National Statistical

It’s more than two years since death certification underwent a major reform. In this blog, David Mais reviews what impact these changes have had on our mortality statistics.

In September 2024, the death certification process in England and Wales underwent its biggest reform in over 50 years. The reforms mean that a medical examiner must scrutinise all deaths not referred to a coroner, as well as collect new data on the characteristics of the deceased. These reforms were led by the Department of Health and Social Care, and implemented by partners from across government and the NHS. 

Today we have published three articles investigating how ONS death registration data has changed since the reforms were introduced, as well as assessing the quality of the new information gathered on pregnancy and ethnicity. 

What has changed? 

Prior to death certification reform, most death certificates would typically be completed based on information provided by the medical practitioner who attended the deceased, or the death would be referred to the coroner for investigation.  

However, this meant there was no independent scrutiny of non-coronial deaths.  Well-publicised reports and inquiries since the Harold Shipman case in 1998 (PDF, 1.45MB) reveal how this can pose a threat to public safety. Without a routine independent review of deaths, serious issues can remain undetected.  

The 2024 reforms mean all deaths must now undergo a form of independent review, and aim to promote appropriate coroner referrals, improve bereaved people’s experiences, and enhance data quality.  

For the ONS, the following were the key changes in terms of their potential impact on the death registration data which underpins all of our mortality statistics:  

• All deaths not referred to a coroner must now undergo statutory independent review by a medical examiner before registration. 

• Added space on the death certificate for additional information on the cause of death, in line with international guidelines. 

• Recording of further information about the deceased, including whether they were pregnant, if that contributed to the death, and their ethnicity. 

Assessing how the reforms affect data quality helps us to ensure continued understanding and interpretation of trends for users of the statistics. 

What do the data show? 

The first of our latest articles: Death Certification Reform, England and Wales: 2025, explores the impact of the reforms on time taken to register deaths, patterns in who certified the death, and changes in the quality of the cause of death information recorded. We compared the last full data year under the old system (2023) with the first complete data year after the reforms (2025), and 2026 data up to the end of August. 

Our analysis showed that it did take longer for deaths to be registered in 2025 than 2023. In 2023, it took an average of 7 days for doctor certified deaths to be registered, which rose to 10 days in England and 14 days in Wales in early 2025.  

By 2026, however, it had fallen to 7 days in England and 9 days in Wales, likely reflecting people becoming more experienced with the new system. As a result, the time taken to register deaths is now only slightly longer than before the reforms. These small increases are to be expected given the additional time required for medical examiners to fulfil their role under the new legislation. In addition, from a data quality perspective, the ONS does not consider the small increase sufficient to materially affect the accuracy, interpretation or reliability of mortality statistics. 

We also observed changes in the pattern of referrals to coroners. Although the number of deaths referred to a coroner has decreased, this is mostly made up of a decline in the number of deaths where the coroner did not believe there was a need for an inquest. This suggests there has been a decline in unnecessary referrals, and more appropriate involvement of a coroner. Also, the number of unknown or uncertified deaths has dropped from 13,676 deaths in 2023 to just 4 deaths in 2025. We expect this to improve data quality by ensuring the most appropriate certification route is followed to accurately record causes of death.  

Finally, has the quality of cause of death information been affected? This can be difficult to determine as every death is different, and certifiers can record a wide range of conditions or sequence of conditions as causes of death. One way we can measure quality is by assessing the proportion of deaths where causes are considered ‘ill-defined’ (such as recording a death as due to “old age” alone). We found that a similar proportion of deaths were ill-defined deaths before and after the reforms, suggesting quality had remained consistent. While it is important for specific causes of death to be identified, we also found that the majority of deaths with an ill-defined cause are in people aged over 80, where sometimes this is just not possible. This was also the case both before and after the reforms, which suggests that quality had remained consistent. 

What about the new information that is collected? 

Beyond changes to the certification process, the reforms introduced the collection of new information on pregnancy and ethnicity on death certificates. We have explored the quality of these new data and what they could be used for in mortality statistics. 

Our second article: Mortality by pregnancy status following Death Certification Reform, England and Wales: 2025 assesses the quality of new pregnancy fields included on death certificates. These record if the deceased was pregnant, and if so, did the pregnancy contribute to the death, which can enable consistent monitoring and reporting of deaths occurring during or within a year of pregnancy and be used to supplement existing maternal death surveillance.   

There are some signs that the new pregnancy information is being recorded accurately. In almost every case where the pregnancy was recorded as having contributed to the death, the separate cause of death data supported this conclusion. This shows that deaths linked to pregnancy on the new forms are being recorded accurately, but it’s harder to know whether any deaths are missed by this process. 

There are also some cases where pregnancy was recorded but not linked to cause of death, and here there is evidence that there may be errors in some cases. For example, where women aged 65 years and over and men had been recorded as having been pregnant in the previous year. The potential introduction of an electronic system in future could help catch and prevent such errors.  

Our final article: Mortality by ethnic group following Death Certification Reform, England and Wales: 2025 examines the reliability of the data on ethnicity now being recorded on death certificates. Better ethnicity data can help identify population-level disparities and emerging risks. We securely linked the new ethnicity information recorded on death certificates with other de-identified sources, such as Census 2021, to check they matched. We found that in most cases, we could match the deceased with their census record, and in over 90% of these cases, the ethnicity recorded was the same.  

When this analysis was disaggregated by ethnic group, the agreement between sources was variable. The level of agreement was highest for White British, Bangladeshi, Pakistani, Chinese and Indian ethnic groups. In addition, around one in five certificates did not record the specific ethnicity of the deceased. Nevertheless, these findings suggest that death registration data can provide a useful source of information on ethnicity, while also highlighting areas where data quality can continue to improve. 

Looking ahead 

The new data collected on ethnicity and pregnancy has the potential to provide additional insights into patterns of mortality across the population. We will continue to review the quality of this data as more becomes available to determine if and how we can use it to produce further breakdowns in our mortality data, such as deaths by ethnicity. 

These reforms represent a significant change to how deaths are certified. However, the research published today shows the data can continue to provide the high-quality information needed to produce our vital statistics and potentially expand them. As more data become available, we will continue to monitor these changes as part of our regular mortality statistics production.