Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Innovation is key for advancing the science of biomedical and health informatics and for publishing in JAMIA

Innovation is key for advancing the science of biomedical and health informatics and for... Innovation is key for scientific advancement. One of the most common reasons that well-written manuscripts are rejected from Journal of the American Medical Informatics Association (JAMIA) is lack of innovation from the perspective of biomedical and health informatics. Oxford defines innovation (in something) as “a new idea, way of doing something, etc. that has been introduced or discovered.”1 Innovation in biomedical and health informatics can take multiple forms (eg, conceptual, topical, methodological, or application domains) and is relevant across manuscript types. In this editorial, I highlight 5 articles that illustrate different aspects of innovation. A research study by Kuo et al2 reflects innovation in its development of a framework that combines level-wise model learning, blockchain-based model dissemination, and a hierarchical consensus algorithm to construct generalizable predictive models using cross-institutional approaches. The framework is designed to take advantage of the privacy-preserving characteristics of blockchain ledger technology while considering the topology of large-scale research enterprises, which the authors characterize as a network of networks. As compared to centralized server privacy-preserving approaches, the peer-to-peer blockchain approach has advantages related to provenance as well as immutability and transparency of the models. They created an implementation of the framework called HierarchicalChain (Hierarchical privacy-preserving modeling on blockChain) and evaluated it using 3 healthcare and genomic datasets comparing HierarchicalChain’s predictive correctness, learning iteration, and execution time with a state-of-the-art method designed for flattened network topology. The authors found that HierarchicalChain improves the predictive correctness for small training datasets and provides comparable correctness results with the competing method with higher learning iteration and similar per-iteration execution time. Two articles in this issue reflect innovation in the application domain of critical care.3,4 Catling and Wolff3 applied a combination of temporal convolutional network for longitudinal data and a feedforward neural network for prediction of subsequent events following critical care admission. As compared with many studies that focus on mortality, length of stay, or a single event, their analysis focused on prediction of important clinician interventions such as up- and down-titrating FiO2, extubation, intubation, starting noradrenaline, and fluid challenge. Their multitask learning approach for longitudinal data generates a single representation of each patient-hour, which is used to predict the different clinical interventions. Such predictions provide the foundation for anticipating interventions and intervening in a timely manner. The temporal convolutional network for longitudinal data and feedforward neural network model performed similarly to the recurrent neural network for some events and outperformed it for others. In a Brief Communication, Ibrahim et al4 demonstrate the value of background knowledge about the type of organ dysfunction observed in patients who develop sepsis in the critical care setting to distinguish patients who develop sepsis in that setting from those who do not. Classification experiments using random forest, gradient boost trees, and support vector machines showed that features selected using sepsis subpopulations as background knowledge yielded a superior performance regardless of the classification model used. Their findings suggest the need for more personalized models for complex conditions and the value of background domain knowledge in creating such models. Deep learning approaches are increasingly the most common machine learning method in JAMIA submissions. In a Review, Wu et al5 report the abstraction of 25 variables across 212 papers on deep learning for natural language processing in the clinical domain to summarize the methods, scope, and context of current research. They found that the number of publications on deep learning in clinical natural language processing has at least doubled on an annual basis over the last 5 years. In addition, the majority of the literature reviewed used existing deep learning models on well-known extraction tasks in English clinical notes. The analysis facilitated novel insights: for example, deep learning methods are increasingly compared with each other rather than only with a traditional machine learning method, suggesting the acceptance of deep learning as a baseline model. It is also of note, given the increased focus on transparency, reproducibility, and replicability in science, that the authors found that although few articles provided information to access the software used in their analyses, a majority provided sufficient methodological detail including hyperparameters so that the methods could potentially be replicated. A Perspective by Reddy et al6 addresses the hot topic of artificial intelligence (AI) in health care. The innovative aspects of this article are conceptual and include the proposed Governance Model for AI in Health Care to address the ethical and regulatory issues that arise from application of AI in health care and an explication of the integration of the model components into clinical workflow. The model includes 4 main components: fairness, transparency, trustworthiness, and accountability. The authors make specific recommendations for each component and also delineate their integration with clinical workflow. Consistent with the opinion nature of a JAMIA Perspective, the authors hope that the article will motivate further debate and actions on this important topic. If you are an author submitting to JAMIA, I hope that these examples will help you to delineate the innovation in your work. If you are a JAMIA reader, I hope that JAMIA will continue to be your go-to resource for innovation in biomedical and health informatics as we strive toward the goals of health and health equity. CONFLICT OF INTEREST STATEMENT None declared. REFERENCES 1 Oxford Learner’s Dictionary. https://www.oxfordlearnersdictionaries.com/us/definition/english/innovation? q=innovation Accessed January 2, 2020 2 Kuo T-T , Kim J , Gabriel RA. Privacy-preserving model learning on blockchain network-of-networks . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 343 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Wu S , Roberts K , Datta S , et al. . H. Deep learning in clinical natural language processing: a methodical review . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 457 – 70 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Catling FJR , Wolff AH. Temporal convolutional networks allow early prediction of events in critical care . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 355 – 65 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Ibrahim Z , Wu H , Hamoud A , Stappen L , Dobson R. On classifying sepsis heterogeneity in the ICU: insight using machine learning . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 437 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Reddy S , Allan S , Coghlan S , Cooper P. A governance model for the application of AI in health care . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 491 – 97 . Google Scholar PubMed WorldCat © The Author(s) 2020. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the American Medical Informatics Association Oxford University Press

Innovation is key for advancing the science of biomedical and health informatics and for publishing in JAMIA

Download PDF
2 pages

Loading next page...
 
