Scenario-based multi-parametric QC for quality control of complex algorithms used in clinical care

Complex algorithms and prediction models are increasingly being developed, investigated and applied for clinical use in medical laboratories and beyond. Proper algorithmic functioning in a clinical setting is of the utmost importance to ensure patient safety. Recently, a tumor marker-based, longitudinal, machine learning model (mSTOP) was developed to predict which non-small cell lung cancer patients would not respond to (chemo-)immunotherapy early in the treatment process. This model was then clinically implemented in a hospital setting for which ICT infrastructure was developed to allow for the model's automated and real-time application. To ensure the appropriate functioning of the ICT infrastructure and the algorithm, a quality control (QC) strategy was developed and executed similarly to normal internal QC for clinical tests performed routinely in the laboratory. This quality control strategy entitled; scenario-based multiparametric quality control (MPQC) is based on the input and output values of different pre-specified scenarios obtained from the original clinical validation study. This perspective aims to justify and elucidate the concept of this QC strategy for quality control of complex algorithms operated in a clinical setting.