Statistical errors and their effect on electrodiagnostic medicine.

Abstract

Failure to demonstrate abnormalities on diagnostic tests in patients felt certain to have a disease is all too common, especially when symptoms are mild to moderate. One strategy used in such cases is to perform multiple tests hoping that one will confirm the clinical suspicion. An example of this strategy is found in carpal tunnel syndrome where multiple tests are often used to confirm the diagnosis. Unfortunately, this approach is often erroneous. Since no diagnostic test can perfectly discriminate between normal and abnormal, critical values are set at levels where most subjects are correctly classified. Using this methodology, approximately 2.5% of patients who are normal will be mistakenly called abnormal. This type I error, is additive when additional independent tests are performed. Since no two tests for a single condition are completely independent, the total error is less than the sum of the individual errors for each test. Even after accounting for interdependency, the total error of combined tests may be unacceptably high. If a single, highly discriminating, test is not available and multiple tests are used, abnormalities in more than a single test are needed to distinguish between normals and abnormals.