Method, and what to distrust

Five metrics — life expectancy at birth, healthy life expectancy at birth, HALE at age 60, infant mortality and maternal mortality — are each converted to a percentile rank across the country pool, then averaged with equal weight. A score of 100 would mean best-in-world on all five.

Why percentile ranks rather than z-scores. Mortality data is heavily right-skewed: a handful of countries with very high maternal mortality would dominate a z-score composite and compress every rich country into an indistinguishable cluster. Percentile ranking is insensitive to that skew.

The pool. 160 countries with population over one million; microstates are excluded because a country of 40,000 people produces unstable vital statistics and would crowd the top of every list. 157 of those have complete data on all five metrics and enter the composite. For internal consistency, every per-metric rank quoted on this site uses that same 157-country pool.

No organisation publishes a "share of the health delivery system that is government-owned" figure, so it was built from the four OECD series that report ownership sector: hospital beds (33 countries), hospital establishments (33), long-term care beds (24) and long-term care facilities (31). Each country's index is the mean of whichever components it reports, and the component count is displayed against every country because an index from one series is not comparable to one from four. 19 countries report all four; 39 have at least one.

Primary care and clinics are not in the numeric index. OECD publishes no ownership breakdown for ambulatory providers, and the WHO Global Health Observatory has no facility-ownership indicator for any country. The index therefore measures acute and residential capacity only.

Primary care is instead covered qualitatively, using a signal in OECD's remuneration series (DSD_HEALTH_REAC_EMP@DF_REMUN). OECD reports GP pay split by worker status — ICSE93_1 for salaried employees, ICSE93_2T5 for the self-employed — and which of those series a country publishes indicates its predominant model. 33 countries classify as salaried (18), independent contractor (9) or mixed (6). This is an indicator of the dominant arrangement, not a headcount of doctors, and it describes employment rather than premises: whether a salaried Finnish GP works from a municipally owned health centre is documented in the country literature, not in any dataset.

Documented estimates. The United Kingdom does not report to the OECD ownership tables, and omitting it would badly distort the picture, since NHS trusts own essentially all acute capacity. It is included in the hospital-beds series at an estimated 95% — derived from an independent sector holding roughly 2% of acute beds and 6% of all hospital beds in England (2018), against 131,862 consultant-led NHS beds in England in Q1 2024/25 — and is flagged as an estimate everywhere it appears, never merged into the reported figures.

• WHO's catastrophic health spending indicator (SDG 3.8.2) returned no usable data through the API, so financial-hardship measures are absent from these rankings.
• The World Bank has archived its UHC service coverage indicator, so that column comes from WHO instead.
• There is no international dataset on public ownership of pharmaceutical manufacturing or health supply chains. That axis is assembled from primary sources and is the least systematic part of this research.
• The UK, Sweden and the Netherlands do not report to the OECD hospital-ownership table, so they are missing from the hospital-beds ranking despite the UK and Sweden being among the most publicly delivered systems in the world.
• The combined three-axis ownership list is a judgment, not an arithmetic result — the three axes have different country coverage and cannot be cleanly averaged.

Every chart and ranking on this site can be narrowed to the 38 OECD member countries using the OECD only switch in the navigation bar. The setting persists as you move between pages. The United States is an OECD member, so it never drops out of a filtered view; nor do Japan, Korea, the United Kingdom or any European comparator. What the filter removes is the long tail of non-member countries that widens the world rankings but is not the comparison most readers want — Cuba, China, India, Brazil and the Gulf states among them.

The delivery-ownership data is drawn from OECD sources and so barely changes under the filter; the financing and outcome rankings, which cover 155–160 countries, change substantially.

Four Python scripts produce everything on this site, and are committed alongside it: pull_wb.py fetches the World Bank indicators, pull_who.py the WHO GHO series, pull_delivery.py the OECD ownership series and the delivery index, pull_gp_model.py the primary-care employment classification, compose.py builds the percentile composite and efficiency rankings, and build_data.py emits the JSON the charts read. The OECD hospital-ownership table is fetched over SDMX in one command. Re-running them against the live APIs will produce updated figures as the underlying databases are revised.