RESEARCH DOSSIER
Medical Cost Trend Decomposition and Inflation Pass-Through
Abstract. Medical cost trend is the change over time in the cost of covered health benefits for a defined population and period. Its apparent simplicity obscures several distinct phenomena: the unit prices negotiated or administered for clinical services; the frequency of encounters; the technological and clinical intensity of care; changes in the distribution of services and patients; the level and composition of insured exposure; diagnostic coding; contractual benefits; and the distribution of expenditure among patients, insurers, and public purchasers. This monograph establishes a source-critical method for disentangling these mechanisms, links price-volume accounting to causal interpretation, and traces their financial transmission into insurance premiums, incurred-claim estimates, and technical provisions. The argument employs established index-number mathematics, expenditure accounting, exposure standardization, observational research design, and established insurance accounting rather than claiming a new theoretical system. National health accounts, commercial claims, provider payments, employer premium surveys, health-specific purchasing power parities, and actuarial valuation data identify different objects; therefore, their growth rates must be interpreted on explicitly specified denominators. Evidence from the United States, international OECD statistics, and supervisory standards illustrates the distinctions. A worked arithmetic decomposition and a reproducible validation protocol demonstrate that the measured contribution of price, utilization, case mix, coding, and insurance contract design depends on classification, population stability, data maturation, and observation timing. The principal conclusion is that credible medical trend attribution is a reconciled chain of measurements and economic mechanisms, extending from clinical units and patient populations through payment contracts and ultimately to insurer obligations and premiums.
Keywords: Medical cost trend; health expenditure growth; medical inflation; price-volume decomposition; medical utilization; service mix; morbidity; diagnosis coding; health insurance; claims inflation; premium pass-through; insurance reserves; health accounting; actuarial science; risk adjustment; index numbers; national health accounts; provider contracting; health economics.
Introduction
A statement that medical costs have increased by a given percentage contains an implicit choice of what is being priced, for whom, by which payer, over which interval, and under which insurance contract. An increase in charges submitted by hospitals may coexist with stable negotiated allowed amounts. Greater use of previously inaccessible care may raise insured expenditure while improving health protection. A change in the share of high-acuity cases may increase average claim severity without any change in the price of a given clinical unit. A modification of diagnosis documentation may increase reported morbidity scores and change risk-adjusted funding while leaving delivered care unchanged. An increase in a family insurance premium may reflect prior-year incurred claims, the design of future benefits, capital charges, administrative costs, taxes, or expected risk-pool composition.
This study analyzes how those processes interact. The primary object is incurred covered medical expenditure and its associated insurer obligation, normalized to a defined measure of exposure. National health expenditure and insurance premium growth are examined as related but differently constructed observations. The analysis spans privately insured markets, public insurance arrangements, regulated provider-payment systems, and international comparative statistics. The four approved Science relations are Health Insurance, Insurance Economics, Insurance Pricing and Reserving, and Long-Term Care and Disability Insurance. The latter provides a related context for duration, care-intensity changes, and future benefit obligations, while this dossier retains its own core object of medical trend attribution.
The work relies on an explicit evidentiary hierarchy. Published accounting identities and regulatory requirements establish definitions; official statistical methods establish the measurement processes of their series; peer-reviewed analyses provide empirical results for their specified samples; and prospective projections remain conditional forecasts. The source universe includes the System of Health Accounts framework (OECD, Eurostat, and WHO, 2017), the CMS national expenditure methodology (2026), publicly documented commercial claims research, and insurance reporting standards. Original authorial contribution consists of integrating their findings into a coherent, auditable interpretation of medical cost trend, specifying the boundaries of valid comparisons, and deriving decision-relevant consequences from existing methods. The analysis makes no claim to have estimated a new empirical causal parameter.
The units of inference
At the clinical-transaction level, the objects are a completed service, a reimbursed episode, a drug dispensation, a claim line, or a quality-adjusted service bundle. At the insurance level, objects include an enrolled member-month, covered life-year, insured encounter, completed accident or service period, and future contracted benefit. At the national-account level, the objects are sectoral output and expenditure as classified by services, financing schemes, and funding sources. A study may validly describe each object; none is interchangeable with another merely because all are expressed in currency.
The OECD financing-scheme statistics classify spending according to financing arrangements, while the WHO expenditure database documentation provides another official framework for tracing health financing internationally. Such aggregates are essential to system-level comparison but cannot reveal the price paid for an otherwise equivalent claim without supplementary service-level information. The distinction between a payer's contractual obligation and an economy's expenditure on care also determines which changes should enter insurer ratemaking and reserving.
Analytical questions
Three questions organize the investigation. First, how much of a measured increase reflects changes in unit prices versus service frequency, clinical intensity, and the composition of services? Second, how much can be attributed to changing risk and insurance exposure rather than coding, selection, or benefit definitions? Third, by what routes and lags does expenditure growth pass through provider contracts and insured benefits into prospective premiums and outstanding claims liabilities? The answer requires a common measurement architecture but different methods at each stage.
Foundations and Source-Critical History
Expenditure, payment, and premium
An expenditure observation can be formulated as the sum of quantities multiplied by the applicable unit payments. For an insurer, the relevant price may be the contracted allowed amount, the insurer-paid share after cost sharing, or the net cost after a rebate. The provider's billed charge is another observable, while the cost of production, a household's out-of-pocket payment, and a national-account imputed price are additional constructs. A decomposition must choose one financial definition and maintain it across periods.
