Thread by @EinsteinsAttic

I see #NHSEx has published the slides from its @HDR_UK “Data Access & Discovery” event, charmingly subtitled “A Forest Through the TREs” - on YouTube, if you want to listen:



__
*I thought the phrase was “cannot see the forest for the TREs”, but hey...

Let’s take a look:

hdruk.ac.uk/wp-content/upl…

One reason we use the category term #TRE to refer to Trusted Research (or any other use) Environments that meet the #FiveSafes is that everyone wants to call their own TRE something different, so no-one really knows what they mean...

Call yours what you like (e.g. for @ONS it's their Secure Research Service) but if it only does four #Safes, or three Safes - and if it isn’t also #consensual and fully #transparent - then it's not a #trustworthy TRE...

Bottom right: one's a logo and one's a face… odd.

Interesting iconography from an outfit that clearly seems to think *everyone orients to it*! The #NHSX brand may be dead, but DHSC and NHSE’s cosy ‘joint working’ arrangement continues in the Service Transformation Directorate.

“Major projects have lowered public buy-in”. No shit!

Whose brilliant idea was it that the body (in some cases, the people) who pushed #caredotdata & #GPDPR - not to mention long-held #DataLake/#TargetArchitecture* ambitions - gets to define the rules?

*medconfidential.org/wp-content/upl…

The pandemic may have felt like the end of the world at times, but it certainly wasn’t the beginning of the story. The draft data strategy published in June 2021, for example, rehashed large parts of the "framework for action" @tkelsey1 published in 2015:

kingsfund.org.uk/sites/default/…

Hmm. Why talk about the £200M for “healthcare data #research” but not mention the Prior Information Notice @NHSEngland has put out for its £240M (or will it be £360M?) ‘#FederatedDataPlatform’?

find-tender.service.gov.uk/Notice/008755-…

#Planning is still #SecondaryUse, not #DirectCare...

Mutually #accredited #TREs are what we’ve said all along - but that everyone *accessing* NHS patients’ data must follow the same #rules is as important as those *providing access* following the same #standards.

That means no ‘VIP lanes’ for anyone - including @NHSEngland.

"shifting from data sharing" is one way to say "ending #dissemination", but claims for "increased #transparency" from @NHSEngland are barely credible. It refused to answer FOIs* on @PalantirTech #projects; what we do know only came out because of a PQ...

*whatdotheyknow.com/request/commun

Given they go on to refer to the #FiveSafes* later in the deck - i.e. #SafePeople, #SafeProjects, #SafeSettings, #SafeData & #SafeOutputs - this definition of a 'Secure Data Environment' is more interesting for what it omits than what it says...

*blog.ons.gov.uk/2017/01/27/the…

Now to the meat of it!

I’ve had to fix this slide 👇 for them, BTW - adding the logo @NHSEngland is clearly reluctant to mention.

(It can try hiding behind "COVID-19 Data Store" all it likes, but by now everyone knows what it means...)

So let’s compare these #national #TREs:

@NHSDigital’s is sustainable, was funded in March 2022 (amount unknown), and replicates the @ONS model which has been proven to work for researchers and analysts...

And @OpenSAFELY’s been going since 2020, cost £??m, has #published 20+ papers and details of ALL of its 84 approved projects: opensafely.org/approved-proje…

It’s open code, and exemplifies modern ways of working.

The one that had no logo doesn’t pretend to be a #TRE. The Data Store’s just a giant Azure bucket; the ‘smarts’ is #Palantir.

NHSE's spent £25 million on six declared projects: questions-statements.parliament.uk/written-questi… - all secretive, no independent oversight, some straying beyond #COPI uses...

[While Tim Ferris suggested 'TREs aren't for #planning' to @CommonsSTC 'cos that’s using statistics, ANY use of patient-level data must be inside the TRE - where stats can be extracted . #RAPs allow that to be done routinely and automatically - just as frequently as #dashboards.]

On the other #TREs, we’re aware of 6 regional pilots - four with no path to sustainability, two of which are questionable. @HDR_UK "invested" £6.4m but has no code, no papers, no platform, no capability, and high risks.

(Nonetheless, HDR HQ awarded them a prize...)

[Bizarrely, HDR has implied the ONS & NHSD clouds may 'run out of capacity', so research will have to rely on HDR. If both Google Cloud (ONS) and AWS (NHSD) were to run out of prioritisable cloud capacity, there would be FAR bigger problems than that...

... so are HDR actively funding the same #availability assurances that NHS national bodies have in their cloud contracts? And how many of its ‘hubs’ actually have #TREs that can be used by someone outside the HDR cabal?]

#Genomic & #imaging data present their own challenges; we’d expect them to run specialist #TREs (@GenomicsEngland sorta already does).

#Consented cohorts like @uk_biobank are more about #linkage; constantly copying patients’ data across defeats the point of the national #TRE(s).

Policy as product - well, OK.

But if the Service Transformation Directorate thinks it’s going to tell a bunch of other data controllers / statutory bodies how to #accredit THEIR platforms & #TREs, things could get bumpy...

(As we’ve said, #mutual accreditation’s the only way.)

“the EXTENT to which we should implement open data practices”?! I thought @dhscgovuk accepted the Goldacre Recommendations; why try walking them back? And we're not talking about #OpenData, rather #OpenCode & #WaysOfWorking - and telling EVERY patient how their data's been used.

For those wondering about #GPDPR, this timeline 👇 is your (implicit) answer:

Because "a TRE has been developed and implemented" is a *prerequisite* to any new #GPdata collection...

...that @NHSTransform says it won't even *begin* #drafting its "Implementation Roadmap" for #TREs until the Autumn, it's probably safe to say #GPDPR #Round2 won’t kick off until 2023 at least.

Fingers crossed 🤞

Addendum on #architecture:

https://twitter.com/EinsteinsAttic/status/1522879134506098689