International Comparison

International Comparison of Health Systems

International comparison in health economics sets countries side by side on what their health systems spend, the prices they pay, the volume of care they deliver and the health their populations achieve. It is the main way analysts ask whether a high-spending system buys more care or pays more for the same care, and whether a reform that worked abroad could work at home. Most of the effort goes into making the figures comparable before any lesson is drawn, because currency conversion and population age structure can each reverse a ranking. Health technology assessment bodies meet a narrower form of the problem when they judge whether a cost-effectiveness result from one country holds in another.

Purposes and design of a comparison

Countries rarely run controlled experiments on their own health systems, so variation between countries is one of the few sources of evidence on how financing, payment and delivery arrangements perform at scale. A comparison can describe how countries differ, place one system against its peers, try to explain a difference or test whether a policy could transfer. Setting a national system against peers is a form of benchmarking applied to whole systems rather than to hospitals. Explaining a difference needs a credible account of confounding, and policy transfer also needs evidence that the mechanism would operate under domestic health system organisation and financing.

The question should be fixed before any indicator is chosen, whatever an international database happens to publish. Country selection should be stated in advance, because a peer group chosen after seeing the results can make almost any system look good or bad. The country is not always the right unit either. Where policy is set by regions or insurers, a national average hides the variation that matters.

Making spending comparable

Spending comparisons rest on a shared accounting framework. A System of Health Accounts 2011, published jointly by the OECD, Eurostat and WHO (revised edition 2017), records each health care transaction on three axes, the function of the care consumed, the provider that delivered it and the financing scheme that paid for it. Its headline aggregate is current health expenditure, which leaves out capital formation. The manual proposes adding capital transfers as an extra aggregate and calls for research on whether this would improve comparability. Ratios built on these accounts, such as health spending as a GDP share, have their own page.

Spending in national currencies must then be converted into a common unit. Eurostat describes purchasing power parities (PPPs) as indicators of price level differences across countries and notes that market exchange rates reflect more than price differences, including financial transactions and expectations in currency markets. Economy-wide PPPs convert spending into a common measure of the resources a country gives up. Health-specific PPPs remove differences in the prices of health goods and services, so the converted figure approximates the volume of care.

$$Q_j = \frac{S_j}{\text{PPP}^{h}_{j}}$$

where $Q_j$ is the volume of health care per person in country $j$ at the prices of the reference country, $S_j$ is health spending per person in national currency and $\text{PPP}^{h}_{j}$ is the health-specific PPP, the units of country $j$'s currency that buy the same health care as one unit of the reference currency.

Health PPPs are harder to measure than most. The SHA 2011 manual explains that publicly provided services often have no meaningful market price, so their parities are usually built from input prices such as staff wages. It warns that this input method ignores productivity differences, so a country that delivers care more efficiently would show no sign of it.

Splitting a spending gap into prices and volumes

Once spending is in a common currency, a ratio between two countries can be split into a price part and a volume part. A gap caused by high prices calls for action on pay, medicine prices and administration, while a gap caused by volume concerns how much care is delivered.

$$\frac{E_A}{E_B} = \frac{P_A}{P_B} \times \frac{Q_A}{Q_B}$$

where $E_A$ and $E_B$ are spending per person in countries A and B converted at economy-wide PPPs, $P_A$ and $P_B$ are relative health price levels (the health PPP divided by the economy-wide PPP) and $Q_A$ and $Q_B$ are volumes of care per person.

Papanicolas, Woskie and Jha compared the United States with ten high-income countries, mainly on data for 2013 to 2016. In 2016 the United States spent 17.8% of its gross domestic product on health care, while the others spent between 9.6% (Australia) and 12.4% (Switzerland). Utilisation rates were largely similar, and the authors concluded that prices of labour and goods, including pharmaceuticals, and administrative costs appeared to be the major drivers of the difference.

