Cost-sensitive decision tree model for optimal identification of need for right ventricular support

Right ventricular (RV) failure following implantation of a left ventricular assist device (LVAD) is a significant complication that is often difficult to predict. To address this problem, we developed a cost-sensitive decision tree to evaluate the risk of RV failure, hence the need for a right ventricular assist device (RVAD). The model was developed based on retrospective analysis of 183 patients who received an LVAD at the University of Pittsburgh Medical Center. It exhibited a specificity of 87% (identifying 136 out of 156 patients who did not receive an RVAD). With respect to the need for RVAD support, the model however identified 20 patients (13%) as RVAD-dependent which in fact did not receive an RVAD. These patients however exhibited significantly lower 90-day survival as those on which the model and expert agreed as RVAD independent. Future development of this model is intended to provide decision support to both aid and translate the judgment of clinical experts.