Accuracy is a function of sensitivity and specificity.
Sensitivity is a measure of true positives (ie the test correctly identifies that someone with cancer has cancer)
Specificity is a measure of true negatives (ie the test correctly identifies that someone without cancer does not have cancer)
The issue with 86% accuracy is that, if specificity is around that mark, for every large sample of people tested, 14% will be incorrectly identified as having cancer.
If 100,000 people are tested, and 1% of the population has cancer (for illustrative purposes) then we can expect that 1,000 people with cancer will take the test. With 86% sensitivity, 860 of these people will correctly be identified as having the disease, while 140 will be told they don't have the disease while actually having it.
For the remaining 99,000 test subjects without cancer, 85,140 people will correctly get a negative result, while 13,860 will be identified as having cancer despite not having the disease (assuming 86% specificity).
Therefore of the 14,720 people identified with cancer, only 860 of them actually will have it.
This is why 86% accuracy is not great for such tests.
RAP's beauty is its low cost and low intrusion. Such solutions can afford to have accuracy around these marks (at current rates it exceeds this). If a test is highly intrusive or costly, 86% accuracy is too low to justify the test unless specificity itself is very high, resulting in fewer false positives.
Expand
(20min delay)