גרף ROC (Receiver Operating Characteristic)

A receiver operating characteristic (ROC) plot is a graphical tool that assesses diagnostic test performance by plotting sensitivity against 1 - specificity across various threshold settings. It illustrates the trade-off between correctly identifying diseased individuals and excluding those without disease. The ROC curve helps clinicians and researchers identify optimal diagnostic thresholds and compare the accuracy of different classification models using sensitivity, specificity and predicted value.

Adjusting diagnostic thresholds shifts the balance between sensitivity and specificity. Lowering the threshold increases sensitivity (catches more diseased individuals) but decreases specificity (more false positives). Raising the threshold increases specificity but decreases sensitivity. The ROC curve visualizes this trade-off across all possible thresholds, helping determine the optimal cutoff for a specific clinical context.

The area under the ROC curve (AUC) quantifies a test's ability to distinguish between diseased and healthy individuals. An AUC close to 1.00 indicates excellent diagnostic accuracy, while an AUC near 0.50 suggests the test performs no better than random guessing. The AUC serves as a single, comprehensive performance measure for comparing diagnostic tools across different thresholds.

When an ROC curve aligns with the diagonal line, it indicates that sensitivity and 1 - specificity are equally distributed, meaning the predictor variable has no association with disease status. This demonstrates the model performs no better than chance at distinguishing between diseased and non-diseased individuals, suggesting the diagnostic test lacks discriminatory power.

Comparing ROC curves from different diagnostic tests reveals their relative accuracy across all threshold settings. A curve that deviates farther from the diagonal line indicates greater diagnostic accuracy. The test with the larger area under its curve (AUC) demonstrates superior ability to distinguish between diseased and non-diseased individuals, making it the more reliable diagnostic tool.

Balancing sensitivity and specificity is crucial because different clinical scenarios require different priorities. High sensitivity is essential for screening to avoid missing diseased individuals, while high specificity is critical for confirmatory testing to avoid unnecessary treatment. The ROC curve helps clinicians select the optimal threshold that best matches their clinical needs and disease consequences.

The ROC curve plots all possible sensitivity-specificity combinations across different thresholds, allowing clinicians to visually identify the threshold that maximizes diagnostic accuracy for their specific context. By examining where the curve deviates most from the diagonal and considering clinical priorities, practitioners can select the optimal cutoff using statistical methods for analyzing epidemiological data.