Abstract:
|
We describe some benefit-risk analyses of diagnostic tests that appear useful. First, we'll consider a descriptive table to compare a new test with a standard test on diagnostic yield (proportions of true and false positives) and the risk of adverse consequences in false positive subjects due to unnecessary downstream invasive testing or treatment. Next, to compare test-based patient management algorithms, we consider benchmarking the risk of the clinical condition of interest associated with each recommendation made by an algorithm (Katki et al, J Lower Genital Tract Disease, 2013, 17, S28-S35). Finally, we show how an ROC plot can be used in a decision analysis to compare new test with a standard test, basing the seriousness (loss) of a false positive error relative to a false negative error on the slope of tangent line to the ROC plot at the operating point of the new test. Generally, benefit-risk measures for diagnostic tests can have inadequacies. Some may not be consistent with any decision theoretic model. Others do not separate the evaluation of the test from the efficacy of treatments recommended in test-based algorithms.
|