| Hubris | One File per Patient |
Current electronic records cannot be moved from practice to practice, even at present as unformatted text, although the TextBase project may loosen that up a little.
The trouble and expense of removing one patient's entries from a relational database EMR appear considerable, and the effects on the validity of the audit trail interesting.
Certain parts of the medical record are both formatted and organised, blood pressure measurements for instance, and immunisation records, and these would be best dealt with as structured elements embedded in the file. Consider as an example an Excel worksheet with the immunisations, with a bookmark to locate it from the table of contents.
The internal structure and size of Word files may not make them the first candidate for such systems, but they are excellent for thought experiments.
There is an audit trail in there somewhere, the question is can it be made as tamper proof as the existing audit trails.
The huge number of structured text files on the web are indexed separately by search engines which actually have many interesting features for medical records purposes. The Excite "documents more like this" feature is a particularly interesting one. Imagine pulling up the record of a man who has just had an early death, and asking the search engine for "documents more like this", and being presented with the notes of the patients who the search engine finds to have notable similarities to the departed one.
Far fetched? Artificial Intelligence fantasy rather than good solid computer engineering? Certainly at present not wholly reliable, but neither are the coding schemes we partially impose upon some of ourselves some of the time in order that our databases may return all patients with CVS disease, or Diabetes, or impotence (not a commonly coded symptom, and just about to become a useful one to keep count of with Viagra arriving.
Weblike search engines searching for precise Read Code phrases picked from a dictionary and placed in a text/XML/Word file among other text would work as well as searches on the Read Code index in our current databases. And in addition, if the relevant wors had been typed in freehand they would be picked up.
It is entirely possible that the combination of free text entry (with optional code selection) and fuzzy searching would be more congenial to GPs, and therefore used more effectively with an end result of an increased accuracy of disease registers and the demand forecasting that we desire in order to run the health service more effectively.
If we find ourselves using a browser front end, running over a TCP/IP network, perhaps even an NHS wide intranet, displaying information from a variety of practice systems, then even if it only shows a picture of the old text-based screen we will have defined a de facto standard for notes format, the HTML packages being sent over the net.
From there to using that format to store the notes in is a relatively small step.
And it is a step which we should continue to consider