One way to test lots of patients where a) there's a low probability that an individual patient will have the disease and b) the test is expensive is to first mix some of the blood of each of N patients and do one test on the mix. If the batch tests negative then all patients are negative and you've only paid for one test. If the batch tests positive then you have to repeat the test on the remaining blood of each of the N patients to determine which were positive. Thus, with a high probability you only pay for 1 test, and with a low probability you have to pay for N+1 tests. The value of N is easily computed to minimize the overall cost, given the cost of each test and the percentage of patients that have the disease.