5 ms·

This is the data in question: http://www.hscic.gov.uk/hes http://www.hscic.gov.uk/hes It is anonymised [1], publicly licensable data. Here are a list of users

by crb 13y ago

This is the data in question: http://www.hscic.gov.uk/hes http://www.hscic.gov.uk/hes It is anonymised [1], publicly licensable data. Here are a list of users and uses. [2] [1] "We apply a strict statistical disclosure control in accordance with the HES protocol, to all published HES data. This suppresses small numbers to stop people identifying themselves and others, to ensure that patient confidentiality is maintained." [2] http://www.hscic.gov.uk/media/10495/Users-and-uses-of-HES/pdf/HES_Users_and_Uses.pdf http://www.hscic.gov.uk/media/10495/Users-and-uses-of-HES/pd...

SideburnsOfDoom 13y ago
> It is anonymised Not meaningfully, no. A UK postcode covers 20 households or less. If you have that plus gender and date of birth (as seems to be the case here), you almost always have a unique individual.
- lotsofmangos 13y ago
  NHS number, date of birth, postcode, ethnicity and gender. Is about as anonymous as wearing a different hat.
  - dhoulb 13y ago
    NHS number is clearly ridiculous. But, I can see why all the other data are statistically relevant in analysis. Postcode, so you can track viral outbreak etc with a fair degree of locational accuracy. Date of birth, so you could analyse if people born in certain seasons had higher incidences of various diseases. (You could probably drop the day and just have month, but there could be cases where there were more hospital deaths for babies born on a Monday, etc). Gender and ethnicity are obvious.
    - andygates 13y ago
      A hash of the NHS number would give a UID without compromising the person's details. Providing a UID that is one-way linked would allow HSCIC to go to the actual data if a care situation warranted it. I'd argue that the first half of the postcode and month of birth are ample for outside-HSCIC use. Monday-deaths is the kind of thing that's done internally anyway.
  - linlea 13y ago
    NHS England’s Chief Data Officer says that with the pseudonymous dataset (he calls it amber data) patient’s identifiers are removed, including "(their date of birth, postcode, and so on)". Also, I think it might be hospital number rather than NHS number in the data http://www.england.nhs.uk/2014/01/15/geraint-lewis/ http://www.england.nhs.uk/2014/01/15/geraint-lewis/
- isp 13y ago
  Tiny nitpick[0]: a UK postcode corresponds to a minimum of 1, and maximum of 100 (not 20), delivery points[1] (not households). Typically about 15 delivery points. I fully agree with your main point, though, that the data are not meaningfully anonymised. [0] http://www.poweredbypaf.com/wp-content/uploads/2013/11/Programmers_guide_20131104.pdf http://www.poweredbypaf.com/wp-content/uploads/2013/11/Progr... (page 19, "Small User Postcode") [1] https://en.wikipedia.org/wiki/Delivery_point https://en.wikipedia.org/wiki/Delivery_point
collyw 13y ago
I have begin collaborating in a rare diseases project. So imagine you have one or two people in the country with certain symptoms. How anonymous is that?
- sentenza 13y ago
  Do you still work in that field? If so, you're probably one of the few people that IMO actually should have some access to that data. Maybe you'll get lucky and the data will become a researcher free-for-all, like the Enron emails did. (Not likely, though.)
  - collyw 13y ago
    No we are discussing all the security implications before we even get close to collecting the data.
    - sentenza 13y ago
      I am happy to hear that. From my experience, some inroads have been made in universities regarding the ethics of research projects. Security of the computer systems used to store and handle tha data, on the other hand, is way to often still not considered.
- linlea 13y ago
  Small numbers (cells containing values less than 6) are supposed to be removed from the data or obscured, to prevent breaches of confidentiality. Word document: http://www.hscic.gov.uk/media/1879/NHSIC-small-numbers-terms-and-conditionsdoc/doc/NHSIC_small_numbers_terms_and_conditions.doc http://www.hscic.gov.uk/media/1879/NHSIC-small-numbers-terms...

by crb 13y ago

SideburnsOfDoom 13y ago

> It is anonymised Not meaningfully, no. A UK postcode covers 20 households or less. If you have that plus gender and date of birth (as seems to be the case here), you almost always have a unique individual.

lotsofmangos 13y ago
NHS number, date of birth, postcode, ethnicity and gender. Is about as anonymous as wearing a different hat.
- dhoulb 13y ago
  NHS number is clearly ridiculous. But, I can see why all the other data are statistically relevant in analysis. Postcode, so you can track viral outbreak etc with a fair degree of locational accuracy. Date of birth, so you could analyse if people born in certain seasons had higher incidences of various diseases. (You could probably drop the day and just have month, but there could be cases where there were more hospital deaths for babies born on a Monday, etc). Gender and ethnicity are obvious.
  - andygates 13y ago
    A hash of the NHS number would give a UID without compromising the person's details. Providing a UID that is one-way linked would allow HSCIC to go to the actual data if a care situation warranted it. I'd argue that the first half of the postcode and month of birth are ample for outside-HSCIC use. Monday-deaths is the kind of thing that's done internally anyway.
- linlea 13y ago
  NHS England’s Chief Data Officer says that with the pseudonymous dataset (he calls it amber data) patient’s identifiers are removed, including "(their date of birth, postcode, and so on)". Also, I think it might be hospital number rather than NHS number in the data http://www.england.nhs.uk/2014/01/15/geraint-lewis/ http://www.england.nhs.uk/2014/01/15/geraint-lewis/
isp 13y ago
Tiny nitpick[0]: a UK postcode corresponds to a minimum of 1, and maximum of 100 (not 20), delivery points[1] (not households). Typically about 15 delivery points. I fully agree with your main point, though, that the data are not meaningfully anonymised. [0] http://www.poweredbypaf.com/wp-content/uploads/2013/11/Programmers_guide_20131104.pdf http://www.poweredbypaf.com/wp-content/uploads/2013/11/Progr... (page 19, "Small User Postcode") [1] https://en.wikipedia.org/wiki/Delivery_point https://en.wikipedia.org/wiki/Delivery_point

collyw 13y ago

I have begin collaborating in a rare diseases project. So imagine you have one or two people in the country with certain symptoms. How anonymous is that?

sentenza 13y ago
Do you still work in that field? If so, you're probably one of the few people that IMO actually should have some access to that data. Maybe you'll get lucky and the data will become a researcher free-for-all, like the Enron emails did. (Not likely, though.)
- collyw 13y ago
  No we are discussing all the security implications before we even get close to collecting the data.
  - sentenza 13y ago
    I am happy to hear that. From my experience, some inroads have been made in universities regarding the ethics of research projects. Security of the computer systems used to store and handle tha data, on the other hand, is way to often still not considered.
linlea 13y ago
Small numbers (cells containing values less than 6) are supposed to be removed from the data or obscured, to prevent breaches of confidentiality. Word document: http://www.hscic.gov.uk/media/1879/NHSIC-small-numbers-terms-and-conditionsdoc/doc/NHSIC_small_numbers_terms_and_conditions.doc http://www.hscic.gov.uk/media/1879/NHSIC-small-numbers-terms...