Agreement of measurements and applications in medicine

Reproducibility of the measurements in medical studies has an important role both in reducing the error of experiment and in comparison of the results with results of similar research. Measurement of agreement is the criterion of the consistency between two clinicians? ratings of a patient or consistency of a clinician?s ratings of a patient over a number of occasionsAssesment of the agreement for different raters or repeated measurements by the same raters, depending on whether measurements are categorical or continuous require use of different measures. This study summarizes the measures of basic and general agreement indices such as p0, ps, kappa, weighted kappa, Fleiss kappa, Kendall?s W, Goodman and Kruskal?s ?, Kendall?s ?, ICC and Krippendorff?s alpha and evaluates them by using data from literature. Additionally data from a study conducted in the department of dermatology at the medical school of Cukurova University is used in the evaluation and comparison of different measures of agreement.Alternative measures of agreement were applied on and compared for data from literature. Comparisons of these measures are made depending on the number of raters and types of variables.The study concludes with three main findings: Kendall W does not reflect true agreement in some cases, because it uses rank of measurements instead of the originals. Thus calculating Krippendorff alfa and comparing it with Kendall W are recommended. For ratings made on nominal, ordinal or continuous scale for multiple raters with missing data, Krippendorff alfa has the advantage that it uses information from individuals with missing ratings. And lastly, the association measures Goodman and Kruskal ?, Kendall ?, Pearson and Spearman correlation coefficient that should be used with care becuause they may not reflect true agreement.