The Scraps for the Bucket

The model used for existing systems is old, tired and will not stretch to the end of the century. So much for Y2k problems. (2002 note: So young, so optimistic.)

Even ignoring for the moment the need to have the evidence-base software running continuously, already important and perhaps the major function in the next few years, we need the notes and prescriptions to be reliable.

We also all need the ability to read and make entries in different places.

The model of the British GP computerised medical records in use at present is that of a single, central database which is updated in real time by the user, directly.

It either works, or doesn't, and when it doesn't there is no fall-back, one must wait until it is restored to working order.

The most common fault is corruption in a single record in an index. Until the index is remade, which may require all indexes to be remade as often the error trap does not report the specific index at fault, still less automatically repair it, no further use can be made of the entire system. This was reasonable when small quantities of data were redundantly held on computer systems and therefore reindexing was quick, but is not good enough now.

Using a resilient or journalling file system such as found on Windows NT or the Reiser and Ext 3 files systems on Linux with transaction tracking to protect against partial updates of the database merely reduces the problem which is implicit in the basic model. Time to change the model.

Doing this turns out to meet many of our other wants and needs half-way, as well.

What is required is a simple client program running on the desktop, the portable, or on the server hardware to drive a terminal, which generates messages to update the main database.

This means that should the database go down, the messages can be stored and played in when it is fixed. It also allows the laptop to be taken for walk and work to continue.

Moving to a model where all information is fed to the main database as messages will affect links to other practices, suddenly much easier, and to laboratories, ditto, and less obviously but also very usefully improves back-up and archiving arrangements.

Conceptually the stream of incoming messages, together with inspection of the stored information, fits the manner in which we do medicine better than the current models. The lesson from many computerised systems is that making the computer model similar to real life is not a necessary condition for happiness, but it is intuitively attractive and can make life easier for users.