For service category i in period t, let q(i,t) denote clinical units and p(i,t) denote the corresponding allowed price. Gross allowed spending is A(t) = sum over i of p(i,t) q(i,t). Net insurer spending is not generally equal to A(t). It must also reflect covered service status, deductibles, copayments, coinsurance, stop-loss recoveries where relevant, subrogation, refunds, and pharmaceutical rebates according to the stated accounting convention. A plan may alter its contribution to costs while total resources paid to providers remain unchanged. Thus an empirical decomposition should first establish a consistent allowed-spending measure and then bridge it to payer-specific net spending.
The US national expenditure fact sheet and Hartman and colleagues' 2024 health spending study report national spending of approximately $5.3 trillion for 2024, a 7.2 percent increase over 2023 and 18.0 percent of GDP. This is a comprehensive national expenditure aggregate. It includes elements that are neither paid by a private health insurer nor covered under a particular employer policy. It is therefore incorrect to substitute the 7.2 percent national number for a specific private insurer's annual claim trend.
Why the price-volume problem emerged
The price-volume question appears wherever monetary expenditure changes over time and the content of the purchased basket evolves. Healthcare makes the problem especially difficult because quality, service availability, diagnosis, technology, site of care, and treatment pathways alter the very unit being compared. Expenditure growth from new high-cost treatment may reflect a new service whose earlier-period price is undefined. Growth from a higher share of difficult cases may be genuine intensity rather than unit-price inflation. Simple averaging of reimbursement across heterogeneous admissions confounds both.
Health accounting developed by extending national accounts into service- and funding-specific categories. The revised international health accounts manual records standardized classification of resources and financing, while CMS's revised national accounts require careful attention to historical methodological revisions. The CMS methods paper explains the construction of personal health care price deflators, including the use of different consumer and producer price indexes for distinct service categories. A statistical deflator is the product of a defined measurement procedure; it is not an observed bilateral price for every clinical service in an insurance portfolio.
The primary source problem is accordingly one of reconciliation: claims databases contain rich transactional detail but may omit uninsured treatment and off-claim transfers; national accounts contain broad expenditure coverage but lack claim-level clinical granularity; employer surveys observe premium contracts but not the same period's finalized incurred benefits. A sound inference links sources only after specifying their measurement interfaces.
Three different meanings of inflation
Economywide inflation denotes a change in a general price index. Medical service price inflation denotes a change in a defined index of healthcare transactions or service prices. Medical cost trend denotes growth in insurer-relevant spending per eligible exposure, potentially incorporating prices, utilization, service intensity, risk, benefits, and financial arrangements. Inflation pass-through denotes the causal or contractual transmission of input cost and medical service price changes into subsequent provider payments, insurer costs, and premiums. These definitions describe connected but separately observable variables.
The BLS medical-care CPI methodology documents the consumer-price measurement approach, including insurance-specific treatment. The health insurance CPI improvements concern a retained-earnings measure, not an annual average employer premium quote. A medical CPI series can inform a trend analysis if its population, basket, payer perspective, and service definitions correspond sufficiently to the research question. It cannot replace the claims-price series simply because both use the word medical.
Empirical constraints
Research on commercial prices demonstrates why a price-only narrative is incomplete. Cooper, Craig, Gaynor, and Van Reenen's private hospital price analysis found substantial price dispersion and important price contributions to variation in commercial hospital spending in the studied US data. Their analysis rests on specific years, hospitals, transactions, and markets; it is evidence of market pricing and bargaining mechanisms, rather than a universal international estimate of the price contribution in every year. A national payer with administered schedules and a private purchaser negotiating local network rates face different transmission processes even when providers experience similar input-cost pressures.
An analytical history is useful only when it preserves these changing institutional objects. Provider price setting, service classification, policy benefit obligations, and insurance accounting must be reconstructed for the precise period under review. The same label, especially medical inflation, does not guarantee an invariant time series.
Unit Price and Service Mix Effects
Exact expenditure accounting
For a stable set of comparable services, aggregate spending satisfies:
For two periods, an exact symmetrical decomposition is:
The first term is a midpoint-weighted price contribution; the second is a midpoint-weighted quantity contribution. Expanding the right-hand side recovers exactly p(i,1)q(i,1) minus p(i,0)q(i,0) for every service, so no residual remains. This is standard symmetric index-number arithmetic, used here as a transparent accounting identity. It assigns half of each cross-period interaction to each driver; an alternative reference basket assigns the interaction differently. The total spending change is determined by the data, while the attribution between contributions depends on the chosen valid decomposition.
At fixed initial quantities, the Laspeyres price index is the ratio of current to initial prices applied to the initial service basket. The Paasche price index uses current quantities as weights. Their geometric mean is the Fisher price index:
The associated Fisher quantity index is constructed by exchanging the roles of prices and quantities. These indices support expenditure reconciliation, and chaining adjacent periods reduces dependence on a distant historical basket. However, services entering and exiting the basket require an explicit imputation, matched-service method, hedonic or quality adjustment, or disclosure that a component is outside the comparable universe. Infinite precision in arithmetic cannot recover the price of a nonexistent prior treatment.
