I have heard about this project on a few podcasts and it came up during the recent post about the UK BioBank leak on Github [0].

In this case, Ben is working on making GP information available for researchers in a protected manner, where they submit their analysis but never have access to the underlying records. I haven't look in deeply about how they avoid really specific queries for de-anonymisation.

I believe there's also been a lot of work going in to mapping 'events' (like a prescription of a drug) to things like conditions (like diabetes/high blood pressure etc). It was surprising to me that this was so hard to extract.

[0] https://news.ycombinator.com/item?id=47875843