity; home care; nursing home; benefit inflation protection; reimbursement insurance; indemnity insurance; reserves; premium adequacy; multi-state actuarial model; experience study; IFRS 17; actuarial standards; solvency; consumer protection.
Introduction
An insurance promise extending several decades into the future is exposed simultaneously to changes in human dependency and to changes in the resources required to meet that dependency. A policyholder may reach a given age with less functional disability than earlier cohorts, yet encounter higher hourly care prices, fewer available family caregivers, a different distribution of care settings, or a contract whose nominal benefit fails to purchase the same service. Conversely, a deterioration in disability incidence may not translate proportionately into insured expenditures when daily limits, elimination periods, or aggregate benefit pools bind. These possibilities are essential to the actuarial interpretation of long-term care (LTC) experience.
The OECD affordability investigation (OECD 2024) places long-term care at the intersection of functional need, households' financial resources, formal and informal care supply, and public benefits. Its projection of substantially greater demand through 2050 is a demographic and policy scenario rather than an estimated rate of insured morbidity deterioration. That distinction determines the research problem here. A growing number of people requiring care can result from population aging even if age-specific dependency probabilities improve. A higher dollar amount paid for covered claims can reflect more intensive utilization, higher service prices, changed claims practice, or contractual inflation protection without a deterioration in underlying health.
Long-term care is not one uniform insured event. A patient requiring short-term skilled nursing following hospitalization, an older person receiving daily assistance with bathing and dressing, an individual needing continuous supervision because of severe cognitive impairment, and a person receiving public residential support can occupy different clinical, legal, and insurance states. The World Health Organization's ICF classification framework (WHO 2001, maintained online) supplies a useful language of bodily functioning, activity, participation, and environment. The insured definition of dependency remains the contract's own eligibility test, situated within applicable national law. Insurance models must preserve this distinction when projecting morbidity.
The analysis is organized around three questions. First, how can age- and risk-adjusted morbidity trend be measured without conflating exposure composition, disability prevalence, claim incidence, and claim termination? Second, how can the price of care be separated from changes in labor inputs, service intensity, setting, quality, and insurer benefit utilization? Third, how do their joint effects pass through contract terms into premiums, cash flows, accounting liabilities, solvency, and policyholder financial protection?
The method is documentary and analytical. It combines original or officially published actuarial experience reports and research protocols, official statistics and care-price survey methodology, professional standards, and peer-reviewed scholarship. A qualitative and mathematical synthesis applies established frequency-severity decomposition, multi-state actuarial methods, competing risks, and discounted expected cash flow techniques. Illustrative calculations use explicitly invented assumptions to test mechanisms rather than to assert market-wide estimates. The research cutoff is November 15, 2026. Where the available evidence concerns a specific reporting period, country, or insured product class, the claim retains that boundary.
Conceptual Foundations and Historical Development
The risk-bearing unit and the care episode
A transparent analysis begins by identifying the unit exposed to risk. In a stand-alone individual LTC insurance portfolio, the denominator for incidence ordinarily begins with insured lives or policy-years that are eligible to become new claimants during a specified interval. A claim can arise only after medical or functional criteria and applicable waiting rules have been satisfied; the insurer's record of claim payment may begin later than the onset of care need. In a population survey, disability prevalence is instead a cross-sectional or longitudinal property of a defined household or institutional sample. These measures answer distinct questions.
For an insured person, a care episode may involve a sequence of states: active without a payable claim; functionally limited but below the contractual trigger; eligible for benefits but in an elimination period; receiving paid home or community care; assisted living; nursing home care; recovery to nonclaim status; exhaustion of a benefit pool; and death. Transitions can recur. A discharge from a nursing home may mean recovery, transfer to home care, movement to a different paid setting, or death. Treating all such changes as equivalent claim terminations loses information about benefit obligations.
The National Association of Insurance Commissioners' long-term care insurance overview (NAIC, current online guidance) explains the U.S. combination of home and institutional care benefits, historic repricing, and regulatory attempts to coordinate rates. The association specifically identifies past understatement of incidence and claim duration and unexpectedly low lapse rates among issues affecting older policy generations. These are explanations of particular historic pricing experience, not evidence that every modern policy shares the same assumption error.
Experience measurement and the changing knowledge base
Long-duration insurers begin with incomplete lifetime experience because many insured people have not reached high claim ages and some claims remain open. Results mature as policy cohorts age; changes in underwriting, distribution, lapse, contracts, and care environments alter subsequent experience. The SOA long-term care study library (SOA Research Institute, online collection) documents the evolution of separate mortality, lapse, incidence, and other insurance-experience investigations. An industry study cannot retroactively make early pricing certain; it can expose the assumptions that were historically most sensitive.
The SOA's 2000–2016 aggregate database program (SOA 2020) provides separately organized claim incidence and termination datasets and a data-definition report. The distinction matters because a claim start process is fundamentally different from the process that determines how long a claim remains open. The program used multiple contributing insurers; its aggregates and the record definitions must be assessed before transferring rates to a particular portfolio. Its accompanying report documents limitations of insurer contributions and eligibility of records. The data are especially valuable for defining an auditable starting point for experience analysis, while underwriting, benefit design, and historical observation periods remain potential sources of noncomparability.
