Jon Patrick wrote: > sorry no papers on it just yet, we only finished the work last week. > However Tim Churches and Peter McIsaacs attended today's presentations > and might be prepared to make comments to the list, critical or otherwise > cheers > jon > > syan tan wrote: >> Is there a paper describing the ed system on the net? >> On Mon, 2006-11-20 at 17:42 +1100, Jon Patrick wrote: >>> I hope the list will excuse a small advertisment for our upcoming R&D >>> Showcase. .... >>> 9:30-10:00 Document-Centric Workflow Management System for an Emergency >>> Department - William Chau >>> 10-10:30 Generic server for terminologies, ontologies and >>> classifications - >>> Peter Budd
Sorry for the delay in responding to this. I would sum up what Jon's students have done as being very, very cool, necessarily incomplete as yet due to the very modest time and resources devoted to it so far, but definitely worthy of further development. Three components were discussed. One is a data collection form definition/back-end database module, which allows data collection forms to be defined using a simple XML format (a nice point-and-click, user-friendly editor for these XML definitions is eminently feasible), which then allows these data forms to be rendered on a portable notepad computer - one of those flat laptops on which you can write directly on the screen using a stylus, and it does handwriting recognition of what you enter (rather well, after some training). All data is stored in a shared back-end repository, with the notepad PCs linked by local area wireless to it. Also hooked in is a version of the text-to-SNOMED engine which Jon has re-drawn our attention to on this list last night. So, if you scrawl (or type - there is an optional keyboard) some notes in a text box on a data collection form, not only is your handwriting converted into and stored as typed text (a service provided by the notepad PC, not by Jon's students' work), but the text is also automatically dredged for SNOMED CT terms and these are added as annotations to the text, and stored with it. All the text processing is done by a remote compute cluster, which can be a collection of a few, up to hundreds of low-cost desktop computers harnessed together as a form of cheap supercomputer. This means that very sophisticated text processing and SNOMED CT term recognition algorithms, which tend to be very CPU intensive, can be used because the processing does not happen locally on the notepad computer. For example, we have been experimenting with fully-Bayesian spelling correction, which works almost supernaturally well at guessing what word was actually intended (given the context), even if spectacularly misspelt or abbreviated, and correcting the spelling, but which requires a large amount of computing power (or a lot of time to think on a single computer - too much time for interactive use). Finally, all these data collection forms are hooked into a patient-centric framework, which allows the data about a particular patient, in the form of completed or partially-completed forms, files or other documents, to accumulate, and for forms to be virtually routed between users via a workflow engine. This workflow engine uses XML definitions of non-deterministic Petri nets (no, not Petri dishes - see http://en.wikipedia.org/wiki/Petri_net ) to determine what should be done next with respect to the forms. All up, very impressive work, and although far from complete or ready for production use, it still seems far, far more advanced in its ability to capture information that is useful to both human clinicians and is computable for decision support, epidemiological and other research purposes than the large, very expensive US-sourced information systems which are currently being rolled out in public hospitals in NSW and other states. Although the students' work centred mainly on EDs, the system would, I think, be ideally suited for community-based shared care systems, which of course involve GPs. Indeed, it could also evolve into a primary care system for use within a group GP practice, as well as for facilitating shared care amongst a wider group of community-based health professionals. Clearly investment is needed to develop such a system - which would be open sourced, of course - Jon's group is fully behind the idea of such work being funded in a way that permits open sourcing as both the immediate and long-term goals - as the aim is to create local R&D and support work opportunities for his IT school's students and graduates, not to make zillions of (almost certainly elusive) dollars in software licensing fees for the university. I can't help thinking that the investment required is measured only in the millions of dollars. Finally, there is a lot of overlap of the work described above with openEHR - but unlike the openEHR group, which has employed a very theoretic, thought-experiment approach, but which still doesn't have a complete set of software tools available which can be used to build functional openEHR-based systems, Jon's students have taken a more middle path, tempering theoretical design work with an empirical "let's try it out and see if it works" approach. As a result, they have been able to come a long way in a short time. There is also a lot of overlap and commonality of approach with the Web-based, open source public health data collection and management systems which I and my colleagues have been working on, on and off, for the last several years - more on these in the next few weeks as we prepare for a Version 1.0 release - they may be of some direct interest for GP research projects, as well as of indirect interest from a technology point of view for future GP information systems. I hope that there will be opportunities in 2007 to converge our public health work with Jon's stdents' work, and also make use of the very significant body of theoretical development made available by the openEHR people. Let's see, but these seem like exciting and very hopeful developments to me. Tim C _______________________________________________ Gpcg_talk mailing list [email protected] http://ozdocit.org/cgi-bin/mailman/listinfo/gpcg_talk
