An in-depth psychometric analysis of the Connor-Davidson Resilience Scale: calibration with Rasch-Andrich model

Abstract

Background: The Connor-Davidson Resilience Scale (CD-RISC) is inarguably one of the best-known instruments in the field of resilience assessment. However, the criteria for the psychometric quality of the instrument were based only on classical test theory.

Method: The aim of this paper has focused on the calibration of the CD-RISC with a nonclinical sample of 444 adults using the Rasch-Andrich Rating Scale Model, in order to clarify its structure and analyze its psychometric properties at the level of item.

Results: Two items showed misfit to the model and were eliminated. The remaining 22 items form basically a unidimensional scale. The CD-RISC has good psychometric properties. The fit of both the items and the persons to the Rasch model was good, and the response categories were functioning properly. Two of the items showed differential item functioning.

Conclusions: The CD-RISC has an obvious ceiling effect, which suggests to include more difficult items in future versions of the scale.

Fig. 1

Distributions of items in function of infit mean square (left) and outfit mean square (right) values. The diameter of a bubble represents the magnitude of standard error. The shadow depicts the range of optimal fit. Items 3, 9 and 20 were suppressed

Fig. 2

Wright Map (persons and items distributions). M = mean; 1S = standard deviation; 2S = Two standard deviations

Fig. 3

Residual saturation (first contrast)

Fig. 4

Category probability curves. Vertical lines indicate Rasch-Andrich thresholds (the points at which adjacent categories are equally probable). Parenthesis represents standard error

Fig. 5

Test Information Function. The curve represents the information provided by complete test (information) at different levels of the latent variable (measure)