Receiver Operating Characteristic (ROC) Curves: The Basics and Beyond.

Area of Science:

• Medical Statistics
• Diagnostic Test Evaluation
• Clinical Epidemiology

Background:

• Diagnostic tests and clinical prediction rules are essential for estimating disease probability.
• Receiver operating characteristic (ROC) curves and area under the ROC curve (AUROC) quantify test discrimination.

Purpose of the Study:

• To review and highlight underappreciated features of ROC curves and AUROC interpretation.
• To provide a deeper understanding of diagnostic test performance evaluation.

Main Methods:

• Review of ROC curve properties and AUROC interpretation.
• Discussion of 5 underappreciated ROC curve features.
• Illustration of concepts using published study data.

Main Results:

• ROC curve slope equals likelihood ratio over a test result interval.
• Optimal test cutoffs depend on pretest probability and harm-benefit analysis.
• AUROC measures discrimination, not probability accuracy.
• AUROC can be misleading with non-monotonically decreasing ROC curve slopes.
• AUROC can be artificially inflated by including low-risk individuals.

Conclusions:

• A nuanced understanding of ROC curves and AUROC is crucial for accurate diagnostic test assessment.
• Beyond discrimination, factors like likelihood ratios and pretest probability influence clinical utility.
• Awareness of AUROC limitations prevents misinterpretation of diagnostic test performance.