John Gage wrote: >> Anyone who has ever worked on developing data models for an >> application will tell you that it involves one agonisingly arbitrary >> decision after another...there is either no method for determining >> the correctness of these decisions before actually building a working >> prototype...or if there is a way of determining correctness, time and >> cost pressures prevent it from being used > > > This is not quite accurate. The level of uncertainty in medicine > makes each data model as valid as another. Hence, the VA data model > is as valid as any other and will remain so for the next 10 to 20 years.
No, I think the proof of a data model is that it works in practice without collapsing under its own shortcomings or internal contradictions, or being supported by 100,000 lines of code behind the scenes (almost any data model can be made to appear to work with enough code...). The VistA models are valuable because they have clearly stood the test of time and real-life usage without requiring inordinate amounts of code to prop them up. The data models of commercial systems of VistA's ilk are usually closely guarded and rigourously protected by licensing agreements which forbid reverse engineering. So VistA's open data models are a rare thing which extend its value to places where VistA implementation is not possible (due to lack of local MUMPS expertise, for example). Tim C
