DECISION-ANALYSIS AND ASSESSMENT OF THE CLINICAL IMPACT OF DIAGNOSTIC-TESTS

Comprehensive assessment of the clinical value of diagnostic tests requires demonstration of technical capacity, diagnostic accuracy, and therapeutic impacts. A test shows technical capability if it detects a condition with reproducible results. Sensitivity, specificity, and receiver operating characteristic statistics permit assessment of tests according to standards of accuracy. Therapeutic efficacy assessment addresses whether a test improves therapy or outcomes more successfully than its predecessors. Therapeutic efficacy may be assessed through a randomized clinical trail or may be modeled using decision analysis. The threshold model is a special case of decision analysis with a focus on the change in expected utility (value) of test information as the probability of disease changes. Decision thresholds are specific probabilities of disease where the preferred course of action changes from observation to testing or from testing to initiating treatment. Thresholds depend on the utilities assigned to outcomes and on the accuracy statistics associated with relevant tests, but do not depend on the prior probability of disease. Prior probability estimates are derived from features of the history or physical findings, and comparison of prior probabilities to the threshold probabilities determines the most useful course of action. Assessment of diagnostic tests goes beyond determination of accuracy. For complex tests or rare diseases, decision analytic models offer useful assessments when randomized trials may not be feasible.