Expected Value of Sample Information, Net Benefit of Sampling
Information has a value in utility terms. Consider a diagnostic test. It provides data, which, when duly interpreted, become information that may allow treatment to be individualized and expected outcome utility to increase. Three qualifications include the following:
First, a test may be too uninformative to influence treatment: VOI (valueofinformation) = 0. However, expected utility cannot decrease: VOI is never negative.
Second, these are average statements. New diagnostic tests, even when correctly interpreted, will often cause outcome utility to decrease for some patients. When population screening is introduced, false positives pay a price.
Third, misinformation does carry negative value. Utility may suffer when decisions rest on biased research or when diagnostic test results are wrongly interpreted, for example, due to overly optimistic ideas concerning sensitivity or specificity.
In this ...