In their excellent overview of a common statistical measure of agreement, Thomas McGinn and colleagues1 suggest in Table 1 that values for the kappa statistic range from 0 to 1. However, negative values of kappa are also possible.2 Although unusual in practice, a negative kappa statistic results when agreement occurs less often than predicted by chance alone. This may indicate genuine disagreement, or it may reflect a problem in the application of a diagnostic test. Readers and researchers who encounter a negative kappa statistic should be aware of its implications, rather than blaming mathematical or typographic errors or computer “gnomes.”
David N. Juurlink Sunnybrook and Women's College Health Sciences Centre Toronto, Ont. Allan S. Detsky Mount Sinai Hospital Toronto, Ont.
References
- 1.McGinn T, Wyer PC, Newman TB, Keitz S, Leipzig R, Guyatt G, for Evidence-Based Medicine Teaching Tips Working Group. Tips for learners of evidence-based medicine: 3. Measures of observer variability (kappa statistic). CMAJ 2004;171(11):1369-73. [DOI] [PMC free article] [PubMed]
- 2.Kramer MS, Feinstein AR. Clinical biostatistics. LIV. The biostatistics of concordance. Clin Pharmacol Ther 1981;29:111-23. [DOI] [PubMed]