What makes organizations so similar? We contend that the engine of rationalization and bureaucratization has moved from the competitive marketplace to the state and the professions. Once a set of organizations emerges as a field, a paradox arises: rational actors make their organizations increasingly similar as they try to change them. We describe three isomorphic processes-coercive, mimetic, and normative—leading to this outcome. We then specify hypotheses about the impact of resource centralization and dependency, goal ambiguity and technical uncertainty, and professionalization and structuration on isomorphic change. Finally, we suggest implications for theories of organizations and social change.

The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields (Chinese Translation)

Builders of medical informatics applications need controlled medical vocabularies to support their applications and it is to their advantage to use available standards. In order to do so, however, these standards need to address the requirements of their intended users. Over the past decade, medical informatics researchers have begun to articulate some of these requirements. This paper brings together some of the common themes which have been described, including: vocabulary content, concept orientation, concept permanence, nonsemantic concept identifiers, polyhierarchy, formal definitions, rejection of "not elsewhere classified" terms, multiple granularities, multiple consistent views, context representation, graceful evolution, and recognized redundancy. Standards developers are beginning to recognize and address these desiderata and adapt their offerings to meet them.

/pdf/desiderata-for-controlled-medical-vocabularies-in-the-twenty-4ti3v14w3l.pdf

Desiderata for Controlled Medical Vocabularies in the Twenty-First Century

Deep learning methods exhibit promising performance for predictive modeling in healthcare, but two important challenges remain: - Data insufficiency: Often in healthcare predictive modeling, the sample size is insufficient for deep learning methods to achieve satisfactory results. Interpretation: The representations learned by deep learning methods should align with medical knowledge. To address these challenges, we propose GRaph-based Attention Model (GRAM) that supplements electronic health records (EHR) with hierarchical information inherent to medical ontologies. Based on the data volume and the ontology structure, GRAM represents a medical concept as a combination of its ancestors in the ontology via an attention mechanism. We compared predictive performance (i.e. accuracy, data needs, interpretability) of GRAM to various methods including the recurrent neural network (RNN) in two sequential diagnoses prediction tasks and one heart failure prediction task. Compared to the basic RNN, GRAM achieved 10% higher accuracy for predicting diseases rarely observed in the training data and 3% improved area under the ROC curve for predicting heart failure using an order of magnitude less training data. Additionally, unlike other methods, the medical concept representations learned by GRAM are well aligned with the medical ontology. Finally, GRAM exhibits intuitive attention behaviors by adaptively generalizing to higher level concepts when facing data insufficiency at the lower level concepts.

https://dl.acm.org/doi/pdf/10.1145/3097983.3098126

GRAM: Graph-based Attention Model for Healthcare Representation Learning

The Logical Observation Identifier Names and Codes (LOINC) database provides a universal code system for reporting laboratory and other clinical observations. Its purpose is to identify observations in electronic messages such as Health Level Seven (HL7) observation messages, so that when hospitals, health maintenance organizations, pharmaceutical manufacturers, researchers, and public health departments receive such messages from multiple sources, they can automatically file the results in the right slots of their medical records, research, and/or public health systems. For each observation, the database includes a code (of which 25 000 are laboratory test observations), a long formal name, a "short" 30-character name, and synonyms. The database comes with a mapping program called Regenstrief LOINC Mapping Assistant (RELMA(TM)) to assist the mapping of local test codes to LOINC codes and to facilitate browsing of the LOINC results. Both LOINC and RELMA are available at no cost from http://www.regenstrief.org/loinc/. The LOINC medical database carries records for >30 000 different observations. LOINC codes are being used by large reference laboratories and federal agencies, e.g., the CDC and the Department of Veterans Affairs, and are part of the Health Insurance Portability and Accountability Act (HIPAA) attachment proposal. Internationally, they have been adopted in Switzerland, Hong Kong, Australia, and Canada, and by the German national standards organization, the Deutsches Instituts fur Normung. Laboratories should include LOINC codes in their outbound HL7 messages so that clinical and research clients can easily integrate these results into their clinical and research repositories. Laboratories should also encourage instrument vendors to deliver LOINC codes in their instrument outputs and demand LOINC codes in HL7 messages they get from reference laboratories to avoid the need to lump so many referral tests under the "send out lab" code.

