Preliminary feasibility assessment of CDM-based active surveillance using current status of medical device data in medical records and OMOP-CDM

In recent years, there has been an emerging interest in the use of claims and electronic health record (EHR) data for evaluation of medical device safety and effectiveness. In Korea, national insurance electronic data interchange (EDI) code has been used as a medical device data source for common data model (CDM). This study performed a preliminary feasibility assessment of CDM-based vigilance. A cross-sectional study of target medical device data in EHR and CDM was conducted. A total of 155 medical devices were finally enrolled, with 58.7% of them having EDI codes. Femoral head prosthesis was selected as a focus group. It was registered in our institute with 11 EDI codes. However, only three EDI codes were converted to systematized nomenclature of medicine clinical terms concept. EDI code was matched in one-to-many (up to 104) with unique device identifier (UDI), including devices classified as different global medical device nomenclature. The use of UDI rather than EDI code as a medical device data source is recommended. We hope that this study will share the current state of medical device data recorded in the EHR and contribute to the introduction of CDM-based medical device vigilance by selecting appropriate medical device data sources.

Introduction of Systematized Nomenclature of Medicine–Clinical Terms Coding Into an Electronic Health Record and Evaluation of its Impact: Qualitative and Quantitative Study

Background This study describes the conversion within an existing electronic health record (EHR) from the International Classification of Diseases, Tenth Revision coding system to the SNOMED-CT (Systematized Nomenclature of Medicine–Clinical Terms) for the collection of patient histories and diagnoses. The setting is a large acute hospital that is designing and building its own EHR. Well-designed EHRs create opportunities for continuous data collection, which can be used in clinical decision support rules to drive patient safety. Collected data can be exchanged across health care systems to support patients in all health care settings. Data can be used for research to prevent diseases and protect future populations. Objective The aim of this study was to migrate a current EHR, with all relevant patient data, to the SNOMED-CT coding system to optimize clinical use and clinical decision support, facilitate data sharing across organizational boundaries for national programs, and enable remodeling of medical pathways. Methods The study used qualitative and quantitative data to understand the successes and gaps in the project, clinician attitudes toward the new tool, and the future use of the tool. Results The new coding system (tool) was well received and immediately widely used in all specialties. This resulted in increased, accurate, and clinically relevant data collection. Clinicians appreciated the increased depth and detail of the new coding, welcomed the potential for both data sharing and research, and provided extensive feedback for further development. Conclusions Successful implementation of the new system aligned the University Hospitals Birmingham NHS Foundation Trust with national strategy and can be used as a blueprint for similar projects in other health care settings.

Mapping the Korean National Health Checkup Questionnaire to Standard Terminologies

Objectives: An increasing emphasis has been placed on the integration of clinical data and patient-generated health data (PGHD), which are generated outside of hospitals. This study explored the possibility of using standard terminologies to represent PGHD for data integration.Methods: We chose the 2020 general health checkup questionnaire of the Korean Health Screening Program as a resource. We divided every component of the questionnaire into entities and values, which were mapped to standard terminologies—Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) version 2020-07-31 and Logical Observation Identifiers Names and Codes (LOINC) version 2.68.Results: Eighty-nine items were derived from the 17 questions of the 2020 health examination questionnaire, of which 76 (85.4%) were mapped to standard terms. Fifty-two items were mapped to SNOMED CT and 24 items were mapped to LOINC. Among the items mapped to SNOMED CT, 35 were mapped to pre-coordinated expressions and 17 to post-coordinated expressions. Forty items had one-to-one relationships, and 17 items had one-to-many relationships.Conclusions: We achieved a high mapping rate (85.4%) by using both SNOMED CT and LOINC. However, we noticed some issues while mapping the Korean general health checkup questionnaire (i.e., lack of explanations, vague questions, and overly narrow concepts). In particular, items combining two or more concepts into a single item were not appropriate for mapping using standard terminologies. Although it is not the case that all items need to be expressed in standard terminology, essential items should be presented in a way suitable for mapping to standard terminology by revising the questionnaire in the future.

