This paper is hard work, but rewarding. You HAVE to read it.
The NHS aspires to be a data driven organisation. The proof is delivery. In @EBMDataLab we have up-close experience of working on NHS data. We hit barriers at every step: those barriers show what needs to change. 2/n
-
-
Prikaži ovu nit
-
First up, the prescribing data itself. Best practice is that this should be easily discoverable, accessible without human intervention, made available at addresses that do not change, and documented so the relevant concepts are clearly explained. The reality....
Prikaži ovu nit -
This data randomly changes location. Every time the location of the data changes, it breaks the software we have written to automatically download it. There are two versions of the data: they don't cross reference each other. One is no longer in NHS England's "Data Catalogue"...
Prikaži ovu nit -
To download one NHS open dataset, that changes monthly, you need to fill out a CAPTCHA. So it can never be automated. The NHS should be helping this work be done by machines. But every month an engineer has to stop work to prove - manually and needlessly - that we are humans...
Prikaži ovu nit -
Then there is the content of the data. The British National Formulary data schema changes from time to time. Who knew? Nobody. There's no mention from the NHS anywhere that such changes are possible. We inferred it following user enquiries about apparently disappearing drugs...
Prikaži ovu nit -
We now email a person in the NHS annually for a spreadsheet detailing the changes. This is not okay. BUT we do not complain mindlessly. Our paper is full of detailed recommendations. For example....
Prikaži ovu nit -
The fact that the BNF data schema changes regularly should be documented. And BNF code changes should be published monthly as a mapping. Oh, I didn't mention that: we have inferred the NHS BNF data schema changes MONTHLY, but we can only get ANNUAL changes from NHS BSA...
Prikaži ovu nit -
Here's an interesting side note. We don't just share 44,000 lines of code on GitHub, we also share about ONE THOUSAND issues so you can see how we work, and how we work around NHS data barriers. Here's the BNF data schema one:https://github.com/ebmdatalab/openprescribing/issues/456 …
Prikaži ovu nit -
Okay, I have to stop now to work on the press release for our next big paper out this week. Come back tomorrow for more hellish tales of bad practice around NHS data, and read the paper!https://www.jmir.org/2020/1/e15603/
Prikaži ovu nit -
I really mean it. Anyone can read and understand this paper. Its contents will make you WEEP at the chaos of how the NHS handles data. It will make you YEARN for change. It will make you stand on a table and CHEER for our recommendations.https://www.jmir.org/2020/1/e15603/
Prikaži ovu nit -
RIGHT, we return to our paper on BAD PRACTICE around DATA from the NHS, led by our mighty
@sebbacon RECAP Building@openprescribing we hit barriers: NHS data that moves location, contains errors, shifts format, undocumented, no way to report errors...https://www.jmir.org/2020/1/e15603/Prikaži ovu nit -
So, lots of examples of problems with NHS data in the thread above. Let’s have some new ones. To build http://OpenPrescribing.net we need to know the size of the patient list at every NHS GP practice in England. This is available as open data… kind of.
Prikaži ovu nit -
The location of this file jumps all over the internet. So we can’t automate downloading it. Until 2018, we automated a search for the phrase “Number of Patients Registered at a GP Practice” on the
@NHSDigital website...Prikaži ovu nit -
From July onward, the data were moved to a different location. The location has changed twice more since then. This is very, very unhelpful, and falls short of best practice around data management. Files should retain a stable location.
Prikaži ovu nit -
But this list size doesn’t just change location. The format of the dates changed between June and July. This is very, very unhelpful, and falls short of best practice around data management. Files should retain a stable format.
Prikaži ovu nit -
It gets worse. There’s no proper documentation for this file. It contains errors. And there’s no clear way to report those errors. For example: we found lots of care homes and violent patient services classified incorrectly as standard settings. Mistakes happen. We wanted to help
Prikaži ovu nit -
We were told errors can only be corrected by CCGs. But they couldn’t give us contact details for the CCG people to get them to notify the NHS of the errors in their data. It’s not clear that any part of the NHS considers itself responsible for maintaining accurate data here.
Prikaži ovu nit -
Moving on. Often we find errors in the data for closures and mergers of practices. This matters, as it affects denominators for measures on http://openprescribing.net Users (GPs, CCGs) often inform us of changes, and ask us to correct the data.
Prikaži ovu nit -
This puts is in a hard position: we don’t want to modify canonical NHS data. We’re hardly resourced to manage a new canonical dataset. The NHS must surely have a way to handle CCGs and practices informing them of closures and mergers?
Prikaži ovu nit -
In fact they do: they have two systems. Every practice must fill out at least two nearly identical forms to notify NHSBSA and Primary Care Support England (PCSE). We requested any data resulting from both these forms in FOI requests.
Prikaži ovu nit -
PCSE eventually supplied us with a spreadsheet that appeared to bear little relationship to real closures. https://github.com/ebmdatalab/openprescribing/issues/932 … (Look
that's called an ISSUE on GitHub: we don't just share our code, we share our workarounds, barriers hit, how they are overcome)Prikaži ovu nit -
NHS BSA told us “Prescriber and practice/cost centre amendments are only held as paper forms”. So the NHS BSA ask practices to send in those forms and then they are… filed? In a... filing cabinet? On… paper?
Prikaži ovu nit -
As a result, every April (when boundary changes happen), our developers have to make educated guesses about which contemporary CCGs the patients of a closed practice now belong to, and at a practice level, there is nothing we can do to amend the data correctly.
Prikaži ovu nit -
Why are we sharing all this? We use NHS open data to build
@OpenPrescribing used by 130,000 users a year. Our mission is “fix problems for us, fix them for everyone”. We put our code on GitHub (45,000 lines to date): we also document all our barriers and workarounds there.Prikaži ovu nit -
Here’s the thing: you only really spot problems in data when you USE it to DO SOMETHING that is concrete, like build a service. It's VITAL that the NHS has an INBOX that can receive blockage reports like the ones we're sharing.
Prikaži ovu nit -
We know we're not alone. We know that private + public sector innovators working with patient-level NHS data hit these problems too. We know they feel less able to speak up about it. Because we are discussing open data, it’s okay to discuss the problems. Another benefit of Open!
Prikaži ovu nit -
Tomorrow, if you’re very lucky, I’ll explain what happened when we tried to find out whether NHS England has an email address for every NHS GP practice in the country. More in our paper, here by t’mighty
@sebbaconhttps://www.jmir.org/2020/1/e15603/Prikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.