1. The patient record should be held by the patient's registered GP. Not everyone will have a registered GP but this should not be an impediment to getting something underway, if they don't get an electronic record this is no different to the existing situation of not having a paper record and, if necessary, a separate procedure can be put into place if, for instance, they turn up at A&E.
2. In reality this will mean that the GP's service provider such as TPP.
3. Access from other practitioners such as A&E, consultants to whom the patient is referred gen access via a broker which checks for appropriateness, the broker to be run by NHS. If necessary such practitioners would be able to add to the record.
4. The service providers, including those used by those with access in 3. will have different systems. A common exchange format would be required. My suggestion for this would be XML. This allows processing aong the lines of Provider1 data > Provider1 XML >(XSLT)> Common XML >(XSLT)> Provider 2 XML > Provider 2 data
5. Start simple. One reason for suggesting XML is that it's eXtensible. Get a system under way with a minimal common data format ven if it's only a list of appointments and free text notes. Add more to the schema later. Trying to produce a bells and whistles version from the start will ensure it never happens.
6. It should take into account the patient's other registered practitioners such as dentists and opticians. If they have suitable systems they should have automatic rights to access and contribute a suitable subset of the data but again this need not be in the initial version.
The negotiation for access at 3. is critical for maintaining confidentiality as well as ensuring essential access. It's something that should be run in-house, not farmed out to the likes of Palantir nor should it be in a public cloud.