Count monthly medications of pharmacy first events

[milanwiedemann]

Closes #49

This code aims to count all medications prescribed during a Pharmacy First consultation. However, there are two main limitations: (1) only the first prescribed medication is selected per consultation, and (2) only one event per month per patient is captured – I think this is acceptable for this?

To test the measure definitions, I added some dummy data to help me verify that the code works as intended. However, I wish there were an easier way to test and extract this data?!

• Patients 1 to 10: PF event (1577041000000109) and medication (10000000000000000) with same consultation ID
  • I expect these 10 patients to show up in my ouput
• Patients 11 to 20: PF event (1577041000000109) and medication (11111111111111111) with same consultation ID
  • I expect these 10 patients to show up in my ouput
• Patients 21 to 30: PF consultation (1577041000000109) and no medication with same consultation ID
  • I'm a bit confused why these patients also show up with a missing dmd_code, but I'm not too bothered about it
• Patient 31: PF consultation (1577041000000109) and medication (22222222222222222) with same consultation ID
  • I expect this code to show up but replaced as 0 because of statistical disclosure control
• Patient 32: No PF event and medication (99999999999999999) with same consultation ID
  • I dont expect to see this code in the output

opensafely exec ehrql:v1 generate-measures analysis/measures_definition_pf_medications.py --dummy-tables dummy_tables

[viv3ckj]

Nice approach to this ticket, approved!

[viv3ckj]

Could also be helpful to look at last medication (last_for_patient()), to see if this approach is missing out on medication counts.

In addition, another idea could be to look at the first and last medication for the first consultation and first and last medication for the last consultation of every patient.

[LFISHER7]

This looks good to me! Nice to have the tests. Patients 21-30 show up in the measure because they are included in the measure denominator. If you don't care about the denominator or the ratio calculated here, you could just set the denominator to the same as the numerator and then that group wouldn't be included