Making outcomes comparable

Crude death rates partly reflect how old a population is, so an older country can look worse even when its health system performs well. Direct age standardisation applies each country's age-specific rates to one common standard population.

$$R^{\text{std}}_{j} = \sum_{i=1}^{n} w_i \, r_{ij}$$

where $R^{\text{std}}_{j}$ is the age-standardised rate for country $j$, $r_{ij}$ is the death rate in age group $i$ in country $j$, $w_i$ is the share of the standard population in age group $i$ (the shares sum to 1) and $n$ is the number of age groups.

The standard is a convention, and changing it changes the rates. Eurostat's task force on the European Standard Population, reporting in 2013, replaced the 1976 standard with the unweighted average age structure of the EU-27 and EFTA countries. The report noted that a standard giving more weight to older ages, where mortality is higher, should be expected to raise standardised death rates, but its tests found that the rank order of countries for overall mortality did not change significantly between the candidate standards.

Mortality from all causes reflects much more than health care. Nolte and McKee therefore compared deaths considered amenable to health care before age 75 in the United States and eighteen other industrialised countries. Between 1997-98 and 2002-03 amenable mortality fell by 16% on average, against 4% in the United States. Equity needs separate indicators, such as catastrophic health expenditure and the universal health coverage index, because a national average says nothing about who bears out-of-pocket costs.

Composite rankings and the World Health Report 2000

League tables of whole health systems attract attention because they reduce many dimensions to one number. The World Health Report 2000 used five indicators, covering the level and distribution of health, the level and distribution of responsiveness, and the distribution of financial contribution, to assess the health systems of 191 member states. WHO's announcement placed France first and the United States 37th.

The weights behind such an index are value judgements. WHO derived them from a survey of 1,006 respondents in 125 countries, half of them its own staff, and gave health 50% of the weight, responsiveness 25% and fairness of financial contribution 25%. Alan Williams challenged the exercise in Health Economics in 2001, asking whether the report was science or marketing, and Murray and colleagues replied in the same journal. A composite rank is therefore best shown alongside its components, weights and uncertainty, since small changes can reorder countries that sit close together.

Association and causation across countries

An observed link between a system feature and an outcome across countries does not show that the feature caused it. Rich countries tend to spend more, live longer and have different institutions, so a regression on a few dozen countries cannot separate one feature from everything else that varies with income and history.

Stronger designs use change over time. A difference-in-differences analysis compares the change in a reforming country with the change in countries that did not reform, and assumes that the two groups would have kept parallel trends without the reform, which similar trends before it make more credible. A synthetic control builds a weighted combination of non-reforming countries that tracks the reforming country before the policy, and an interrupted time series uses the country's own earlier trend.

Transferring evidence between jurisdictions in HTA

Health technology assessment meets international comparison when evidence produced in one country is used in another. The ISPOR Good Research Practices Task Force report on transferability by Drummond and colleagues notes that a growing number of jurisdictions request economic data for pricing and reimbursement, and that there are many reasons why cost-effectiveness might differ between jurisdictions. Unit costs, resource use, clinical practice, comparators and the local reference case can all differ. The task force recommended strategies based on individual patient data and on decision-analytic modelling, in which local inputs replace foreign ones. Prices are compared across borders for another purpose in external reference pricing, where the conversion problems above also apply.

Worked example with two countries and two corrections

All numbers in this example are illustrative. Country A is the reference country and spends 6,000 USD per person on health. Country B spends 120,000 units of its own currency per person. The market exchange rate is 24 units of B's currency per USD, the economy-wide PPP is 30 and the health-specific PPP is 20.

1. Convert at the market exchange rate. B spends 120,000 divided by 24, or 5,000 USD per person, so A appears to spend 6,000 divided by 5,000, or 1.2 times as much.

2. Convert at the economy-wide PPP. B spends 120,000 divided by 30, or 4,000 USD per person in purchasing power terms, so A spends 1.5 times as much of its economic resources on health.

3. Convert at the health-specific PPP. B's volume of care is 120,000 divided by 20, or 6,000 USD per person at A's health prices, the same as A's. The volume ratio is 1.0.

