You asked
Purpose: Understand the underlying source data (raw) for CV-19 deaths.
Issue: ONS data here which feeds Worldometers and other sites, appears to be just an xls spreadsheet with manual data entries. No raw source data is given linking the totals to the various categories, subcategories by area/city.
Please answer the following questions:
1-Where is the raw data which supports the published official ONS statistics on Covid deaths? For example, if in week 44, Nottingham records 100 dead from CV-19 in hospitals, where is the source data for the 100 dead and the various hospitals in Nottingham which recorded this?
2-How is the underlying raw source data verified? For example, if in one hospital in Nottingham 50 people died in week 44 from CV-19, where can one view the clinical proof of such deaths which has been verified (and by whom)?
3-How does the ONS or official government numbers aggregate the information from across the country? Do you use a centralised database which is fed the numbers from regional NHS databases? If so, how is data quality checked and what systems or databases are you using?
Or is it a manual process? Can you provide a description or business process of how the numbers are aggregated and confirmed?
We said
Thank you for your request.
1
We are responsible for the production of mortality statistics in England and Wales, this is driven by information collected on the death certificates at death registration. More information on this process can be found in section 2 of our User guide to mortality statistics.
All of the conditions mentioned on the death certificate are coded using the International Classification of Diseases, Tenth Revision (ICD-10) the coded mortality data available up to end of December 2019 is coded to version 2014 of the ICD10 framework. From all of these causes an underlying cause of death is selected using ICD-10 coding rules. The underlying cause of death is defined by WHO as:
- the disease or injury that initiated the train of events directly leading to death, or
- the circumstances of the accident or violence that produced the fatal injury
Deaths in England and Wales are registered with an area of usual residence. The area of usual residence is attributed with a lookup code in line with our National Statistics Postcode Lookups (NSPL) references, this allows us to break down the death by different geography levels.
The raw data from registrations is provided to ONS from the General Registrar Office and then processed and coded against the subjects described in the user guide for mortality statistics, such as underlying cause of death, geography, occupation. From this we can safely aggregate the data and remove personal identifiers and publish accurate statistics.
We have produced several streams of (provisional) analysis for deaths that have occurred and registered this year in response to the pandemic. However, we are unable to provide you with the raw data in line with our disclosure control policy: Policy on protecting confidentiality in tables of birth and death statistics. Section 39 of the Statistics and Registration Service Act 2007 (SRSA) makes it an offence to release personal or potentially identifiable information.
If you require access to the raw data then you may be able to access the data via ONS secure research service as an accredited researcher: https://www.ons.gov.uk/aboutus/whatwedo/statistics/requestingstatistics/approvedresearcherscheme
2
Section 6 details the process undertaken by the doctor or coroner who will complete the medical certificate of cause of death.
When a death occurs, a doctor must complete a Medical Certificate of Cause of Death (MCCD), the doctor must have seen the deceased within the last 2 weeks of life. For deaths where the doctor had not seen the deceased within the last 14 days of life, these cases must be referred to a coroner. In the example you gave for 50 people who died of COVID-19 in one Nottingham hospital, it would likely be the certified doctors there completing the death certificates. However, if the cause of death was undetermined, this would be passed onto a coroner for further investigation i.e. autopsy.
Several changes have been made to the death certification and registration process under the Coronavirus Act 2020. Notable changes:
- The length of time for a doctor to have seen the deceased has increased from 14 days to 28 days during the emergency period.
- Another doctor may be able to complete the MCCD if the patients attending doctor is unable to as result of sickness, self-isolation or some other reason. Providing they can certify the cause to the best of their knowledge.
- If there is no doctor who can certify the cause, or the cause appears to be unknown, the registrar will be obliged to refer the death to the coroner before they can register the death.
For more information on these changes, please see Quality of mortality data during the coronavirus pandemic, England and Wales: 2020
These changes were largely implemented to allow more resource for treating patients and protect the workforce from COVID-19.
We do not hold the death certificates as the data is provided to us by the GRO. As stated in response to your first question, section 39 of the Statistics and Registration Service Act 2007 prevents us from publishing identifiable data obtained from the death certificates.
NHS Digital will hold any treatment plans and medical records and may be better placed to assist with this part of your enquiry. Alternatively, you can apply to access the data via the accredited researcher scheme.
3
Section 14 details how the mortality data is processed and verified by ONS Titchfield. Mortality data is loaded into our centralised database 'Life Events Data Repository' (LEDR) which are autonomously processed. ONS also code paper-based registrations manually at Titchfield. Section 14 also describes the process used to quality check the data produced before publishing to ensure accuracy. This includes extensive monthly examinations from our cause coders who look for inconsistencies in the outputs and potentially spot any errors in the variables. We also essentially reproduce the dataset and independently carry out checks on the data to ensure the data matches what has been produced by LEDR. Additionally, ONS have a large amount of process-based checks which involve checking row counts, unique variable counts and format checks which ensures product accuracy and consistency with previous publications.