The use of machine learning for the prediction of response to follow-up in spine registries

Alice Baroncini, Andrea Campagner, Federico Cabitza, Francesco Langella, Francesca Barile, Pablo Bellosta-López, Domenico Compagnone, Riccardo Cecchinato, Marco Damilano, Andrea Redaelli, Daniele Vanni, Pedro Berjano

Abstract: Highlights•Registries are gaining importance both from a clinical and scientific perspective.•Maintaining a high follow-up rate over time is one of the main challenges when compiling a registry.•Machine learning models, Cautious AI models based on a Random Forest model, can successfully predict which patients will comply to follow-up.•The main characteristics that drive compliance with follow-up are length of follow up, the level of the main pathology and extent of surgery, the SF-36 and the BMI.•Implementing ML models in the clinical practice might lead to an increase of follow-up compliance and to a reduction in the costs required to maintain the registry.