/pdf/loinc-a-universal-standard-for-identifying-laboratory-5am36aum4m.pdf

LOINC, a Universal Standard for Identifying Laboratory Observations: A 5-Year Update

Motivation: Despite the central role of diseases in biomedical research, there have been much fewer attempts to automatically determine which diseases are mentioned in a text—the task of disease name normalization (DNorm)—compared with other normalization tasks in biomedical text mining research. Methods: In this article we introduce the first machine learning approach for DNorm, using the NCBI disease corpus and the MEDIC vocabulary, which combines MeSH and OMIM. Our method is a high-performing and mathematically principled framework for learning similarities between mentions and concept names directly from training data. The technique is based on pairwise learning to rank, which has not previously been applied to the normalization task but has proven successful in large optimization problems for information retrieval. Results: We compare our method with several techniques based on lexical normalization and matching, MetaMap and Lucene. Our algorithm achieves 0.782 micro-averaged F-measure and 0.809 macroaveraged F-measure, an increase over the highest performing baseline method of 0.121 and 0.098, respectively. Availability: The source code for DNorm is available at http://www. ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/DNorm, along with a webbased demonstration and links to the NCBI disease corpus. Results on PubMed abstracts are available in PubTator: http://www.ncbi.nlm. nih.gov/CBBresearch/Lu/Demo/PubTator

DNorm: disease name normalization with pairwise learning to rank.

Two large health care reference terminologies, SNOMED RT and Clinical Terms Version 3 , are in the process of being merged to form a comprehensive new work referred to as SNOMED Clinical Terms. The College of American Pathologists and the United Kingdom s National Health Service have entered into a collaborative agreement to develop this new work. Both organizations have extensive terminology development and maintenance experience. This paper discusses the process and status of SNOMED CT development and how the resources and expertise of both organizations are being used to develop this new terminological resource. The preliminary results of the merger process, including mapping, the merger of upper levels of each hierarchy, and attribute harmonization are also discussed.

SNOMED clinical terms: overview of the development process and project status.

Updating the Read Codes: User-interactive Maintenance of a Dynamic Clinical Vocabulary

Symbolic anatomic knowledge representation in the Read Codes version 3: structure and application.

SNOMED RT and Clinical Terms Version 3 are two large, controlled medical terminologies that are being merged to form a new work titled SNOMED Clinical Terms (SNOMED CT). One of the first steps in this process was to create maps between semantically equivalent and proximate concepts in the two terminologies. Same-as and is-a relationships were used to map the descriptions from one terminology to concepts in the other terminology. The objectives were to identify semantically equivalent concepts in the two terminologies, to find the most semantically proximate is-a relationships for non-equivalent concepts, and to evaluate the synonymy in the source terminologies. The results suggest that the rate of semantic overlap between descriptions in SNOMED RT and CTV3 is approximately 28%. This article discusses the methodology, issues, and findings of the description mapping process.

Mapping between SNOMED RT and Clinical terms version 3: a key component of the SNOMED CT development process.

A European pre-standard and an intermediate representation facilitated exchange of two independently authored compositional knowledge bases: one formal and automatically classified, the other manually classified. The exchange highlights different strengths and weaknesses in each approach, and offers a mechanism for partial, mutual quality assurance.

http://www.opengalen.org/download/Validating-Clinical-Terminology-Structures.pdf

Colin Price

Papers

SNOMED clinical terms: overview of the development process and project status.

Updating the Read Codes: User-interactive Maintenance of a Dynamic Clinical Vocabulary

Symbolic anatomic knowledge representation in the Read Codes version 3: structure and application.

Mapping between SNOMED RT and Clinical terms version 3: a key component of the SNOMED CT development process.

Validating clinical terminology structures: integration and cross-validation of Read Thesaurus and GALEN.