A Rasch analysis of the Conley Scale in patients admitted to a general hospital

Purpose: The Conley Scale (CS) is a widely used tool for assessing the risk of falling for inpatients. The purpose of this study was to assess its unidimensionality, internal construct validity, targeting and reliability using Confirmatory Factor Analysis (CFA) and Rasch analysis (RA). Methods: The CS was administrated to a sample of 58,370 subjects admitted to a general hospital. Results: The CFA supported the unidimensionality of the CS (Root Mean Square Error of Approximation (RMSEA)?=?0.040) only after adjusting for local dependency between two items. The scale did not fit the Rasch model (?(2)(18)?=?4688.5; p?=?0.0000) and this was confirmed notwithstanding adjusting for type-I error (by creating 10-subsample of 250 subjects) and extensive post-hoc modifications. The analysis of targeting showed a marked floor effect (47.1%), whereas the reliability appeared adequate for group measurement (0.800) only after adjusting for the skewed distribution of the calibrating sample. Conclusion: The results of this study suggested that the CS, although unidimensional, could not provide interval-scale measurement of the risk of falling, had a measurement range that mismatched the ability range of the population being measured, and had a reliability inadequate for individual person measurements. Given these findings, the use of the CS to identify inpatients at risk of falling is not recommended.