Dear Valerio, I think you can use Diagnostic decision trees. I think that some statistical software such as SAS or SPSS perform this kind of predictive modeling analyses.

I am not sure we (others and me) others understand your question.  I often hear of the difference between statistical significance and economic significance.  For example, when assessing a program's effectiveness, some researchers report a statistically significant increase from the pretest to the post test.  Using hypothetical data I created for this answer, let's say scores increase from a 35.5 percent correct to a 38.7 percent correct and that this is statistically significant at the 0.05 level.  However, some might say that spending $x million dollars is economically significant, or some might call it economically inefficient because the gains are nearly what were expected.  

I suspect my answer will help you do describe what you are seeking, which very well could have nothing to do with my answer.  Regardless, I hope it helps.

Rob

Taking the limulus test for endotoxins as a reference point, perhaps sensitivity of the test and the time taken to get a reliable result are the two main evaluative criteria for economic efficiency of diagnostic tests, as only a small amount of test material is required (tests-sensitivity), while the shorter the time, the quicker the intervention can be administered. In a general sense, the economics of diagnostics boils down to space (volume) and time (reaction). There is also a silver lining to the limulus case, since the horseshoe crabs are bled and not killed for their blood, "no animals were harmed" in the final testing, unlike a dead, diseased lab animal, when a negative result is obtained. So being "renewable" is also a great plus point for economic efficiency.

https://www.ncsu.edu/project/bio402_315/Lecture%20two/limulus.html

Article Annals of Nuclear Medicine and Molecular Imaging | A compari...