4. Split the gap. B's relative health price level is 20 divided by 30, or 0.667, against 1.0 for A, so the price ratio is 1.0 divided by 0.667, or 1.5. The ratio of 6,000 to 4,000 equals 1.5 × 1.0, and all of A's extra spending reflects higher health prices rather than more care.

5. Standardise mortality. In A the death rate is 2.0 per 1,000 under age 65 and 40.0 per 1,000 at 65 and over, and 20% of the population is 65 or over. In B the rates are 2.5 and 45.0, with 10% aged 65 or over. A's crude rate is 0.8 × 2.0 + 0.2 × 40.0 = 9.6 per 1,000 and B's is 0.9 × 2.5 + 0.1 × 45.0 = 6.75, so B looks healthier. With a standard population of 85% under 65 and 15% aged 65 or over, A's standardised rate is 0.85 × 2.0 + 0.15 × 40.0 = 7.7 and B's is 0.85 × 2.5 + 0.15 × 45.0 = 8.875. A's rate is 7.7 divided by 8.875, or 0.868 of B's, about 13% lower, and the crude ranking is reversed.

6. Interpret. A spends 50% more of its resources on health for the same volume of care and has lower standardised mortality. The figures do not show that A's higher prices buy better care, since an input-based health PPP would miss any quality or productivity difference and mortality depends on much besides health care.

Where international comparisons mislead

Most errors come from treating a published figure as if it measured the same thing in every country, and six of them recur.

1. Exchange rates in place of PPPs. Currency movements change the apparent gap without any change in health care.
2. Total spending read as inefficiency. A gap can reflect prices, volumes or accounting boundaries, and only a price and volume split shows which.
3. Crude rates across different age structures. Older populations look less healthy until rates are standardised.
4. Ranks without uncertainty. A composite rank inherits the value judgements in its weights and the measurement error in its components, so close ranks may not differ.
5. Correlation read as cause. A few dozen countries that differ in income and history cannot isolate one system feature.
6. Copying a visible programme. A practice that works abroad may depend on financing, workforce or regulation the importing country lacks.

Sources

• Drummond M, Barbieri M, Cook J, Glick HA, Lis J, Malik F, Reed SD, Rutten F, Sculpher M, Severens J. Transferability of economic evaluations across jurisdictions: ISPOR Good Research Practices Task Force report. Value in Health. 2009;12(4):409-418. https://doi.org/10.1111/j.1524-4733.2008.00489.x
• Eurostat. Glossary: Purchasing power parities (PPPs). Statistics Explained, accessed 28 October 2026. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Purchasing_power_parities_(PPPs)
• Eurostat. Revision of the European Standard Population: report of Eurostat's task force. Methodologies and Working Papers, 2013 edition. https://ec.europa.eu/eurostat/documents/3859598/5926869/KS-RA-13-028-EN.PDF
• Murray CJ, Frenk J, Evans D, Kawabata K, Lopez A, Adams O. Science or marketing at WHO? A response to Williams. Health Economics. 2001;10(4):277-282. https://doi.org/10.1002/hec.623
• Nolte E, McKee CM. Measuring the health of nations: updating an earlier analysis. Health Affairs. 2008;27(1):58-71. https://doi.org/10.1377/hlthaff.27.1.58
• OECD, Eurostat, World Health Organization. A System of Health Accounts 2011: Revised edition. OECD Publishing; 2017. https://ec.europa.eu/eurostat/documents/3859598/7985806/KS-05-19-103-EN-N.pdf
• Papanicolas I, Woskie LR, Jha AK. Health care spending in the United States and other high-income countries. JAMA. 2018;319(10):1024-1039. https://doi.org/10.1001/jama.2018.1150
• Williams A. Science or marketing at WHO? A commentary on 'World Health 2000'. Health Economics. 2001;10(2):93-100. https://doi.org/10.1002/hec.594
• World Health Organization. The World Health Report 2000. Health systems: improving performance. Geneva: WHO; 2000. https://www.who.int/publications/i/item/924156198X
• World Health Organization. World Health Organization assesses the world's health systems. News release, 2000. https://www.who.int/news/item/07-02-2000-world-health-organization-assesses-the-world%27s-health-systems