Expressiveness of an International Semantic Standard for Wound Care: Mapping a Standardized Item Set for Leg Ulcers to the Systematized Nomenclature of Medicine–Clinical Terms

Background Chronic health conditions are on the rise and are putting high economic pressure on health systems, as they require well-coordinated prevention and treatment. Among chronic conditions, chronic wounds such as cardiovascular leg ulcers have a high prevalence. Their treatment is highly interdisciplinary and regularly spans multiple care settings and organizations; this places particularly high demands on interoperable information exchange that can be achieved using international semantic standards, such as Systematized Nomenclature of Medicine–Clinical Terms (SNOMED CT). Objective This study aims to investigate the expressiveness of SNOMED CT in the domain of wound care, and thereby its clinical usefulness and the potential need for extensions. Methods A clinically consented and profession-independent wound care item set, the German National Consensus for the Documentation of Leg Wounds (NKDUC), was mapped onto the precoordinated concepts of the international reference terminology SNOMED CT. Before the mapping took place, the NKDUC was transformed into an information model that served to systematically identify relevant items. The mapping process was carried out in accordance with the ISO/TR 12300 formalism. As a result, the reliability, equivalence, and coverage rate were determined for all NKDUC items and sections. Results The developed information model revealed 268 items to be mapped. Conducted by 3 health care professionals, the mapping resulted in moderate reliability (κ=0.512). Regarding the two best equivalence categories (symmetrical equivalence of meaning), the coverage rate of SNOMED CT was 67.2% (180/268) overall and 64.3% (108/168) specifically for wounds. The sections general medical condition (55/66, 83%), wound assessment (18/24, 75%), and wound status (37/57, 65%), showed higher coverage rates compared with the sections therapy (45/73, 62%), wound diagnostics (8/14, 57%), and patient demographics (17/34, 50%). Conclusions The results yielded acceptable reliability values for the mapping procedure. The overall coverage rate shows that two-thirds of the items could be mapped symmetrically, which is a substantial portion of the source item set. Some wound care sections, such as general medical conditions and wound assessment, were covered better than other sections (wound status, diagnostics, and therapy). These deficiencies can be mitigated either by postcoordination or by the inclusion of new concepts in SNOMED CT. This study contributes to pushing interoperability in the domain of wound care, thereby responding to the high demand for information exchange in this field. Overall, this study adds another puzzle piece to the general knowledge about SNOMED CT in terms of its clinical usefulness and its need for further extensions.

A Fine-Tuned Bidirectional Encoder Representations From Transformers Model for Food Named-Entity Recognition: Algorithm Development and Validation

Background Recently, food science has been garnering a lot of attention. There are many open research questions on food interactions, as one of the main environmental factors, with other health-related entities such as diseases, treatments, and drugs. In the last 2 decades, a large amount of work has been done in natural language processing and machine learning to enable biomedical information extraction. However, machine learning in food science domains remains inadequately resourced, which brings to attention the problem of developing methods for food information extraction. There are only few food semantic resources and few rule-based methods for food information extraction, which often depend on some external resources. However, an annotated corpus with food entities along with their normalization was published in 2019 by using several food semantic resources. Objective In this study, we investigated how the recently published bidirectional encoder representations from transformers (BERT) model, which provides state-of-the-art results in information extraction, can be fine-tuned for food information extraction. Methods We introduce FoodNER, which is a collection of corpus-based food named-entity recognition methods. It consists of 15 different models obtained by fine-tuning 3 pretrained BERT models on 5 groups of semantic resources: food versus nonfood entity, 2 subsets of Hansard food semantic tags, FoodOn semantic tags, and Systematized Nomenclature of Medicine Clinical Terms food semantic tags. Results All BERT models provided very promising results with 93.30% to 94.31% macro F1 scores in the task of distinguishing food versus nonfood entity, which represents the new state-of-the-art technology in food information extraction. Considering the tasks where semantic tags are predicted, all BERT models obtained very promising results once again, with their macro F1 scores ranging from 73.39% to 78.96%. Conclusions FoodNER can be used to extract and annotate food entities in 5 different tasks: food versus nonfood entities and distinguishing food entities on the level of food groups by using the closest Hansard semantic tags, the parent Hansard semantic tags, the FoodOn semantic tags, or the Systematized Nomenclature of Medicine Clinical Terms semantic tags.

