MESH IO Student Youyun (Peter) Zheng wins Magna Cum Laude Abstract Award at the American Society of Emergency Radiology
Predicting imaging utilization based on initial triage information would enable great strides in dynamic workflow management. A new abstract presented from the MESH Innovation & Operations Research Center (MESH IO) aims to leverage natural language processing to do just that. The abstract, entitled “Next-level Imaging Triage: Predicting Imaging Utilization based on Emergency Department Initial Triage Notes using Machine Learning and Natural Language Processing (Abstract CID: 3598332) was presented by first author, Harvard Medical student Youyun (Peter) Zheng, Sept 21 at the annual ASER meeting in Tampa Bay, FL.
Zheng analyzed thousands of CT and radiograph data, the dominant imaging modalities, from a quaternary care academic ED over a period of 4 months. CT and radiograph usage were summed within 8 hours of ED arrival. Corresponding nursing triage notes were then collected for each ED visit and a bigram bag of word model was applied on tokenized notes to construct features upon which we predicted whether or not patients received imaging workup. Of 43390 Emergency Department patient visits during the study period (M 51.9%, F 48.1%, age 47.0 ± 22.5), 13705 CTs and 19620 X-rays were performed, with 23.6% and 33.7% of patients having at least one scan of each of the modalities respectively. For CT utilization prediction, our model achieved 77% accuracy, 52% precision, 11% recall and 18% F1 score. For X-ray utilization prediction, our model achieved 70% accuracy, 61% precision, 28% recall and 38% F1 score.
Senior study author Marc Succi, MD, notes “Leveraging NLP to predict, before a patient walks into the hospital, the certainty with which they will receive imaging holds great promise to optimize pre-hospital and early triage workflows, especially for emergency departments”. Study coauthors include Youyun (Peter) Zheng, Will Ge, David Whitehead MD MBA, Ben White, MD, Tarik Alkasab, MD, Sanjay Saini MD, Michael Lev, MD, Matthew D. Li, MD, Marc Succi, MD.
—————————–
About MESH IO
The Medically Engineered Solutions in Healthcare Innovation in Operations Research Center (MESH IO) is a unique inter-disciplinary research center across Mass General Brigham, Harvard Medical and Harvard Business School comprised of specialty leaders, including physicians and executives, in innovation and operations in healthcare. This creative marriage of both innovation and operations researchers is dedicated to investigating key aspects of business in healthcare through original research and thought-leading papers for the medical community. The group is comprised of faculty and executives from different specialties, including medicine, radiology, emergency medicine, surgery and more.
For more information, contact: meshincubator@partners.org