International Comparison

How international comparison supports health-system learning

International comparison examines differences and similarities in health outcomes, spending, access, policy and system organisation across countries. This page explains how countries and indicators should be selected, how differences in definitions and context affect interpretation, and how comparisons can support learning without assuming that a policy successful in one country will work identically elsewhere.

The purpose may be descriptive, evaluative or explanatory. A comparison can show where countries differ, assess performance against peers or investigate how financing, delivery and policy choices may contribute to observed outcomes.

Defining the comparison question

A useful comparison begins with a clear question rather than a collection of available country statistics. The question determines which countries, indicators, years and contextual factors are relevant.

The specification should identify:

• The health outcome, cost, policy or system feature being compared.
• The countries or jurisdictions included.
• The population and time period.
• The analytical perspective.
• The reason countries are considered comparable.
• The hypothesised mechanisms linking system features to outcomes.
• Whether the aim is description, benchmarking, explanation or policy transfer.

Comparing total spending across countries answers a different question from comparing what health outcomes are achieved for that spending. The intended interpretation should be stated before indicators are selected.

Selecting countries

Country selection affects the conclusion. Selecting only countries known to support a preferred argument can create a misleading benchmark.

Countries may be selected because they share:

• Income level and fiscal capacity.
• Demographic and epidemiological profiles.
• Healthcare-system structure.
• Geography or regional institutions.
• Data availability and quality.
• Policy exposure or timing.
• A relevant reform or natural experiment.
• Similar starting conditions but different policy choices.

Selection should be justified and reproducible. A broader comparison can describe international variation, while a carefully matched comparison may be more useful for examining policy mechanisms.

Understanding the unit of comparison

The country is not always the most informative unit. Health policy, financing and delivery may vary substantially between states, provinces, regions or insurance systems within the same country.

The analysis should identify whether the unit is:

• A sovereign country.
• A devolved national system.
• A state, province or region.
• A payer or insurance arrangement.
• A provider network.
• A population subgroup within each country.

National averages can conceal regional variation and inequality. Comparisons should use the level at which the relevant policy is designed and implemented whenever possible.

Ensuring indicators represent the same concept

An indicator with the same label may be defined or collected differently across countries. Comparability requires examination of the numerator, denominator, coding rules, population coverage and data source.

For every indicator, the analysis should confirm:

• The precise definition.
• The population included and excluded.
• The measurement unit.
• The data source and collection method.
• The reference year.
• Whether the value is observed, estimated or modelled.
• Whether coding or reporting practices differ.
• Whether revisions have been applied consistently.

Hospital admission rates, for example, can vary because of disease burden, admission thresholds, primary-care access, bed supply and coding practice. They should not be interpreted as a direct measure of system quality without examining these mechanisms.

Comparing health outcomes fairly

Countries differ in age structure, disease prevalence, socioeconomic conditions and other determinants of health. Crude outcome rates may therefore reflect population composition rather than healthcare-system performance.

Age standardisation can improve comparability when age distributions differ. For age groups indexed by (i), a directly standardised rate can be expressed as:

$$ Standardised\ rate = \sum_{i=1}^{n} w_i r_i $$

where:

• (r_i) is the age-specific rate in group (i).
• (w_i) is the proportion of the standard population in group (i).
• (n) is the number of age groups.

The standard population and age groups should be stated because different standards can produce different values. Standardisation addresses measured compositional differences but does not remove every contextual influence.

Comparing expenditure and costs

Currency conversion is necessary when expenditure is reported in different national currencies. Market exchange rates reflect the price of currencies in financial markets, while purchasing power parities adjust for differences in domestic price levels.

The appropriate method depends on the question:

• Market exchange rates may be relevant for internationally traded goods and financial flows.
• General purchasing power parities support comparisons of overall economic resources.
• Health-specific purchasing power parities may better reflect the prices of healthcare inputs when reliable estimates are available.
• National-currency values may be most appropriate when assessing affordability within each country's own budget.

Expenditure should also specify whether it is total, public, private or out of pocket and whether it is expressed per person, as a share of gross domestic product or relative to another denominator.

For per-person spending converted using purchasing power parity:

$$ Spending\ per\ person = \frac{National\ health\ expenditure}{Population \times PPP\ conversion\ factor} $$

The base year, price year and conversion method should be reported. Mixing nominal values, constant prices and different years can create artificial differences.

Comparing healthcare prices, quantities and intensity

Higher spending can result from higher prices, greater quantities of care, greater treatment intensity or a different mix of services. Total expenditure alone cannot identify which mechanism is responsible.

An approximate expenditure relationship is:

$$ Expenditure = Price \times Quantity $$

In practice, each service has its own price, volume and intensity. Analysts may need to compare workforce pay, medicine prices, hospital activity, diagnostic use, length of stay and treatment complexity separately.

Differences in accounting also matter. One country may include long-term care or capital expenditure in health spending while another reports it elsewhere.

Describing health-system context

Country outcomes arise from institutions and policies that operate together. A comparison should describe the system context rather than attribute differences to one visible feature.

Relevant characteristics include:

• Revenue collection and risk pooling.
• Population coverage and entitlements.
• Public and private financing.
• Provider ownership and organisation.
• Purchasing and payment methods.
• Workforce supply and distribution.
• Primary-care and referral arrangements.
• Medicine pricing and coverage.
• Public-health capacity.
• Social care and long-term care.
• Data and governance systems.

The same payment method can produce different effects under different ownership, regulation and workforce conditions. System features should therefore be analysed as interacting components.

Benchmarking performance

Benchmarking compares a country's result with peers, an average, a frontier or a policy target. The benchmark should be appropriate to the decision and should not be selected merely because it makes one country appear favourable or unfavourable.

