The One Model Falacy

Just about every standards organization starts from a core or reference information model these days.  From this model all things exchanged are eventually derived.  The models themselves are most often expressed in a collection of one page UML diagrams.

A model is a description of a thing meant to show important properties and to enable others to study its characteristics.  Multiple viewpoints need to be accounted for in modeling.  Just like in architecture where we need to show the structural steel and concrete in one view, and the electrical power runs in another view, and the final artists rendering of the building in yet another view.  Each of these views "hides" some important component of the finished product, making others clearer and more visible so that the interested parties can see what is important to them.  But each view is necessarily incomplete in order to show what is normally hidden behind other parts of the system.

In developing requirements for a use case (such as transfers of care), we being to develop a "picture" (another word for model) describing what is needed.  The desire to put these into a model is hard to avoid, and in fact, provides some needed rigor to the process (which often starts off as an Excel table with three word phrases in it).  The evolving "pictures" often look pretty familiar. There are already a lot of models of information in healthcare, and in a healthcare activity involving 100 people, I can assure you that at least 50 of them have already been involved in other activities.  Having built the table, sometimes we look to adopt a particular model.   Model envy ensues, often descending into Ken-L Ration justifications for use specific models.

Right now, there is no single commonly accepted model of healthcare information.  There are some common components which just about everyone agrees upon.  There are some ways of putting things together that have majority appeal.  After that, especially at higher levels of detail, it descends not quite into chaos, but certainly into a variety of approaches defined by a number of different methodologies argued for by experts.  These experts are comfortable with one model "brand" or another, and sometimes even formulate their own possibly even based on the work of those others so as to include some viewpoint or feature omitted or not specified in enough detail.

Why are there so many models?  Why isn't there just one?  Modeling is often a subjective exercise.  The model is "complete" when the viewpoint being expressed is understood (and agreed to) by the experts expressing it.  Understanding is based on skills, education, and having been present for the discussions where the model is created.  There is often no objective criteria to establish completeness, to judge appropriateness, et cetera.  It's still largely an art rather than a science, much like gardening.

One of the questions I asked the Clinical Information Modeling team is whether they had been able to establish objective criteria for what information belonged in the "Core" of the exchange packet.  As yet, we don't have any.  It is based on the subjective experience of the model developers.  Perhaps there might be some evidence-based approach we could use to define essential, relevant, et cetera, but we haven't found it yet.

We have the same problem with interoperability.  What is the objective measurement of interoperability?  The definition doesn't really help.

Bringing some objectivity into the discussion might help us all to move forward.  Think about that the next time you build an model.  What is your objective criteria for inclusion?  Completeness?