Hybrid predictive modelling: Thyrotoxic atrial fibrillation case

Abstract

In this work, we propose a new approach to predictive modelling of disease complications development. This approach is based on hybrid methods that have several advantages in comparison with classic methods. The main advantage is the inclusion of the complex information about the dynamics of a patient’s conditions using pathways analysis and graph-based predictive modelling method. Hybrid approaches integrate results of classic machine learning (ML) models and dynamic analysis methods for better modelling and prediction.We present this method’s application to the practical case of predictive modelling of Thyrotoxicosis Atrial Fibrillation (TAF) development. Medical specialists need tools to estimate the level of risk of developing TAF. Using the proposed predictive modelling method, our team developed such a tool. The method was validated using common ML metrics and expert evaluation and can be used as part of a decision support system for medical staff who work with thyrotoxicosis patients.This manuscript presents an extended version of the work described in the paper [1]. In this work, we proposed several methods for calculating the probability of TAF development. Our methods include arterial fibrillation risk questionnaire for use in practical diagnostic tasks and tools for analyzing TAF dynamic. The extended study presents further development of the approach within the hybrid modelling approach.