A Retrospective Comparative Study of Three Data Modelling Techniques in Anticoagulation Therapy

Abstract

Three types of data modelling technique are applied retrospectively to individual patients' anticoagulation therapy data to predict their future levels of anticoagulation. The results of the different models are compared and discussed relative to each other and previous similar studies. The conclusions of earlier papers, that machine learning could help anticoagulation clinicians achieve better results, are reinforced here using an extensive data set. Continuously-updating neural network models are shown to predict future INR measurements best of the three types of models presented here.