Desiderata for a clinical terminology server.

Clinical terminology servers are distinguished from more broadly based terminology servers intended for nomenclature development or mediation across classifications. Focusing upon the consistent and comparable entry of clinical observations, findings, and events, key desiderata are enumerated and expanded. These include 1) word normalization, 2) word completion, 3) target terminology specification, 4) spelling correction, 5) lexical matching, 6) term completion, 7) semantic locality, 8) term composition and 9) decomposition. Comparisons of this functionality to previously published models and specifications are made. Experience with a clinical terminology server, Metaphrase, is described.

Metaphrase: an aid to the clinical conceptualization and formalization of patient problems in healthcare enterprises.

Patient descriptors, or "problems," such as "brain metastases of melanoma" are an effective way for caregivers to describe patients. But most problems, e.g., "cubital tunnel syndrome" or "ulnar nerve compression," found in problem lists in an Electronic Medical Record (EMR) are not comparable computationally--in general, a computer cannot determine whether they describe the same or a related problem, or whether the user would have preferred "ulnar nerve compression syndrome." Metaphrase is a scalable, middleware component designed to be accessed from problem-manager applications in EMR systems. In response to caregivers' informal descriptors it suggests potentially equivalent, authoritative, and more formally comparable descriptors. Metaphrase contains a clinical subset of the 1997 UMLS Metathesaurus and some 10,000 "problems" from the Mayo Clinic and Harvard Beth Israel Hospital. Word and term completion, spelling correction, and semantic navigation, all combine to ease the burden of problem conceptualization, entry and formalization.

MEME-II supports the cooperative management of terminology.

Health care enterprises need enterprise-wide terminologies to compare, reuse and repurpose health care descriptions. But once they are created, these terminologies need to be maintained and enhanced to sustain their utility and that of the descriptions encoded with them. MEME II (Metathesaurus Enhancement and Maintenance Environment, Version II) supports the required activities and enables enterprises to leverage their investment in terminology and descriptions by permitting remote-extra-enterprise-enhancements to terminology to be incorporated locally, and local-intra-enterprise-enhancements to be shared remotely. MEME II represents all changes to terminologies as data, or "actions," that can be interpreted by an "action engine." These actions, or messages, represent semantic "units of work" that can be interpreted by other copies of MEME II. The exchange of update messages increases the likelihood that the comparability of terminology-based health care descriptions can be sustained.

Toward reusable software components at the point of care.

An architecture built from five software components -a Router, Parser, Matcher, Mapper, and Server -fulfills key requirements common to several point-of-care information and knowledge processing tasks. The requirements include problem-list creation, exploiting the contents of the Electronic Medical Record for the patient at hand, knowledge access, and support for semantic visualization and software agents. The components use the National Library of Medicine Unified Medical Language System to create and exploit lexical closure-a state in which terms, text and reference models are represented explicitly and consistently. Preliminary versions of the components are in use in an oncology knowledge server.

Merging terminologies.

A terminology is a systematic, authoritative collection of concept names, or terms, in some domain. No single terminology names all the important concepts in biomedicine. One approach to creating a more comprehensive biomedical terminology is to merge existing biomedical terminologies, as the UMLS( Metathesaurus( has done for the last six years. Because existing terminologies may overlap--for example, one terminology may name a concept also named by another terminologyQthe terminologies in the Metathesaurus must be merged. Some terminologies suggest merges through their structure or content e.g., they suggest synonyms or connections to other terminologies; other merges can be suggested by algorithm. Regardless, all merges in the Metathesaurus must be approved by a human editor with appropriate domain knowledge. By the time Meta-U96 is released early in 1996, one prototype and seven released versions of the Metathesaurus will have been produced by a sequence of four qualitatively different methods, named for the way in which they merge terms: #1 "Term Rewrite Rules, #2 "Transitive Closure on Facts," #3 "Fact-at-a-Time Concept Merging," and #4 "Action-at-a-Time Object Processing." The development of each method has been constrained by the annual Metathesaurus release schedule. The first two methods made the best use of limited computational resources, and the last two make better use of human editing resources.

Identifying concepts in medical knowledge.