An illustrative audit of the arithmetic
Consider a synthetic portfolio in which 100 clinically identical procedures cost $200 each in year zero and 110 procedures cost $220 each in year one. Spending rises from $20,000 to $24,200, an increase of 21 percent. The matched-service price grows 10 percent; the number of procedures grows 10 percent; their multiplicative combination is 1.10 times 1.10 equals 1.21. In midpoint dollar attribution, the price contribution is $20 times 105 procedures, or $2,100, and the quantity contribution is 10 procedures times the $210 midpoint price, another $2,100. Each contributes 10.5 percentage points to the observed 21 percent change relative to the initial base. No underlying medical insurer data were used in this example.
Now add one extra distinction. If the 110 later procedures were performed for 110 covered people instead of 100 while each person still received exactly one service, volume per covered person would be stable. If enrollment remained 100 people, utilization per person would rise 10 percent. These are different exposure-normalized conclusions from the same claim total. Therefore, provider transaction growth, utilization per member, and spending per member must always be separately reported.
Services, intensity, and clinical pathways
The service mix changes when patients shift among inpatient care, outpatient surgery, specialist visits, generics, branded drugs, home healthcare, or telemedicine. Within an episode, case complexity may change the intensity and number of line items. A single average charge per hospital admission cannot separate a higher reimbursement for the same admission type from a growing fraction of complex admissions. Episode grouping or clinically stable service classes are necessary.
The service unit must reflect actual care. A diagnostic image, hospital admission, clinical episode, and patient-year each answer different questions. Counting visits may misstate activity when technology substitutes home monitoring for inpatient days. A shift from more invasive hospital treatment to a lower-cost outpatient pathway may reduce spending even if the outpatient service's unit price increases. For an insurer, the relevant economic outcome involves both total covered costs and the care outcomes purchased.
Service mix and intensity can be decomposed by stratifying episodes into clinically comparable cells. Let k index a care episode and j its constituent services. The episode cost is the sum of prices and quantities across j. Movement in the distribution of k generates a between-episode mix effect; movement within episodes generates a within-episode cost effect. Changes in patient classification can affect both. In practice the hierarchy should be disclosed before estimating coefficients, because post hoc regrouping can manufacture an apparently large or small mix component.
Reimbursement arrangements and economically relevant prices
Fee-for-service contracts make per-unit allowed amounts relatively observable; bundled payments shift the unit to an episode; capitation changes the observable payment to a covered member-period; global budgets can decouple marginal service volume from the annual total paid to a provider. Each arrangement changes how a provider input shock appears in payer data. A workforce wage increase can raise a hospital's production cost without immediately increasing a fixed fee schedule, while a multi-year contract may transmit that increase only after renegotiation. An indexed agreement may produce a more direct and timelier pass-through. A hospital's nominal list charge may increase yet have little effect on the negotiated price under an existing contract.
An insurer analysis must therefore distinguish billed charges, negotiated rates, actual allowed payments, public administered tariffs, covered benefit payments, and any retrospective settlement. The corresponding service set and risk exposure have to remain comparable over time. Differences in institutional negotiation or bargaining power are relevant when interpreting observed price trends; they should not be assigned a causal magnitude without identification from comparable markets or contract changes.
Price growth in observed commercial claims
The HCCI 2019 claims report examined an employer-sponsored insured population younger than age 65. It reported per-person spending growth of 2.9 percent in 2019 alongside average prices increasing 3.6 percent and utilization declining 0.7 percent. Those estimates are report-specific measurements under the HCCI service classifications, rather than components of the CMS national accounts. Manufacturer rebates were excluded from the expenditure amount examined. The difference illustrates how a modest total trend can coexist with faster service price growth when utilization contracts, and why pharmacy rebate omissions should be made explicit.
The source also demonstrates the importance of composition controls. Employer-sponsored commercial insured samples can differ from public insurance, uninsured care, other private coverage, and individuals who switch insurers. When weights standardize age, sex, or geography, they address those dimensions rather than all differences in diagnosis, provider market, negotiated price, and benefit design. Cross-portfolio generalization requires a new assessment of representativeness.
Medical technologies and net prices
For drugs, the list price, wholesale acquisition cost, allowed charge, plan payment, patient payment, and net cost after rebates can follow different trajectories. A drug's launch adds an entirely new treatment option, and a shift toward high-cost specialty drugs can raise average spending through mix. Comparative analyses that rely on gross claims and omit manufacturer rebates should state that they measure pre-rebate payer amounts. An analysis of rebate impact requires access to contract-specific rebate timing, allocation, and accrual; otherwise it remains a sensitivity scenario.
For hospital and physician care, quality adjustment is equally material. If a new technology reduces complications at a higher initial payment, the benefit per dollar may improve despite higher nominal spending. The present decomposition is a financial accounting of insured expenditure and should not silently equate expenditure reductions with efficiency gains or expenditure increases with waste.
Population Risk and Coding Changes
Denominators: members, lives, and exposure duration
Health insurance experience is observed over people exposed to covered risk. For a population with total eligible member-months M, annual member-years equal M divided by 12. A per-member-per-month claim cost is incurred covered spending divided by member-months. A per-member-year cost is spending divided by member-years. A count of unique enrollees at any point during the year is a different denominator when membership churns or coverage duration varies.
