Example Graph
A Simple Method for Evaluating Prediction Models, Diagnostic Tests, and Molecular Markers
Diagnostic & prognostic models are evaluated by accuracy measures (e.g. AUC) that don’t address clinical consequences. Decision-analytic techniques address those consequences, but only with extensive information, and are not easily applicable to models with percent risk estimates. DCA incorporates clinical consequences, addressing model benefits and harms with limited data.
For more information visit: http://decisioncurveanalysis.org/
Code (R, Python, Stata, and SAS)
Tutorial & Walkthrough (R, Python, Stata, and SAS)
Other Resources
Peer-Reviewed Literature
Original Paper on DCA
Extensions to DCA
- Applications to Survival Time Data among others
-
DCA & Treatment Response
- Model Evaluation: https://pubmed.ncbi.nlm.nih.gov/17550609/
- Trial Design: https://pubmed.ncbi.nlm.nih.gov/17022818/
Introductory Papers and Guides
- General Introduction to DCA
- How to Understand a Decision Curve
- Guide for Investigators on Reporting Decision Curves
Discussion Papers and Theoretical Background
- General Theoretical Background to DCA
- Mathematical Underpinnings of DCA through Relationship to Relative Utility (Similar Method)
- Widely Cited First Prediction Model Evaluation Paper to Introduce 3-Step Approach of Discrimination, Calibration, Clinical Utility
Editorials and Commentaries Recommending DCA
- Editorials recommending DCA Published in Major Journals
- Other Editorials Recommending DCA
