Palantir: NHS faces legal action over data firm contract
bbc.com
bbc.com
- Palantir's U.K. head of sales meets with NHS Chief in 2019 over dinner. Apparently there were some fancy cocktails involved...how sensational!
- Pandemic rolls around; Palantir--among other tech companies--awarded contract to do data analytics work on early Covid response
- Palantir's contract now extended. Ostensibly NHS is pleased with their work?
Maybe of more interest https://investors.palantir.com/governance/board-of-directors...
It at least makes me wonder if this selection was intentional & what the relationship between the two & the intelligence communities of both the U.S. and the U.K. are
I hope that helps.
https://www.bmj.com/content/367/bmj.l6326?__cf_chl_jschl_tk_...
In fact its non-anonymised data never hits the "Datastore" (Palantir Foundry) or leaves NHS Systems.
Morally sketchy extension to the contract? Yes. Illegal? No.
[1] https://www.england.nhs.uk/wp-content/uploads/2020/06/202004...
Edit: Corrected Palantir will have access to data, not be given it.
This is why sharing even redacted datasets isn't generally done.
[0] https://www.wired.com/2007/12/why-anonymous-data-sometimes-i...
Edited to correct the star (89->87, also Americans not Brits, I thought it was Brits)
The document I linked says they do have these controls in place but having done my fair slog in the NHS I'm doubtful.
Does something similar apply here? Can it be formally proven that there's no way to de-anonymise the data?
For example - drug manufacturers want to know demand so they can plan ahead. But in the UK, doctors, NHS, DOH etc don't share that data of usage as would be granular enough to be possibly identifiable - so how do they get the data.
Well, chemists/hospitals are supplied by a few companies(Think Amazon for drugs/medical supplies) - they can from ordering terminals from those customers, give hourly totals level of demand to the manufacturers if they will pay for it. See, the distributors are and have been in a position for decades to cash in upon selling such data at a level in which you will not identify any single individual ever. Also was secure back then on many levels for transfer. Oh and those rare drugs that only few people have - that and other area's of data is never sold. What data is sold is you everyday drugs, as manufacturing really does have to plan ahead more with some drugs than others, more so when the shelf life various as does storage and then some customers will only take drugs with at least X time left until expiry. So planning supply and demand really does get very complicated, very quickly. Hence knowing as real-time
I've worked upon such a system at a large EU based company 20 odd years ago, also worked at the UK DOH during the transition days from Conservative to Labour and ironical just the other day found the old Tony Blair letter introducing the new future ahead to DOH staff.
Now in this case they say aggregated data, which can be done in a way that if you can pick out an individual, you would of already mastered unscrambling eggs.
Here's the practice-level prescribing data for GPs in England: https://digital.nhs.uk/data-and-information/areas-of-interes...
Here's all the GP prescribing for (god-awful) co-proxamol: https://openprescribing.net/chemical/0407010Q0/
Here's co-proxamol prescribing by CCG[1]: https://openprescribing.net/analyse/#org=CCG&numIds=0407010Q...
Here's co-proxamol prescribing by NHS Gloucestershire CCG: https://openprescribing.net/ccg/11M/lpcoprox/
Here's co-proxamol for one GP practice (Royal Well Surgery) within Gloucestershire: https://openprescribing.net/measure/coproxamol/practice/L840...
After my time that, as mentioned this was 20 years ago, but interesting. Had a quick play, certainly few aspects come to mind and that would be looking at methadone usage in an area in regards to buying a house and seeing the trend. Which does raise the whole aspect of which on-line estate agent will be first to link such data and analysis it in ways that would indicate the health of the area and insight as well into general lifestyle above and beyond crime data from the police.
So for me, the question is - should open data have some control upon how it is used. Certainly don't want Domino's Pizza doing leaflet bombing of area's based upon an increase in that area of people with known eating disorder medications. Or a spoon company targeting an area because there is an increase in methadone usage. Certainly some dark marketing aspects that such data does open itself towards, if it's not granular enough. Surgery, is too detailed for public IMHO. Also be flawed as only some chemists/surgery's can prescribe certain medications, so you would see artificial bunching in some area's due to various reasons. Which without such information to go with that data to interpretive it correctly, you will make mistakes in processing without that area weighting aspect factored in. Things like that skew the data in a way that sets it up for interpretation abuse. Wait until Estate Agents start tapping into this - they already include school data and police data, so will only take one and then the rest will follow as it would be a competitive edge otherwise.
All this reminded me - Census in the UK this year, I'll pop Jedi as religion again, good for the credit rating one day I hope.
"The datastore is the largest pool of patient data in UK history."
No idea what this current one does precisely but they do say they aggregate the data - to what extent I do not know.