Discrimination and Calibration of Clinical Prediction Models

TLDR

Accurate prognosis is essential for optimal care, and prediction models that integrate multiple prognostic factors and provide absolute risk estimates must discriminate between event and non‑event patients while being well calibrated; this guide complements prior guides on model development and validation. The guide aims to help clinicians understand metrics for assessing discrimination, calibration, and relative performance of prediction models, thereby enabling optimal use of existing models. It explains how to evaluate discrimination, calibration, and relative performance of prediction models using available metrics.

Abstract

Accurate information regarding prognosis is fundamental to optimal clinical care. The best approach to assess patient prognosis relies on prediction models that simultaneously consider a number of prognostic factors and provide an estimate of patients' absolute risk of an event. Such prediction models should be characterized by adequately discriminating between patients who will have an event and those who will not and by adequate calibration ensuring accurate prediction of absolute risk. This Users' Guide will help clinicians understand the available metrics for assessing discrimination, calibration, and the relative performance of different prediction models. This article complements existing Users' Guides that address the development and validation of prediction models. Together, these guides will help clinicians to make optimal use of existing prediction models.