What Is an ROC Curve?

A receiver operating characteristic (ROC) curve shows how a test's sensitivity and specificity change across all possible cut-off values for a continuous or ordinal result. It plots sensitivity (true positive rate) on the vertical axis against 1 − specificity (false positive rate) on the horizontal axis.

How to read it

• A curve closer to the top-left corner indicates better discrimination.
• The diagonal line represents a test no better than chance.
• Each point on the curve is a threshold: moving along it trades sensitivity for specificity.

Area under the curve (AUC)

The AUC summarises discrimination: 0.5 means chance, 1.0 means perfect separation. It can be interpreted as the probability that a randomly chosen person with the disease has a higher test value than a randomly chosen person without it. Report the AUC with a confidence interval. Descriptive labels such as "acceptable" or "excellent" are conventions, not rules; clinical usefulness depends on the setting.

Choosing a threshold

Several approaches exist, such as maximising the sum of sensitivity and specificity (Youden's index) or choosing a threshold that achieves a required sensitivity or specificity for the clinical goal. A threshold chosen from the same data used to estimate performance is optimistically biased, so validate in new data.

Illustrative example

Invented for teaching. A biomarker is measured in 200 patients with and 300 without a condition. At a cut-off of 10 units, sensitivity is 85% and specificity is 70%; at 15 units, sensitivity is 70% and specificity is 88%. The ROC curve joins such points, and the AUC is 0.84 (95% CI 0.80 to 0.88). The clinical team chooses a cut-off by deciding which error is more costly.

Limitations

AUC ignores prevalence and the consequences of errors, and may mislead when classes are very imbalanced. Calibration and clinical utility need separate assessment.

Common pitfalls

• Selecting the best cut-off in the same data and reporting its performance as if validated.
• Comparing AUCs without a suitable test.

Key takeaways

• ROC curves show trade-offs; AUC summarises discrimination.
• Validate thresholds and consider clinical consequences.

This page is general educational information about research methods. It is not medical, legal or regulatory advice; follow your institution's ethics board and your target journal's instructions.