NHS data mining: stupidity and cleverness

[Adrian Midgley]

I suspect that in the same way that only a proportion of people can
get their minds around the idea of a pointer in C there are only a
proportion of people involved in the administration of the NHS and the
DoH who can get their minds around the idea that one might send a
question to a host of peripheral systems, and receive back the
answers, rather than collect all the data from all those peripheral
systems, and then run searches on those.

I doubt that all the possible questions can be answered by the first
method, but then I doubt they can be, legally or accurately, by the
second either.

And yet again we see a row over "nonymisation" which turns out not to work.


Is it time to state that MIQUEST was the way to go, and that questions
should be sent out, and only answers returned, or is that approach
also too vulnerable to inference and cross referencing?


(MIQUEST is a subset of a superset of SQL called Health Query
Language, with software that runs on the GP automation systems, and
mechanisms to export, check, import and control queries, and
presumably to aggregate the results at the centre.  It is a beast, and
locally in Devon the admin people seems convinced that their appraoch
involving large Excel spreadsheets must be better.)

-- 
Adrian Midgley   http://www.defoam.net/