Statistics & Data
Why a Positive Test Result Might Still Mean You're Probably Fine
A medical test that is 99% accurate can still be wrong the majority of the time — and understanding why reveals a flaw baked into the way human minds handle probability.
The Idea
Here is the trap. A disease affects 1 in 1,000 people. A test for it is 99% accurate — meaning it correctly identifies both the sick and the healthy 99% of the time. You test positive. Most people's instinct: there is a 99% chance you have the disease. The actual answer: roughly a 9% chance. This is base rate neglect in action. The 'base rate' is the background frequency of something in the population — how common the disease actually is. When that number is very low, even a highly accurate test will produce far more false positives than true ones, simply because it is being applied to a sea of healthy people. For every 1,000 people tested: roughly 1 genuinely has the disease (and the test likely catches them). But the test also wrongly flags about 10 healthy people as positive. So among 11 positive results, only 1 is real. The math is straightforward; the psychology is the problem. We are wired to update on vivid, specific information — a positive test, a dramatic news story, a friend's anecdote — and to unconsciously ignore the dull statistical backdrop against which that information sits. Kahneman and Tversky named this tendency in the 1970s, and it has never really gone away. Base rate neglect is not a beginner's mistake; it routinely trips up doctors, lawyers, and policy makers. The fix is not to become a statistician — it is to ask one quiet, powerful question before reacting to any new information: how common is this thing to begin with?
In the World
In the early months of widespread COVID-19 rapid antigen testing, a version of this problem played out at population scale. In the United Kingdom, when community transmission was relatively low — say, roughly 1 in 500 people infected at a given moment — epidemiologists pointed out that a lateral flow test with a specificity of around 99.9% would still generate a meaningful proportion of false positives among the enormous number of people testing themselves asymptomatically. This was not a flaw in the tests exactly; it was base rate arithmetic made visible. The tests were designed for symptomatic people or high-exposure contexts, where the base rate of actual infection is far higher, making positive results far more reliable. The moment you apply a test more broadly to a low-prevalence population, the signal-to-noise ratio shifts. A similar dynamic shaped a quieter controversy in cancer screening. When mammography screening was extended to lower-risk age groups, the debate was never really about the test's accuracy in isolation — it was about what a positive result actually meant statistically for a woman whose base rate of having breast cancer was already quite low. Decades of anxiety, invasive follow-up procedures, and sometimes unnecessary treatment flowed partly from this misunderstanding. The radiologist who reads the scan can be excellent. The statistician's concern lives somewhere upstream of that — in the question of who we are scanning, and why.
Why It Matters
Base rate neglect is not confined to medicine. It shapes how we assess risk in almost every domain — personal, professional, political. When you read that a certain food 'doubles your risk' of something, the first question to ask is not 'doubles from what?' as a rhetorical dismissal, but genuinely: what is the baseline? Doubling a 0.01% risk is not the same as doubling a 10% one. When a news story reports a spike in some alarming phenomenon, ask how many cases were expected in the first place — context that is almost never provided, because it is less compelling than the spike. Carrying this habit does not make you cynical or dismissive of evidence. It makes you a more accurate reader of the world. It gives you a way to notice when your emotional reaction to a vivid fact is outrunning the underlying mathematics. The practical upshot is simple: before you update your beliefs based on new information, spend a moment with the prior — the boring background frequency of the thing in question. That number does most of the real work.
A Question to Ponder
Think of something you currently believe is likely or unlikely — about your health, your industry, your future — and ask yourself: do you actually know the base rate, or have you been reasoning entirely from a single vivid data point?
Get a new one of these every morning.
Start learning with Thinkable