OK, @Dominic2306 -
That doesn’t matter. We really need to talk about data “anonymisation”.
It is not water-tight at all and re-identification is very easy especially with health data (when combined with other datasets). This is a huge issue that needs to be discussed publicly.
That doesn’t matter. We really need to talk about data “anonymisation”.
It is not water-tight at all and re-identification is very easy especially with health data (when combined with other datasets). This is a huge issue that needs to be discussed publicly.
Re-identification is a fascinating field, where it becomes increasingly easy with more and more publicly available online data about people.
Here’s a fun example about “anonymous DNA” (individuals) re-identification just using publicly available data.
biorxiv.org/content/biorxi…
Here’s a fun example about “anonymous DNA” (individuals) re-identification just using publicly available data.
biorxiv.org/content/biorxi…
Furthermore, when holding vast databases on health (or any other personal) data for time... you have to ask what it is being used for.
Eg if you had NHS database for Covid, you could analyse it for another disease, develop algorithms, then pitch for contract on that.
Eg if you had NHS database for Covid, you could analyse it for another disease, develop algorithms, then pitch for contract on that.
Fine, you might think - that’s exactly what we want!
But then you get companies using privilege for new purposes to lock in algorithms (that they may own/ patent) and developing lock in on analysis & policy advice - and then govt policy becomes driven by private interests.
But then you get companies using privilege for new purposes to lock in algorithms (that they may own/ patent) and developing lock in on analysis & policy advice - and then govt policy becomes driven by private interests.
So you cannot be casual with handing out access to huge databases without clear time and remit restrictions. There are lots of risks of unwanted behaviour.
These are poorly understood by govt and definitely need open public, expert & commons/lords debate on data management.
These are poorly understood by govt and definitely need open public, expert & commons/lords debate on data management.
There are ways to handle data better...
...for example where analysts are not given single-record level databases.
Rather, they can analyse through an interface that gives aggregate results only (eg N>30 only) and the interface records all analyses requested/run.
...for example where analysts are not given single-record level databases.
Rather, they can analyse through an interface that gives aggregate results only (eg N>30 only) and the interface records all analyses requested/run.
That way, the govt better retains its own database, ensures no copying, protects individuals at individual record level absolutely —- & itself monitors all analyses done with patient data.
That’s the responsible way to get large private corp help. Sharing dbases v irresponsible.
That’s the responsible way to get large private corp help. Sharing dbases v irresponsible.
A colleague just messaged me after seeing the thread.
Makes an excellent point about GP and hospital visit dates.
Health is complex - and you NEED complex data for any good analysis... but single-line records (ie individuals) also open huge reidentification opportunities.
Makes an excellent point about GP and hospital visit dates.
Health is complex - and you NEED complex data for any good analysis... but single-line records (ie individuals) also open huge reidentification opportunities.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
