Another way to think about test results
I was pleasantly surprised to stumble across this video by 3Blue1Brown explaining how to use the Bayes Factor (aka Likelihood ratio) to make quick estimations about test results. Starting with a population of 100 and converting to odds is a simple yet accurate way to make quick guesses about likelihood of disease based on pretest probability, sensitivity and specificity.
Like most people, I have a lot of trouble understanding the logic of how a getting a positive result with a test with good sensitivity and specificity doesn’t necessarily mean a high chance of disease. This video does a great job of breaking down the paradox of how highly accurate tests are not necessarily highly predictive of disease.
It all depends on prior (or pretest) probability. For example, if you use a highly sensitive and specific test on someone with only a 1% probability of having disease, only 1/11 positive tests will only mean truly diseased. The other 10 will be false positives!
Questions? Watch the video or play around with the numbers on our website.