Medical False Positive: HIV, Mammogram, and the Citizen Panic

"A positive test for a rare disease usually doesn't mean what you think. The math is simple; the citizen panic is universal."

The Setup

Suppose a disease affects 1 in 1,000 people in a population. A test for the disease has 99% sensitivity (P(positive | disease) = 0.99) and 95% specificity (P(negative | healthy) = 0.95, so the false-positive rate is 5%). You test positive. The citizen instinct: 'the test is 99% accurate, so I almost certainly have it.' The Bayesian answer: about 2%.

The Calculation

Imagine 100,000 people from the population. About 100 have the disease; about 99,900 do not. Apply the test:

99 of the 100 sick are correctly positive (99% sensitivity).
~4,995 of the 99,900 healthy are falsely positive (5% false-positive rate).

Total positives: ~5,094. Of those, only 99 are actually sick. P(sick | positive) ≈ 99 / 5,094 ≈ 1.94%. Not 99%. About 2%.

The arithmetic is straightforward. The reason citizens get it wrong is they confuse sensitivity (P(positive | sick)) with posterior (P(sick | positive)). The two are conditional probabilities in opposite directions, related by Bayes' rule and the prior.

Why This Matters in the Clinic

Routine screening recommendations are tuned to this Bayesian reality. Mammograms for women under 40 are not routinely recommended despite breast cancer's seriousness, because the prior probability is low enough that positive screens produce more anxiety and unnecessary biopsies than they catch real cases. HIV screening is recommended for high-risk groups (high prior) but not for the general population (low prior) for the same reason.

The policy is not 'we don't care about catching cancer.' It is 'given the prior, the screening tool produces more harm than benefit at this population level.' The citizen who reads the policy as 'doctors don't care' is misreading the Bayesian math the policy is built on.

The Operating Principle

For any positive test result, ask: what is the prior probability of the condition in my situation? If the prior is low, even a sensitive test produces a posterior far below the test's nominal accuracy. Test accuracy is the likelihood; what you want is the posterior; the prior is the bridge. Without the prior, the test result tells you about the test, not about you. This is one of the few places where bringing math into a clinical conversation can change the patient's behavior in a directly health-improving way.

Code

Concrete tabulation of the medical false-positive paradox·python

# Concrete numbers for the classic medical false-positive case.
N_population = 100_000
base_rate = 0.001              # 1 in 1,000 has the disease
sensitivity = 0.99             # P(+ | sick)
specificity = 0.95             # P(- | healthy)
false_pos_rate = 1 - specificity

sick = int(N_population * base_rate)
healthy = N_population - sick
true_positives = int(sick * sensitivity)
false_positives = int(healthy * false_pos_rate)
total_positives = true_positives + false_positives

print(f"Population:                  {N_population:>8,}")
print(f"Actually sick:               {sick:>8,}")
print(f"Actually healthy:            {healthy:>8,}")
print(f"True positives (sick + +):   {true_positives:>8,}")
print(f"False positives (healthy + +): {false_positives:>8,}")
print(f"Total positives:             {total_positives:>8,}")
print(f"\nP(sick | positive) = {true_positives} / {total_positives} = {true_positives / total_positives:.4f}")
print(f"Notice: the test is 99% accurate, but the posterior is only ~2%.")

Exercise

Pick a screening or diagnostic test you (or someone close to you) might encounter (mammogram, prostate screening, COVID test, dietary allergy panel). Find or estimate: (1) the base rate of the condition in your demographic; (2) the test's sensitivity; (3) the test's specificity. Compute the posterior P(condition | positive). Then ask: would you behave differently if you knew the posterior rather than the sensitivity? Most people would.

Hint

Sensitivity and specificity are advertised; base rates are hidden. Look up your demographic's base rate before reasoning about any positive result.

Medical False Positive: HIV, Mammogram, and the Citizen Panic

The Setup

The Calculation

Why This Matters in the Clinic

The Operating Principle

Code

External links

Exercise

Progress

Comments 0