Friday, November 06, 2009

the problem with NPfIT is the "NP" bit

Something interesting about the NHS NPfIT project. During my recently completed two-week conference binge, I spoke to people from a British telecommunications company who were fresh, if that's the word, from tangling with the NHS IT Zombie, and had apparently escaped before it ate their brains with a spoon. I also heard people from a French telecommunications company who had been working in the same field speak.

They agree on this; national healthcare institutions are too complicated for any one organisation to build the kind of comprehensive, end-to-end workflow system that NPfIT envisaged. This is partly because of the incredible complexity of their business processes; an episode of care can span anything from a GP appointment that ends by the patient being told there is nothing the matter with them, or an immunisation being administered in a single visit by a nurse, to 20 years of treatment for a cancer and subsequent surveillance. There are a hell of a lot of other organisations that interact with the NHS, and who aren't part of the project.

In fact, if they were, the scale and scope of NPfIT would increase to the point at which it encompassed most of the public sector; it would have to integrate with the social security system, and because of all those benefits that are delivered as tax credits, with the Revenue as well, and (because the NHS provides the armed forces' medical care) with the MOD's personnel system and even with tactical communications systems in the RAF, because Selly Oak receives casualties direct from the war. Of course, it no doubt already needs to talk to the Treasury's systems. You might as well just ask the ghost of Stafford Beer to build us a Cybersyn for the whole economy.

But that wasn't the worst of it. The real problem, according to my source, was that the designers of NPfIT believed that there was an organisation called the NHS. In fact, this was a bit like modelling a blue whale as a homeogenous sphere to make the maths easier. The killer wasn't that medicine is complicated; it was that the NHS isn't a monolithic organisation. It is, of course, an institution - a set of social, political, and economic expectations and relationships, a recognisable culture, a way of understanding the world. But it's far from being a single organisation.

Instead, it's an ecosystem, made up of many organisations that sometimes play similar roles (it's a hospital; it's a GP practice) but differ dramatically in their internal structure, rather as a dolphin and a Humboldt squid are both social, pelagic, fast-swimming predators in the subtropical ocean. However, only one of them is even a tetrapod, and only a real idiot would assume they were both sufficiently described by the concept of "shark". And the interactions between the creatures in this ecosystem are deeply complicated.

In that sense, it's quite a lot like the Internet. That, too, consists of a grab-bag of diverse organisations that cooperate with varying success on the basis of a few rules and a rough common culture, which is often honoured more in the breach than the observance. That also has a lot of odd emergent features that arise from its complexity, and would almost certainly be impossible to design as a single organisation. Indeed, an old staple of Internet-related mailing lists is the question of what the word "Internet" actually means.

Cue facile libertarian woofing. Yadda yadda Hayek privatise the BBC. Spare me. Neither does this mean the NHS is disorganised; it may well mean that it's better for its geographically and functionally diverse components to work differently. It would be surprising if they all shared a single optimal strategy. Of course, there is a perfectly good paradigm for building effective information systems in circumstances like these (and another one). What's really deeply depressing about this is that after all the blundering about and the money, there's still not the key element that makes a Web-like approach possible - standard data formats and interconnection procedures.

How much would it have cost to sponsor an effort to fix that, coming up with an XML standard or a Semantic Web ontology and some NHS standards, setting down for example where the canonical data would live and who could get at it in what circumstances?

1 comment:

yorksranter said...

this comment is entirely free of content

kostenloser Counter