, Volume 45, Issue 5, pp 381-384

Internal consistency reliability of the fractionated and whole University of Pennsylvania Smell Identification Test

Abstract

The internal consistency reliability (ICR) of the 40-item University of Pennsylvania Smell Identification Test (UPSIT) and its 10-, 20-, and 30-item fractions was explored, as well as the relationships between the fractions and the entire 40-item test. Pearson correlation coefficients (rs) were computed among all independent combinations and permutations of the four 10-item UPSIT booklets using data from 774 subjects. The median r values of the 10- and 20-item combinations were used to establish the ICRs of the 10- and 20-item tests. The ICRs of the 30- and 40-item tests were estimated using the Spearman-Brown formula and the medianrs of the 20-item combinations. Additional ICR estimates of the 40-item UPSIT were obtained from nonsymmetrical fractions using the Horst formula. The ICRs for the UPSIT and its 10-, 20-, and 30-item fractions were 0.922, 0.752, 0.855 and 0.898, respectively. No major sex differences emerged. Estimates of correlations between (1) single booklets and two-booklet combinations and (2) the 40-item UPSIT using Guilford’s (1953) correction for nonindependence ranged from 0.812 to 0.871. Overall, these results indicate that (1) the UPSIT and its 10-, 20-, and 30-item fragments have very high ICRs and (2) individual UPSIT booklets or their combinations can be used to assess smell function in a reliable manner where extreme time constraints are present (e.g., in surveys and in brief neuropsychological test batteries).

This work was supported by Grant NS 16365 from the National Institute of Neurological and Communicative Disorders and Stroke