For claims of fixed accounting scope, define C(t) = S(t)/E(t), where E(t) is covered member-years. Then the exact identity is S(1)/S(0) = [E(1)/E(0)] × [C(1)/C(0)]. In rates, total growth combines exposure growth and per-exposure cost growth with a multiplicative interaction. A comparison of spending totals without exposure reconciliation can mistake coverage expansion or enrollment attrition for medical inflation.
The CMS 2024 spending infographic provides a particularly clear illustration. Aggregate 2024 national health expenditure growth of 7.2 percent was split into 2.5 percentage points of medical price growth and 4.7 percentage points from nonprice factors including use, mix, population growth, and demographic mix. Separately, private-health-insurance-financed expenditure grew 8.8 percent while associated enrollment grew 3.5 percent and spending per enrollee grew 5.2 percent. Medicaid expenditure rose 6.6 percent while enrollment declined 8.6 percent and spending per enrollee increased 16.6 percent. The three series have distinct denominators and are not evidence of the same insurer pricing or utilization mechanism. The percentages are published rounded statistics and need not reproduce exact unrounded multiplicative identities.
Age, morbidity, geography, and benefit eligibility
An insurer's claim distribution is sensitive to enrolled age, diagnosed conditions, socioeconomic circumstances, geographical markets, provider access, and length of insured exposure. These characteristics can alter the frequency of illness, the intensity of treatment, and reimbursement negotiated in each location. Changes in risk mix affect the observed annual average even with stable prices and stable within-stratum utilization.
Direct standardization provides a useful diagnostic. Divide the population into clinically or demographically justified strata g. Let c(g,t) denote mean spending per standardized exposure in stratum g at time t. Let w(g,ref) denote fixed reference-population weights that sum to one. Then:
Changes in the standardized rate measure within-stratum evolution under a fixed reference distribution. Differences between the observed and standardized rates can be interpreted as composition effects only in relation to the specified strata and reference weighting method. They do not by themselves prove a causal effect of underlying disease burden. Unmeasured severity, provider intensity, insurance selection, and diagnostic conventions may remain confounded.
A robust study reports both crude and standardized trends. It should also report the reference population's period, age structure, geographic scope, and eligibility definition. Applying current risk weights to the past may produce a different historical attribution than applying the past risk weights to the present. Both are mathematically possible; neither automatically represents a unique causal counterfactual.
Claims frequency and severity
Frequency-severity analysis decomposes expected claim cost per exposure into the number of covered events per exposure and cost per event. For homogeneous events, expected cost per member-year equals expected frequency times expected severity. When events differ substantially in clinical intensity, further classification is required; the apparent severity effect may be an unmeasured service mix effect. Claims that bundle multiple services can increase apparent frequency or severity depending on how episode boundaries are drawn.
The statistical estimation also depends on the timing of claims. Paid claim count, incurred service events, denied claims, encounters lacking an insurer payment, and adjusted or voided claims require separate treatment. A rise in utilization induced by expanded coverage may represent fulfillment of previously unmet healthcare needs, while greater coding completeness may mechanically increase the number of recorded diagnoses. Their implications for benefits, access, and financial forecasting differ.
Coding intensity and risk scores
Diagnosis coding translates clinical information into administrative classifications, often used for risk adjustment. It affects reimbursement when funding formulas depend on recorded morbidity. The CMS Medicare risk adjustment materials document that context for Medicare Advantage. An additional diagnosis can change a risk score even if no new service is delivered, so growth in recorded morbidity need not be equivalent to growth in biological disease incidence or clinical utilization.
The original risk-adjustment coding incentives study (Carlin, Feldman, and Jung, 2024) investigates incentives arising from risk adjustment architecture. The Medicare Advantage coding variation study (Curto and colleagues, 2025) examines differences in coding intensity across organizations. An updated coding-pattern analysis published in 2026 provides estimates that differ from other prominent approaches after accounting for policy and methodological changes. This is an analytically consequential disagreement about estimated magnitude, not a basis for averaging incompatible estimates into a single point number. The studies depend on comparator populations, diagnosis data, risk models, selection adjustments, and study periods. Their results should be presented together with these identification assumptions.
The relevant operational distinction has three layers: actual morbidity burden, documented clinical diagnosis, and payment-relevant coded risk classification. A plan can improve coding capture without changing underlying health status, while an epidemic may change health status before recorded diagnoses fully capture it. Changes in classification systems or reporting requirements can cause discontinuities. Each layer requires independent validation.
Selection, churn, and differential retention
Enrollment transition affects medical trend because the people who remain insured may differ systematically from those who enter or leave. A high-cost individual's termination can lower an insurer's observed PMPM cost without any improvement in clinical care prices or efficiency. Conversely, expanded coverage can increase observed utilization during the initial period as deferred needs are met. Provider network changes may redirect members toward providers with different prices, producing a mix effect linked to insurer design.
Claims for people continuously enrolled in both periods provide a useful matched-cohort view. That cohort is itself selected by stable enrollment and must not be confused with the full risk pool. A second all-member analysis with proper exposure normalization is necessary. The difference between cohort results is an object for interpretation, rather than a mechanical adjustment that proves selection caused the entire difference.
