What the child and infant mortality statistics cover, how we produce them, and their quality and comparability. Includes definitions and latest, past and upcoming changes.
We produce annual statistics on child and infant mortality in England and Wales. These include characteristics of babies and parents, and analysis of risk factors in relation to deaths. We produce these by linking information from birth registrations, birth notifications, and death registrations in England and Wales.
It is a legal requirement to register births and deaths (and for healthcare providers to submit birth notifications) within a certain time and with correct information. This means the data is generally complete.
We and the registrars carry out thorough validation and quality assurance checks on the data to ensure our statistics are as accurate as possible.
These are accredited official statistics.
We updated this guide on 8 April 2025. Important changes to quality and methods include:
For more information on these latest changes, as well as any past and upcoming changes, go to Changes and their effects on comparability over time.
We present statistics on child and infant mortality for England and Wales in two separate datasets:
An infant death (a death that occurs within a year of birth) can occur in the same calendar year as the birth, or the subsequent calendar year.
Some users need deaths information for a group of infants born in a particular calendar year, while some need it for a group of infants that died in a particular calendar year.
The child and infant mortality (by year of death): 2023 dataset contains information on all infants that died in 2023 before their first birthday, that were born in 2022 or 2023.
The infant mortality before their first birthday (by year of birth): 2022 dataset contains information on all infants that were born in 2022, that died in either 2022 or 2023 before their first birthday.
Both datasets include breakdown by characteristics of the baby and mother, and age at death. These allow us to make useful comparisons.
These categories allow us to compare death rates – for example, between younger and older mothers, and between babies with low and normal birth weights.
We use different classifications for place of delivery between the two datasets.
For the child and infant mortality (by year of death) dataset, we get place of delivery information from birth registrations and birth notifications. We include breakdowns by the following places of delivery:
For the infant mortality before their first birthday (by year of birth) dataset, we only get place of delivery information from birth notifications. We include breakdowns by the following places of delivery:
Excel spreadsheet (XLSX format, 26KB)
Both datasets exclude:
Sub-divisional breakdowns (for regions, counties, metropolitan counties, and local authorities) also exclude births and deaths of non-England and Wales residents registered in England and Wales. However, these are included for the total England and Wales figures.
A parent must register a birth with a registar within 42 days.
Birth registrations include information about:
Midwifery staff present at the birth must submit a birth notification within 36 hours:
Once the baby's birth notification is available, the PDS notifies us.
A birth notification includes information that is not included on the birth registration, such as gestation length and the baby’s ethnicity.
A death must be registered by “the informant” (a relative, someone present at the death, the occupier of or an official from the public building where the death occurred, or the person arranging the funeral) within five days of it occurring.
However, sometimes death registrations can be delayed if they are referred to a coroner for investigation.
Death registrations include information about:
We link death registrations with corresponding birth registrations. This allows us to get more social and biological information about the baby and parents than we could from death registrations alone.
Where possible, we’ve also linked death registrations with corresponding birth notifications back to 2007 – the earliest year we can apply this type of linkage. This allows us to get more information about the baby, such as gestational age and ethnicity, than we could from death registrations and birth registrations alone.
Having a low gestational age and being from a Black or Asian ethnic group are two of the main risk factors associated with deaths in the perinatal (stillbirths and early neonatal) period. Linking these helps to provide better analysis of the relationship between gestational age and ethnic group with deaths.
The registrar links birth notifications with birth registrations, which creates a unique sequence number. We use this number to carry out our own re-linkage.
If the registrar has not been able to link a birth registration and birth notification (meaning the unique sequence number is not available), we use probabilistic linkage. This means we link birth registrations with birth notifications based on the likelihood of them being for the same baby.
We extract the total number of deaths occurring in a calendar year (known as the “data year” – for example, 2023) from the deaths registrations database. We do this approximately 9 to 10 months after the end of the data year to allow for late registrations.
We do not publish the small number of registrations registered after we extract the data. This is an appropriate balance between accuracy and timeliness.
We then link these deaths with birth registrations and birth notifications.
We usually cannot link around 3% of deaths (9.5% for the 2020 data year because of the coronavirus (COVID-19) pandemic – read more in Strengths and limitations). The main reasons are not being able to find a birth record, or the birth was registered outside of England and Wales.
We extract the total number of births occurring in a calendar year (known as the “data year” – for example, 2022) from the birth registrations database and birth notifications database. We do this approximately six to seven months after the end of the data year and then link the two.
Next, we extract the infant deaths occurring in the same year as the birth, and in the following year (for example, 2022 and 2023, respectively). We extract data for both years because an infant born near the start of the data year may die during that year before their first birthday, and an infant born near the end of the data year may die in the following year before their first birthday. We do this 10 to 11 months and 22 to 23 months after the end of the data year, respectively.
This means that the infant mortality before their first birthday (by year of birth) dataset is less timely than the child and infant mortality (by year of death) dataset. However, this time lag allows us to provide additional analysis.
We then link the birth and death occurrences to decipher how many infants that were born in a certain data year died before their first birthday.