A further methodological step appears in Researching Long-Term Care Insurance Incidence Rates Over Time (SOA 2018), which describes observation-based and actual-to-model approaches for detecting calendar-year effects. These methods have different vulnerabilities. A raw calendar-year comparison is exposed to changes in portfolio age and other characteristics. An actual-to-expected comparison is exposed to the suitability and vintage of the expected basis. Neither method becomes a causal estimate merely by including the calendar year as a variable.
The SOA and LIMRA's 2000–2023 data request (SOA Research Institute 2025) specifies claim incidence, termination, and utilization for stand-alone individual and group LTC insurance, excluding combination and accelerated-benefit products from that particular program. The August 2025 joint experience-study announcement reports participation plans and a broader initiative by SOA, LIMRA, and NAIC. These documents evidence research design and planned collection, not completed findings for the 2000–2023 period. Until released study results are independently inspected, the project cannot be cited as having established a new morbidity trend.
An additional practitioner perspective is supplied by Milliman's 2025 LTCI claims projection (Milliman 2025). It describes incidence, utilization, continuance, recovery, disabled-life mortality, underwriting selection, and distinct annual cost assumptions for care settings. Its illustrative portfolio projection applies proprietary model assumptions and is analytically informative, while the projection is not a substitute for independently verified insurer-level data. Its published methodological description is useful precisely because it exposes the assumptions whose divergence can compound in a long-duration block.
Population disability, insured selection, and demand
The population experiencing severe limitations is the source of potential care demand, yet the insured subset differs by resources, health at purchase, underwriting rules, geography, and attitudes toward risk. Stallard and Yashin's population morbidity improvement study (2016) examines HIPAA-related activities-of-daily-living disability and cognitive impairment using 1984–2004 National Long Term Care Survey observations. It provides a historically important population-trend analysis. Its observation years, disability definitions, and population sampling do not by themselves establish contemporary morbidity improvement within a private LTCI block.
The importance of selection is strengthened by Finkelstein and McGarry's (2006) study of private information, which found offsetting dimensions of private information associated with health risk and insurance preference in the U.S. market. It is therefore unsafe to assume that insured experience is simply the population experience multiplied by a constant underwriting factor. The risk and preference composition of buyers can change at different rates, and the observed association between purchase and realized claims need not have a uniform sign.
Brown and Finkelstein's (2007) analysis of LTC insurance demand documents the importance of benefit design, loads, and demand-side behavior in the historical U.S. market. The study helps explain why the insured population and the general older population differ; its published historical market outcomes should not be treated as contemporary national market-share statistics. In research on morbidity, these market-selection mechanisms are part of the observation process. In research on care costs, the same mechanisms affect which benefits and services become reimbursed rather than paid privately.
International institutional context
LTC financing spans private indemnity and reimbursement contracts, public entitlement systems, social insurance, and means-tested support. The OECD care affordability report (OECD 2024) compares countries with widely different financing and service configurations. Its observation that older people may face substantial out-of-pocket burdens despite public support has direct relevance for insured benefit adequacy. Still, public spending per older person, household spending on home care, insurer-paid claims, and cost per professional care hour are not interchangeable price series.
This institutional setting also explains why the same aging trajectory can lead to different insurer cash flows. A country that expands subsidized home care may reduce privately reimbursed facility utilization; another that tightens public eligibility may shift expenses toward private insurance and households. Regulatory changes to claim assessments can alter recorded incidence without changing underlying functional need. An international model must therefore hold both clinical and contractual definitions visible rather than reporting a single global LTC morbidity coefficient.
Morbidity Trend Decomposition
What a morbidity trend measures
Morbidity trend is an observed or projected change in the probability, timing, severity, or persistence of a defined functional or cognitive state after accounting for the relevant risk population. A claim incidence trend is narrower: it concerns new qualifying events under the specified coverage, risk set, calendar period, and claim convention. A dependency prevalence trend also depends on survival and duration. A severity trend may change hours of care per dependent person while incidence is unchanged. The terms must be attached to observable variables before a trend assumption can be identified.
Let E[a,c,t] be eligible active insured person-years at attained age a, issue cohort c, and calendar year t. Let N[a,c,t] be the count of new payable episodes recognized under a documented incidence rule. An empirical central incidence rate is
i[a,c,t] = N[a,c,t] / E[a,c,t].
An insurer may require prospective entry counts and exposure measured to claim initiation, with adjustments for delayed reporting or incomplete observation. The formula presupposes compatible numerator and denominator definitions. Counting reopened claims as new incidence, including time already on claim in active exposure, or mixing paid claims with unmet need creates a different estimator. Policy-year, attained-age, issue-age, sex, duration since underwriting, benefit trigger, geography, and contract generation can all require stratification.
