On misconceptions about the Brier score in binary prediction models
Preprint
Abstract
The Brier score is a widely used metric evaluating overall performance of predictions for binary outcome probabilities in clinical research. However, its interpretation can be complex, as it does not align with commonly taught concepts in medical statistics. Consequently, the Brier score is often misinterpreted, sometimes to a significant extent, a fact that has not been adequately addressed in the literature. This commentary aims to explore prevalent misconceptions surrounding the Brier score and elucidate the reasons these interpretations are incorrect.Keywords
All Related Versions
This publication has 13 references indexed in Scilit:
- nhanesA: achieving transparency and reproducibility in NHANES researchDatabase: The Journal of Biological Databases and Curation, 2024
- Performance Metrics for the Comparative Analysis of Clinical Risk Prediction Models Employing Machine LearningCirculation: Cardiovascular Quality and Outcomes, 2021
- The Integrated Calibration Index (ICI) and related metrics for quantifying the calibration of logistic regression modelsStatistics in Medicine, 2019
- A systematic review shows no performance benefit of machine learning over logistic regression for clinical prediction modelsJournal of Clinical Epidemiology, 2019
- Using simulation studies to evaluate statistical methodsStatistics in Medicine, 2019
- The Brier score does not evaluate the clinical utility of diagnostic tests or prediction modelsDiagnostic and Prognostic Research, 2017
- Use of Brier score to assess binary predictionsJournal of Clinical Epidemiology, 2010
- Measuring the accuracy of prognostic judgments in oncologyJournal of Clinical Epidemiology, 1997
- Assessing predictive accuracy: How to compare brier scoresJournal of Clinical Epidemiology, 1991
- External correspondence: Decompositions of the mean probability scoreOrganizational Behavior and Human Performance, 1982