Morbidity in long-duration healthcare promises
Long-term care and disability arrangements introduce service duration and care-state progression. Cost may grow because prices per day of care rise, because the probability of entering a benefit state changes, or because the duration and intensity of an existing benefit spell increase. For such products, a claim event can extend over multiple calendar years. Standardized claim incidence and care duration are analytically distinct. An insurer's liability may be sensitive to future wage and facility costs even when reported current-year paid expenditure is stable. This connection to Long-Term Care and Disability Insurance broadens the relevance of medical inflation attribution without redefining that field's morbidity, longevity, or disability-specific research questions.
Data, Measurement, and Validation
Source hierarchy and reconciliation
The evidence base is deliberately heterogeneous. CMS national accounts report US expenditure flows and an official price-volume methodology. OECD and WHO accounts support international sector and financing comparisons. HCCI reports commercial claims by service and insured population. KFF observes employer insurance premiums. BLS measures consumer prices. Academic studies investigate negotiated provider rates and coding incentives. Insurers' internal claim triangles and contracting data, when available, can link these aggregate measures to a particular liability. No individual source by itself identifies the full causal chain.
A research table should attach to each observation an institutional payer, country and jurisdiction, period, currency and inflation reference, product, coverage definition, exposure unit, service unit, provider sector, data maturity, and accounting treatment of risk transfers and rebates. This is essential because two reported series may have the same nominal label but different contents. Vintage control is especially important: the CMS NHEA methodology and revision record explains that the historical accounts undergo comprehensive revisions that can update entire earlier time series. Estimates from two publication vintages should not be combined as though they came from a single unrevised panel.
Actuals, price deflators, and forecasts
The 2024 national expenditure findings describe historical expenditure data and a 2024 aggregate growth rate of 7.2 percent. The companion CMS infographic attributes 2.5 percentage points to medical price change and 4.7 points to nonprice factors. These are descriptive national-account components. They do not separately identify the effects of insurer premiums, changes in diagnosis coding, or any one class of insured people. A decomposition that forces each of these micro-level factors into the CMS nonprice total would require new bridging data.
The 2025–2034 national expenditure projections published in June 2026 forecast almost $9.0 trillion of US national health spending by 2034, approximately 20.6 percent of GDP, under the modeled assumptions. These are projections, not realized outcomes for 2034 and not evidence that a specific insurer must experience a particular rate. Their usefulness here lies in demonstrating how demographic shifts, utilization, legislative assumptions, and prices can be simulated over time. Uncertainty intervals and alternative policy assumptions should accompany any operational use of such projections.
For international comparison, the OECD health-sector price comparison uses healthcare-specific purchasing power parities and a standardized basket to distinguish price-level differences from measured health-service volume across countries. It reports that the healthcare price level in the United States and Switzerland was 52 percent above the OECD average in the stated 2023 comparison. This is a cross-sectional price-level statement rather than an annual US medical inflation rate. The same publication discusses standardized healthcare volumes, which also depend on comparability and service definition. Exchange-rate conversions would answer a different question and should not be substituted for health-specific PPP analyses without an explicit rationale.
A research-ready claims data architecture
A reproducible data extract should contain, subject to applicable privacy and governance requirements: a stable deidentified member identifier; eligibility start and end dates; product and plan identifiers; service date and paid date; claim version or reversal flags; service and episode codes; diagnosis and risk-score fields; provider identifiers and locations; billed, allowed, plan-paid, and patient-paid amounts; units and place of service; pharmacy product identifiers; benefit and network status; and documented net rebate or settlement adjustments where available. Deriving quantities solely by counting rows is unsafe because a single episode may contain many lines and a single claim may be later adjusted.
The production sequence is: confirm data authority and privacy protections; freeze data vintage; identify relevant calendar/service periods; construct consistent membership exposures; reconcile claim totals to general ledger or actuarial triangles; deduplicate and process reversals; stabilize service code crosswalks; adjudicate exceptional payments; compute price and volume indices within common matched services; distinguish newly introduced categories; standardize member risk; estimate uncertainty; and trace recognized insurer-paid amounts into pricing and reserving. Each stage should write a machine-readable reconciliation output and exceptions log.
Uncertainty and validation
Several forms of error warrant explicit separate treatment. Sampling uncertainty arises when observed providers and members represent a larger population. Claims runout uncertainty arises because claims incurred late in the observation window have yet to be paid or reported. Misclassification can arise from changed coding standards or service definitions. Model uncertainty reflects alternative risk groups, index weights, episode definitions, or price concepts. Parameter uncertainty is the finite-sample error in estimated risk and pass-through coefficients. Structural uncertainty concerns the relevance of historical relationships to future institutions, technologies, or policy conditions.
For a matched-price index, a bootstrap should resample at a dependence-aware unit such as member, employer, provider, or market, rather than individual claim lines when lines from one episode are correlated. Weighting uncertainty and rare costly events may require stratification or appropriate heavy-tail robust intervals. For a reserves study, a stochastic model of claim reporting and settlement should be validated against holdout accident or service periods and tested for sensitivity to assumption changes. For a pass-through regression, credible inference requires adequate observations over contract renewal cycles and common shocks, with appropriately clustered uncertainty estimates.
