Predictive Model for Determining the Probability of Producing Non-Conforming Medical Systems

Abstract

Medical systems used for in vitro diagnostics are complex instruments that require precise manufacturing process controls to ensure accurate and reliable results. Existing methods for selecting manufacturing process controls for new medical systems may not adequately measure the final product functionality, leading to the high production of non-conforming systems. This can limit new system launches, which can be costly for the manufacturer and prevent caregivers from accessing valuable medical technology. A predictive model that can use data from the manufacturing process to predict if a medical system is non-conforming can give engineering managers and stakeholders confidence that the system being released for use will work as intended and allow commercialization activities to start with minimal restrictions. This could reduce costly low-yield issues observed during product launches. This Praxis describes the process of developing a predictive model for predicting the probability of producing non-conforming medical systems based on a multivariable logistic regression model. The model was developed using information and data generated during the design transfer and early manufacturing phase for a new medical system. The study aimed to identify the control system parameters that were significant contributors to the outcome of the system’s analytical test, which was used to determine non-conformance. Results generated during the research indicate that the predictive model developed was effective at predicting the non-conformance disposition of a system based on the testing data generated as part of the manufacturing process.