These accredited official statistics were independently reviewed by the Office for Statistics Regulation in February 2013. They comply with the standards of trustworthiness, quality and value in the Code of Practice for Statistics and should be labelled "accredited official statistics”.
The Office for National Statistics (ONS) has developed Guidelines for measuring statistical quality based on the five European Statistical System (ESS) Quality Dimensions. These are:
We have integrated these considerations into the guide.
Over the years, we have changed how we produce child and infant mortality statistics. Sometimes this is to improve the statistics, and sometimes it is because of external changes (such as software advancements, changes to policy, or changes to legal definitions).
Changes can sometimes affect comparability of the statistics over time.
On 8 April 2025, we changed the name of the child mortality (death cohort) tables to “child and infant mortality (by year of death)”. This explains more clearly what we mean by “death cohort” (that the statistics are organised by year of death), and that the dataset includes statistics for infants as well as children. This name change applies to the 2023 data year onwards.
On 8 April 2025, we also changed the name of the infant mortality (birth cohort) tables to “infant mortality before their first birthday (by year of birth)”. This explains more clearly what we mean by “birth cohort” (that the statistics are organised by year of birth), and that the dataset is specifically about babies that die before their first birthday. This name change applies to the 2022 data year onwards.
The statistics themselves have not changed, so this does not affect comparability over time.
These changes are ordered by date, with the most recent first.
From the 2021 data year, we stopped classifying live births at 23 and 24 weeks as stillbirths. This had no significant effect on the stillbirth rate, so does not affect comparability over time.
On 1 October 1992, the definition of a stillbirth changed from 28 or more weeks’ gestation to 24 or more weeks’ gestation (Still-Birth (Definition) Act 1992). This means our statistics on stillbirths from the 1993 data year are not directly comparable with previous releases.
We have used different softwares to code cause of death over the years. From the 2021 data year, we updated our software from to Multicausal and Unicausal Selection Engine (MUSE) 5.5 to MUSE 5.8. In line with this, we updated the hierarchical classification system we use to classify ONS cause of death groups for stillbirths and neonatal deaths for statistics.
This means ONS cause of death groups from the 2021 data year onwards are not directly comparable with 2014 to 2020.
Read more about the impact of implementing Iris.
From the 2018 data year, we updated our software from Iris to MUSE 5.5. MUSE 5.5 increased the automation of coding and operated based on internationally agreed decision tables to reflect the ICD-10.
This change slightly reduced comparability with previous releases, but increased accuracy in coding causes of death.
Read more about the impact of implementing MUSE 5.5.
On 1 January 2014, we updated our software from Medical Mortality Data Software (MMDS) to Iris. Iris incorporated official updates to the ICD-10 approved by the WHO, helping to improve international comparability.
For stillbirths and neonatal deaths, Iris coded maternal condition mentioned on the death certificate to chapter XVI (certain conditions originating in the perinatal period) rather than chapter XV (pregnancy, childbirth and the puerperium).
These changes limited the comparability of the “antepartum infections” and “infections” cause of death groups from the 2013 data year with previous releases.
In 2021, we updated baby ethnic groups to better align with the groups used in Census 2021 back to the 2019 data year, meaning they are not directly comparable with previous releases.
In 2020, we introduced three new tables in the child and infant death (by year of death) dataset on the number and rates of live births, stillbirths, and infant mortality by ethnicity. These were previously available in our Births and deaths by ethnicity dataset for 2007 to 2019 .
This does not affect comparability of the statistics over time.
We have made several changes to how we present unlinked deaths (deaths that could not be linked to a birth) over the years.
For the 2020 data year, we created a separate category for unlinked deaths.
For the 2019 data year, we included unlinked deaths in all tables in the by year of death dataset. We included them in a “not stated” category in tables that analysed birth registration variables, and presented them separately in tables that analysed death rates for linked deaths.
Before this, we had left unlinked deaths out of some of the tables completely.
Since the 2017 data year, we’ve used the latest geographic boundaries for sub-divisional breakdowns.
Any changes to geographic boundaries can affect the comparability of sub-divisional breakdowns over time. Find specific changes in our Data sources guide to mortality statistics.
From the 2015 data year, we began extracting death registrations 10 to 11 months and 22 to 23 months after the end of the data year, respectively.
The change meant we captured 5% to 6% more late registrations, and that statistics from the 2015 data year are not comparable with previous releases.
Read more about our current approach in When we extract the data and carry out linkage for the child and infant mortality (by year of death) dataset.
From the 2015 data year, we combined our Birth cohort tables for infant deaths dataset and Pregnancy and ethnic factors influencing births and infant mortality dataset into one dataset: infant deaths before their first birthday (by year of birth) (previously known as infant mortality (birth cohort) tables in England and Wales). We made this change based on feedback from the infant mortality user consultation.
From May 2012, birth registrations started including the number of previous live births and stillbirths all mothers had had, rather than just those married mothers had had with the current or former husband. This change came about because of amendments to the Population (Statistics) Act 1938.
This simplified the question and led to improved accuracy in responses, but means statistics from the 2014 data year onwards are not directly comparable with previous releases.
