Diagnostic thresholds

Diagnostic Thresholds - The Cutoff Conflict

• The cutoff value determines a test's sensitivity (Sn) & specificity (Sp).
• An inverse relationship exists: ↑Sn = ↓Sp, and vice-versa.
• Receiver Operating Characteristic (ROC) curve: Plots Sn (true positive rate) vs. 1-Sp (false positive rate) for various cutoffs.
  • Area Under the Curve (AUC) measures accuracy.
  • AUC = 0.5: Chance (useless test).
  • AUC = 1.0: Perfect test.

⭐ For severe diseases with effective treatment (e.g., cancer screening), a lower cutoff is used to maximize sensitivity and not miss cases (ruling out disease, SNOUT).

ROC Curves - Accuracy at a Glance

• A Receiver Operating Characteristic (ROC) curve graphically plots a test's diagnostic performance across all possible cut-off thresholds.
• It plots Sensitivity (True Positive Rate) vs. 1-Specificity (False Positive Rate).
  • Y-axis: Sensitivity = $TP / (TP + FN)$
  • X-axis: 1 - Specificity = $FP / (FP + TN)$

• Interpretation:
  • The curve shows the trade-off between sensitivity and specificity.
  • A more accurate test has a curve that "bows" closer to the top-left corner.
• Area Under the Curve (AUC):
  • Represents the test's overall accuracy; its ability to correctly distinguish between diseased and non-diseased individuals.
  • AUC = 1.0: Perfect test.
  • AUC = 0.5: No discrimination (the diagonal line).

⭐ The ideal cut-off point on the ROC curve is often the point closest to the top-left corner (0,1), as it maximizes the Youden Index (Sensitivity + Specificity - 1).

Clinical Strategy - Screen or Confirm?

• Screening Tests: Prioritize high sensitivity (Sn).
  • Goal: Rule-out disease (📌 SNOUT: SeNsitivity rules OUT).
  • Use when missing a case is unacceptable (e.g., serious but treatable conditions).
  • Accepts higher false positives (FP).
• Confirmatory Tests: Prioritize high specificity (Sp).
  • Goal: Rule-in disease (📌 SPIN: SPecificity rules IN).
  • Use after a positive screen to confirm diagnosis.

⭐ In low prevalence settings, a test's Negative Predictive Value (NPV) is maximized, making a negative result highly reliable for ruling out disease.

Predictive Values - The Prevalence Effect

• Positive Predictive Value (PPV) is directly proportional to prevalence.
  • As disease prevalence ↑, PPV ↑.
• Negative Predictive Value (NPV) is inversely proportional to prevalence.
  • As disease prevalence ↑, NPV ↓.
• Sensitivity and Specificity are intrinsic to the test and are not affected by prevalence.

⭐ High-prevalence populations yield higher PPVs, making screening tests most useful when targeted at high-risk groups.

High-Yield Points - ⚡ Biggest Takeaways

• Altering the diagnostic threshold creates a trade-off between sensitivity and specificity.
• ↓ threshold: ↑ sensitivity (more true positives), ↓ specificity (more false positives). Good for screening tests.
• ↑ threshold: ↑ specificity (more true negatives), ↓ sensitivity (more false negatives). Good for confirmatory tests.
• ROC curves plot sensitivity (y-axis) vs. 1-specificity (x-axis) for various thresholds.
• The Area Under the Curve (AUC) measures a test's overall diagnostic accuracy.