/lp/oxford-university-press/innovation-is-key-for-advancing-the-science-of-biomedical-and-health-d5PirzZNfe

References (6)

Publisher
Oxford University Press
Copyright
© The Author(s) 2020. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
1067-5027
eISSN
1527-974X
DOI
10.1093/jamia/ocaa002
Publisher site
See Article on Publisher Site

Abstract

Innovation is key for scientific advancement. One of the most common reasons that well-written manuscripts are rejected from Journal of the American Medical Informatics Association (JAMIA) is lack of innovation from the perspective of biomedical and health informatics. Oxford defines innovation (in something) as “a new idea, way of doing something, etc. that has been introduced or discovered.”1 Innovation in biomedical and health informatics can take multiple forms (eg, conceptual, topical, methodological, or application domains) and is relevant across manuscript types. In this editorial, I highlight 5 articles that illustrate different aspects of innovation. A research study by Kuo et al2 reflects innovation in its development of a framework that combines level-wise model learning, blockchain-based model dissemination, and a hierarchical consensus algorithm to construct generalizable predictive models using cross-institutional approaches. The framework is designed to take advantage of the privacy-preserving characteristics of blockchain ledger technology while considering the topology of large-scale research enterprises, which the authors characterize as a network of networks. As compared to centralized server privacy-preserving approaches, the peer-to-peer blockchain approach has advantages related to provenance as well as immutability and transparency of the models. They created an implementation of the framework called HierarchicalChain (Hierarchical privacy-preserving modeling on blockChain) and evaluated it using 3 healthcare and genomic datasets comparing HierarchicalChain’s predictive correctness, learning iteration, and execution time with a state-of-the-art method designed for flattened network topology. The authors found that HierarchicalChain improves the predictive correctness for small training datasets and provides comparable correctness results with the competing method with higher learning iteration and similar per-iteration execution time. Two articles in this issue reflect innovation in the application domain of critical care.3,4 Catling and Wolff3 applied a combination of temporal convolutional network for longitudinal data and a feedforward neural network for prediction of subsequent events following critical care admission. As compared with many studies that focus on mortality, length of stay, or a single event, their analysis focused on prediction of important clinician interventions such as up- and down-titrating FiO2, extubation, intubation, starting noradrenaline, and fluid challenge. Their multitask learning approach for longitudinal data generates a single representation of each patient-hour, which is used to predict the different clinical interventions. Such predictions provide the foundation for anticipating interventions and intervening in a timely manner. The temporal convolutional network for longitudinal data and feedforward neural network model performed similarly to the recurrent neural network for some events and outperformed it for others. In a Brief Communication, Ibrahim et al4 demonstrate the value of background knowledge about the type of organ dysfunction observed in patients who develop sepsis in the critical care setting to distinguish patients who develop sepsis in that setting from those who do not. Classification experiments using random forest, gradient boost trees, and support vector machines showed that features selected using sepsis subpopulations as background knowledge yielded a superior performance regardless of the classification model used. Their findings suggest the need for more personalized models for complex conditions and the value of background domain knowledge in creating such models. Deep learning approaches are increasingly the most common machine learning method in JAMIA submissions. In a Review, Wu et al5 report the abstraction of 25 variables across 212 papers on deep learning for natural language processing in the clinical domain to summarize the methods, scope, and context of current research. They found that the number of publications on deep learning in clinical natural language processing has at least doubled on an annual basis over the last 5 years. In addition, the majority of the literature reviewed used existing deep learning models on well-known extraction tasks in English clinical notes. The analysis facilitated novel insights: for example, deep learning methods are increasingly compared with each other rather than only with a traditional machine learning method, suggesting the acceptance of deep learning as a baseline model. It is also of note, given the increased focus on transparency, reproducibility, and replicability in science, that the authors found that although few articles provided information to access the software used in their analyses, a majority provided sufficient methodological detail including hyperparameters so that the methods could potentially be replicated. A Perspective by Reddy et al6 addresses the hot topic of artificial intelligence (AI) in health care. The innovative aspects of this article are conceptual and include the proposed Governance Model for AI in Health Care to address the ethical and regulatory issues that arise from application of AI in health care and an explication of the integration of the model components into clinical workflow. The model includes 4 main components: fairness, transparency, trustworthiness, and accountability. The authors make specific recommendations for each component and also delineate their integration with clinical workflow. Consistent with the opinion nature of a JAMIA Perspective, the authors hope that the article will motivate further debate and actions on this important topic. If you are an author submitting to JAMIA, I hope that these examples will help you to delineate the innovation in your work. If you are a JAMIA reader, I hope that JAMIA will continue to be your go-to resource for innovation in biomedical and health informatics as we strive toward the goals of health and health equity. CONFLICT OF INTEREST STATEMENT None declared. REFERENCES 1 Oxford Learner’s Dictionary. https://www.oxfordlearnersdictionaries.com/us/definition/english/innovation? q=innovation Accessed January 2, 2020 2 Kuo T-T , Kim J , Gabriel RA. Privacy-preserving model learning on blockchain network-of-networks . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 343 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Wu S , Roberts K , Datta S , et al. . H. Deep learning in clinical natural language processing: a methodical review . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 457 – 70 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Catling FJR , Wolff AH. Temporal convolutional networks allow early prediction of events in critical care . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 355 – 65 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Ibrahim Z , Wu H , Hamoud A , Stappen L , Dobson R. On classifying sepsis heterogeneity in the ICU: insight using machine learning . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 437 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Reddy S , Allan S , Coghlan S , Cooper P. A governance model for the application of AI in health care . J Am Med Inform Assoc 2020 ; 27 ( 3 ): 491 – 97 . Google Scholar PubMed WorldCat © The Author(s) 2020. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Journal of the American Medical Informatics Association – Oxford University Press

Published: Mar 1, 2020

There are no references for this article.