Evaluating a model to predict protocol performance.

BACKGROUND A test protocol is created when individual tests are combined. Even with a few tests, many different protocols are possible. Often, the clinician must select a protocol without information as to the performance of the possible protocols. A model to predict protocol performance could help in this selection process. PURPOSE To evaluate the validity and accuracy of a mathematical model for predicting protocol performance. RESEARCH DESIGN Predictions of the model are compared to actual data on protocol performance. RESULTS With complete information, there was almost perfect agreement between predicted and actual data. With partial information, the model still made very accurate estimates of protocol performance. CONCLUSIONS Even with incomplete information, which is frequently the case in a clinical situation, the model can eliminate many protocols from consideration and aid in the selection of an appropriate protocol.