*MEDICINE*: vaccination status

[sebbacon]

Vaccination status - get from both Read codes and vaccination table. https://github.com/ebmdatalab/tpp-sql-notebook/blob/cohort-extraction/notebooks/flu-vacc-covariate.ipynb

In adults:

• Pneumococcal
• Flu vaccine (time-sensitive: within current flu season)
• Shingles

In children:

• See NHS website

[alexwalkerepi]

Example using flu vaccines is here, but only uses Read codes so far, as the current build of the dummy dataset doesn't contain the vaccination table yet. https://github.com/ebmdatalab/tpp-sql-notebook/blob/cohort-extraction/notebooks/flu-vacc-covariate.ipynb

[annaschultze]

@alexwalkercebm we're aiming to include this as a covariate in the ICS study (at least flu vaccination) - is there a separate vaccination table in the data as of now? @ianjdouglas may know if we have existing code lists from lshtm for Read and/or Immunisation codes from CPRD studies (but not sure whether the coding systems for immunisations would be the same)

[ianjdouglas]

@brianmackenna mentioned he has ways of looking for vaccines :-)

[brianmackenna]

The database schema has separate fields for

• Vaccination
• VaccinationReference

@chris-tpp has also mentioned on slack

Bit of a mix. There is a native vaccination table which the majority of entries are in. This contains reference data to antigen content so you can pretty easily create the list you are interested in. There will also be some in Read codes that are entered as coded entries e.g. 65E.. is a flu vaccination code that does get used in bulk after flu jab clinics.

@chris-tpp you might have answers to some of these questions?

• how is the vaccination structured (e.g. is it dm+d) so I can create list of vaccines?
• Will the dummy data contain all the vaccines so I can work backwards off that?
• Or do you have technical specification of the vaccine fields somewhere that I can read?
• You mention a nit of a mix - does that mean we need to build a rule for the medication issue fields as well as the vaccination part?

[chris-tpp]

Nothing in medication but there are vaccinations as both codes and as a native structure. Don’t need to do any DM+D work.

I’d get the vaccination reference data sent to you - it’s a reference table of all possible native vaccs with a field for antigen content. Only a couple of hundred entries, I think.

There are some vaccs recorded as codes in the main coded event table too. This tends to be for things like bulk addition of flu vaccs, if the GP practice use this functionality to add codes instead. That is a SNOMED search. I’d find the highest level parent vacc/ imm code and then we’ll port it back to get you a list for validation.

[brianmackenna]

@chris-tpp Will build SNOMED search - presume there are a lot of 955691000000108 Seasonal influenza vaccination given by pharmacist

From what are the vaccine reference tables built?

Reason I ask is that 10k flu vaccines appear to go through tradsitional FP10 as opposed to vaccine FP34. Would this 10k get picked up in the vaccine ref table?

[brianmackenna]

Discussion and investigations today. There are three places for recording a vaccine

• Vaccination table: This is built off the place where practices should issue vaccines. This is where the vast majority of vaccines should appear.
• Medication issue: GPs may still issue FP10s for vaccines or sometimes they may be recorded here. TPP data shows approximately 360k flu vaccines here in last few seasons.
• Clinical codes: Snomed/READ may also be used to record vaccines. This may be more common when operating various types of clinics. We suspect this might increase this year with inicreasing numbers of vaccines administered in places that aren't GP e.g. community pharmacy

We propose to use a combination of the thee i.e. “Vaccine tables + medication issues + CTV3 codes”

Depending on data we could use a hierarchy or presence of each.

[brianmackenna]

Creating separate issues for each vaccine as a result of this complexity.

pneumococcal flu

[brianmackenna]

After further work we have divided into a pragmatic rule option to identify vaccines administered and an optional second option with clinical codes depending on the research question/resource

1st Pragmatic Option: Patient appears in TPP vaccination table & dm+d medication issue (include all strains of pneumococcal for this study)

or 

2nd Additional Option: Presence of clinical code indicative of vaccination (where more than one code appears in the record we use the algorithm on the sheet "comments" from excel file) 

@evansd will have some ides but way to query TPP dataset is

• dm+d medication issue: This can be queried in same way as other medicines (example influenza codelist)
• Clinical codes: This can be queried in same way as other clinical codes
• Vaccination table:
  • Query in a similar manner to medication issue by generating a codelist (example TPP vaccine table influenza codelist
  • Build into a query into the python library to look for all vaccines associated with pathogen/disease (vaccinationcontent). Here we may also want to add limits around date of issue, where in schedule it occurs, which particular product (important for some vaccinations).