Enhancing the Efficiency of Confrontation Naming Assessment for Aphasia Using Computer Adaptive Testing

Purpose: In this study, we investigated the agreement between the 175-item Philadelphia Naming Test (PNT; Roach, Schwartz, Martin, Grewal, & Brecher, 1996) and a 30-item computer adaptive PNT (PNT-CAT; Fergadiotis, Kellough, & Hula, 2015; Hula, Kellough, & Fergadiotis, 2015) created using item response theory (IRT) methods. Method: The full PNT and the PNT-CAT were administered to 47 participants with aphasia in counterbalanced order. Latent trait-naming ability estimates for the 2 PNT versions were analyzed in a Bayesian framework, and the agreement between them was evaluated using correlation and measures of constant, variable, and total error. We also evaluated the extent to which individual pairwise differences were credibly greater than 0 and whether the IRT measurement model provided an adequate indication of the precision of individual score estimates. Results: The agreement between the PNT and the PNT-CAT was strong, as indicated by high correlation (r = .95, 95% CI [.92, .97]), negligible bias, and low variable and total error. The number of statistically robust pairwise score differences did not credibly exceed the Type I error rate, and the precision of individual score estimates was reasonably well predicted by the IRT model. Discussion: The strong agreement between the full PNT and the PNT-CAT suggests that the latter is a suitable measurement of anomia in group studies. The relatively robust estimates of score precision also suggest that the PNT-CAT can be useful for the clinical assessment of anomia in individual cases. Finally, the IRT methods used to construct the PNT-CAT provide a framework for additional development to further reduce measurement error.