Novel metrics for risk stratification with nuclear cardiology.

Abstract

Risk prediction through integration of clinical variables and imaging data into a prognostic model can identify those factors with a high impact on patient outcome. The development of a prognostic model traditionally includes the evaluation of the independent contribution of variables with consideration of covarying factors that may confound risk stratification. Several methods and metrics are available to assess the performance of a prediction model. Traditional measures for binary and survival outcomes include the concordance (C) statistic for discriminative ability, and goodness-of-fit statistics for calibration. Different prognostic approaches, including variants of the C statistic for survival data, reclassification tables, net reclassification improvement, and integrated discrimination improvement have been recently proposed to evaluate the contribution of new markers when they are added to a defined risk model and to evaluate how these variables are able to change the class of risk of the patient and the resulting diagnostic and therapeutic managements. Moreover, decision-analytic measures including decision curves analysis have been introduced to assess the net benefit obtained by making clinical decisions based on model predictions. This review provides an overview of various metrics available for risk stratification using nuclear cardiology procedures, underlying the need of choosing the appropriate prognostic model in order to justify the referral decision.