Research will be undertaken with the long-term objectives of improving the means by which the results of multiple diagnostic tests are combined and of improving diagnostic strategies for the use of those tests. The current proliferation of new imaging modalities that provide different amounts and kinds of information relative to a given disease and organ, often expensively and sometimes with risk to the patient, makes apparent the need for well-founded strategies should be founded on valid empirical analyses of the accuracies of the tests and on valid theoretical and empirical analyses of the gains in accuracy that can be achieved from their combination. Such strategies should also take into account the utilities -- i.e., the costs and benefits -- of using the tests singly and in combination. The project proposed here has two general aims. The first is to make an illustrative empirical analysis of the accuracy of multiple tests for a given organ and disease, used singly and in combination, under various modes of combination. Physical examination, radiography, and either diaphanography or thermography in breast cancer detection will be studied here. The second general aim is to develop further and to compare to the data a theory of optimal combination of test results, and then to use that theory to predict optimal diagnostic strategies. Our empirical analyses will be carried out with RDC measurements, which acknowledge that single or combined tests will have sensitivities and specificities that may be covaried inversely over a large range. This empirical study will be more systematic and extensive than others undertaken to date. Our theoretical analysis also will be based in RDC theory and will be the first theoretical approach capable of handling both continuous (or graded) test outcomes and combinations of correlated (i.e., partially redundant) tests. These capabilities represent a substantial advance over conventional decision analysis, which is severely limited by its assumptions of binary test outcomes, fixed sensitivities and specificities of each test, and independent outcomes of alternative tests.
| Swets, J A (1992) The science of choosing the right decision threshold in high-stakes diagnostics. Am Psychol 47:522-32 |
| Metz, C E; Shen, J H (1992) Gains in accuracy from replicated readings of diagnostic images: prediction and assessment in terms of ROC analysis. Med Decis Making 12:60-75 |
| Dorfman, D D; Berbaum, K S; Metz, C E (1992) Receiver operating characteristic rating analysis. Generalization to the population of readers and patients with the jackknife method. Invest Radiol 27:723-31 |