A source-level audit must distinguish published data from inferred interpretations. The original HCCI employer-claims report and CMS national accounts can both illustrate historical US healthcare spending trends, but a formal difference between their price estimates is not a valid test of accuracy when their populations and expenditure scope differ. The right reconciliation reconstructs both quantities under a common service basket and payer definition before comparing them, or documents why such reconstruction is impossible.
Sensitivity specification
The minimum sensitivity suite has three sets of choices. The population suite uses all eligible member-months, continuously enrolled cohorts, and age-sex-geography standardized populations; differences indicate compositional dependence. The claims suite compares paid versus incurred runout-adjusted experience, pre-rebate versus net-rebate spending when rebate data are available, inpatient versus outpatient episodes, and matched-service versus full-basket trends. The financial suite examines realized allowed rates, plan-paid benefits, prospective premium indications, and liability remeasurement at explicit lags.
Each sensitivity scenario should preserve the observed total under its own coherent data definition and show how allocations change. A large difference among decompositions is itself a result about identifiability and data architecture. A single preferred index without sensitivity bounds may conceal service mix and quality changes that are economically more important than the reported unit-price effect.
Observational identification and its limits
Identifying causality requires variation in an input that is plausibly independent of the potential outcome except through the mechanism being investigated. Provider-contract changes may be negotiated in response to anticipated utilization. Benefit changes may be introduced because a plan expects future claims to increase. Diagnoses are not assigned at random, and risk adjustment can affect coding behavior. A regression of medical spending on price changes is therefore vulnerable to simultaneity, omitted variables, and post-treatment controls.
Potential designs include policy discontinuities, staggered contract renewals with valid comparison groups, exogenous fee schedule changes, transparent difference-in-differences assumptions, and event studies with pretrend checks. Each design estimates an effect in a defined population and interval. The document's index equations, by contrast, are exact descriptions of observed changes, not causal estimators. Maintaining the division between identity, statistical association, and causal effect prevents false precision in the interpretation of medical inflation.
International Synthesis and Scientific Conclusions
Healthcare institutions determine the visible form of pass-through
Cross-country medical expenditure comparison begins with the financing boundary. Public provision, compulsory insurance, social funds, private supplementary coverage, and voluntary insurance can produce different observed price series and payment arrangements. OECD's health accounts framework identifies financing schemes at system level; it does not require all markets to operate through private annual insurance premiums. A negotiated provider tariff increase can be transmitted through insurer plan payments, a social contribution, a budget transfer, or a service volume limit according to the institutional design.
A national medical price index can be stable while a private insurer's claims trend rises because the commercial payer faces a different market price and mix. Conversely, an insurer's current claims may be stable while aggregate care expenditure rises owing to public program coverage, uninsured care, or a demographic shift. Comparative conclusions about inflation should therefore be stated separately for health-system price levels, national real volumes, and risk-pool-specific claims.
The underlying economics of medical trend
The accounting identity identifies where spending growth appears. Economic explanation asks why the parameters change. Provider bargaining and market concentration shape negotiated prices. Technology and treatment standards change service intensity and outcomes. Insurance coverage changes the point-of-care price faced by patients, influencing utilization subject to need and access constraints. Population ageing, morbidity, diagnostic accuracy, and enrollment selection change the composition of insured exposure. Regulation determines benefits, allowable rating, accounting, and the distribution of costs. These mechanisms can operate simultaneously and interact.
The international implication is that policymakers and actuaries need an explicit translation between clinical, financial, and statistical units. Importing an external benchmark trend into a portfolio without adjustment for payer type, service mix, geography, age, and contract terms can produce a misleading rate indication or reserve assumption. Importing a commercial claims price signal into a national public financing debate without accounting for institutional coverage can be equally misleading.
Reconciling divergent evidence
The apparently divergent evidence on US medical trend is informative. CMS national spending increased 7.2 percent in 2024; HCCI's historical employer-sponsored series registered 2.9 percent spending growth in 2019, with stronger price than utilization growth; KFF reported 5 percent family-premium growth for 2026; and updated Medicare Advantage coding studies do not agree on the magnitude of coding differences. These observations describe different years, populations, measures, and identification assumptions. They therefore provide complementary evidence about mechanisms rather than numerical estimates of one universal medical cost trend.
The necessary synthesis preserves the original estimand in each source. A premium survey measures a contract price paid for coverage. A health account measures financial expenditure for services and administration. A claims dataset can measure paid or allowed medical transactions. A risk-score study may estimate coding-related differences in administrative diagnoses. None should be substituted into another without a documented bridge.
Implications for insurer governance
An insurer's actuarial trend committee should maintain a source-traceable decomposition by service category, payer-paid perspective, population risk, and contract cohort. The pricing review should document trend selection and its lag between the experience and rating periods. The reserve review should assess sensitivity to reporting development, future settlement costs, and contractual inflation clauses. Governance should distinguish historical outcome, projected assumptions, and release of uncertainty. Where an insurer has limited data, prudent uncertainty presentation should replace claims of precise source attribution.
For supervisors, a trend review is more informative when it distinguishes provider price growth, utilization, risk mix, benefit changes, and expense margins. Financial controls should preserve underwriting and accounting boundaries and avoid relying on a general CPI number as the sole proxy for medical benefits. The application of international principles should be mapped to the actual local regulatory authority and version.
