Countries

The United Kingdom counts deaths three times. England and Wales register through the General Register Office and publish through the Office for National Statistics. Scotland counts through National Records of Scotland. Northern Ireland counts through its Statistics and Research Agency. The three use compatible definitions but different timetables. A UK figure is the sum of three series.

Deaths in England and Wales are counted by date of registration, in the weekly figures and the annual figures alike. Tables by date of death are published alongside. The difference matters. Deaths referred to a coroner — sudden, unexplained, or in custody — can take months to register. Weekly counts are therefore always incomplete at the end, and are revised upwards later. Cause of death is coded by software applying the ICD rules. An update to those rules in 2011 moved many deaths from stroke to vascular dementia, and the switch to new coding software, IRIS, in 2014 produced a second step. Dementia is the series most visibly affected. Public health surveillance sits with the UK Health Security Agency, which publishes weekly respiratory reports through the winter. Its announced releases appear in our release calendar.

Population estimates come from the 2021 census in England, Wales and Northern Ireland, and from Scotland's 2022 census, with mid-year estimates in between. The ONS publishes national population projections on a two-year cycle. The most recent is 2024-based, released in April 2026, and assumes long-run net migration of 230,000 a year, down from 340,000 in the previous round. The ONS moved its excess-mortality method to an age-standardised, trend-based approach in 2024, and revised its figures. Healthy life expectancy is published from the Annual Population Survey in three-year rolling windows, with a separate disability-free measure.

Three quirks are worth knowing: the registration delay, the three-nation structure, and the split of England's health statistics between the ONS, NHS England and the Office for Health Improvement and Disparities. Those three bodies do not always agree on definitions.

The United States publishes multiple-cause data going back decades, which is why much of the methodological literature on coding uses American data. Deaths are registered by state and compiled nationally through the National Vital Statistics System at the National Center for Health Statistics, part of the Centers for Disease Control and Prevention. Cause of death is coded centrally with automated software. Provisional counts appear within weeks and are revised as states report. Final annual data follow about a year later.

Life expectancy fell before the pandemic, in 2015, 2016 and 2017, in a way no other high-income country matched. Drug overdoses, suicide and a stall in cardiovascular improvement drove the fall. It is published annually. The pandemic fall was larger, and there has been a recovery since. Population estimates come from the Census Bureau, which also publishes projections. The 2020 census was conducted during the pandemic, and its quality has been questioned for some groups.

Health statistics differ widely between states, and national figures do not show that. Respiratory disease is tracked through the CDC's weekly reports. Vaccine effectiveness is estimated through networks of participating health systems.

Three quirks are worth knowing: the state-by-state reporting lag; the absence of a national identifier for linkage; and the decline before the pandemic, which makes the United States an outlier in any comparison of mortality improvement.

Statistics Netherlands publishes weekly deaths within days of the end of the week. That made it one of the first countries to show the pandemic's excess deaths in near real time. It also maintains a linked population register in the Nordic style. Doctors certify cause of death and the statistical office codes it, and multiple-cause data are published.

The National Institute for Public Health and the Environment runs infectious disease surveillance. It publishes weekly respiratory reports and effectiveness estimates, and it advised on the country's pandemic response. It publishes largely in Dutch, with English summaries. Long-term care sits under a separate insurance act, with spending among the highest in the OECD as a share of GDP. Its statistics are unusually complete: who receives what care, and where.

Population projections are published annually. Migration is the dominant assumption, and the most revised.

Excess-mortality estimates from Statistics Netherlands and from the institute have differed by method. The country also had a high care-home population before the pandemic, which shaped its mortality pattern in ways that complicate comparison.

Every resident has a personal identification number from birth or arrival. That number links the population register, the cause-of-death register, the patient register, the prescription register and the cancer registry, amongst others. That is why so many of the strongest studies of whole populations come from Denmark. Data is complete and can be easily joined.

Deaths are registered within days, so weekly counts are near-complete quickly. Cause of death is certified electronically and coded centrally. Statistics Denmark publishes population, deaths and projections. The Danish Health Data Authority holds the health registers, and its announced releases sit in our calendar. Statens Serum Institut runs infectious disease surveillance.

Known quirks: Access to the registers for research requires approval and is easier for Danish institutions, so the evidence is abundant but the analysts are few. The population is under six million, so rare events produce wide intervals.

Germany's statistics are federal in every sense. Deaths are registered locally, compiled by the sixteen state offices and published by the Federal Statistical Office. The lag has shortened since the pandemic. It remains longer than in neighbouring countries: complete national figures arrive about four weeks after the deaths, with modelled weekly estimates published in the meantime. There is no national mortality register linking deaths to health records, and no personal identifier joining datasets. As such, the population-scale linked studies routine in the Nordic countries are not possible. German evidence tends to come from insurance claims data, regional cohorts or the statutory health insurers' research institutes.

Cause of death is coded by the state offices. Some code automatically with Iris software. Many still code by hand. The Robert Koch Institute runs infectious disease surveillance and publishes in German. We do not read it for papers, and its figures reach the site only through the trackers. The federal office publishes population projections on a multi-year cycle, with variants rather than a single central path.

Known quirks are the lag, the absence of linkage, and the variation in coding between states. That variation can produce regional differences in cause-specific mortality that are partly administrative.

Japan reports the world's highest life expectancy and one of its lowest fertility rates, and both are measured well. The Ministry of Health, Labour and Welfare publishes the vital statistics, drawn from the family register system. Deaths are registered promptly, and cause of death is coded centrally. Monthly provisional figures appear quickly. The annual vital statistics report is the definitive series.

Infectious disease surveillance moved in April 2025. The National Institute of Infectious Diseases merged with the National Center for Global Health and Medicine to form the Japan Institute for Health Security, which publishes weekly reports. Each round of the population projections has revised them towards a smaller, older population. They come from the National Institute of Population and Social Security Research, on a five-year cycle tied to the census. Healthy life expectancy is an official target. It is measured from a survey question about limitation in daily life, which differs from the European definition.

The long-term care insurance system has been mandatory since 2000, and is the longest-running such system anywhere. It generates unusually detailed statistics on care use by level of need.

Known quirks: much of the primary material is in Japanese. The healthy-life measure is not comparable with Europe's. The census-to-projection cycle means assumptions can be five years old.