This document discusses clinical prediction models. It defines a clinical prediction model as using multiple clinical predictors to calculate the probability of a patient outcome in order to help with diagnostic and treatment decisions. An example given is the Framingham risk score, which predicts cardiovascular disease risk within 10 years. The document notes that developing accurate prediction models requires adequate sample sizes and appropriate statistical methods. It describes calibration and discrimination as important measures of a model's accuracy, with discrimination referring to a model's ability to differentiate outcome classes using tools like ROC curves. Clinical prediction models can be used for targeting preventive interventions and informing surgical decision-making.