To isolate a calendar effect, compare standardized rates over a fixed age-and-cohort distribution, or estimate an experience model with those characteristics. A simple directly standardized rate is I[t] = Σ_g w[g] × i[g,t], with fixed, documented weights w[g] summing to one across comparable cells g. The raw rate Σ_g E[g,t] i[g,t] / Σ_g E[g,t] instead changes whenever the portfolio's exposure distribution changes. Both are valid descriptive statistics, but they quantify different objects.
Age, period, and cohort identification
Attained age, calendar period, and birth cohort obey an exact linear relation. A fully unrestricted model containing all three effects is therefore unidentified without constraints or additional structure. Product issue cohort adds further dependencies because calendar duration equals calendar year minus issue year. Estimating a coefficient labeled 'morbidity improvement' does not resolve this identification problem. A responsible trend analysis documents which dimensions are fixed or constrained, the reason for the constraint, and the sensitivity of conclusions to plausible alternatives.
A plausible age-standardized decline can represent health improvement, a change in which insured lives remain in force, favorable underwriting cohorts, a shift to policies with stricter triggers, or a change in claim notification behavior. Particularly at advanced ages, small exposure and selective mortality can produce unstable apparent trends. Analysts should present cell counts, confidence or uncertainty intervals, and diagnostics for low-exposure groups. The SOA's two incidence-trend approaches are useful precisely because their disagreement can reveal sensitivity to how expectations and composition are specified.
Multi-state dependency and competing exits
A single 'disabled' state may hide the main drivers of care expenditures. Consider states A (active and independent), D (dependent but not contractually payable), H (payable home care), L (assisted living), N (nursing care), R (recovered or below trigger), and X (dead). In continuous time, transition intensities q[j,k](a,t,d) may depend on age, calendar time, and time d already spent in a state. The generator matrix has off-diagonal entries q[j,k] ≥ 0 and each diagonal equal to the negative sum of departures. The row-vector state distribution satisfies dπ(u)/du = π(u) Q(u) in a time-inhomogeneous Markov formulation.
A semi-Markov specification is often more credible for active claims because recovery and mortality intensities can vary materially with elapsed claim duration, even at the same attained age. An equivalent duration-explicit transition model may be used. A claim may also transfer from H to N without ending; a cap or benefit exhaustion may stop payment even if dependency continues. The relationship between clinical transitions and insurer-paid transitions is therefore mediated by the contract.
For a simplified open-claim state with competing recovery and death exit intensities μ_R(u) and μ_X(u), the probability of remaining in that state through duration d is S(d) = exp(-∫_0^d [μ_R(u) + μ_X(u)]du), when those are the only exits. The expected open-claim time is ∫_0^∞ S(d)dd, provided the integral exists. If transfers, exhaustion, administrative closure, or requalification are relevant, they must enter the event system, and recorded termination cannot be interpreted solely as clinical recovery. The SOA aggregate experience database illustrates why distinct incidence and termination records are essential to that analysis.
Observed incidence and actual-to-expected ratios
For each comparable cell g,t, define A[g,t] as new observed events and M[g,t] = E[g,t] i_0[g,t] as expected events under a predeclared reference basis i_0. A portfolio-level actual-to-expected ratio is R[t] = Σ_g A[g,t] / Σ_g M[g,t]. Its interpretation depends on the basis. A rising ratio may indicate deteriorating experience relative to that basis, but it can also arise because the expected model is stale, benefit triggers changed, or reporting developed unevenly. If expected events are very low, ratios become unstable.
A reference basis can be re-estimated on training years, with subsequent data reserved for validation. It should not be silently updated in every comparison year if the objective is to measure change relative to a fixed benchmark. Analysts should report both raw outcomes and standardized A/E ratios; suppressing one hides an important check against compositional artifacts. Onset-year and reporting-year triangles help reveal development delays and shocks to claim processing.
Morbidity improvement and its limits
An improvement assumption is typically encoded through a reduction in age-specific incidence or a favorable shift in recovery and severity. Such a factor requires clinical and statistical definition. For example, a one-percent annual decrease in incidence applied to an unchanged age-specific baseline over 20 years gives 0.99^20 ≈ 0.818 of the baseline rate, all else equal. It does not imply an 18.2 percent reduction in the total number of claims, because future insured exposure, mortality, lapses, and benefit mix also change. Nor does it imply a similar decrease in insured cost if claimant survival or care intensity rises.
The 1984–2004 evidence investigated in Stallard and Yashin's morbidity improvement study deserves scientific attention as a defined longitudinal historical observation. It cannot identify future effects of dementia prevalence, obesity, disability treatment, pandemic aftereffects, or home-care policy without additional evidence and a new observation window. A forward-looking estimate should be a scenario whose assumptions can be falsified against later cohorts.
Causal and descriptive conclusions
Aggregate insured experience frequently supports risk-adjusted description more readily than causal attribution. Better clinical prevention could reduce functional dependency; stricter claim assessments could reduce payable incidence while leaving needs unchanged. Greater survival with impairment could raise disabled-life duration. Increased family assistance could reduce formal insured utilization without changing dependency. A defensible interpretation enumerates such mechanisms, identifies the data that would distinguish them, and avoids converting an estimated calendar coefficient into a causal conclusion about national health improvement.
