At 08:06 PM 5/12/2002 -0400, Bruce Slater, MD wrote: >Thanks for all the information from Thomas Beale, Tim Churches, Adrian >Midgley and Falbal. > >Concerning EMPI: >Do people use CORBAmed? I am sure it has to be as complex as it seems to be >generalizable across platforms, but it creates too much of a barrier for me >right now to understand all that stuff.
I guess complexity is in the eye of the beholder. >Specifically I was hoping for an algorithm in pseudo-code in which the >looked up name is parsed into similar, sound-alike, frequent mis-keyed >alternatives, middle and first names or last names reversed alternatives and >then compared to the database of names similarly parsed. This would result >in multiple candidate matches and could rank them by likelihood of being the >actual match. There were a lot of theoretic papers at one site that probably >contain something like that, but were too dense for me to understand. There are many ways for you to accomplish this within the framework of the OMG PIDS service. The purpose of the spec was not to define such an algorithm, but to enable such an algorithm to be accessed through a fairly simple interface enabling vendors or implementors to compete on their algorithms. The find_candidates method of IdentifyPerson component of the PIDS service is where this would be implemented. If you are trying to implement a complete EMPI I think you will find the PIDS service to be fairly simple and a component out of which you could build an EMPI. It encapsulates many of the fundamental operations used in an EMPI. If you look at the conformance points, you will find that you can start with a Simple PIDS and build up from there. The complexity has little to do with the cross platform issue but with the specification of the many functions that are used in an EMPI. If you look at our AMIA 2000 paper you will find an example of how to use the PIDS specification in a Federated PIDS service. I hope this helps. >Thanks for the replies. > >Bruce >----- Original Message ----- >From: "Thomas Beale" <[EMAIL PROTECTED]> >To: <[EMAIL PROTECTED]> >Cc: "Fabiane Bizinella Nardon" <[EMAIL PROTECTED]> >Sent: Sunday, May 12, 2002 1:21 AM >Subject: Re: new subject > > > > > > > > Bruce Slater, MD wrote: > > > > >In thinking about ways to build modular functionality, there are some >things > > >that would be helpful and could possibly be built for other purposes and > > >used for open source EMR development. > > > > > >I am thinking of an Enterprise Master Patient Index - to identify all the > > >patients, providers and staff unambiguously in a large institution or > > >network for positive identification and authentication. Specifically, >does > > >anyone know a good algorithm for comparing a typed in name against a > > >database of names, given all the world wide variations in naming > > >conventions? The Soundex is a simplistic one. There must be a systematic >way > > >that has been written about in the literature. > > > > > I have seen a good implementation from the people at the Heart Institute > > in Brazil, and I believe they would share it. Try mailing Fabiane > > Bizinella at [EMAIL PROTECTED] > > > > The reason they have done this is that Brazil has all kinds of people, > > with german, spanish, chinese, indigenous etc names - and it is common > > for people to know how to say someones name but not spell it. So they > > have some sophisticated phonetic matching happening. > > > > - thomas beale > > > > > > > > > > > >
