On 20 May 2020 the RSS Official Statistics Section (OSS) in collaboration with the Health Statistics User Group (HSUG) delivered a webinar which focussed on sharing experiences and discussing approaches to the methods being used to produce the death statistics for the current COVID-19 pandemic in the four nations of the UK.
This was the first in a series of webinars organised jointly by OSS and HSUG on topics relating to Covid-19. The webinar was split into two segments, the first represented by speakers who were producing the data while the second half moved on to the user perspective and Q&As.
The following is a general overview of the presentations by each speaker representing those producing the data on the day (Individual presentations available on the HSUG website):
Nick Andrews - PHE (Public Health England)
Myer Glickman - ONS (Office for National Statistics)
Myer outlined the coding definitions for Covid-19 and issues around the coding of confirmed and suspected Covid-19. He spoke about the different reports that were produced to various timescales and for different purposes. Graphs on the cumulative numbers of deaths reported by the various methods were shown. The data flows for death registration in England and Wales were shown and future analysis plans presented. The work on excess deaths was explained (current deaths vs five-year average) and more work was currently being done on this about ‘what it actually tells us’.
Nick explained some of the key outputs including the weekly all-cause mortality and the Euromomo (the European mortality monitoring) system. He outlined the following three sources the PHE data series on deaths is drawn from:
- Deaths occurring in hospitals, notified to NHS England by NHS trusts
- Deaths notified to local PHE Health Protection Teams in the course of outbreak management
- Laboratory reports where a person has had a laboratory confirmed COVID-19 test linked to death reports from electronic hospital records.
The data from each of these sources are validated and merged into a single dataset, removing duplicate records. Nick presented some of the outputs from the system and highlighted some of the strengths and limitations of the approach.
Scott Heald - PHS (Public Health Scotland)
Julie Ramsay - NRS (National records for Scotland)
Scott outlined the organisational arrangements in Scotland and the recent formation of PHS from several organisations. The different data producers in Scotland work closely to ensure statistics are aligned and coherent. Each organisation signposts the others. He said PHS received daily lab confirmed diagnoses of those who have died. Graphs of the daily lab confirmed deaths were presented too.
Julie explained that NRS was monitoring suspected (but not confirmed) as well as confirmed cases of Covid-19 and as a result the time series was higher. The difference was shown graphically, and further graphs were presented that showed the following:
- Deaths by week of registration, showing the excess weekly deaths compared with the average of the last 5 years
- Excess deaths by underlying cause
- Death rates by deprivation quintiles, illustrating a rate 2.3 times higher in the most deprived areas compared with the least deprived areas
- Covid-19 death rates by urban/rural classifications
Gareth John - NHS Wales Information Service
Gareth explained the two major source of information used in Wales; ONS death registrations, and the Public Health Wales rapid surveillance data. Use was also made of the electronic Master Patient Index (e MPI) and the mortality surveillance e-form. He presented graphs highlighting some of the differences and referenced the blog post written by the chief statistician recently that explained the various mortality data sources. He expressed although registration data are reasonably timely there are still delays, and it is important to produce estimates of the up to date picture to inform where Wales is on the trajectory. Gareth outlined a capture-recapture method he was using and set out the formulae. The method was producing better information than just relying on the source data, but required the partial data being representative of the whole. He said the situation was being monitored closely and ideally a third source would help validate the method.
The session then moved onto the user perspective and a Q&A (the responses to the Q&A will be posted on the HSUG website).
Neil Bendel from Manchester Health and Care Commissioning was invited as lead discussant to give a general user perspective and his personal view on aspects of the data he saw as important from a local user perspective in England. He outlined the following vital six points:
- Timeliness: it was important to improve timeliness of data for local leaders and elected members.
- Comparability and consistency: concerns about different descriptions and coverage although more to address this is now emerging. It is important to compare local with national trends.
- Granularity: data needed to a greater degree of detail e.g. PHE data not available by Local Authority (LA). It is also important to obtain individual data locally on a regular basis.
- Content: important to get more detail on local perspective e.g. by Black, Asian and Minority Ethnic (BAME) groups, linking to local populations.
- Flexibility: requires the flexibility of record level data in order to examine key local issues.
- Freedom: to share and link data to add value to analyses.
Overall, the webinar was very informative and the contributions from each speaker were well received, especially since this is a very busy period for them and therefore their time was much appreciated. Details of further webinars will be publicised and posted on the HSUG website.
Written by Hira Naveed and Richard Willmer
Hira Naveed is the RSS Official Statistics Section (OSS) Secretary and a Data and Research Analyst at Cancer Research UK. She has been nominated to stand for the RSS Council for the 2020 election.
Richard Willmer is a freelance consultant working with the Health Statistics Users' Group (HSUG).