Distributional and welfare implications
The insured medical expenditure trend is a financial measure. Its welfare consequences depend on access, health outcomes, quality, risk sharing, and the ability to pay. A decline in costly avoidable admissions may reflect better care. A decline in necessary care following increased member cost sharing may reduce claims while worsening protection. A price increase for a more effective treatment can change both expenditure and health value. Accordingly, medical cost containment policies should evaluate clinical effects and financial protection alongside decomposition of spending.
The consequences also differ by household income, employer contributions, provider region, and risk pool. Premium pass-through may strain affordability, and a provider cost shock that is temporarily absorbed by a public budget may later affect taxes, coverage, or capacity. Quantifying those distributional outcomes requires additional data beyond the price-volume decomposition developed here.
Research agenda grounded in identified measurement gaps
Further empirical work would benefit from linked, privacy-preserving insurer and provider data containing actual allowed payments, insurance enrollment histories, contracting dates, clinical episodes, quality measures, and rebate settlements. Multi-payer comparisons need a common service and patient-risk classification alongside independent institutional funding records. Cross-country measurement needs stable PPP conventions and service mix comparability. Long-term care insurance requires explicit modeling of care duration and future service intensity. None of these proposals establishes a proprietary theory: they identify datasets and research designs needed to answer empirical questions left open by the currently available evidence.
Conclusion
Medical cost trend measures changes in the expense of defined healthcare benefits for defined insured exposures and periods. It is generated by multiple linked processes: the unit prices of comparable services, service quantities, clinical intensity, composition of care, population risk, coding, contract benefits, and payer-specific financial shares. Total expenditure can be decomposed exactly into price and quantity effects for a stable basket, but neither the assignment of interaction terms nor the construction of service categories is unique. Interpretation therefore requires transparent index choices and sensitivity analyses.
The study establishes five substantive conclusions. First, all reported cost growth rates must be attached to explicit financial objects and exposure denominators. Second, service price inflation and aggregate medical cost trend are separate measurable quantities, and their difference can reflect utilization, service mix, morbidity, and contract design. Third, risk adjustment and diagnostic coding introduce important changes in measured morbidity that should be compared with independent clinical and utilization evidence. Fourth, the path from provider cost to insurance premium and reserve is mediated by negotiated payment arrangements, benefit rules, experience periods, reporting lags, regulatory frameworks, and financial margins. Fifth, international comparison is scientifically meaningful when health-system accounts, purchasing power, insurance financing arrangements, and the specific sampled portfolios are reconciled.
The central analytical contribution is a unified evidentiary reading of established expenditure, index-number, risk-adjustment, and actuarial valuation methods. Its practical result is an auditable interpretation protocol: define the object; reconcile price, quantity, composition, and exposure; evaluate coding and data completeness; trace contractual payment; and validate prospective premium and reserve implications. This completes the source-based scientific analysis within the documented limits of available data, without advancing an untested new theory or implying that an empirical decomposition establishes causal effects.
Technical Appendix: Accounting Identities, Worked Examples, and Review Protocol
Identity and index checks
For the synthetic one-service illustration, initial spending is 100 × $200 = $20,000 and final spending is 110 × $220 = $24,200. The nominal change is $4,200, and the growth factor is 1.21. The exact midpoint-weighted price component equals ($220 − $200) × (100 + 110)/2 = $2,100. The midpoint-weighted volume component equals (110 − 100) × ($200 + $220)/2 = $2,100. The components sum to $4,200. Dividing each by initial spending yields 10.5 percentage points. Multiplicative growth reporting instead yields a 10 percent price factor and a 10 percent quantity factor, with a 1 percent interaction on the initial base. Both presentations are internally correct; mixing their contributions is not.
Reconciliation of exposures and insurer shares
An analyst should report, for each service month and claim cohort, the number of fully and partially enrolled member-months, total allowed amounts, contractually eligible costs, paid amounts, patient payments, and recognized recoveries. The relationship among these components should be documented with signed adjustments and a deterministic reconciliation to the financial record. A change in insurance plan-paid share may create a measurable plan-cost difference even if the total allowed payment is unchanged. A decline in membership with greater concentration of high-cost patients can raise PMPM expenses while reducing total expenditure.
Standardization assumptions
A direct standardization requires nonnegative weights, a fixed comparison population, mutually comprehensible strata, and sufficiently observed within-stratum mean costs. Small strata, changes in coding, and unobserved risk can destabilize estimates. A sensitivity table should compare the unadjusted rate, direct age-sex standardized rate, morbidity-adjusted rate, and a matched continuous-enrollment rate. Each must state its estimand. A regression model does not confer causal identification merely through additional covariates.
Premium and liability boundary checks
Historical incurred benefits and future projected benefits are different periods. Rate filings may include indicated trend, benefit and risk changes, expenses, and allowed margins. An accounting liability includes only the cash flows and contract obligations under the applicable framework. For claims with multi-year care benefits, the projection must evaluate future payment rates; for already priced, short-tail claims, inflation may have a different effect. Scenario design should document expected nominal inflation, real utilization, service mix, covered exposure, and payment delay separately.