The barrier word method of identifying nominal phrases in text, using a very long barrier word list, was evaluated in two different sets of text. In a sample of 10 paragraphs from the Medical Knowledge Self-Assessment Program of the American College of Physicians, the yield of nominal phrases as a percent of total chunks isolated was 66%. Some 500,000 chunks were isolated from Principles and Practice of Oncology (PPO). 38% of these chunk-occurrences were of chunks which matched to 10,000 concept names in Meta-1.4, the most recent version of the UMLS Metathesaurus. 50 paragraphs from PPO were chosen at random. Co-occurrences of concepts in those paragraphs were reviewed. 42 of the paragraphs had unique or infrequently occurring co-occurrences which described closely the major thrust of the paragraph.

Accessing oncology information at the point of care: experience using speech, pen, and 3-D interfaces with a knowledge server.

Oncologists' information needs arise at diverse times and settings. For example: "Is superior vena cava syndrome a medical emergency?" Our collaborative group is developing a system that supports an interface with combinations of spoken, gestural, and simulated three-dimensional manipulation to help an oncologist focus on the information need, not the system. The system requires a small amount of input from the oncologist, and then anticipates what information is pertinent to the patient at hand, based on the sources it has available. The system makes use of a "Knowledge Server" to find relevant information. The Knowledge Server uses selected data for the particular patient from a Computer-based Patient Record (CPR) to provide context for the information needs. The Knowledge Server leverages the Unified Medical Language System (UMLS) resources as well as relevant communications standards. A layered, interaction protocol is used to help manage the fulfillment of information needs. Each of the oncology knowledge sources is transformed into a uniform representation that utilizes both its formal schema (e.g., its table of contents) and its concepts and words indexed through the UMLS Metathesaurus. Our focus on the appropriate use of information from a CPR, and on anticipating oncologists' information needs, resulted from our study of several longitudinal patient scenarios. We believe that our use of scenario-based design techniques will help to ensure the system's success.

Semantic visualization of oncology knowledge sources.

Visualization of knowledge sources can have a substantial impact on the use of such sources at the point of care. This is because barriers to use at the point of care include hours required to master the electronic interfaces to those sources, and minutes required to master the electronic interfaces to those sources, and minutes required to accomplish any one retrieval. For a system to be used regularly at the point of care, therefore, it must be intuitive and fast. This paper presents a three dimensional interface to oncology knowledge sources that aims to meet this challenge.

Formal properties of the Metathesaurus.

The Metathesaurus is a machine-created, human edited and enhanced synthesis of authoritative biomedical terminologies. Its formal properties permit it to be a) exploited by computers, and b) modified and enhanced without compromising that usage. If further constraints were imposed on the existence and identity of Metathesaurus relationships, i.e., if every Metathesaurus concept had a "genus" and a "differentia," then the Metathesaurus could be converted into an "Aristotelian Hierarchy." In this sense, a genus is a concept that classifies another concept, and a differentia is a concept that distinguishes the classified concept from all other concepts in the same class. Since, in principle, these constraints would make the Metathesaurus easier to leverage and maintain computationally, it is interesting to ask to what degree the maintenance and enhancement procedures now in place are producing a Metathesaurus that is also an "Aristotelian Hierarchy." Given a liberal interpretation of the current Metathesaurus schema, the proportion of the Metathesaurus that is "Aristotelian" in each annual version is increasing in spite of dramatic concurrent increases in the number of Metathesaurus concepts. Without formality there is no modifiability nor scalability. [1] We need formal methods and computer-based tools that can help us with the task [of controlled medical vocabulary construction]. We need research in which controlled vocabulary development is the focus rather than a stepping stone for work on other theories and applications. [2]

The UMLS Metathesaurus: representing different views of biomedical concepts.

The UMLS Metathesaurus is a compilation of names, relationships, and associated information from a variety of biomedical naming systems representing different views of biomedical practice or research. The Metathesaurus is organized by meaning, and the fundamental unit in the Metathesaurus is the concept. Differing names for a biomedical meaning are linked in a single Metathesaurus concept. Extensive additional information describing semantic characteristics, occurrence in machine-readable information sources, and how concepts co-occur in these sources is also provided, enabling a greater comprehension of the concept in its various contexts. The Metathesaurus is not a standardized vocabulary; it is a tool for maximizing the usefulness of existing vocabularies. It serves as a knowledge source for developers of biomedical information applications and as a powerful resource for biomedical information specialists.

Recognizing new medical knowledge computationally.

