Researchers, scientists, and clinicians in the Department of Pediatrics have been working on a revolutionary new technology for years, and earlier this month, their work culminated in a publication in the Journal of the American Medical Association. Titled “Development and Validation of an Automated Classifier to Diagnose Acute Otitis Media in Children,” this work represents a significant step forward in the accurate testing and diagnosing of acute otitis media (AOM) in pediatric patients with the use of an AI-powered application. 
The result of a cross-disciplinary collaboration among a group of experts from the Division of General Academic Pediatrics (GAP) in our Department of Pediatrics, the Tandon School of Engineering at New York University, the Bosch Center for Artificial Intelligence, and Dcipher Analytics in Stockholm, Sweden, this new tool will increase the accuracy in diagnoses of AOM, thereby reducing the unnecessary overmedication of pediatric patients, while significantly reducing healthcare costs per year.
Lead researchers on this publication from the Department of Pediatrics include Nader Shaikh, MD, Shannon J. Conway, BSc, Timothy R. Shope, MD, Mary Ann Haralam, CRNP, Catherine Campese, CRNP, Matthew C. Lee, BA, all of GAP, as well as Director of the Division, Alejandro Hoberman, MD.
The key question of this work focused on investigating whether an artificial intelligence decision support tool can be used in a primary care setting to enhance accuracy in the diagnosis of AOM in children.
While AOM is frequently diagnosed in children, the accuracy of this diagnosis has been consistently low over the last several decades. Despite advances in diagnostic tools and clinician training and skill, accuracy rates have improved only marginally.
This study took place at two outpatient clinics in Pennsylvania between 2018 and 2023, and involved an analysis of otoscopic videos of the tympanic membrane captured with a smartphone. Over 600 children were included as eligible participants who presented for sick or wellness visits.
After training the AI tools to predict features of the tympanic membrane and the diagnosis of AOM and no AOM, two different decision-making networks were developed in order to test accuracies across different approaches.
Ultimately, after reviewing 1,151 videos across the participating children, both decision-making networks had nearly identical accuracy in the mid 90th percentiles. Given this incredibly high accuracy, this medical-grade AI application could reasonably be used in a primary care or acute care settings to assist with the diagnosis of AOM in children and its related treatment.
This groundbreaking step forward in clinical diagnostic care was also featured in Science Daily and other national publications.
Join us in congratulating all of the researchers and clinicians involved in this exciting development in the otologic care of pediatric patients.
Follow Pitt Pediatrics on Twitter and Instagram for more updates from our GAP Division and other groundbreaking publications.