Youden Index Calculator

Compute Youden's J statistic (J = Sensitivity + Specificity − 1) from a 2×2 diagnostic table to score how well a binary test separates cases from non-cases.

Quick Facts

Formula
J = Sensitivity + Specificity − 1
Sensitivity = TP/(TP+FN); Specificity = TN/(TN+FP). J runs from 0 (chance) to 1 (perfect).

Your Results

Calculated
Youden's J index
-
Sensitivity + Specificity − 1
Sensitivity (TPR)
-
TP / (TP + FN)
Specificity (TNR)
-
TN / (TN + FP)
Interpretation
-
Discriminating power

Ready

Enter the four cell counts of your 2×2 table and calculate.

What the Youden Index measures

Youden's J statistic, introduced by W. J. Youden in his 1950 paper "Index for rating diagnostic tests," condenses a binary diagnostic test into a single number:

J = Sensitivity + Specificity − 1

Sensitivity (the true-positive rate) is the fraction of actual positives the test correctly flags; specificity (the true-negative rate) is the fraction of actual negatives it correctly clears. J adds the two and subtracts 1, so it ranges from −1 to +1. In practice a usable test lands between 0 and 1: J = 0 means the test performs no better than a coin flip, while J = 1 means it is perfect — it catches every case and never raises a false alarm.

Computing J from a 2×2 table

Most tests are summarized in a confusion matrix with four counts: true positives (TP), false negatives (FN), true negatives (TN), and false positives (FP). From those:

  • Sensitivity = TP / (TP + FN) — of everyone who truly has the condition, how many tested positive.
  • Specificity = TN / (TN + FP) — of everyone who is truly free of the condition, how many tested negative.
  • J = Sensitivity + Specificity − 1.

Worked example: with TP = 90, FN = 10, TN = 160, FP = 40, sensitivity = 90/100 = 0.90 and specificity = 160/200 = 0.80, so J = 0.90 + 0.80 − 1 = 0.70.

The ROC connection and the optimal cut point

Plot sensitivity on the y-axis against 1 − specificity on the x-axis and you get an ROC curve. At any single threshold, J is exactly the vertical distance from the ROC point up to the 45° chance diagonal. Because of this, the threshold that maximizes J — the point on the ROC curve farthest above the diagonal — is called the Youden optimal cut point. It is the classifier threshold that treats a missed case and a false alarm as equally costly. If false positives and false negatives carry very different real-world costs, a weighted or cost-based cut point may be preferable to the plain Youden point.

Reading the number

There is no universal cutoff for "good," but common informal bands are: below 0.2 is near chance, 0.2–0.4 weak, 0.4–0.6 moderate, 0.6–0.8 good, and above 0.8 excellent. Always report J alongside the raw sensitivity and specificity, because the same J can come from very different balances — J = 0.6 could be 90%/70% or 80%/80%, and which one you want depends on whether missing a case or a false alarm hurts more.

Frequently Asked Questions

What is the Youden Index (Youden's J)?
Youden's J statistic is J = Sensitivity + Specificity − 1. Introduced by W. J. Youden in 1950, it summarizes a binary diagnostic test in one number. J ranges from −1 to 1: 0 means the test is no better than chance, and 1 means it is perfect (no false positives and no false negatives).
How does the Youden Index relate to the ROC curve?
On an ROC plot of sensitivity versus 1 − specificity, J equals the vertical distance between the curve and the 45° chance diagonal at a given cutoff. The threshold that maximizes J — the point farthest from the diagonal — is the Youden optimal cut point, which weights sensitivity and specificity equally.
Can the Youden Index be negative?
Mathematically yes: if sensitivity + specificity is below 1, J is negative, meaning the test is worse than random. In practice you can flip the test's decision rule (call positives negative and vice-versa) to turn a J of −0.3 into +0.3, so a genuinely negative J usually signals a labeling or threshold error rather than a useful test.
Does the Youden Index depend on disease prevalence?
No. Because sensitivity and specificity are each computed within the actual-positive and actual-negative groups separately, J is independent of how common the condition is in your sample. That makes it stable across populations — unlike predictive values (PPV/NPV), which shift with prevalence.

Related calculators