Labor and Care Cost Inflation
The service purchased versus the price paid
Care cost inflation is the change in the price of a specified care service at comparable quality, location, time unit, and payment arrangement. Total care expenditure also changes with the quantity and composition of purchased services. Write a simplified service budget for care setting s as C[s,t] = U[s,t] × P[s,t], where U is intensity or units (hours, days, visits) and P is price per comparable unit. For several service types, C[t] = Σ_s U[s,t]P[s,t]. Changes in total spending may therefore reflect different quantities, relative prices, or a shift among settings even if the underlying service-specific inflation rates are unchanged.
The payer's price may differ from a publicly advertised private-pay median because of negotiated reimbursement, provider networks, taxes, room-and-board treatment, care intensity, location, and insurance policy restrictions. A nursing-home daily price is not commensurable with an hourly personal-care rate without an assumed schedule and scope. A single annual 'LTC inflation' figure should consequently disclose its basket and weights. If weights shift in response to price or disability, a fixed-weight index and a current-weight index measure different effects, and a decomposition should show the interaction or allocate it by a disclosed conventional method.
CareScout service prices and what they establish
The CareScout 2025 Cost of Care Survey reports national private-pay median prices for several U.S. service categories, based on provider surveys conducted between July and November 2025. It reports a $35 hourly median for nonmedical in-home caregiving, a $6,200 monthly assisted-living median, $315 per day for a semiprivate nursing-home room, and $355 per day for a private nursing-home room. On the survey's illustrative service schedules, these translate to $80,080 a year for 44 hours of home care each week, $74,400 for 12 months of assisted living, $114,975 for a 365-day semiprivate room, and $129,575 for a private room. These are market survey medians under stated use assumptions, not average insurer-paid amounts.
CareScout's survey release and methodology (CareScout 2026) reports more than 25,000 collected rates and year-over-year increases around 1–5 percent in most reported settings, with adult day health care declining. The survey site displays some measures in monthly rather than daily units. Minor differences between rounded daily growth and monthly comparisons can reflect rounding and construction, so annual rates should be quoted in the unit actually used. Geographic cost dispersion and provider sampling limit a national median's predictive power for an individual portfolio. Most critically, a change in the sample mix of providers can influence a repeated survey median even if a constant-service transaction price is unchanged.
The correct actuarial use of such statistics is as a source for external care-price triangulation, calibration checks, and scenarios after adjusting for geography, care setting, provider contract, and product wording. A $35 hourly median times a 24-hour assumption would radically overstate the specific annual figure based on 44 hours per week; conversely, an insurance benefit of $200 per covered day need not pay the actual number of billed hours. Units and utilization assumptions must accompany every monetary figure.
Labor supply, wages, productivity, and quality
LTC services are labor intensive. Personnel costs arise from direct care hours, supervision, scheduling, training, benefits, staffing regulations, agency overhead, and recruiting costs. The same nominal wage change can translate into a different provider cost change because staffing intensity, productivity, overtime, and nonwage expenses change. An approximate cost index for a provider with labor share α is g_C ≈ α g_W + (1−α) g_O + g_Q, where g_W is labor compensation change, g_O nonlabor cost change, and g_Q an explicitly defined quality, productivity, or composition term. The approximation is bookkeeping, not an empirically fitted law.
The U.S. Bureau of Labor Statistics' second-quarter 2026 ECI wage table documents wage and salary changes by industry, including nursing and residential care. It reports a 3.0 percent 12-month change for nursing and residential care facilities in the published June 2026 table. The ECI tracks employer labor costs for a defined employment sample; it is not a transaction-price index for nursing-home resident days, and it excludes important sources of price variation. The distinction is especially important when labor costs are one of several components of the provider's charge.
The BLS occupational outlook for home care aides (2026) projects substantial 2025–2035 employment growth in the U.S. occupation. A projection of worker demand is not a forecast of wage inflation: actual rates will depend on supply, immigration, training, reimbursement, technology, and regional labor conditions. It does, however, establish why the assumption that sufficient care hours can always be purchased at an indexed historical price deserves explicit stress testing.
The OECD's 2025 LTC workforce comparison reports an average of approximately five formal LTC workers per 100 people aged 65 and older in the 31-country comparable group for 2023, with wide national dispersion. This is a workforce density indicator, not a direct ratio of available caregiver hours to insured care needs. The OECD's working conditions investigation (2023) also identifies labor shortages, poor working conditions, and public financing constraints as determinants of care-sector capacity. These findings support scenarios involving higher wage costs, unmet demand, and substitution between providers rather than a mechanically fixed wage-to-price pass-through.
Price, utilization, access, and informal care
Care needs may be met through professional home care, family assistance, community services, and residential institutions. An increase in the opportunity cost of informal caregiving can raise paid service use at unchanged medical severity; a shortage of paid caregivers can push demand back toward families or leave needs unmet. Both mechanisms can cause a departure between measured insured claims and the welfare cost of dependency. The OECD analysis of care affordability explicitly identifies constrained family-care capacity and high out-of-pocket exposure.
