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).
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.
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.