Minimum scientific audit record
A complete internal audit record identifies: the source URL and reviewed claim; source type and vintage; the observed variable and unit; the inference performed; denominator and data-period definition; applicable actuarial or legal framework; open factual uncertainties; consistency of chapter citations with the full bibliography; and the checks actually executed. Results should be recorded as tested or untested rather than being inferred from the existence of a file. The original source hashes must remain stable after final approval; a substantive text change requires renewed review and a new fingerprint.
References
OECD, Eurostat, and World Health Organization. (2017). A System of Health Accounts 2011: Revised Edition. OECD Publishing, Paris. https://doi.org/10.1787/9789264270985-en
OECD. (2025). Prices in the health sector. 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/prices-in-the-health-sector_962656ec.html
OECD. (2025). Health expenditure by financing scheme. 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/health-expenditure-by-financing-scheme_074bead0.html
World Health Organization. (2026). Global Health Expenditure Database: Documentation Centre. WHO, Geneva. https://apps.who.int/nha/database/DocumentationCentre/en/
Centers for Medicare & Medicaid Services. (2026). National Health Expenditure Fact Sheet: 2024 Historical Data. CMS Office of the Actuary. https://www.cms.gov/data-research/statistics-trends-and-reports/national-health-expenditure-data/nhe-fact-sheet
Centers for Medicare & Medicaid Services, Office of the Actuary. (2026). National Health Expenditure Accounts (NHEA): Methodology Paper, 2024—Definitions, Sources, and Methods (updated January 2026). https://www.cms.gov/files/document/definitions-sources-methods.pdf
Centers for Medicare & Medicaid Services, Office of the Actuary. (2026). National Health Expenditures: 2024 Highlights. CMS. https://www.cms.gov/files/document/nhe-infographic.pdf
Hartman, M., Martin, A. B., Lassman, D., Catlin, A., and the National Health Expenditure Accounts Team. (2026). National health care spending increased 7.2 percent in 2024 as utilization remained elevated. Health Affairs, 45(2). https://doi.org/10.1377/hlthaff.2025.01683
Fiore, J. A., Sisko, A. M., Poisal, J. A., Smith, S. D., Cuckler, G. A., Madison, A. J., Keehan, S. P., Rennie, K. E., and Feehley, N. J. (2026). National health expenditure projections, 2025–34: Strong utilization growth initially, legislative impacts later. Health Affairs. https://doi.org/10.1377/hlthaff.2026.00642
U.S. Bureau of Labor Statistics. (2026). Measuring Price Change in the CPI: Medical Care. BLS. https://www.bls.gov/cpi/factsheets/medical-care.htm
U.S. Bureau of Labor Statistics. (2026). Improvements to the CPI Health Insurance Index. BLS. https://www.bls.gov/cpi/additional-resources/improvements-cpi-health-insurance-index.htm
Health Care Cost Institute. (2021). 2019 Health Care Cost and Utilization Report. HCCI. https://healthcostinstitute.org/all-hcci-reports/2019-health-care-cost-and-utilization-report/
Cooper, Z., Craig, S. V., Gaynor, M., and Van Reenen, J. (2019). The price ain't right? Hospital prices and health spending on the privately insured. Quarterly Journal of Economics, 134(1), 51–107. https://doi.org/10.1093/qje/qjy020
KFF. (2026, October 22). 2026 Employer Health Benefits Survey. Kaiser Family Foundation. https://www.kff.org/health-costs/2026-employer-health-benefits-survey/
Centers for Medicare & Medicaid Services. (2026). Rate Review Data. CMS Marketplace. https://www.cms.gov/marketplace/resources/data/rate-review-data
Centers for Medicare & Medicaid Services. (2026). Medical Loss Ratio. CMS Marketplace. https://www.cms.gov/marketplace/private-health-insurance/medical-loss-ratio
Centers for Medicare & Medicaid Services. (2026). Medicare Advantage Risk Adjustment. CMS. https://www.cms.gov/medicare/payment/medicare-advantage-rates-statistics/risk-adjustment
Carlin, C. S., Feldman, R., and Jung, J. (2024). The mechanics of risk adjustment and incentives for coding intensity in Medicare. Health Services Research, 59(3), e14272. https://doi.org/10.1111/1475-6773.14272
Curto, V. E., Politzer, E., Anderson, T. S., Ayanian, J. Z., Souza, J., Zaslavsky, A. M., and Landon, B. E. (2025). Coding intensity variation in Medicare Advantage. Health Affairs Scholar, 3(1), qxae176. https://doi.org/10.1093/haschl/qxae176
Albanese, J., Aramanda, A., Brooks, J., and Klomp, C. (2026). An updated analysis of coding pattern differences in Medicare Advantage. Health Affairs Scholar, 4(1), qxag010. https://doi.org/10.1093/haschl/qxag010
IFRS Foundation. (2026). IFRS 17 Insurance Contracts: Standard and Supporting Overview. https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
European Union. (2009, as amended). Directive 2009/138/EC, Article 77: Calculation of Technical Provisions, EIOPA Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2160_en
International Association of Insurance Supervisors. (2026). Insurance Core Principles and ComFrame: Online Tool. IAIS. https://www.iais.org/icp-online-tool/
National Association of Insurance Commissioners. (2026). Statutory Accounting Principles. NAIC. https://content.naic.org/insurance-topics/statutory-accounting-principles