Can new medical knowledge be recognized computationally? We know knowledge is changing, and our knowledge-based systems will need to accommodate that change in knowledge on a regular basis if they are to stay successful. Computational recognition of these changes seems desirable. It is unlikely that low level objects in the computational universe, bits and characters, will change much over time, higher level objects of language, where meaning begins to emerge, may show change. An analysis of ten arbitrarily selected paragraphs from the Medical Knowledge Self-Assessment Program of the American College of Physicians was used as a test bed for nominal phrase recognition. While there were words not known to Meta-1.2, only 8 of the 32 concepts new to the primary author were pointed to by new words. Use of a barrier word method was successful in identifying 23 of the 32 new concepts. Use of co-occurrence (in sentences) of putative nominal phrases may reduce the amount of human effort involved in recognizing the emergence of new relationships.

Toward an interim standard for patient-centered knowledge-access.

Most care-giver "knowledge" needs arise at the point of care and are "patient-centered." Many of these knowledge needs can be met using existing on-line knowledge sources, but the process is too time-consuming, currently, for even the computer-proficient. We are developing a set of public domain standards aimed at bringing potentially relevant knowledge to the point of care in a straight-forward and timely fashion. The standards will a) make use of selected items from a Computer-based Patient Record (CPR), e.g., a diagnosis and measure of severity, b) anticipate certain care-giver knowledge needs, e.g., "therapy," "protocols," "complications," and c) try to satisfy those needs from available knowledge sources, e.g., knowledge-bases, citation databases, practice guidelines, and on-line textbooks. The standards will use templates, i.e., fill-in-the-blank structures, to anticipate knowledge needs and UMLS Metathesaurus enhancements to represent the content of knowledge sources. Together, the standards will form the specification for a "Knowledge-Server" (KS) designed to be accessed from any CPR system. Plans are in place to test an interim version of this specification in the context of medical oncology. We are accumulating anecdotal evidence that a KS operating in conjunction with a CPR is much more compelling to users than either a CPR or a KS operating alone.

The homogenization of the Metathesaurus schema and distribution format.

The third version of the UMLS Metathesaurus, Meta-1.2, to be released in October 1992, will have a simpler schema and simpler distribution formats than the first two versions, Meta-1.0 and Meta-1.1 released in October 1990 and 1991, respectively. For one thing, it will have only a single kind of entry (Concept), rather than three (Concept, Related, and Synonym). Further, the Relational Format, will consist of four logical relations, or tables, instead of the nearly three score different tables used to represent the same kind of information in Meta-1.1. These four tables will contain, respectively, (1) the names of each concept, (2) the relationships between concepts, (3) attributes of the concepts, and (4) a word-based index into the concept names. We argue that the new schema and formats provide a better conceptual model of the Metathesaurus, and represent the information contained there more uniformly. Even though these changes are incremental and evolutionary, both users and software developers should find the Meta-1.2 significantly easier to understand, and the information contained in it significantly easier to use.

The Meta-1.2 engine: a refined strategy for linking biomedical vocabularies.

This paper presents a preliminary description of the database schema and associated procedures that are the foundation for the "engine" that will produce Meta-1.2. Meta-1.2 is the next incarnation of the Metathesaurus, which is one of the principal components of the National Library of Medicine's Unified Medical Language System (UMLS). We use the word "engine" as a generic term that includes a database and the programs that operate on it. While this design builds heavily upon previous work, it incorporates some major changes in philosophy. A major hypothesis is that the simple representation described here is suitable for any controlled vocabulary in the biomedical domain. Indeed, this hypothesis is central to a strategy for producing future versions of the Metathesaurus and for supporting collaboration with people who wish to contribute additional terms and relationships to the Metathesaurus. Another change involves the representation of classes and relationships. The revised database schema includes an explicit representation of the source or "authority" for relationships, which is analogous to the way that the sources of terms have been represented since the first version of the Metathesaurus. A sequence of steps utilizing the new representations to produce the Metathesaurus is presented.

The semantic structure of the UMLS Metathesaurus.

