Estimating reliability of evoked potential measures from residual scores: an example using tibial SSEPs

A normative study of tibial SSEPs was performed in 74 healthy subjects, and the effects of the variables sex, age and height on SSEP parameters were assessed. In a subgroup of 20 subjects a test-retest study was also performed, which allowed us to estimate the reliability of the different parameters by means of the intraclass correlation coefficient. We demonstrated that the intraclass correlation coefficient may be biased by predictable effects of subject-related variables (such as age, height and sex), if it is computed from raw original values. This bias can be eliminated by estimating reliability indices on residual scores calculated as the differences between observed values and those predicted by subject's age, height and sex.