Barnhart HX, Lokhnygina Y, Kosinski AS, Haber M. Comparison of the concordance coefficient and the individual concordance coefficient in the assessment of conformity. J Biopharm Stat. 2007;17 (4):721-38. RAP is supported in this work by NHS Lothian through the Edinburgh Clinical Trials Unit. We thank Professor Michael Haber (Emory University) for providing examples of R and SAS program code from one of his publications [27] which we modified and adapted to apply the Coefficient of Individual Agreement method to our copD example in a previous draft of this document. Schluter PJ. Bayese`s multi-arched hierarchical approach to measuring concordance in comparison studies of repeated measurement methods. BMC Med Res Methodol. 2009;9(1):6.

In our assessment of the tuning methods, we implicitly assume that the available sample size is sufficient to achieve convergence of models. In cases where the number of patients or repeated measurements are low, the methods may not work properly, and this is a means for future research. In addition to the above methods, other methods have been used to assess compliance, although some of them are inappropriate. A systematic review [36] of compliance studies reported between 2007 and 2009 showed that about 10% of the studies used inappropriate methods to assess compliance, including standard correlation coefficients, the coefficient of determination from regression analysis (square R) and comparative methods (e.g. B t to detect differences in mean value). Carrasco JL, Caceres A, Escaramis G, Jover L. Distinction and correspondence with continuous data. Stat Med. 2014;33 (1):117-28. Barnhart HX, Haber MJ, Lin LI. An overview of conformity assessment for continuous measures.

J Biopharm Stat. 2007;17 (4):529-69. Lin L, Hedayat AS, Sinha B, Yang M. Statistical methods for evaluating similarities: models, problems and tools. J Am Stat Assoc. 2002;97(457):257–70. The need for confidence intervals, in addition to correspondence limits, is strongly mentioned in the literature, and rightly so. But we think it`s just as important – if not more important – to report the different components of variance (e.g. B variance between subjects and variance within the subject) and bias estimates, in addition to convergence indices, because they shed light on the source of disagreement. In addition, it is important to know that discrepancies between devices, as observed in correspondence indices, can mask differences in accuracy and measurement errors between devices, as well as reflect underlying average distortions that cannot be adequately modeled by absolute differences in average. That is why it is essential to examine, beyond differences of opinion, the underlying causes, in order to critically assess the results of the agreement. Rubio N, Parker RA, Drost EM, Pinnock H, Weir CJ, Hanley J, et al.

. .