Read more about the impact of this change on our birth statistics:
In 2012, we started categorising the statistics by the most advantaged parent’s socio-economic classification (a “household-level classification”). Previously, we had automatically categorised the statistics based on the father. This was because many mothers did not have a paid occupation or chose not to state their occupational details on the birth registration.
In 2011, we rebased the NS-SEC on the new Standard Occupational Classification (SOC 2010) to categorise socio-economic class. This change meant classifications were based on employment conditions instead of skills to more accurately describe modern occupations.
Both of these changes means statistics from the 2012 data year onwards are not directly comparable with previous releases.
From 1 September 2009, two females in a same-sex couple became able to register a birth under the Human Fertilisation and Embryology Act 2008.
From the 2010 data year, we categorise births registered to same-sex couples:
The numbers of births registered to same sex-couples is relatively small, so this change has a negligible effect on comparability over time.
We include footnotes in the datasets stating the number of births registered to same-sex couples included in the numbers.
We have used different revisions of the International Statistical Classification of Diseases (ICD) to code cause of death over the years.
From the 2009 data year, we switched from using ICD-10 v2001.2 to ICD-10 v2010.
From the 2000 data year, we switched to using the International Statistical Classification of Diseases 10th revision (ICD-10) to code causes of death, based on the recommendation of the WHO. This replaced the ICD-9, which we had used since 1979.
These changes limited comparability with previous releases, but resulted in:
In 1986, England and Wales introduced new certificates for registering neonatal deaths (infants that died aged under 28 days) and stillbirths in line with a recommendation from the World Health Organization’s (WHO).
These are different to standard death certificates – they do not assign a single underlying cause of death, and instead give certifiers more flexibility to report on multiple relevant diseases in both the infant and the mother.
This meant we could no longer directly compare a single underlying cause of death for neonatal deaths and stillbirths with postneonatal deaths (infants that died aged between 28 days and 1 year).
To address this, we developed a Classification system of broad Office for National Statistics (ONS) cause of death groups (PDF, 72KB), where a computer algorithm directs any mention of a “mechanism” that led to death on both death certificate types to the first appropriate cause group.
This allows us to directly compare stillbirths and neonatal deaths with postneonatal deaths, despite the differences in registration.
We currently have no plans to change the methods in the near future.
Mothers and Babies: Reducing Risk through Audits and Confidential Enquiries across the UK's (MBRRACE-UK) Infant mortality statistics are not comparable with ours because of the following differences.
| ONS | MBRRACE-UK |
|---|---|
| Includes deaths of children up to 15 years | Includes infant deaths only |
| Includes all live-born babies, including those born before 24 weeks | Excludes live-born babies born before 24 weeks |
| Includes stillbirths from 24 weeks' gestation in line with the legal definition | Includes stillbirths from 22 weeks' gestation |
Northern Ireland Statistics and Research Agency's (NISRA) Child and infant mortality statistics are not comparable with ours because of following differences.
| ONS | NISRA |
|---|---|
| Based on date of death occurrence – gives a more accurate representation of deaths in a given year and helps track trends over time | Based on date of death registration – reduces the delay in producing statistics, but result in less complete statistics because late registrations are not captured |
National Records Scotland's (NRS) Child and infant mortality statistics (part of their vital events statistics) are not comparable with ours because of the following differences.
| ONS | NRS |
|---|---|
| Based on date of death occurrence – gives a more accurate representation of deaths in a given year and helps track trends over time | Based on date of death registration – delays are shorter in Scotland, so little difference between death occurrence and registration dates |
A range of organisations and government departments use our child and infant deaths statistics to inform services and policies, and ensure targets are met. For example:
Other users include academics, researchers, charities, and the media.
For definitions of how we categories different ages at death, go to Age at death categories.
The matching of death registration records to their corresponding birth registration record and/or birth notification record, and birth registration records to their corresponding birth notification record.
The number of:
The number of deaths according to the date on which the deaths were registered, and the number of births according to the date on which the births were registered.
“The disease or injury which initiated the train of morbid events leading directly to death, or the circumstances of the accident or violence which produced the fatal injury”, in accordance with the rules of the International Classification of Diseases (excludes deaths that occurred under age 28 days).
Live births, stillbirths and linked infant deaths occurring annually in England and Wales, and associated risk factors.
Annual statistics on births and infant deaths based on babies born in a calendar year that died before their first birthday linked to their corresponding birth notification and their corresponding death registration.
Stillbirths, infant and childhood deaths occurring annually in England and Wales, and associated risk factors.
Supporting information for mortality statistics, which present figures on deaths registered in England and Wales in a specific week, month, quarter or year.
Supporting information for birth statistics, which present figures on births that occur and are then registered in England and Wales. Figures are based on information collected at birth registration.
Office for National Statistics (ONS), updated 8 April 2025, ONS website, quality and methods guide, Child and infant mortality In England and Wales quality and methods guide
Population Health Monitoring Group
health.data@ons.gov.uk
Telephone: +44 1329 444 110
Monday to Friday, 9am to 4pm