How the health system should work
I decided to blog this here since its a bit involved for a comment on a politics blog. Suffice to say I’ve been argumentatively agreeing with LPUK.This is also the second time, I’ve advocated the semantic web to solve market place failure. There is no particular reason for this, its just that modern life seems to gel with it, or something, and once you are familiar with it you can’t help but apply it when thinking about IT issues. It’s also a very open liberal system of working.
I’m going to get really specific here, because that’s the the best way to avoid being vague:
I’d choose to use Linked Data expressed in RDF+N3 to represent information about my health, and something not unlike FOAF+SSL for authentication, since I’d like to be able to use more than one service provider at the same time so it needs to be a format with native features that enable integration of data. RDF happens to also resemble EAV/CR which is a medical design pattern. The data would be stored wherever, whenever, by any number of arbitrarily chosen organisations and would be brought together ad-hoc via a tools in the Linked Data tradition. Integration tools would also be selected from an open market for doing exactly that job. If I chose to use one provider for every medical service obviously I wouldn’t need this extra bit, but having that allows a more diverse market, more privacy. Importantly, emphasising data integration as a feature leaves scope for organisations to add integration features to whatever system they already have, which includes retaining human procedures speeding up the evolution of this ecosystem.
If I were someone with specific conditions likely to need it then I’d carry a card designed to permit rapid access at A&E departments. I’d buy these cards from a similar marketplace of providers, but probably all of them would be eventually forced to catch up with the state of the art circa 1994 and support content negotiation such that once the URL is accessed whatever the doctor needs is delivered to them over HTTP. All the standard authentication options for HTTP, including FOAF+SSL would be available and may or may not be used in deciding to serve up the data. Imposing a standard protocol and format here wouldn’t be too bad, but the state needed bother compatibility with A&E is the core feature.
The method of formatting the card would be decided by an industry organised standards body, but need only be a URI. There is nothing scary now about URIs! It will contain a very long random number – too long to bother guessing it – and after first use, this URI can only be accessed for a few days. The server will know its serving emergency data and can take care of procedure matters, like waking up your mom, if that’s what you want.
The system is essentially a Summary Care Record resource but hosted by the person I chose to host it, and containing whatever I decided to put on it. If I remain in a coma, the care record will name someone to come sort out access to data, probably a relative or a staffer from one of the many organisations I might buy services or insurance from. Possibly the expired record will still provide that data, just in case.
XML, having at least the ability to be unambiguous and machine verified would be my 2nd choice of format. Automated integration is not a feature, but there is a good selection of tools and an experienced workforce. Stuff like SOAP might make things harder – too many variables – but a proper REST implementation would evolve as a norm (Linked Data is RESTful). With content negotiation the syntax doesn’t actually matter that much, because providers of emergency care cards will be incentivised to run really good software to handle syntax issues, as would the ad-hoc integration providers used in every other circumstance. Obviously then, data integration features are the deciding factor for consumers, and whatever data-integration techniques work best will rise to the top in an open market.
That said, if people want to squirt pigments through feathers onto bits of reconstituted tree to depict vague and inconsistently applied words and move the resulting “information” around using horse and cart, then they should be free to do that. Want to use something properly stunted like JSON? Sure, but those organisations doing so should also be perfectly at liberty to loose customers to competitors doing it properly.
Insurance providers would insure against the cost of transferring data out of systems when providers go bust and will set premiums according to how well the chosen providers operate, taking into account things like security, off-site back-up procedures as well as the quality of implementation details. If I choose to rely on paper records, I pay a bit more. If I am foolish enough to use a record-keeping service that uses JSON then I will pay a lot more – obviously 
Abody of shared knowledge will be created about the quality of each company – known as its “reputation”, remember them? – that will include horror stories about data coming out wrong and user interfaces being good, bad or ugly just as search engines or price comparison sites have reputation today for the same features. Obviously consumers won’t ask “is that SOAP or REST?”, or “do you have a comprehensive OWL ontology for health records?” but they’ll get to know the consequences of those technology options. Just like with search engines there will be default choices that people make, and times when you want something different, or more complex to suit your needs, so no-one will be greatly inconvenienced.
Anyone too stupid to want to own their own records could just be handed the existing dead tree words or digital records on CD and told to keep them safe or suffer their fate. A kinder alternative would be to apply the kind of opt out system advocated for education (not the one for health, but I don’t feel strongly on that) in The Plan, to the NHS until such point that all the pathetic losers that can’t be bothered to think about staying alive end up dead and the NHS is mercy slain in 2060.
It goes without saying, that side issues like access to anonymous data by academics will also be subject to market forces and people will vote with wallets.
