Previous Next

Limitations of the traditional approach to medical terminologies

The first major problem with enumerative classification is that an apparently modest increases in the expressive power of a scheme can produces a large increase in the number of terms and relationships. Consider a scheme in which there are 1,000 terms for diseases. If we introduce the idea of the severity of a disease, and limit this to three degrees of severity (mild, moderate, and severe), then the total number of terms representing diseases becomes:

1,000 disease x 3 severities + 1000 original terms = 4,000 terms

In order to increase the expressive power to cover the severity of a disease the number of terms has quadrupled. If we now introduce the progress of a disease (better, same, worse) the result is 16,000 terms.

As well as a dramatic increase in the number of terms there is a corresponding and potentially more serious increase in the number of relationships between those terms. In order to cope with multiple uses it is necessary to accommodate multiple hierarchies and non-hierarchical relationships. Thus for each term it is necessary to enumerate at least one, and possible several, hierarchical relationships. Thus even for the small example above, the modest addition to the clinical scope which led from 4000 to 16000 terms could easily require a further 50,000 relationships to be added. This is the ‘combinatorial explosion’.

The problem with the combinatorial explosion is not just one of the rapid growth in the numbers of terms and relationships. Medicine is large and complex, and hence anything we do to represent medical concepts will in some respect be large and complex. The difficulty with the traditional schemes is that they rely solely on human effort to enumerate and interpret pre hoc all of the possible terms and their relationships. It is perhaps just conceivable that such an effort could enumerate all the required terms for a broad area of medicine. It is inconceivable that unaided human effort could enumerate all of the required hierarchical and other relationships between those terms. Traditional schemes have until, recently been of a size and complexity that could be comprehended by one or two workers. Schemes of hundreds of thousands of terms with multiple classifications and various ad hoc qualifiers are beyond unaided human comprehension. GALEN recognises this as a fundamental problem with representing medical concepts using traditional schemes and is developing an alternative approach.

Previous Next
Making the impossible very difficult, ©OpenGALEN.org, All rights reserved