COVID Vaccines - How to identify Shielding/CEV and At Risk/CV groups

[HelenCEBM]

Background

There were two groups of clinically vulnerable people in relation to COVID and the vaccination programme.

1. "Shielding"

AKA Clinically Extremely Vulnerable (CEV) people who were advised to shield to avoid exposure to the virus.

• A shielding ("high risk from COVID-19") flag was added to their GP record through various possible routes, for example through the initial nationally applied algorithm (as far as I know the details are not available), by their GP or hospital doctor, or later by the Q-COVID algorithm (February 2021).
• Patients may also have been removed by various routes, e.g. through their condition improving. These patients have a "low risk" or "medium risk" flag added to their GP record to indicate they were no longer considered "high risk".
• The CEV group is therefore relatively simply defined as: having a flag for being at increased risk from COVID and NOT having a more recent flag for being at medium or low risk.
• For vaccination purposes, this group were in group 4 of the 9 initial vaccine priority groups, and covered ages 16-69 (many older people 70+ were also shielding, but they had already become eligible for vaccination due to their age).

2. "At Risk"

AKA Clinically Vulnerable (CV) people were not advised to shield but may have been at increased risk from COVID-19 compared to others of the same age due to less severe conditions.

• The CV/at-risk group is much more complex because it looks at various conditions, taking into account prescriptions, hospitalisations, diagnoses and "resolved" codes for each one.
  • e.g. patients are excluded from the diabetes group if they have a "diabetes resolved" code more recently than a diabetes code;
  • asthmatics may have certain diagnosis codes, a series of prescriptions OR hospitalisation; etc.
• For vaccination purposes, this group were in group 6 of the 9 initial vaccine priority groups. This covered ages 16-64 (many older people 65+ were also At Risk, but they had already become eligible for vaccination due to their age).

Identifying Shielding/CEV And CV/At Risk Groups in Opensafely

Codelists

• Codelists from the official specification are here https://www.opencodelists.org/codelist/primis-covid19-vacc-uptake/
• Note these codelists are correct as of ~January 2021, we have not included the various more recent versions which have since been released.

Example Repositories

We have implemented the codelists using a few slightly different structures in different repos, though they should ultimately achieve the same results.

Shielding / CEV

• For our weekly report we create the CEV (shielding) flag entirely within the study definition, comparing the latest date of any high risk and lower risk flags for each patient (link).
• For our paper which combined TPP and EMIS we only extract the dates of the "high risk" and "lower risk" records in the study definition and then use separate Python scripts to compare the dates (here and here). This may be quite a complex set of scripts to follow but the reason for this structure was that we auto-generated the scripts from the specification document to reduce manual labour.

CV/At risk

• We have not implemented this group in our weekly report.
• For our paper, as for CEV, we only extract the date of each individual event (e.g. prescriptions, diagnoses) in the study definition and then use separate Python scripts to implement the logic for each condition in turn and then make a combined at-risk group (here and here).
• In some ongoing work we have instead implemented most of the date comparisons in the study definition then made the combined CV/at risk group in R; this was done this way to fit it in around the existing analysis structure.

[elsie-h]

Thanks @HelenCEBM - this is really helpful.

I've got a question about defining JCVI group 1 (older adults resident in care homes). Is this primis code list considered sufficient for defining that group, or should any additional conditions on age be added? For example, I wondered if the code list alone might identify younger people with learning disabilities who are in long-term residential care?

[HelenCEBM]

Yes you will identify some younger people (but not all, I think) so I would say it's a good idea to limit to 65+. We did this in our vaccine-declined report https://www.medrxiv.org/content/10.1101/2021.08.05.21259863v1.full-text (see "Priority groups for vaccination" and Box 1).

It's also useful to limit the pregnancy preg / pregdel flags to females of childbearing age, if you're using those.