At 10:23 PM 5/15/2002 -0400, [EMAIL PROTECTED] wrote: >Andrew points well taken. >But to raise the ongoing issue, how do we ever get to determine a uniform >identifier that can be attached to a patient's medical record/chart if we >don't come to >some determining of a format for that identifier.
We don't need a uniform identifier. We only need a uniform way of expressing identifiers. That is what the QualifiedPersonId is for in the PIDS specification. It includes a domain along with the "unique" identifier. We also need a uniform way of correlating identifiers, which is what the CorrelationMgr is for in PIDS. It is practically impossible to come up with a unique identifier. This has been discussed in the US ad nauseum by the NCVHS. >That identifier also has to have easily searchable features in order to make >it useful to research demographics etc. This is an assumption but isn't necessary for an EMPI. The PIDS specifications is an example of why this is not necessary. One has a identity which has attached to it an arbitrary number of traits. Each trait is a name,value pair. One could find all the people based on a set of demographics quite easily using this approach. At the same time, if some demographics changes (like an error in a date of birth, or a sex change operation), the identifier doesn't change. >While we can't make everyone document blood pressures the same way, hmmm I >thought we all did that, we have to have some standards for a number. >Social security numbers are what the government uses and we all know how >useful those are. >If for instance the Stamford IPA physicians, at least a group of them, all >independent business men and women with solo, group practices decide to use a >central enterprise server a more definitive way to come up with a Enterprise >Universal Patient Identifier needs to be coordinated amongst all those >separate entities. The mechanism of doing this coordination is what the Person Identification Service (PIDS) is all about. >That is the only way to assure that demographics on the same patients don't >get entered more than one, or at least it is what the goal should be. We have a paper presented at AMIA 2000 on how to build a federated person identification service on a national scale, utilizing different local identifiers. >Guess we have to start somewhere and with some standardized conventions >rather than letting the computer just pick sequential or random numbers! Please try to read the Person Identification Service specification. It describes how to do this coordination whether it be at an enterprise level, regional level, federal level, or international level. (The problem is a little harder internationally because of variants in the way of expressing traits, but is still doable). You can pick a sequential or random number for the identifier. The only requirement is that it be unique in the specified domain. See the note by Wayne Wilson on the importance of correlation management. Dave >Falball