An insurer may experience higher claims when beneficiaries substitute more expensive formal hours for family care. It may also experience temporarily lower cash utilization when formal care cannot be accessed, with eventual adverse outcomes from delays or higher acuity. Therefore lower claims during labor shortages should not automatically be interpreted as improved morbidity. The relevant evidence includes unmet-need surveys, hours billed, hours authorized, time until provider placement, and movement among care settings.
A provider price increase can be quality related. Regulation requiring greater staffing may increase cost while also improving resident safety. A fixed-service price index must identify whether the care package has changed. At the other extreme, providers may reduce service duration or quality to hold a quoted rate constant. A monetary trend cannot, by itself, reveal the welfare consequences of these adjustments. External consumer price measures and internal claim-level service utilization should be analyzed together.
Inflation protection and the erosion of nominal benefits
Contract terms determine how inflation passes to the insurer. Consider three stylized forms: reimbursement up to a daily cap, a fixed cash indemnity triggered by eligible dependency, and a reimbursement benefit with a cumulative lifetime pool. A claim dollar can increase with provider prices only while an applicable reimbursement ceiling remains unbound. Once the ceiling binds, the insurer's incremental payment may be zero despite a rising household expenditure. Fixed indemnity promises can instead raise insurer payments through the contractual benefit escalation rule, independently of the actual service bill.
For an illustrative current daily care bill of $250 growing at 5 percent annually for 20 years, the bill becomes approximately \$250 × 1.05^20 = \$663.32. If the insured daily maximum starts at $200 and increases by 3 percent annually, the maximum becomes approximately \$200 × 1.03^20 = \$361.22. Assuming eligible full-day use and no other restrictions, the household's daily funding gap then approaches $302.10. These numbers follow only from the assumed rates and cap; they are not survey observations or product quotations. A fixed maximum, simple escalation, compound escalation, attained-age limitations, and a shared benefit pool yield materially different outcomes.
The difference between the cost-growth factor and benefit-growth factor is [(1.05)/(1.03)]^20, approximately 1.47 in this illustration. That ratio expresses the widening relation between care price and benefit ceiling when both have an initial comparable unit; it is not a measured inflation shortfall for all long-term care policies. Insurer liabilities and policyholder adequacy can therefore diverge over time, making simultaneous measurement essential.
Data, Legal, and Model Validation
Evidence populations and reproducible observation
Source datasets need a stable codebook covering policy identifier, insured person's age and sex where lawfully usable, issue cohort, underwriting class, contract generation, eligibility trigger, benefit inflation feature, elimination period, annual and lifetime benefit limits, calendar-time exposure, claim onset, notification, payment, care setting, service units, unit prices, claim closure reason, and censoring. Data governance must protect privacy and report how incomplete claim histories or insurer contributions are handled. Experience results should distinguish first-time claims from reopens and policy lapse from mortality termination.
The SOA aggregate dataset documentation supplies a concrete public example of definitions needed when combining carrier experience. The 2025 SOA/LIMRA submission protocol shows why experience updating depends on consistent fields across participating insurers and years. Neither constitutes permission to assume that every insurer's private records share identical codes. A research-quality data extraction should identify exclusions, variable missingness, left truncation, right censoring, and calendar cutoffs explicitly.
Exposure bias, selective survival, and claim development
New issues, deaths, lapses, benefit exhaustions, and policy conversions all alter the denominator for new claims. An age-standardized analysis is insufficient if changes in underwriting, product benefits, or lapse selection remain material. Insureds who voluntarily surrender policies may differ from those remaining; selective mortality may change the risk composition of survivors; early claim reporting can vary when benefit triggers or administrative systems change. Every rate should therefore be tied to a risk set that is continuously reconstructed over observation time.
Recent incurred claims may develop after the data cutoff. An open claim is right-censored for duration measurement and requires the appropriate survival likelihood rather than treating its current duration as final. Delayed onset notification can distort calendar comparisons. Left-truncated claims require conditioning on survival to entry; otherwise long-duration continuing claims can be overrepresented. For competing exits, the probability of recovery by time t depends on the cause-specific recovery hazard and survival from all competing causes. An observed share of terminated claims assigned to recovery is not generally the same quantity as the cumulative recovery probability for a new claimant.
Claims rules and contract boundaries
A contract's insured benefit is a legal and administrative mapping from assessed need and expenditure to insurer payment. Eligibility may involve activities-of-daily-living thresholds, severe cognitive impairment, professional certification, a benefit waiting period, licensed providers, covered locations, or a specified form of home care. Policy generations can differ materially within one carrier. The WHO ICF framework supplies a general classification of functioning, while the NAIC consumer and regulatory guidance describes features of insurance protection. Mapping the former directly into insurer claims without examining contract wording would misstate the legal risk transfer.