Introducing SNOMED-CT* Coding into an Electronic Health Record: Impact on Clinicians, Data Sharing and Research potential *Systemised Nomenclature of Medicine Clinical Terminology (Preprint)

BACKGROUND This study describes the conversion within an existing Electronic Health Record (EHR) from the coding system International Classification of Diseases version 10 (ICD-10) to the Systematized Nomenclature Of MEDicine - Clinical Terms (SNOMED-CT), for collection of patients’ history and diagnoses. The setting is a large acute hospital, designing and building its own EHR. Well-designed EHRs create opportunities for continuous data collection which can be utilised in Clinical Decision Support rules to drive patient safety. Collected data can be exchanged across healthcare systems to support patients in all healthcare settings. Data can be used for research to prevent disease and protect future populations. OBJECTIVE To migrate a current electronic health record, with all relevant patient data, to the coding system, Systematized Nomenclature of Medicine - Clinical Terms, to optimise clinical utilisation and clinical decision support, and facilitate data sharing across organisational boundaries for national programmes, and remodelling of medical pathways. METHODS The study used qualitative and quantitative data to understand the successes and gaps in the project, clinician attitudes to the new tool, and future use. RESULTS The new coding system (“tool”) was well received and immediately widely used in all specialities. It resulted in increased, accurate and clinically relevant data collection. Clinicians appreciated the increased depth and detail of the new coding, welcomed the potential for both data sharing and research, and gave extensive feedback for further development. CONCLUSIONS Successful implementation aligned the Trust with national strategy and can be used as a Blueprint for similar projects in other healthcare settings. CLINICALTRIAL NA

A Fine-Tuned Bidirectional Encoder Representations From Transformers Model for Food Named-Entity Recognition: Algorithm Development and Validation (Preprint)

BACKGROUND Recently, food science has been garnering a lot of attention. There are many open research questions on food interactions, as one of the main environmental factors, with other health-related entities such as diseases, treatments, and drugs. In the last 2 decades, a large amount of work has been done in natural language processing and machine learning to enable biomedical information extraction. However, machine learning in food science domains remains inadequately resourced, which brings to attention the problem of developing methods for food information extraction. There are only few food semantic resources and few rule-based methods for food information extraction, which often depend on some external resources. However, an annotated corpus with food entities along with their normalization was published in 2019 by using several food semantic resources. OBJECTIVE In this study, we investigated how the recently published bidirectional encoder representations from transformers (BERT) model, which provides state-of-the-art results in information extraction, can be fine-tuned for food information extraction. METHODS We introduce FoodNER, which is a collection of corpus-based food named-entity recognition methods. It consists of 15 different models obtained by fine-tuning 3 pretrained BERT models on 5 groups of semantic resources: food versus nonfood entity, 2 subsets of Hansard food semantic tags, FoodOn semantic tags, and Systematized Nomenclature of Medicine Clinical Terms food semantic tags. RESULTS All BERT models provided very promising results with 93.30% to 94.31% macro F1 scores in the task of distinguishing food versus nonfood entity, which represents the new state-of-the-art technology in food information extraction. Considering the tasks where semantic tags are predicted, all BERT models obtained very promising results once again, with their macro F1 scores ranging from 73.39% to 78.96%. CONCLUSIONS FoodNER can be used to extract and annotate food entities in 5 different tasks: food versus nonfood entities and distinguishing food entities on the level of food groups by using the closest Hansard semantic tags, the parent Hansard semantic tags, the FoodOn semantic tags, or the Systematized Nomenclature of Medicine Clinical Terms semantic tags.

DeepSuggest: Using Neural Networks to Suggest Related Keywords for a Comprehensive Search of Clinical Notes

