Classification of System and Software Related FDA Medical Device Recalls with a Three-Level Taxonomy Approach: Defibrillator Case

In the healthcare domain, where safety is paramount, medical device recalls are highly critical events that can pose significant risks to health. The U.S. Food and Drug Administration (FDA) maintains a comprehensive database of recall data, including fields such as product description, product type, recall reason, and termination status. However, the recall reason field lacks a predefined taxonomy for root cause classification, which limits the effectiveness of analyzing and preventing recurring issues. In this study, we aim to classify recalls based on their reasons by proposing a three-level taxonomy to improve the classification process. The first level classification approaches to the problem from a broad perspective, including labels such as operational, environmental, electrical, hardware, software, and mechanical. The second level utilizes the FDA root cause options to provide a more detailed classification. The third level uses the SW91 Classification of Defects in Health Software taxonomy specifically for software, which offers a standardized framework for classifying defects in health software. This taxonomy approach would not only aid in more precise recall classification but also lays the groundwork for subsequent research focused on developing predictive models to prevent recurring defects in healthcare software. By improving classification accuracy, this study aims to increase ways for product safety and enhance regulatory oversight in the medical device domain. In this study, we applied this approach to 271 different defibrillator recall cases reported by the FDA and categorized them based on their recall reasons. This enabled us to better understand recurring issues and allow for a more precise classification of recall reasons. This approach not only aims to analyze and address existing problems more effectively but also seeks to provide a foundation for future research focused on predicting and preventing potential defects.