In order to test the hypothesis that agreement was most limited by a few outlier subjects, a sensitivity analysis was performed separately for each scale, excluding three residents with the lowest and the highest mean extent of agreement between raters http://www.selleckchem.com/products/U0126.html on the item scores of each scale separately and recalculating the kappa estimates for total scale scores of the remaining 31 residents. Graphical analysis was used for comparing interobserver reliability of subjects with and without cognitive impairment for each instrument and in each study condition separately: the pretest intervention group, the pretest control group, the posttest intervention and the posttest control group.

Results Total scale scores At baseline, all kappas referring to the agreement on total scale scores for BES and AGGIR indicated moderate agreement and were not significantly different between the intervention and the control group (Table (Table2).2). At the second assessment, all kappas referring to the agreement on total scale scores were higher than before the intervention, but the agreement was not significantly different between the second and the first assessment for BES and AGGIR total scale scores in both the intervention and the control condition. Table 2 Kappa (��) and its 95% confidence limits and the proportion observed agreement as measures of agreement between multiple raters about Belgian Evaluation Scale assessments and AGGIR assessments Item scores At baseline, agreement on BES and AGGIR item scores was not different between the intervention and the control group: e.g.

for BES washing, �� = 0.43 (95% CI 0.37-0.49) in the intervention group and �� = 0.44 (95% CI 0.38-0.50) in the control group (Table (Table3).3). At the second assessment, only the kappa for the BES item washing (�� = 0.63 [95% CI 0.56-0.70]), was significantly higher than the kappa for washing of the first assessment (�� = 0.44 [95% CI 0.38-0.50]). All other item kappas were not significantly different between two assessments, for either the intervention group or the control group (Table (Table33). Sensitivity analysis After excluding the three residents with the lowest or the highest mean extent of agreement per instrument from the analysis, kappas referring to the agreement on total scale scores were higher and lower respectively, and were not significantly different between the intervention and the control group (data not shown in this paper).

All other trends were similar to trends for the agreement in the total group. Graphical analysis In Figure Figure2,2, the proportions observed agreement (vertical bars) and kappa values with its 95% confidence intervals are represented in adjacent graphs for residents with cognitive impairment (�� 23) and without cognitive impairment (> 23) and for each study condition (intervention group Drug_discovery and control group; first and second assessment).