Skip to content
LatticeMLS, Home

Add your earlier progress to your account?

From before you signed in, this browser has:

Which plan do you want to keep?

You changed your plan before signing in, and those changes don't match the plan saved on your account.

SettingSaved planPlan changes

Study progress is waiting to be saved

Sensitivity, specificity and predictive values

14 min

Skills you'll build3 skills
  • Calculate diagnostic sensitivity against the reference classification
  • Calculate diagnostic specificity against the reference classification
  • Calculate positive and negative predictive values

Try first

Try first

A new rapid test is compared with a reference standard in 200 people. The counts are TP = 45, FN = 5, FP = 30 and TN = 120. What is the test's diagnostic sensitivity?

Keep going to see the reasoning.

Keep going to see the reasoning.

Keep going to see the reasoning.

Right. Keep going to follow the reasoning.

Keep going to see the reasoning.

Get the idea

A diagnostic accuracy study classifies every participant twice: once by the test being studied and once by a reference standard that defines who has the target condition. The four possible pairings form a two-by-two table.1,2

Reference: condition presentReference: condition absent
Test positiveTrue positive (TP)False positive (FP)
Test negativeFalse negative (FN)True negative (TN)

Each measure is a fraction of either one row or one column.

Sensitivity and specificity

Both are calculated from the reference standard's classification.

  • Sensitivity works down the first column. It is the percentage of people classified as positive by the reference standard who also test positive.
  • Specificity works down the second column, among the people classified as negative by the reference standard.

sensitivity = TP ÷ (TP + FN) × 100%

specificity = TN ÷ (TN + FP) × 100%

Predictive values

Predictive values start from the test result a clinician receives.1,3

  • The positive predictive value (PPV) is the share of positive results that are true.
  • The negative predictive value (NPV) is the share of negative results that are true.

PPV = TP ÷ (TP + FP) × 100%

NPV = TN ÷ (TN + FN) × 100%

Effect of prevalence

The two pairs behave differently when the tested population changes.1,3

  • Sensitivity and specificity describe the test at a stated cutoff, in the kind of population studied, and they can still differ from one study setting to another.
  • Predictive values also depend on prevalence, the share of tested people who have the condition. As a condition becomes rarer, false positives make up a larger part of the positive row, so PPV falls and NPV rises, even with the same test and cutoff.

Check the denominator

Before dividing, name the denominator.

  • A column total gives sensitivity or specificity.
  • A row total gives a predictive value.

Follow the bench demonstration

Sources for this step
References
  1. US Food and Drug Administration. Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests: Guidance for Industry and FDA Staff. Final guidance. Issued March 13, 2007. Accessed September 23, 2026.
  2. Clinical and Laboratory Standards Institute. Evaluation of Qualitative, Binary Output Examination Performance. 3rd ed. CLSI guideline EP12. Clinical and Laboratory Standards Institute; 2023. Accessed August 31, 2026.
  3. Bishop ML, Fody EP, Van Siclen C, Mistler JM, Moy M. Clinical Chemistry: Principles, Techniques, and Correlations. 9th ed. Jones & Bartlett Learning; 2023.

Watch one

A screening test has a sensitivity of 95% and a specificity of 90%. It is used in 1,000 people, 10% of whom have the condition. What are the PPV and NPV in this group?

  1. Split the group by the reference standard: 10% of 1,000 = 100 with the condition and 900 without.

    Predictive values need counts in every cell, and prevalence determines how many people fall in each column.

  2. Fill that column: TP = 0.95 × 100 = 95, so FN = 100 − 95 = 5.

    Sensitivity is a fraction of the condition-present column.

  3. Fill the other column: TN = 0.90 × 900 = 810, so FP = 900 − 810 = 90.

    Specificity is a fraction of the condition-absent column.

  4. PPV = 95 ÷ (95 + 90) = 95 ÷ 185 = 51.4%.

    PPV starts from everyone with a positive result, so it uses the test-positive row.

  5. NPV = 810 ÷ (810 + 5) = 810 ÷ 815 = 99.4%.

    NPV starts from everyone with a negative result, so it uses the test-negative row.

PPV is 51.4% and NPV is 99.4%. At 10% prevalence, about half the positive results come from people without the condition.

Practice this skill

Your turn

Problem 1 of 3 · Hints available

A binary test is compared with a designated reference standard in 200 participants: TP = 90, FN = 10, FP = 20, and TN = 80. What is the estimated diagnostic sensitivity?

Incorrect. 80 ÷ (80 + 20) is specificity, using participants without the target condition.

Incorrect. 90 ÷ (90 + 20) is positive predictive value, using all test-positive participants.

Correct. Sensitivity uses participants with the target condition: 90 ÷ (90 + 10) × 100 = 90%.

Hint
  1. Sensitivity is calculated among the participants the reference standard classifies as positive.
  2. Those participants are the true positives and the false negatives.

Review Diagnostic performance

Problem 2 of 3 · Hints available

A test is compared with a reference standard in 200 participants: TP = 72, FN = 8, FP = 12 and TN = 108. What is the diagnostic specificity?

Hint
  1. Specificity looks only at the people without the condition.
  2. Those people are the true negatives and the false positives.
  3. Divide the true negatives by that column's total.
Show the answer

90 %

The reference standard classifies 108 + 12 = 120 participants as negative, and the test classified 108 of them as negative: 108 ÷ 120 × 100 = 90%.

Review Diagnostic performance

Problem 3 of 3 · On your own

In a study of 200 participants, TP = 38, FP = 12, FN = 2 and TN = 148. What is the positive predictive value?

Show the answer

76 %

Fifty participants tested positive (38 + 12), and 38 of those results were true: 38 ÷ 50 × 100 = 76%.

Review Diagnostic performance

For reference

Sources checked