Table 1 Definitions of some common terms used in this review

Clinical prediction models - sometimes referred to as clinical prediction rules, prediction algorithms, or risk scoring tools, are evidence-based tools that can aid in personalized medical decision making

Evaluation - assessment of model validity (statistical performance) and impact (clinical performance)

Updating - adjustment or re-specification of a model, e.g., to improve its performance in new data or incorporate new markers

Overfitting - when model predictions are not valid for new subjects due to parameter uncertainty (i.e., uncertainty in predictor effects) or model uncertainty (e.g., selection of predictors may be biased) in the derivation data

Shrinkage - a correction factor that can be applied to a model to address overfitting

Case-mix - distribution of outcome and subject characteristics

Calibration drift - miscalibration over time due, e.g., to changes in case-mix or clinical practice

Decision analysis - a method used to assess clinical usefulness that takes into consideration different decisions or actions based on a clinical threshold or range of plausible thresholds

Individual participant data (IPD) - raw, not aggregated or summarized, data