Next steps on variation by software vendor

[sebbacon]

How can we control for geographic variation as a confounder?

With a focus on diltiezam (and ciclosporin to a lesser extent)

• Check practices that have changed system
• Add in deciles so you can see in national context
• Consider adding in outer deciles for each vendor also
• Compare side-by-side with greographic variation
• Is a regression that includes CCG as a random effect going to help? (Also multi-level?)
• Go and try out and get screenshots of choice pickers for prescribers in each system

How can we / should we control for decision support software

• optimiserx is built into EMIS but a plugin with SystmOne
• there is an FOI request somewhere with 200 responses about who has what!
• Or we could just ask both optimiserx, PrescQIPP for this data?

[sebbacon]

Have done some of the following, here: https://github.com/ebmdatalab/jupyter-notebooks/blob/master/measures_by_software/diltiezam%20focus.ipynb

[sebbacon]

We're going to write a paper on this. Before I close this issue, I'd like to fix the 50/50 charts so they have the same y-axis. Also consider if there are any other measures worth a deep-dive. Currently blocked on me setting up my "new" laptop