A 99%-accurate test says you have a rare disease. How likely is it real?

Trust your first instinct here, then watch the math pull it apart.

3 quick questions · about 2 min · no sign-up

Question 1 of 3

A disease affects 1 in 10,000 people. A test for it is 99% accurate, whether you're sick or healthy. You test positive. Roughly how likely is it that you actually have the disease?

• About 99% — the test is 99% accurate
• About 1% — most positives are false alarms
• About 50% — it's basically a coin flip either way
• I'm not sure

You said: About 99% — the test is 99% accurate

That reads the accuracy as the answer, but it's actually about 1%. The disease is so rare that false alarms from the huge healthy crowd vastly outnumber the true positives from the tiny sick group. A 99% accurate test on a 1-in-10,000 disease is right only about 1% of the time on a single positive.

You said: About 1% — most positives are false alarms

Exactly — about 1%. Because the disease is so rare, the few real cases get buried under false alarms from the enormous healthy population. The test didn't lie; there are just far more healthy people to generate false positives than sick people to generate true ones.

You said: About 50% — it's basically a coin flip either way

That's the instinct to hedge in the middle, or to think 'either I have it or I don't, so 50/50.' Both ignore how rare the disease is. The real answer is about 1%: false alarms from the huge healthy crowd swamp the handful of true positives.

You said: I'm not sure

No problem — the answer is about 1%. The key: the disease is so rare that false alarms from the enormous healthy population vastly outnumber the true positives from the tiny sick group. A positive is far more likely to be one of those false alarms.

Another way to see it

Another angle: think signal versus noise. Real cases are a faint signal — only 1 in 10,000 people are genuinely sick. The test's 1% error rate is background noise: 1 in 100 healthy people gets a wrong positive, and almost everyone is healthy. So 1-in-100 noise buries 1-in-10,000 signal, roughly 100 times more of it. When your detector lights up, the odds are overwhelming that noise tripped it.

The whole answer hinges on counting the two crowds that test positive. Let's count them.

Question 2 of 3

Take a city of 1,000,000 people and test everyone. About 100 are truly sick, 999,900 are healthy. The test is 99% accurate both ways. Roughly how many of each group test positive?

• About 99 sick and about 9,999 healthy test positive
• About 99 sick and only about 100 healthy test positive
• About 99 sick and almost no healthy people test positive
• I'm not sure

You said: About 99 sick and about 9,999 healthy test positive

Exactly. The test correctly flags about 99 of the 100 sick people. But 1% of the 999,900 healthy people get a wrong positive — about 9,999 false alarms. So roughly 10,000 test positive and only 99 are truly sick: about 1%.

You said: About 99 sick and only about 100 healthy test positive

The sick count is right, but the false alarms are far larger. A 1% error rate on 999,900 healthy people is about 9,999 false positives, not 100. That flood of false alarms is exactly why your positive is only about 1% likely to be real.

You said: About 99 sick and almost no healthy people test positive

The test is wrong 1% of the time on healthy people too, and there are 999,900 of them — that's about 9,999 false positives. They overwhelm the ~99 true positives, which is why a single positive is only about 1% likely to be real.

You said: I'm not sure

Here's the count: about 99 of the 100 sick test positive, and 1% of the 999,900 healthy — roughly 9,999 — get false positives. So ~10,000 test positive, only 99 truly sick. That's about 1%.

So the false-positive pile dwarfs the true-positive pile whenever the disease is rare. One last case to lock it in.

Question 3 of 3

Suppose a SECOND disease is far more common — say 1 in 100 people have it — and you use the same 99%-accurate test. Compared to the rare disease, what happens to a positive result?

• It's now much more likely to be real — the sick pool is far bigger relative to false alarms
• Nothing changes — the test is still 99% accurate, so it's still about 1%
• It gets even less reliable — more people means more false alarms
• I'm not sure

You said: It's now much more likely to be real — the sick pool is far bigger relative to false alarms

Right. With 1 in 100 sick, the true-positive pool grows enormously while the false-alarm pool shrinks, so a positive becomes far more trustworthy. The base rate, how common the disease is, is what drives the answer — not the accuracy number.

You said: Nothing changes — the test is still 99% accurate, so it's still about 1%

The accuracy is the same, but that was never what set the odds. The base rate did. With 1 in 100 sick instead of 1 in 10,000, far more positives are real — so a positive becomes much more trustworthy. Start from how common the disease is.

You said: It gets even less reliable — more people means more false alarms

False alarms don't grow here; the sick pool does. Going from 1 in 10,000 to 1 in 100 sick means far more true positives relative to false alarms, so a positive becomes more trustworthy. The base rate is what's driving the result, not the accuracy.

You said: I'm not sure

It becomes much more likely to be real. With 1 in 100 sick instead of 1 in 10,000, the true-positive pool is far bigger relative to false alarms. The base rate, how common the disease is, drives the answer — not the accuracy number.

The takeaway

Keep this one thing: a test result's trustworthiness is set by the base rate — how common the disease is — not by the accuracy number. The rarer the target, the more a single positive means retest, not panic.

Pull the thread

That positive test meant less than your gut screamed — because the disease was rare. Now flip it: a clue that tells you far more than it looks. A game-show host is about to open a door, and it is not random.