LAUSR

To the Editor: We were interested to read the article by Alsherbini et al.1 published in July 2017 issue of Journal of Clinical Neurophysiology. The purpose of the authors was to evaluate the incidence of stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs) in 52 comatose patients in the neuroscience intensive care unit who underwent a standard stimulation protocol and define inter-reader reliability for continuous electroencephalographic monitoring.1 The stimulation protocol had three sequence parts, including verbal, tactile, and painful stimulation. All the continuous electroencephalographic epochs relevant to the study were reviewed by the principal investigator and then prepared for review by two masked board-certified EEG interpreters. With each test segment clearly marked, the reviewer determined whether an SIRPID occurred after stimulation. Next, the reviewers identified the SIRPID, specifying if they would consider the pattern ictal or not ictal and to report any electrographic seizure during the stimulation interval. If two readers disagreed on the presence of definite or possible SIRPIDs, a third masked reviewer evaluated the continuous electroencephalographic monitoring and a final diagnostic conclusion was made by majority decision.1 They used kappa statistic to assess the inter-rater agreement. Agreement or reliability was considered perfect, substantial, moderate, fair, or slight if the kappa statistic values were 0.81 to 1, 0.61 to 0.80, 0.41 to 0.60, 0.21 to 0.40, and 0.20, respectively. Based on their results, inter-rater agreement in reporting the different EEG patterns as SIRPIDs (present or absent) by three experienced EEG readers performed using the kappa statistic was 0.5 (95% confidence interval: 0.1e0.7); this finding suggested moderate agreement/reliability between raters. However, these results cannot be the most appropriate estimates to evaluate inter-rater agreement (reliability). First of all, it is crucial to bear in mind that, one cannot present a value of kappa that universally indicates a good agreement. In sense of assessing the agreement of a qualitative variable using the kappa value, its two crucial weaknesses should be considered. (1) As it depends on the prevalence in each category, it is possible to have different kappa value while having the same percentage for both concordant and discordant cells. (2) Kappa value depends on the number of categories, so that increasing the categories results in lower amount of kappa value. Therefore, to assess the agreement (reliability) of a qualitative variable especially with more than two categories, in this case 3 observers, Kappa Fleiss or weighted Kappa should be applied.2e4 Finally, the authors concluded that inter-reader reliability for continuous electroencephalographic monitoring was moderate. Such a conclusion may be a misleading message because of the inappropriate use of a statistical test to assess agreement (reliability). Briefly for reliability analysis, appropriate tests should be used by researchers. Otherwise, misdiagnosis and mismanagement of the patients cannot be avoided.5