For each projected payment date, a reimbursement benefit can be represented as Payment = max(0, min(Eligible Expense, Available Daily Benefit, Remaining Policy Pool)), after suitable adaptation for deductibles, elimination periods, co-payments, inflation provisions, and units. This is a simplified example, since some products use fixed indemnities, monthly pools, shared care riders, or different structures. The rule is evaluated at the actual future payment path. If a benefit pool is exhausted, projected payment stops while care need may persist. Recording that event as clinical recovery would contaminate morbidity trend estimation.
Inflation-data comparability and effective dates
External labor and provider-price series are subject to distinct statistical designs. BLS wage indexes follow compensation for sampled employers; CareScout medians reflect sampled provider service quotes; OECD country indicators reflect different national system definitions and currency or purchasing-power treatments. A cross-country comparison should identify currency, deflator, purchasing-power adjustment, service standard, payer, care setting, reference year, and observed versus projected values. It should not compare an employer's annual wage directly with an annual household care bill as if they were identically defined prices.
The CareScout national rate table covers the 2025 survey year, even though its release was announced in 2026. The BLS second-quarter 2026 wage data cover a later 12-month observation window. The OECD 2025 workforce chapter relies largely on 2023 workforce measurements. Each can inform a model, but a valid combination aligns dates and describes the timing differences. Treating the reports' publication years as observation years would create false contemporaneity.
Scenario identification and sensitivity table
The following table defines distinct stresses on a hypothetical uncapped portfolio. All percentages are user-free analytical assumptions, not estimated market outcomes.
| Analytical scenario | Changed assumptions | Gross cost index, baseline 100 | Interpretation |
|---|---|---|---|
| Baseline | No change | 100.0 | Fixed illustrative reference |
| Morbidity incidence | Incidence +10% | 110.0 | More payable claim starts |
| Longer continuance | Duration +15% | 115.0 | Longer average paid episodes |
| Care price | Unit price +20% | 120.0 | Higher comparable service prices |
| Combined deterministic stress | Incidence +10%, duration +15%, price +20% | 151.8 | Multiplied assumptions, before caps |
| Combined with binding benefit caps | Same underlying changes | Product-specific | Cannot infer from gross index |
The table allows direct reproduction of the 151.8 calculation but also demonstrates the limit of mechanical extrapolation. Incidence and duration assumptions are applied as independent multiplicative shocks for illustration, not because their statistical independence is established. A real model should explore adverse dependence, benefit exhaustion, mortality, discount rates, service mix, lapse behavior, and calendar timing.
Sensitivity testing should include alternative care-price baskets, fixed and dynamic utilization weights, severity-specific incidence, calendar-by-cohort trend constraints, structural breaks, alternative mortality and recovery hazards, benefit inflation provisions, and alternative contract interpretations. Backtesting can compare earlier projected claims with later realized claims, separating assumption error from data development and contract changes. One-way sensitivities show local model behavior; multivariate scenarios test joint adverse movement; probabilistic capital exercises require a separately justified distribution and dependence structure.
Evidence hierarchy and open limitations
This dossier directly examines official public research pages, standard-setting descriptions, institutional statistics, and scholarly publication metadata or abstracts. It does not represent a line-by-line audit of all underlying insurer contribution files, confidential claims, every paywalled paper, or full legislative inventories in multiple jurisdictions. Provider surveys establish their stated sampling and central tendency measures, while independent replication of provider responses is outside the available material. The 2000–2023 SOA initiative is treated as a collection program rather than an observed result. These scope boundaries determine the wording of each conclusion and prevent unsupported estimates of a global annual morbidity or LTC price trend.
Integrated Findings and Research Agenda
Interpretation of the joint risk
The evidence supports a disciplined distinction between five phenomena: aging of the insured exposure; change in age- and contract-specific incidence of eligible claims; change in claim-state duration and transitions; change in paid service mix and intensity; and change in the unit price of comparable care. An observed increase in aggregate insurer claims is an outcome of all five, combined with contractual payment limits, policyholder selection and financial discounting. Calling the entire increase 'medical inflation' or 'morbidity deterioration' discards information needed to value the promise.
The distinction has direct consequences for policyholder welfare. A dollar-denominated cap can insulate the insurer from some care-price changes while enlarging the customer's unfunded care requirement. A policy with automatic compound inflation protection can maintain a larger share of care purchasing power while increasing the insurer's long-tail exposure. Greater caregiver scarcity may simultaneously raise wages and reduce access; observed lower paid utilization in that setting can coexist with worse unmet need. Contract design and care-market capacity are therefore inseparable from interpreting the cost of insured protection.
The evidence also suggests why historical pricing problems should be understood jointly. The NAIC's LTCI historical account highlights actual experience that diverged from older assumptions for incidence, continuance, and lapse. Milliman's published projection methods separately parameterize morbidity, care setting, utilization and inflation. Combined, these sources show that the correct object of monitoring is the interaction among assumptions rather than a narrow calendar inflation factor. This is an analytical synthesis of existing actuarial concepts, not a claim to have discovered a new model.
