A machine learning-based medical device recall initiator prediction framework: From supply chain risk management and resilience view

Abstract

Persistent quality problems with medical devices and the associated recall present potential health risks to users, bringing extra costs and disturbances to the supply chain. Classical medical device recall strategy neglects the significance of the failure detection process in the premarket phase, increasing the medical device recall risks. This research first established the theoretical foundation for the medical device recall reasons detection problem by reconstructing the medical device recall strategy from the supply chain risk and resilience view and reinforced the importance of failure detection and quality inspection work in the premarket stage. Moreover, existing medical device failure reason prediction research was limited in practicality and scalability. To address this problem, we developed a machine learning-based medical device recall initiator prediction system framework to conduct proactive failure detection based on the industrial case. By redesigning in dataset, clustering method and input feature selection, an accuracy rate of 88.85% is achieved, which indicates the potential of the proposed framework in assisting manufacturers with asset predictive failure detection for reducing recall. A comparative analysis of prediction performance between our framework and the most similar research that utilized the same prediction algorithms was presented. The comparison results showed that our distinctive design in the dataset, clustering method, and key input features chosen are valid and efficient. Before redesigning the prediction algorithms that require higher technical investment, our elaborate research design in selecting the dataset, cluster method, and key input features can be the antecedents of better prediction performance for manufacturers. The proposed predictive framework obtains higher accuracy, scalability, practicality, with accessibility.