Chief Statistician for Wales update: Explaining Covid-19 mortality data sources for Wales
Glyn Jones is Co-Director at ADR Wales and Chief Statistician for Wales. As Chief Statistician, Glyn regularly updates on the statistical issues in Wales. Here he discusses the different sources of data on coronavirus (Covid-19)-related deaths in Wales.
In this blog I wanted to help provide an understanding of the different sources of data on coronavirus (Covid-19) related deaths in Wales. There has been much public scrutiny of these figures over the last week, and we have today published, for the first time, the overall number of deaths in care homes in Wales occurring over the period of the outbreak.
How are COVID-19 deaths reported?
There are two primary sources of data being published for Covid-19 deaths across the whole population. These data are collected for slightly different purposes and have different levels of coverage.
It is important to reflect there are different ways a death could be attributed to Covid-19. The deceased may have had a positive test for Covid-19. However without a laboratory confirmed test, Covid-19 could still have been considered by clinicians to be an underlying or contributory cause of death. There are also some deaths which may be caused by the wider impact of Covid-19 on society, which cannot be directly related to Covid-19 in published statistics.
Public Health Wales (PHW) carries out rapid surveillance of Covid-19 mortality and published data on a daily basis, based on deaths in confirmed Covid-19 cases. These are the figures reported by the media at 2pm each day. These are collected through PHW’s statutory surveillance function and use a case-definition similar to the one the World Health Organisation suggests for Covid-19 deaths in confirmed cases. The headline figures tend to focus on new cases reported on that day, but these may include deaths which occurred many days previously. PHW also provide a series of data by day of death. They include deaths in hospitals, but also some deaths notified to PHW by residential care home or hospice settings with laboratory confirmed tests (see also below for more information on care homes).
The PHW rapid surveillance data add an important timely picture of the current position of the outbreak but the processes involved mean it will never be a complete picture of all Covid-19 relate deaths. The system does not capture data on Welsh residents if they died in England, nor Welsh residents who passed away at their own home. On a daily basis Welsh residents who passed away in English hospitals will be included in the data published by the Department for Health and Social Care in England. The number of deaths in previous days will also always be subject to revision as new data is reported.
The most complete and authoritative dataset of deaths to Welsh residents is published by the Office for National Statistics (ONS), including for individual local authorities and health boards. The data are underpinned by the legal processes of having to register all deaths. These data are published every Tuesday and are based on all deaths registered involving Covid-19 according to death certification, including those where laboratory testing was not carried out. These numbers take longer to prepare as they have to be certified by a doctor, registered and processed (so data published today, 5 May, includes registrations up to 24 April). These data provide a complete picture of data by local authority of residence, wherever the death occurred, including in England, and can be broken down by whether this was in a hospital, care home or at home.
How different are these data?
The ONS data are expected to be higher for the reasons described above, as illustrated in the chart below. What we have seen in recent weeks is a divergence between the sources, as the percentage of deaths that occurred outside hospital increased.
This chart is based on the day of death, rather than the figures reported daily by PHW. This is important as it provides a better understanding of the curve of the outbreak and whether we are seeing deaths declining overall, rather than it being a function of a delay in reporting which, for example is often the case on the weekend and can lead to volatility in the data. That’s why we have begun using the following chart in our weekly Covid-19 dataset and I would suggest this should be the focus of users and the media. As an example the chart includes two days from past two weeks where the numbers reported were particularly high due to the retrospective reporting of a number of deaths by particular health boards. But in the date of death line these will have been allocated to the appropriate date, providing a more accurate understanding of the trend.
What about deaths in care homes, or to care home residents?
As stated above, ONS capture data on all deaths including by place of death. This means we can measure the number of deaths that took place in care homes, although it does not tell us if a care home resident subsequently passed away in hospital after being infected with Covid-19. PHW are also notified of laboratory confirmed deaths in care homes where outbreaks/incidents are being investigated, and these are included in the totals above.
Today we began publishing a new series based on notifications of death to Care Inspectorate Wales (CIW) from registered settings. As the independent regulator for care home providers in Wales, providers are statutorily required to notify CIW of deaths to residents. This provides a picture of deaths to care home residents, whether that occurred within a care home or in hospital. It also enables us to understand the wider picture of deaths to care home residents where Covid-19 might not have been the underlying cause.
These data are based on the cause of death reported by the care home provider. They are not based on laboratory confirmed tests, and are therefore not directly comparable with the PHW data. Given there is an overlap between the two source, these data cannot be added together. We will be working with PHW and CIW to understand this overlap and seek to quantify how many of these deaths are included in PHW estimates.
Can I trust the rapid surveillance data published by Public Health Wales?
On Tuesday 29 April the Minister for Health and Social Services published a review on the provision of data on deaths through the PHW rapid surveillance process. This review included a number of actions we are now taking forward. The First Minister asked me to provide a system-wide oversight to ensure everyone can have confidence in the data provided by Health Boards and published by PHW. We have begun a series of weekly discussions with HBs to ensure this is the case and reporting is being undertaken on a consistent basis. We have already had assurances from all health boards they are now providing data to the same consistent definition.
As stated above these data are based on a rapid surveillance process and there are aspects which at this time are not captured through these data, and given the pace of data notification there will be revisions to previous days data. Hence the importance of the ONS data to provide a complete picture of Welsh resident Covid-19 deaths wherever they occur (including in England).
We’re constantly trying to improve our understanding of Covid-19 deaths
As well as improving the processes around collating the data we are also trying to work together across organisations to help understand the data better, and in turn help our understanding of the outbreak. This includes:
- The recent introduction of a consistent reporting process for rapid surveillance.
- Through this reporting process, the collection of data on ethnicity, key worker status, smoking, and underlying health issues.
- Work with ONS to reconcile data to understand the coverage of the PHW rapid surveillance data.
- Developing models allowing us to produce more timely estimates of the total number of deaths ahead of the publication of registered deaths.
- Working across CIW and PHW to understand the overlap in recording of deaths to care home residents.
ONS have also published plans for further analysis of Covid-19 related deaths data. This has already included the important analysis last week of deaths by geography and deprivation. In future weeks they will be publishing an analysis of deaths by ethnic group and by background in England and Wales.
This was originally published on Welsh Government's Digital and Data Blog.