Meta-1.1, the UMLS metathesaurus, represents medical knowledge in the forms of names of concepts and links between those concepts. The representations of the semantic neighborhood of a concept can be thought of as dimensions of the property of semantic locality and include term information (broader, narrower, or otherwise related), the contextual information (parent-child, siblings in a hierarchy), the semantic types, and the co-occurrence data (links discovered empirically from concepts used to index the medical literature.) The degree of redundancy of each of these dimensions was investigated by reviewing the extent of multiple presentations of concepts which appear as related to a given concept. The degree of overlap was surprisingly small. While the co-occurrence data finds some of the links represented by other dimensions, those links are but minute fractions of the vast amount of co-occurrence derived links. Because parent-child relationships are often subsumptive (or categorical) in nature, it might be expected that siblings usually share the same semantic types. While true in the aggregate, the wide variance in percent of types shared may reflect the intended usages of the source vocabularies. Noun phrases were extracted from the definitions of 40 concepts in Meta-1 in order to assess systematically the coverage of important concepts by Meta-1, and to assess whether the links between these definitional concepts, which may have special value, and the concept being defined were indeed present. Out of 161 of these definitional concepts, 29 were not represented in Meta-1, and 37 of those represented in Meta-1 had no direct link to the concept they were defining.(ABSTRACT TRUNCATED AT 250 WORDS)

Biomedical database inter-connectivity: an experiment linking MIM, GENBANK, and META-1 via MEDLINE.

The linkage of disparate biomedical databases is an important goal of the Unified Medical Language (UMLS) Project. We conducted an experiment to investigate the feasibility of using UMLS resources to link databases in clinical genetics and molecular biology. References from MIM ("Mendelian Inheritance in Man") were lexically mapped to the equivalent citations in MEDLINE. The MeSH major subject headings by which the citations in a particular MIM entry had been indexed were used to develop a "genetic-disorder-centered view of the world" in Meta-1 (the first official version of the UMLS Metathesaurus). Our hypothesis was that these MeSH subject headings could provide access to a "semantic neighborhood" in Meta-1 that would be relevant to a particular genetic disorder. By browsing in this "semantic neighborhood," a user could select various combinations of terms with which to search MEDLINE through an interface between Meta-1 and Grateful Med. Such searches might retrieve citations that were more recent than those in MIM or that provided useful supplementary information. Since some MEDLINE records contain pointers to entries in GENBANK, information about genetic sequences related to a particular clinical genetic disorder could also be retrieved. This scenario was implemented for a small number of MIM entries, providing a concrete demonstration that linking disparate electronic databases in an important subdomain of biomedicine is relatively straightforward.

From meaning to term: semantic locality in the UMLS Metathesaurus.

The Unified Medical Language System Metathesaurus represents the results of a synthesis of existing biomedical naming systems (thesauri). The naming and other information about the meanings in the Metathesaurus can be used to find the preferred naming of that meaning in the source chosen by the user, by exploiting the property of semantic locality. The aspects of semantic locality in the Metathesaurus which can be thus exploited are the terms, the semantic types, the use of that term in a source context, and the co-occurrence of terms in MEDLINE. To find how a meaning is named in the source of choice, a user must exploit one of these aspects of semantic locality, entering a term somehow related to the term being sought, and navigating to the preferred term. While the first three of these aspects of semantic locality are normative, the last is empirical. Testing of the utility of the aspects of semantic locality in information retrieval would require a uniform interface with 1, no Metathesaurus, 2, the Metathesaurus without the aspects in question, and 3, the Metathesaurus including all the aspects. Other potential uses of empirically derived semantic locality include defining or suggesting potentially relevant concepts in a given situation.

Adding your terms and relationships to the UMLS Metathesaurus.

The National Library of Medicine's Unified Medical Language System [1] Metathesaurus contains the richest single corpus of biomedical names in existence. Yet, developers wishing to make use of the Metathesaurus will be confronted by users who want to add local terminology and relationships not already represented there. We urge developers to fill those needs, while, at the same time, they plan for the many consequences of unilateral Metathesaurus enhancement. Foremost among these consequences is the need to maintain local enhancements across subsequent releases of the Metathesaurus. These problems are illustrated via examples of candidate Metathesaurus enhancement terms in use at the Columbia-Presbyterian Medical Center (CPMC), at the Mayo Clinic, and in Current Disease Descriptions (CDD). Sharing and reuse of Metathesaurus enhancement methods may permit local enhancements to be used at other sites, and it may permit the global Metathesaurus utilization effort to benefit from economies of scale.