A method and system is provided which takes into account actual human data previously collected. The human data is used to predict the probability of a test conducted on a patient later, yielding a result above a threshold level which would be informative to a doctor treating a patient. The test which is conducted is for detecting levels of C-reactive protein, lipoprotein(a) or homocysteine in a patient. Patient data is collected and input into a multivariate function which then yields a percentage probability that test results on such a patient will be above a predetermined level.