A Practical Design for using SNOMED-CT
in Patient Problem Lists

Partners Healthcare System (PHS) engaged Sujansky & Associates to assist in the design of a planned enterprise-wide repository of patient problem lists. The specific project was to analyze whether and how the SNOMED-CT terminology could be used to (1) represent complex clinical problems accurately and in a computable form and (2) efficiently retrieve and analyze encoded problem expressions to drive automated decision support, reporting, and research functions.

The project entailed an in-depth analysis of the SNOMED-CT model for constructing, storing, and analyzing post-coordinated expressions, as well as consideration of the computational performance of alternative implementation approaches. The project also required an understanding of the specific requirements, technical and clinical, for encoding patient problems in information systems.


Sujansky & Associates' role in this project included:

• Review of PHS's high-level requirements for a problem-list repository
  
  
• Research into the various representation methods and subsumption-testing algorithms supported by SNOMED-CT, as well as optimization strategies
  
  
• Formulation of specific recommendations for implementing the models and algorithms in the context of PHS's program objectives.
  
  
• Documentation of analysis and recommendations in a comprehensive report
  
  
• Presentation of recommendations to PHS's technical staff during on-site design meetings

Highlights

Encoded problem lists must support at least the following operations: Human review, automated decision support, search, and statistical abstraction

Medicolegal requirements influence design decisions for electronic problem lists

Depending on the storage formats selected for SNOMED-CT expressions, annual releases of the SNOMED-CT terminology may require updates of the entire
patient database

Certain optimizations of subsumption testing can produce incorrect
query results