A new reliability parameter for automated perimetry: inconsistent responses

A proper interpretation of perimetric results first requires an adequate evaluation of patient reliability. The patient reliability parameters that are currently used require extra testing time and are based on a relatively small sample of the subject's responses. To develop a new, more robust reliability parameter, 160 visual fields of 20 eyes of 20 healthy volunteers were performed with a customized static threshold program of 44 test locations. The responses to the bracketing strategy algorithm were analyzed for inconsistencies that occurred when the subject claimed to see and not see the same intensity stimulus when represented at the same test location. These inconsistencies were summed over the entire visual fields and termed "number of inconsistent responses.'' The number of inconsistent responses correlated significantly with the following parameters: the sum of false-positive and false-negative responses to catch trials (r = 0.62; p = 0.003), fixation losses (r = 0.51; p = 0.020), and total number of stimuli (r = 0.61; p = 0.004). In contrast to conventional reliability parameters, the number of inconsistent responses is based on all the subjects' responses and requires no additional testing time.