Construct Validity of Scores From the Connor–Davidson Resilience Scale in a Sample of Postsecondary Students With Disabilities

Abstract

Although theory posits a multidimensional structure of resilience, studies have supported a unidimensional solution for data obtained from the commonly used Connor–Davidson Resilience Scale (CD-RISC). This study investigated the latent structure of CD-RISC responses in a sample of postsecondary students with disabilities. Furthermore, the validity of CD-RISC scores was examined with respect to career optimism and well-being. The analyses were conducted using confirmatory factor analysis and exploratory structural equation modeling (ESEM). Results supported a bifactor-ESEM representation of the CD-RISC data that accounts for construct-relevant multidimensionality in scores due to the presence of general and specific factors and the fallibility of indicators as pure reflections of the constructs they measure. Although three specific factors showed meaningful residual specificity over and above the general factor, two specific factors were weakly defined with little meaningful residual specificity. However, these factors may retain some utility in the bifactor-ESEM model insofar as they control for limited levels of residual covariance in items. Evidence was also obtained for relations of the general and substantively interpretable specific factors with career optimism and well-being. The results of the study provide validation data for the CD-RISC and clarify recent research converging on seemingly disparate unidimensional and multidimensional solutions.

Resolving Dimensionality Problems With WHOQOL-BREF Item Responses

“…Although the CFF-ESEM fitted the data well and produced theoretically plausible factor correlations, loading estimates were suggestive of the presence of a second source of construct-relevant multidimensionality in item responses. Across all four factors, there were large numbers of sizeable secondary loadings, sometimes exceeding the magnitude of target loadings, which is suggestive of the presence of a general construct underlying the data (Morin, Arens, Tran, et al, 2015; Perera, 2015a; Perera & Ganguly, 2016). Although the CFF-ESEM accounts for construct-relevant multidimensionality due to indicator fallibility, the structure cannot sufficiently control for psychometric multidimensionality attributable to the presence of general and specific constructs underlying item responses.…”

“…For the CFF-ESEM model, even though the factor correlation estimates were theoretically consistent, across the four factors there were 12 instances of sizeable (>.250) cross-loadings. This pattern of large cross-loadings may emerge in correlated first-order ESEM structures where some general factor is unmodeled (Morin, Arens, & Marsh, 2015;Perera, 2015a;Perera & Ganguly, 2016). A BF-ESEM model can accommodate this psychometric multidimensionality due to the putative presence of a general construct underlying all item scores in addition to systematic residual specificity in item sets captured by specific or group factors.…”

Examination of Transition Readiness, Medication Adherence, and Resilience in Pediatric Chronic Illness Populations: A Pilot Study

“…Total scores range from 0 to 100 with higher scores reflecting greater resiliency. In the general population, the mean of the CD-RISC is 80 with an interquartile range of 73-90 [30][31][32][33]. This measure has good reliability and validity and has been validated in adolescents and young adults in a number of settings [30][31][32][33].…”

Propiedades psicométricas de la Escala de Resiliencia de Connor y Davidson (CD-RISC) en población española.

“…However, despite there being a great number of studies on the psychometric properties of the CD-RISC, there is still a great lack of consensus on the internal structure of the scale, since most studies in the literature reveal different factorial structures. Empirical evidence has supported the onedimensional model (Arias-Gonzalez, Crespo-Sierra, Arias-Martinez, Martinez-Molina & Ponce, 2015;Burns and Anstey, 2010;Campbell-Sills & Stein, 2007;Gucciardi, Jackson, Coulter & Mallett, 2011;Notario-Pacheco et al, 2011;Ponce-Cisternas, 2015;Sarubin et al, 2015), the twodimensional model (Fu, Leoutsakos & Underwood, 2013;Green et al, 2014;Jorgensen & Seedat, 2008;Perera & Ganguly, 2016), the three-dimensional model (Karairmak, 2010;Mealer, Schmiege & Meek 2016;Menezes de Lucena et al, 2006;Serrano-Parra et al, 2012;Xie, Peng, Zuo & Li, 2016;Yu & Zhang, 2007), the four-dimensional model (Crespo et al, 2014;Khoshouei, 2009;Lamond et al, 2008;Singh & Yu, 2010;Solano et al, 2016), the five-dimensional model (Fujikawa et al, 2013;Gillespie, Chaboyer, Wallis & Grimbeek, 2007;Jung et al, 2012;Manzano-García & Ayala-Calvo, 2013) and the second-order model (Yu et al, 2011). According to the above list, one could believe that the CD-RISC presents a different factorial configuration between studies, countries or sample types and therefore it would lead one to further believe that, in each case, one is measuring different constructs.…”