Using pooled kappa to summarize interrater agreement across many items

We propose the pooled estimator of kappa, an efficient estimator when summarizing the interrater agreement for qualitative data with many items but few subjects. We evaluate this estimator through a simulation of proposed and alternative (average kappa) estimators and subsequently apply our method to calculate pooled and average kappas over 2,176 rated items from six semistructured interviews with sponsors of the CAHPS. The proposed pooled kappa estimator efficiently summarizes interrater agreement by domain. It is more widely applicable and makes better use of scarce subjects than simply averaging item-level kappas.