Re: Warning of major NHS IT overspend

[Ian Miller <[EMAIL PROTECTED]>]

At 1:04 +0000 2/11/04, Dave Howe wrote:
>Actually, [distributed databases] can be quite simple and reliable -

Sort of.  They can be conceptually simple and ultimately reasonably
reliable, however it is a massive amount of effort to do so.  So much so
that even for a system the size of the NHS they are almost certainly better
using an existing architecture than attempting to build one from scratch.

> how simple and reliable
>depends on how often the distributed data needs to be updated, how
>conflicting updates are resolved, and how often the schema which
>indicates what is stored where is changed.
>DNS (as used by the internet) is a good example of a distributed system

A very good example.  Whereas DNS is principle very simple, DNS servers are
very complex programmes if they are going to do their job properly and
efficiently.  The same is true of a lot of other Internet infrastructure
services.

At 21:54 +0000 1/11/04, Roger Hird wrote:
>I think I'm saying that if you want a system giving full access to records
>from a range of sources (GP, hospitals, etc)

That is an awfully big 'if', which is what most of the argument is about.
I have yet to hear any clinical requirement for this.

>a central database is
>probably still the only way that you can do it reliably.

I don't think that this is necessarily still the case.  There has been an
enormous amount of progress on distributed filing systems, Gnutella,
Eternity-Service, etc.  I don't know enough detail about the existing
systems to say if any would be suitable.  The main problem is probably that
most of the existing work has been anti-censorship and so concentrates on
maximising availability.  Whereas a system handling medical records should
be maintaining a high degree of confidentially.

My own view is that trying to implement a centralised database is an
extremely expensive way of compromising medical confidentially with the
potential to dangerously undermine the public's willingness to confide in
their doctors.  If this results people with dangerous infectous diseases
not consulting their doctors, this could kill a lot of people.  I don't see
any corresponding benefits.

What would make sense is specifying as a requirement for all new medical
systems that they must all be capable of interchanging data in NHS standard
data formats, so that records can easily be transferred around the system
where necessary.  However the decision to transfer any given patients
records should be taken by the clinician responsible for the patient.

Ian

--
Singularis Ltd, 32 Stockwell St, Cambridge, CB1 3ND
Tel:  +44 1223 525088	            Mobile: +44 777 5536663
Fax:  +44 870 0514333	 (e-mail preferred to Fax)