The maximum sum score is 30 points, selleck chemicals Vandetanib indicating excellent cognitive function. A sum score of less than or equal to 23 was chosen as the cut-off for cognitive impairment [20]. Statistical analysis The kappa statistic (��) for multiple raters and its 95% confidence interval (CI) and the proportion of observed agreement were chosen as the measure for interrater agreement [21,22]. If there is complete agreement among the assessors, then �� = 1. If observed agreement is greater than or equal to chance expected agreement then �� �� 0. If observed agreement is less than or equal to chance expected agreement, then �� �� 0. The classification of Landis and Koch [23] was used for the interpretation of the relative strength of agreement associated with kappa statistics.

The following labels were assigned to the corresponding ranges of ��: < 0.00 poor; 0.00-0.20 slight; 0.21-0.40 fair; 0.41-0.60 moderate; 0.61-80 substantial; 0.81-1.00 almost perfect. We calculated the kappa statistic (and the 95% CI) with Microsoft? Excel software. Confidence intervals for kappas were calculated using the standard error. The extent of the agreement among the raters concerning each individual subject was calculated separately over all BES items and over all AGGIR items, using the formula of Siegel [21]. Paradoxes in �� values can be due to differences between two samples in the prevalence of an attribute [24-26]. Two samples can have the same proportion of agreement on a condition between raters but if the prevalence of that condition is higher in one sample and almost all ratings will fall into one category, then �� will typically be lower.

This paradoxical difference of the �� values arises because of the decision to impose a correction for chance agreement, making the assumption that the expected values for agreement should depend on the marginal totals. Since no assumptions are made about the marginal totals, two observers can get low values for �� despite a high percentage of observed agreement [24]. It must be emphasized that in the examples that are given in the publications mentioned above [24-26], the prevalence effect on the �� value demonstrated the effect of the imbalance of marginal totals of two response categories. In the present study, the items of the assessment instruments had three to five response categories.

It is obvious that if there is an imbalance in the marginal totals Cilengitide of multiple response categories, then the �� value will also typically be lower. In the present study, the proportion observed agreement is always presented next to the �� value in order to assess whether low �� value was due to low interobserver agreement or to the prevalence effect. If �� is low but the proportion agreement is high then it might be concluded that the measurement might to some extent be reliable.