Research propositions capable of empirical testing
Several implications can be formulated as empirical research questions without introducing new theoretical systems. Does care-cost growth predict higher insured dollars after controlling for benefit maximums and setting mix? Does the same provider-price change affect capped reimbursement and fixed-indemnity contracts differently? Do cohorts with greater access to informal care experience less paid formal service use conditional on functional dependency, and does the association change with caregiver labor supply? Does a calendar trend estimated from new claims persist when product generation, age, underwriting selection, and claim reporting delays are controlled? Are duration shocks mainly attributable to improved survival during dependency, slower recovery, different severity, or benefit-design differences? Does the insurer-paid share of eligible care expenditure decline over time in products whose benefit escalation lags local service prices?
Each question requires a particular unit of observation and counterfactual. The provider-price question needs transaction prices and service units, while the benefit-cap question also needs complete policy wording and remaining pool balances. The informal care question needs independent measurement of family assistance and functional severity. The claim-incidence question needs risk-set construction and reporting-development adjustment. The duration question requires cause-specific transitions with censoring. A cross-jurisdiction result requires comparable need measures and explicit adjustment for public entitlement design. The underlying limitations are productive: they specify what data could confirm or disconfirm a proposed relationship.
Editorial and actuarial governance implications
Internal actuarial review should require documentation of where source observations end and scenario assumptions begin. At each assumption review, the insurer should retain the data vintage, observation window, contract generation, definitions of incidence and duration, choice of care-price index, alternative scenarios, and the actual change to projected cash flows. Material changes must be traceable to recognized causes rather than unexplained balancing adjustments. When published survey medians or labor indexes are imported, the mapping to the portfolio's paid services requires a recorded justification.
The ASOP No. 18 professional standard offers a U.S. practice reference for this discipline; IFRS 17 and the IAIS supervisory principles supply distinct reporting and risk-governance contexts. They are not identical mandates, and a reporting-basis reconciliation is essential. National regulatory approval of LTC premium changes and contract modifications remains a separate legal inquiry from whether a model's indicated premium is economically sufficient.
For research institutions, useful next evidence would include linked functional-assessment and insurer-claim records with consent and legal protections, extended high-age exposure, claim-state transitions, policy benefit exhaustion, provider-level price and quality panels, and comparable household informal-care measures. Joint modeling of climate disruption, workforce shortages, public eligibility change, and regional care supply could extend the scenario set; these are research directions and established risk interactions rather than proposed proprietary doctrines. Their value lies in testing whether risk is properly transferred and funded for the actual care that people can obtain.
Conclusion
Morbidity trend and care cost inflation shape long-term care insurance through different observable processes whose effects are joined by insured eligibility and benefit design. Mortality and survival determine exposure to disability, functional-state transitions determine the timing and persistence of care, households and providers determine how much formal care is delivered, and service markets determine its price. Insurance contracts determine which part of that expenditure is payable by the insurer, subject to waiting periods, covered settings, daily or monthly caps, inflation protection, and remaining benefit pools. Financial valuation then converts those contingent payments into current obligations under the relevant accounting and regulatory system.
The review establishes three substantive conclusions. First, long-term claims development cannot be explained by a single unadjusted morbidity trend because aging, selection, claims definitions, occurrence, and continuance are different mechanisms. Second, care-price inflation must be observed alongside service units, setting mix, workforce capacity, and quality; headline survey medians, employer wages, and actual insurer reimbursement measure different quantities. Third, joint incidence, duration, and cost shocks can compound materially in projected obligations, but policy benefit limits may reroute the shock into household out-of-pocket expenditure. A reserve projection that is insensitive to provider prices may therefore coexist with deteriorating consumer protection rather than indicating a low social cost of care.
The resulting analytical standard is practical: define the insured state and exposure, preserve cause-specific transitions, measure comparable care units and their prices, apply the complete contractual payment function, test dependent multivariate scenarios, and record data and jurisdictional limitations. This synthesis clarifies the conditions under which morbidity trends and inflation are interpretable for pricing, reserving, solvency, and policyholder adequacy. It neither claims a universal numerical forecast nor introduces a new original theory. The research is complete as a documentary and analytic monograph subject to its stated evidence boundaries.
