UAB Informatics Institute Director and CCTS Co-Director James J. Cimino, MD, summarized the key developments in biomedical informatics at the 2018 American Medical Informatics Association (AMIA) annual meeting in San Francisco. The session, “Year in Review: Putting the Working Groups to Work,” drew an overflow crowd of 500 attendees (roughly a quarter of all attendees). He reprised his talk for the Hub as part of the Institute’s Clinical Informatics PowerTalk Seminar Series on Friday, Nov. 9.
After a review using criteria that included justification, visuals, and “personal opinion of what is significant and cool,” Cimino zeroed in on 67 out of the 106 biomedical informatics papers submitted by AMIA’s Working Groups. He categorized them into knowledge discovery, bioimaging, visual analytics, nursing informatics, dental informatics, consumer informatics, pharmacoinformatics, clinical information systems, clinical decision support, natural language processing (NLP), knowledge representation, mental health, public health, global health, evaluation, intensive care, primary care, and education, including high school and college student pathways into informatics. He also offered a few mini-tutorials along the way, on topics such as neural networks, NLP, and systematic reviews, to make sure the audience would appreciate papers of interest. To keep the project manageable, he excluded genomics and translational bioinformatics and clinical research informatics, since those topics are covered by AMIA's Spring Biomedical Informatics Research Summit.
Cimino presented summaries of the papers with “the greatest impact, broadest interest, and/or highest entertainment value” in a 90-minute multimedia event, with infographics, videos, and even music. Among the clinical informatics highlights were tools and/or methodologies to assess opioid prescribing practices, pinpoint important patterns in patient-reported data, and capture hospital progress notes using mobile voice recognition software integrated with a commercial EHR system.
One paper described an exciting new tool, the “Glucolyzer,” that helps clinicians identify clinically meaningful patterns in patient-generated glucose measurements for those with Type 2 diabetes. Such innovations seek to capitalize on data generated by increasingly popular wearable health monitors, such as Fitbits and the Apple Watch, making them more useful for research and clinical practice.
Neural network papers offered insights on the use of distributed deep learning networks for medical imaging, with some algorithms outperforming human detection of lymph node metastases in women with breast cancer and classification of echocardiograms. Other studies addressed heritability of traits based on family trees drawn from emergency contact information in EHRs and the application of adversarial attacks used in machine learning to EHRs to better understand which parts of EHRs are most sensitive to errors and improve predictive models.
Cimino stressed the intradisciplinary nature of many informatics discoveries, which are no longer restricted to the realm of terminology/ontology. “Deep learning,” he noted, “especially artificial neural networks, is a leading enabler of research.” With one to six papers to synthesize per topic, and only about a minute to convey each paper’s discovery and what made it potentially impactful, he encouraged the audience to consider his presentation “a sampler.”
“I invite you to delve deeper into whichever topical areas interest you most. As a whole, this body of work represents a treasure trove for hypothesis generation, systematic reviews, and training the future informatics workforce,” he said.
A pdf version of Cimino’s AMIA presentation is available here. A bibliography is included. For more about CCTS Informatics at the AMIA Annual Meeting, see “Informatics faculty, staff, and students present at national informatics conference.”