Sculpting the UMLS Refined Semantic Network

Background

The Refined Semantic Network (RSN) for the UMLS was previously introduced tocomplement the UMLS Semantic Network (SN). The RSN partitions the UMLSMetathesaurus (META) into disjoint groups of concepts. Each such group issemantically uniform. However, the RSN was initially an order of magnitudelarger than the SN, which is undesirable since to be useful, a semanticnetwork should be compact. Most semantic types in the RSN representcombinations of semantic types in the UMLS SN. Such a “combinationsemantic type” is called Intersection Semantic Type (IST). Many ISTsare assigned to very few concepts. Moreover, when reviewing those concepts,many semantic type assignment inconsistencies were found. After correctingthose inconsistencies many ISTs, among them some that contradicted UMLSrules, disappeared, which made the RSN smaller.

Objective

The authors performed a longitudinal study with the goal of reducing the sizeof the RSN to become compact. This goal was achieved by correctinginconsistencies and errors in the IST assignments in the UMLS, whichadditionally helped identify and correct ambiguities, inconsistencies, anderrors in source terminologies widely used in the realm of publichealth.

Methods

In this paper, we discuss the process and steps employed in this longitudinalstudy and the intermediate results for different stages. The sculptingprocess includes removing redundant semantic type assignments, expandingsemantic type assignments, and removing illegitimate ISTs by auditing ISTsof small extents. However, the emphasis of this paper is not on the auditingmethodologies employed during the process, since they were introduced inearlier publications, but on the strategy of employing them in order totransform the RSN into a compact network. For this paper we also performed acomprehensive audit of 168 “small ISTs” in the 2013AA versionof the UMLS to finalize the longitudinal study.

Results

Over the years it was found that the editors of the UMLS introduced some newinconsistencies that resulted in the reintroduction of unwarranted ISTs thathad already been eliminated as a result of their previous corrections.Because of that, the transformation of the RSN into a compact networkcovering all necessary categories for the UMLS was slowed down. Thecorrections suggested by an audit of the 2013AA version of the UMLS achievea compact RSN of equal magnitude as the UMLS SN. The number of ISTs has beenreduced to 336. We also demonstrate how auditing the semantic typeassignments of UMLS concepts can expose other modeling errors in the UMLSsource terminologies, e.g., SNOMED CT, LOINC, and RxNORM that are importantfor health informatics. Such errors would otherwise stay hidden.

Conclusions

It is hoped that the UMLS curators will implement all required correctionsand use the RSN along with the SN when maintaining and extending the UMLS.When used correctly, the RSN will support the prevention of the accidentalintroduction of inconsistent semantic type assignments into the UMLS.Furthermore, this way the RSN will support the exposure of other hiddenerrors and inconsistencies in health informatics terminologies, which aresources of the UMLS. Notably, the development of the RSN materializes thedeeper, more refined Semantic Network for the UMLS that its designersenvisioned originally but had not implemented.