References
Actuarial Standards Board (ASB). 2022. Actuarial Standard of Practice No. 18: Long-Term Care. Revised March 2022; effective September 1, 2022. Actuarial Standards Board. https://www.actuarialstandardsboard.org/asops/asop-18-long-term-care/
Brown, Jeffrey R., and Amy Finkelstein. 2007. “Why Is the Market for Long-Term Care Insurance So Small?” Journal of Public Economics 91 (10): 1967–1991. https://doi.org/10.1016/j.jpubeco.2007.02.010
CareScout. 2026. “CareScout Releases 2025 Cost of Care Survey Results.” Genworth Financial investor news release, survey fielded July–November 2025. https://investor.genworth.com/news-events/press-releases/detail/1054/carescout-releases-2025-cost-of-care-survey-results
CareScout. 2026. Cost of Care: 2025 National Median Values and Survey Methodology. CareScout. https://www.carescout.com/cost-of-care
Finkelstein, Amy, and Kathleen McGarry. 2006. “Multiple Dimensions of Private Information: Evidence from the Long-Term Care Insurance Market.” American Economic Review 96 (4): 938–958. https://www.aeaweb.org/articles?id=10.1257%2Faer.96.4.938
Government Accountability Office (GAO). 2008. Long-Term Care Insurance: Oversight of Rate Setting and Claims Settlement Practices. GAO-08-712. U.S. Government Accountability Office. https://www.gao.gov/assets/a277848.html
International Accounting Standards Board (IASB). 2017, as amended. IFRS 17 Insurance Contracts. IFRS Foundation; effective for annual periods beginning on or after January 1, 2023. https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
IFRS Foundation. n.d. “IFRS 17 Insurance Contracts: Key Terms.” Official implementation resource. https://www.ifrs.org/supporting-implementation/supporting-materials-by-ifrs-standards/ifrs-17/key-terms/
International Association of Insurance Supervisors (IAIS). 2024. Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups. Updated December 2024. https://www.iais.org/activities-topics/standard-setting/icps-and-comframe/
Milliman. 2025. “Milliman’s Annual U.S. Industry LTCI Claims Projection.” Milliman, methodology and assumptions publication. https://www.milliman.com/en/insight/annual-us-industry-ltci-claims-projection-2025
National Association of Insurance Commissioners (NAIC). n.d. “Long-Term Care Insurance.” Insurance Topics, current regulator overview and Multistate Rate Review Framework summary. https://content.naic.org/insurance-topics/long-term-care-insurance
National Association of Insurance Commissioners (NAIC). n.d. “Long-Term Care Insurance.” Center for Insurance Policy and Research, topic background and historical assumption experience. https://content.naic.org/cipr_topics/topic_longterm_care_insurance.htm
Organisation for Economic Co-operation and Development (OECD). 2023. Beyond Applause? Improving Working Conditions in Long-Term Care. OECD Publishing, Paris. https://www.oecd.org/en/publications/beyond-applause-improving-working-conditions-in-long-term-care_27d33ab3-en/full-report.html
Organisation for Economic Co-operation and Development (OECD). 2024. Is Care Affordable for Older People? OECD Health Policy Studies. OECD Publishing, Paris. https://www.oecd.org/en/publications/is-care-affordable-for-older-people_450ea778-en.html
Organisation for Economic Co-operation and Development (OECD). 2024. Is Care Affordable for Older People? Full Report. OECD Publishing, Paris. https://www.oecd.org/en/publications/is-care-affordable-for-older-people_450ea778-en/full-report.html
Organisation for Economic Co-operation and Development (OECD). 2025. “Long-Term Care Workers.” In Health at a Glance 2025: OECD Indicators. OECD Publishing, Paris. https://www.oecd.org/en/publications/health-at-a-glance-2025_8f9e3f98-en/full-report/long-term-care-workers_9c3bdbaf.html
Society of Actuaries Research Institute (SOA). 2018. Researching Long-Term Care Insurance Incidence Rates Over Time. SOA Research Institute. https://www.soa.org/resources/experience-studies/2018/researching-ltc-insurance-incidence/
Society of Actuaries Research Institute (SOA). 2020. 2000–2016 Long-Term Care Aggregated Databases. SOA Research Institute. https://www.soa.org/resources/experience-studies/2020/2000-2016-ltc-aggregate-database/
Society of Actuaries Research Institute (SOA). 2025. 2000–2023 Long-Term Care Data Request. SOA Research Institute / LIMRA study collection protocol. https://www.soa.org/research/opportunities/2025/2000-23-ltc-datarequest/
Society of Actuaries Research Institute (SOA). 2025. “Industry Leaders Launch New Long-Term Care Insurance Industry Experience Study.” Joint SOA, LIMRA, NAIC press announcement, August 4, 2025. https://www.soa.org/resources/announcements/press-releases/2025/industry-leaders-launch-ltc-study/
Society of Actuaries Research Institute (SOA). n.d. Long-Term Care Experience Studies. SOA study library. https://www.soa.org/research/topics/ltc-exp-study-list/
Stallard, P. J. Eric, and Anatoliy I. Yashin. 2016. U.S. Population Long Term Care Morbidity Improvement Study. SOA Research Institute. https://www.soa.org/resources/research-reports/2016/2016-ltc-morbidity-improvement-study/
U.S. Bureau of Labor Statistics (BLS). 2026. “Table 9: Wages and Salaries, Employment Cost Index for Private Industry Workers, Second Quarter 2026.” National Compensation Survey. https://www.bls.gov/news.release/eci.t09.htm
U.S. Bureau of Labor Statistics (BLS). 2026. “Home Health and Personal Care Aides.” Occupational Outlook Handbook, 2025–2035 projections. https://www.bls.gov/ooh/healthcare/home-health-aides-and-personal-care-aides.htm
World Health Organization (WHO). 2001, maintained online. International Classification of Functioning, Disability and Health (ICF). WHO. https://www.who.int/standards/classifications/international-classification-of-functioning-disability-and-health/