LO79: Climbing the learning curve teaching the pediatric emergency physician how to interpret point-of-care ultrasound images

Introduction: Point-of-Care Ultrasound (POCUS) is rapidly being integrated into Pediatric Emergency Medicine (PEM), and image interpretation is an important component of this skill. Currently, PEM physicians often rely on case-by-case exposure and feedback by a POCUS expert physician to learn this skill; however, this may not be efficient, reliable or feasible. Thus, there is a pressing need to develop effective POCUS image interpretation learning and assessment tools. We developed an on-line learning platform that allowed for the deliberate practice of images in four POCUS applications [soft tissue, lung, cardiac and Focused Assessment Sonography for Trauma (FAST)], and determined the quantity of participant skill acquisition by deriving performance metrics and learning curves. Methods: This was a prospective cross-sectional study administered via an on-line learning and measurement platform. Images were acquired from a pediatric emergency department and each POCUS application contained 100 still/video images. Final diagnosis of each image was determined via the consensus of three PEM POCUS experts. PEM fellow and attending study participants were recruited from the USA and Canada and were required to complete the cases of at least one application. We aimed to enroll 200 participants who had to complete a minimum of 100 cases which, based on prior work, would provide sufficient raters for item analyses and comparisons between PEM attendings and fellows. To derive reference standard performance metrics and to validate image interpretations, a unique set of five PEM POCUS experts completed each application. Results: We enrolled 225 PEM physicians, 74 fellows and 151 attendings. For all applications, the Cohens d effect size was large at 0.87, and there was no difference between PEM attendings and fellows with respect to summary performance metrics (accuracy, p= 0.29; sensitivity, p=0.13; specificity, p=0.92). Final accuracy soft tissue, lung, cardiac, and FAST for all participants was 86.4%, 89.6%, 81.6%, 88.0%, respectively, and the corresponding accuracy of PEM POCUS experts for each application was 96.0%, 96.0%, 90.0%, and 93.0%. Learning curves show maximal learning gains (inflection point) up until 65 cases for soft tissue, 70 for FAST, 75 for lung, and 85 for cardiac. Conclusion: Deliberate practice of POCUS image interpretation was effective for ensuring broad domain coverage and predictable skill improvement. Specifically, there was a large learning effect after 100 case interpretations, and 65-85 case interpretations were needed to reach an accuracy threshold of approximately 85%.