This #BOPA2023 poster, aside from its interesting (& concerning) findings, originally began as a data analysis exercise. @chloewaterson88 had the goal of learning some #rstats, & while I have a fondness for ๐ง + ๐, I was eager to work with actual data. Check out the thread ๐๐งต
I thought we might be able to do something building on my guest blog with the @BennettOxford about High Dose Dexamethasone: bennett.ox.ac.uk/blog/2019/11/high-dose-dexamethasone/ (BTW: it seems blogging about this might result in product discontinuation!)
We quickly spotted there wasn't an easy way to pick all the drugs we might want, and exclude the ones we didn't want from @openprescribing.
Instead we needed to build a list of drug codes. ๐ก another @BenGoldacre #opensource project: OpenCodeLists.org
For example, you could pick all cytotoxics, but that isn't just cancer treatment. We didn't want to see data for methotrexate for instance.
To ensure we didn't miss anything, we required a decent drug classification. Sadly, the BNF Chapter Numbers from ๐ @PharmPress don't work well for #chemotherapy/#SACT.
So, we turned to the Anatomical Therapeutic Classifications from the @WHO. whocc.no/atc_ddd_index/?code=L&showdescription=yes
We did briefly toy with the idea that we might ask #ChatGPT to write the #rstats code. It tried... It failed.
Getting the #data into :R: in this way was definitely not beginner level coding. But the actual analysis was mostly straightforward #tidyverse.
This entire project was testament to the success of Open Data & the Freedom of Information Act.
I'm sure we all in the #NHS have moments where we wonder why #FOIA exists..
..the answer is so the public can look at data and question things - like why ยฃ1m of #SACT in Primary Care
Without #FOIA @NHSBSA would likely not release the data that @bengoldacre and @BennettOxford use to create @openprescribing.
Why is there no Secondary Care SACT data equivalent? @theosanderson has crafted a mini app to try: x.com/theosanderson/status/1704265751891239024?s=20
I'm keen to play more with the dataset he used and see what more we could create.
Open to collaboration if people want to build a bit of an :R: project...?
& lets not forget @NHSOpenSource #OpenSoftware too. Analysis was performed in #Rstudio from @posit_pbc (@posit@fostodon.org) using @hadleywickham's (@hadleywickham@fostodon.org) fantastic #tidyverse.
Don't understand what I'm talking about? - come & ask... or check out @NHSrCommunity on twitter or @NHSrCommunity@fostodon.org
So anyway... what did @chloewaterson88 find on her #BOPA2023 #poster?
- ยฃ๐๐ฆ ๐จ๐ ๐๐ฑ๐ฉ๐๐ง๐๐ข๐ญ๐ฎ๐ซ๐ ๐จ๐ง ๐๐๐ง๐๐๐ซ ๐ญ๐ซ๐๐๐ญ๐ฆ๐๐ง๐ญ๐ฌ
- ๐๐๐จ๐ ๐ซ๐๐ฉ๐ก๐ข๐ ๐ฏ๐๐ซ๐ข๐๐ญ๐ข๐จ๐ง
- ๐๐ข๐ค๐๐ฅ๐ฒ 17% ๐๐ฆ๐๐ซ๐ ๐๐ง๐๐ฒ ๐ฌ๐ฎ๐ฉ๐ฉ๐ฅ๐ฒ
Do you know why? Are you a PCN pharmacist in one of the dark areas on the map? Did you know this happens in your patch? @pcpa_org
Speaking of the #map - this is another example of the #FOIA in action. The map boundaries come courtesy of @ONS and @OrdnanceSurvey - all released under an #OpenGov License thanks to the #FOIA.
If we return to @chloewaterson88's findings - there are two concerns:
๐ Is it EVER clinically appropriate to be making these supplies? Does a 1 tablet emergency supply of #palbociclib make sense?
๐ The price paid is (a) for a full pack (b) not at NICE discount
#BOPA2023
If emergency supply is appropriate, we need to find better ways to handle them. But maybe we need some national guidance from @BOPACommittee to help? #OncoPharm #PharmRx
@msonetwork do you find this worrying?
Want to do your own number crunching in #rstats? :R:
Our code is all released on @github github.com/Chlo115/Primary-care-SACT-use the code list is on OpenCodeList. (Links on the poster.)
@chloewaterson88 will probably tell you its a bit of a steep learning curve ๐though.