Objective A large amount of clinical data are stored in clinical notes that frequently contain spelling variations, typos, local practice-generated acronyms, synonyms, and informal words. Instead of relying on established but infrequently updated ontologies with keywords limited to formal language, we developed an artificial intelligence (AI) assistant (named “DeepSuggest”) that interactively offers suggestions to expand or pivot queries to help overcome these challenges. Methods We applied an unsupervised neural network (Word2Vec) to the clinical notes to build keyword contextual similarity matrix. With a user's input query, DeepSuggest generates a list of relevant keywords, including word variations (e.g., formal or informal forms, synonyms, abbreviations, and misspellings) and other relevant words (e.g., related diagnosis, medications, and procedures). Human intelligence is then used to further refine or pivot their query. Results DeepSuggest learns the semantic and linguistic relationships between the words from a large collection of local notes. Although DeepSuggest is only able to recall 0.54 of Systematized Nomenclature of Medicine (SNOMED) synonyms on average among the top 60 suggested terms, it covers the semantic relationship in our corpus for a larger number of raw concepts (6.3 million) than SNOMED ontology (24,921) and is able to retrieve terms that are not stored in existing ontologies. The precision for the top 60 suggested words averages at 0.72. Usability test resulted that DeepSuggest is able to achieve almost twice the recall on clinical notes compared with Epic (average of 5.6 notes retrieved by DeepSuggest compared with 2.6 by Epic). Conclusion DeepSuggest showed the ability to improve retrieval of relevant clinical notes when implemented on a local corpus by suggesting spelling variations, acronyms, and semantically related words. It is a promising tool in helping users to achieve a higher recall rate for clinical note searches and thus boosting productivity in clinical practice and research. DeepSuggest can supplement established ontologies for query expansion.

Pancreatic manifestations in rheumatoid arthritis: a national population-based study

Objectives RA is a systemic autoimmune disease characterized by persistent joint inflammation. Extra-articular manifestations of RA can involve different organs including the gastrointestinal (GI) system. Using a large database, we sought to describe the epidemiology of pancreas involvement in RA. Methods We queried a multicentre database (Explorys Inc, Cleveland, OH, USA), an aggregate of electronic health record data from 26 major integrated US healthcare systems in the US from 1999 to 2019. After excluding patients younger than 18, a cohort of individuals with Systematized Nomenclature of Medicine – Clinical Terms (SNOMED–CT) diagnosis of RA was identified. Within this cohort, patients who developed a SNOMED-CT diagnosis of acute pancreatitis (AP), chronic pancreatitis (CP) and primary pancreatic cancer (PaCa) after at least 30 days of RA diagnosis were identified. Statistical analysis for multivariate model was performed using Statistical Package for Social Sciences (SPSS version 25, IBM Corp) to adjust for several factors. Results Of the 56 183 720 individuals in the database, 518 280 patients had a diagnosis of RA (0.92%). Using a multivariate regression model, patients with RA were more likely to develop AP [odds ratio (OR): 2.51; 95% CI: 2.41, 2.60], CP (OR: 2.97; 95% CI: 2.70, 3.26) and PaC (OR: 1.79; 95% CI: 1.52, 2.10). Conclusion In this large database, we found a modest increased risk of AP and CP among patients with RA after adjusting for the common causes of pancreatitis. Further studies are required to better understand this association and the effect of medications used for RA.

Abstract 15300: Epidemiology and Risk of Anxiety and Depression Among Patients With Lown-Ganong-Levine Syndrome

Introduction: Lown-Ganong-Levine (LGL) syndrome is a rare congenital pre-excitation syndrome with a perinodal accessory pathway. Arrhythmic events may lead to, or be influenced by, a myriad of psychological conditions such as anxiety and depressive disorders. There is no epidemiological data on the co-occurrence of these conditions in patients with LGL syndrome. We sought to describe the epidemiology and risk association of anxiety and depression in LGL syndrome. Methods: A multi-institutional database (Explorys Inc, Cleveland, OH, USA), with an aggregate of electronic health record data from 26 major integrated US healthcare systems, was surveyed. A cohort of patients with a Systematized Nomenclature of Medicine—Clinical Terms (SNOMED-CT) diagnosis of “Lown-Ganong-Levine syndrome” between 1999 and 2020 was identified. Demographic characteristics along with relevant clinical and social information were obtained. Results: Of 73,044,190 individuals in the database (1999-2020), 1,180 were given the diagnosis of LGL. Compared to subjects without LGL, patients with LGL syndrome were more likely to have anxiety (OR=5.9, p<0.001) and depression (OR=5.2, p<0.001). Within the anxiety and depression cohorts, LGL patients were more likely to be females, and had a higher association with smoking, alcohol and substance abuse. Anxiety but not depression was more common in young adults (18-65 years old). No difference in supraventricular arrhythmias was found between LGL patients carrying these psychiatric diagnoses with those who don’t. Conclusion: This is the largest study to date demonstrating the significant association between LGL and common psychiatric conditions. Care should be taken in identifying and managing such conditions as they, along with their pharmacologic therapies, may mimic or exacerbate the symptoms of LGL syndrome.