Risk prediction models

Prognostic research focuses on the prediction of the future course of a given disease in probability terms. Prognostication is performed by clinical decision makers by using risk prediction models that allow us to estimate the probability that a specific event occurs in a given patient over a predefined time period conditional on prognostic factors (predictors). Before application in clinical practice, risk prediction models should be properly validated by assessing their discrimination and calibration, or explained variation. Reclassification analyses allow us to evaluate the gain in risk prediction by using a new model compared with an established one. We discuss the concepts of developing and validating risk prediction models by means of two examples, the Framingham risk calculator for prediction of coronary heart disease (CHD), and the recently published Renal Risk Score to predict progression of chronic kidney disease (CKD).