Sensitivity vs. Specificity

Sensitivity measures how well a test identifies those with the disease (avoiding false negatives). Specificity measures how well it identifies those without it (avoiding false positives).

No test is perfect. You trade one for the other.

The Symptom: The Panic of a Positive Result

You take a screening test for a rare disease. It comes back positive. The brochure says the test is "99% accurate" (Sensitivity). You assume there is a 99% chance you are sick.

You are likely wrong. The chance might be as low as 10%.

This is the Base Rate Fallacy, and it happens because we confuse Sensitivity (how good the test is) with Positive Predictive Value (how likely the result is true).

The Mechanism: The Airport Scanner

Imagine a security scanner at an airport.

Scenario A: High Sensitivity (The Paranoid Scanner)

• Goal: Catch every weapon.
• How: It beeps for everything—guns, keys, belts, coins, foil wrappers.
• Result: No terrorists get through (0 False Negatives). But many innocent people are stopped and searched (Many False Positives).
• Medical Equivalent: Screening Tests (e.g., Mammograms, HIV ELISA). We want to catch everyone who might be sick. We accept that we will scare some healthy people.

Scenario B: High Specificity (The Relaxed Scanner)

• Goal: Only stop people with actual guns.
• How: It is calibrated to only beep for large metal objects shaped like a Glock.
• Result: Innocent people walk through freely (0 False Positives). But a terrorist with a ceramic knife might get through (False Negative).
• Medical Equivalent: Confirmatory Tests (e.g., Biopsy, HIV Western Blot). We want to be 100% sure before we start chemotherapy or surgery.

The Math: Why "99% Accurate" Can Be Misleading

This is the number that matters to you: Positive Predictive Value (PPV).

Imagine a disease that affects 1 in 1,000 people. We test 1,000 people with a test that has 100% Sensitivity and 95% Specificity.

1. The Sick Person: 1 person has the disease. The test catches them. (1 True Positive).
2. The Healthy People: 999 people are healthy.
3. The False Alarms: The test has 95% specificity, meaning it has a 5% error rate for healthy people.
   • 5% of 999 is roughly 50 people.
   • These 50 healthy people get a Positive result.

The Result:

• Total Positive Tests: 51 (1 Real + 50 Fake).
• You tested positive. Are you the 1 sick person or one of the 50 healthy ones?
• Your chance of being sick is only 1 in 51 (~2%).

Even with a "good" test, if the disease is rare, a positive result is mostly noise.

The Prescription: SNOUT and SPIN

Doctors use a mnemonic to remember this:

• SnNout: A test with high Sensitivity, when Negative, rules the disease Out. (Trust the negative).
• SpPin: A test with high Specificity, when Positive, rules the disease In. (Trust the positive).

When reading a paper on PaperScores:

1. Don't just look at "Accuracy". Look for Sensitivity and Specificity separately.
2. Check the Context. A test that works in a hospital (where everyone is sick) will have a much lower PPV in the general population (where most people are healthy).
3. Confirm. A positive screening test should always be followed by a specific confirmatory test.

Related Content