BIOMETRICAL MODELING OF HIV SCREENING IN BLOOD AND PLASMA DONORS USING THE COMBINATION OF 2 SCREENING-TESTS - STRATEGIES, VALIDITY, COST AND EFFECTIVENESS

Objective: The strategies for combining two screening tests for HIV in fections in blood or plasma donors are formulated in biometric terms a nd analyzed with respect to their value, i.e. their vaidity, cost and effectiveness. Design: Biometrical modeling using assumptions on the v alidity of the single tests, the conditional correlations between them , as well as on the cost of testing and the consequences of false-nega tive or false-positive test results. Results: If the test combination is defined as positive whenever at least one of the single tests is po sitive, then this rule (the 'believe the positive' rule, BTP), due to its lower specificity, has extremely low positive predictive values. I n case of high prevalence rates of the infection (e.g. 1:1,000), the B TP rule leads to lower total cost than single testing, unless the latt er has very high sensitivity (e.g. 99 %). For smaller prevalence rates (< 1:50,000), which are more typical of the selected group of blood o r plasma donors, combination testing is of little value because the ex tra cost of detecting one additional infection (compared with single t esting) may reach several 100 million DM. Conclusion: The cost for det ecting additional cases of HIV infection by using combination instead of single testing in HIV screening is so high that this decision requi res a public consensus.