Integration of Current Clinical Knowledge with a Data Driven Approach: An Innovative Perspective

Abstract

Cardiovascular diseases are the leading cause of death worldwide. The development of models to support clinical decision is of great importance in the management of these diseases. This work aims to improve the performance exhibited by risk assessment scores that are applied in the clinical practice. This methodology has three main phases: (i) representation of scores as a decision tree; (ii) optimization of the decision tree thresholds using data from recent clinical datasets; (iii) transformation of the optimized decision tree into a new score. This approach was validated in a cardiovascular disease secondary prevention context, supported by a dataset provided by the Portuguese Society of Cardiology ([Formula: see text]). The respective performance was assessed using statistical metrics and was compared with GRACE score, the reference in Portuguese clinical practice. The new model originated a better balance between the sensitivity and specificity when compared with the GRACE, originating an accuracy improvement of approximately 22%.