I sent Vincent Jones a PM because I was too busy to reply earlier, so I am repeating some of that here.

To some extent at least, I think any problem is down to phrasing myself badly. That is, I am afraid, not that rare. I should also say I am not talking from any expertise in the drug or medical industry (though I have some in statistics.)

When I said "You ultimately rely on an individual patient who is taking two or more drugs to report symptoms, that the doctor concerned notices a connection and reports that upwards until eventually that is assessed and recorded as an incompatibility" it was intended to be more or less equivalent to "every contract I have signed with a pharma client in the past twenty years has included an agreement to report - immediately - any adverse event occurring to someone taking one of their drugs": it is that base collection of data that drives the whole process.   Once you have that base data, it is indeed necessary to collect it into larger datasets and hit it with a lot of statistical analyses, confidence measures, multi-variant hypotheses testing and weird and wonderful mathematics to try to extract the most meaningful data, which will sometimes result in statements of the form "patients taking A and B might get side effect S (including how often, how severe and so on). That triggers a process that may ultimately end up in such things as warnings on the leaflets accompanying drugs A and B.

That the NHS England data is an extremely useful data set to analyse the interactions of drugs (especially those with no immediately obvious links) is certain. In particular, it can give us information, using my A and B example, to determine how many people do not have side effects (naturally, people do not report not having a side effect unless explicitly asked.) It can also determine that people who took A and B reported having S without realising it was a side effect at all. Or might even end up revealing the issue is not A and B at all, but A and something else, but the patient assumed it was B because they had recently changed medication.

So I do not under-value the potential benefits of the NHS data at all. Where my concerns arise is with who has got access to such a valuable commodity. The NHS England web site is definite that the data will not be made available to insurance companies, for example. But in truth all they can really say is there is no intention of doing so. In a few years, with another government, and new priorities who can really say? Added to that, we have to recognise everyone is fallible and not everyone is honest. There have been plenty of examples over the years of civil servants leaving secret documents/laptops on trains and such like. Mistakes will happen, because that is human nature. I pick on insurance companies because they would have an obvious interest, but there are many others.

I make the assumption that company C1 has an near-perfect understanding of its own products, but a much more limited knowledge of those of company C2, and vice versa. If there is a company that is especially adept at analysing the data - A1 - then if C1 can form an exclusive relationship with A1 it could significantly alter the market to disadvantage C2. This is why, in my mind, A1 should not be a commercial body, and if it must be it should certainly not be in a monopolistic position.

Now, there is always a level of trust in these systems. If I choose to opt out, I assume that sets a flag on my data. With human fallibility some coder will at some time accidentally fail to check the flag when they extract data for processing. There is nothing I can do about that. All I can do is choose where to draw the boundaries of where I have to take things on trust and I have to do so recognising that if enough people choose not to share their data, it reduces the potential benefits of the data.

It is, as I said, not an especially easy decision to decide whether to opt out or not. Other people's thoughts are welcome.