Letter: Note on Friedman's 'fundamental theorem of biomedical informatics'

needs of practical applications with the importance of the purity and accuracy of the terminology is a very difficult task..’ We would actually state that it is impossible. We would assert that hierarchies are created for a purpose. The purpose could be to support automated maintenance of the terminology, to create definitional relationships, or to allow appropriate inferences. If the different users of the terminology have different purposes, it is not always possible to meet all needs with a single hierarchy, no matter how much care is taken. Even with the clarifications offered by the respondents, the fact remains that the hierarchical relationships represented in SNOMED CT do not ideally support public health reporting use cases. The desired case roll-up behavior is not supported by the unmodified SNOMED CT hierarchies and concepts. While we asked for the development of a hierarchy that is exclusively for human conditions, a better statement of the need would have been the development of a hierarchy that is valid for public health inferencing about human disease. The problem is not with SNOMED CT per se. The root of the problem is that all purposes cannot be met by a single hierarchy. We need to support many purpose-specific hierarchies. To this end, we support the respondents’ suggestion for increasing creation of subsets. We look forward to learning more about how subsets can be created and exchanged to meet specific clinical needs.