Benchmarking may examine:

• Population health.
• Access and waiting times.
• Quality and safety.
• Patient experience.
• Equity.
• Financial protection.
• Efficiency and productivity.
• Workforce and capacity.
• System resilience.

Ranks can attract attention but often exaggerate small differences and ignore uncertainty. The underlying values, definitions and confidence intervals should be reported alongside any ranking.

Assessing efficiency across countries

Efficiency analysis examines the relationship between resources used and outcomes produced. Cross-country productivity or frontier methods can identify variation, but the result depends heavily on which inputs and outcomes are included.

Challenges include:

• Health outcomes are influenced by factors outside healthcare.
• Spending may purchase improvements that are not captured by the selected indicators.
• Data quality and accounting rules differ.
• Countries may prioritise equity, resilience or patient experience differently.
• Current outcomes may reflect policies and investments from earlier periods.

A country identified as efficient under a narrow model should not automatically be treated as a best-practice system. The analysis should test alternative inputs, outcomes and time lags.

Comparing equity and financial protection

National averages can improve while inequalities remain large. International comparison should examine distributions within countries when equity is relevant to the question.

Measures may include:

• Outcomes by income, education, region, ethnicity or another relevant characteristic.
• Unmet need.
• Differences in waiting time or service use relative to need.
• Out-of-pocket spending.
• Catastrophic or impoverishing health expenditure.
• Coverage gaps.
• Rural and urban access.

Group categories may not be directly equivalent across countries. The analysis should preserve nationally meaningful categories while explaining the limits of cross-country comparison.

Explaining rather than merely describing differences

An observed association between a system feature and an outcome does not prove that the feature caused the difference. Countries vary in many characteristics that can influence both policy choices and health outcomes.

Potential confounders include:

• Income and fiscal capacity.
• Population age and disease burden.
• Education and employment.
• Inequality and social protection.
• Geography and population density.
• Public-health conditions.
• Culture and health behaviour.
• Historical institutions.
• Data quality.

Regression and panel-data methods can adjust for measured characteristics, but the number of countries may be small and unmeasured confounding may remain. Causal language should match the study design.

Learning from policy reforms

Policy changes can create opportunities to compare outcomes across countries or over time. Stronger designs examine pre-policy trends and use an appropriate comparison group rather than attributing every subsequent change to the reform.

Methods may include:

• Interrupted time-series analysis.
• Difference-in-differences analysis.
• Synthetic control methods.
• Comparative case studies.
• Process tracing.
• Mixed-methods evaluation.

The analysis should document implementation timing and intensity. A policy may be formally adopted in the same year across countries but implemented at different speeds or with different resources.

Identifying transferable practice

International comparison is often used to identify best practices, but a favourable outcome does not by itself show which component produced it or whether the practice can be transferred.

Transferability depends on:

• Similarity of the policy problem.
• Institutional and legal compatibility.
• Workforce and infrastructure.
• Financing and administrative capacity.
• Public and professional acceptability.
• Complementary policies.
• Implementation capability.
• Expected costs and opportunity costs.

A practice should be described as promising or transferable only after its mechanism and enabling conditions are understood. Copying a visible programme without the supporting system can produce different results.

Data quality and missing information

International databases often harmonise indicators, but harmonisation cannot eliminate every difference in national reporting. Metadata are therefore part of the evidence rather than optional documentation.

Important checks include:

• Completeness across countries and years.
• Breaks in time series.
• Changes in definitions or coding.
• Estimated or imputed values.
• Differences in survey response.
• Underreporting or delayed reporting.
• Revision policies.
• Whether data refer to residents, users or providers.

Excluding countries with missing data can change the comparison systematically. The analysis should report exclusions and examine whether data availability is related to system capacity or performance.

Combining quantitative and qualitative evidence

Quantitative indicators show patterns and magnitudes, while qualitative and institutional evidence can explain how those patterns developed. Combining them is particularly important when the number of countries is small or system features cannot be represented accurately by a single variable.

Useful contextual evidence includes legislation, policy documents, implementation studies, stakeholder interviews and historical analysis. The sources should be selected systematically and interpreted with the same attention to bias as numerical data.

A simplified example

Suppose three countries report avoidable hospital-admission rates of 800, 650 and 500 per 100,000 people. The country with the lowest rate also reports stronger primary-care access and lower out-of-pocket costs.

The comparison suggests a possible relationship, but it does not establish that primary-care organisation caused the lower admission rate. The populations may differ in age, disease burden, admission practice, coding and community-service capacity.

A stronger analysis would standardise the rates, verify definitions, examine several years, compare primary-care policies and investigate other explanations. Only then could the lower-rate country provide a credible policy lesson.

Common misunderstandings

International comparison is not a competition in which one country is declared the best from a small set of indicators. Performance depends on definitions, priorities, context and the outcomes selected.

Common misunderstandings include:

• Identically named indicators are not always measured identically.
• Higher spending does not automatically mean greater inefficiency.
• Lower utilisation does not automatically mean better health or efficiency.
• Purchasing power parity and market exchange rates answer different questions.
• A national average does not describe equity within the country.
• Correlation between a policy and outcome does not prove causation.
• A high rank may reflect a small or uncertain numerical difference.
• A successful policy is not automatically transferable.
• More countries do not compensate for poor comparability.

Interpreting an international comparison

The findings should be interpreted using the comparison question, country-selection logic, indicator definitions, time period and system context. The analysis should state which conclusions are descriptive, which are explanatory and which remain hypotheses.

A useful international comparison makes differences understandable rather than merely visible. It identifies plausible mechanisms, tests comparability and explains the conditions under which another country could adapt a policy or practice without assuming that health systems are interchangeable.