Psychometric Evaluation of the Comprehensive Score for Financial Toxicity Scale Among Iranian Cancer Patients
Thoughts
Connects with: @desouza2014 @zafar2013
Annotations
sharif2020 - p. 4
. EFA was conducted on the rst dataset and CFA was performed on the second dataset to avoid over tting. Table 1 reports the results of conducting EFA with promax oblique rotation and CFA on the Persian version of COST. The KMO value was .791 and Bartlett’s sphericity test was signi cant (p < .001, 1050.172, df = 45) indicating that sampling was adequate. Using the eigenvalue greater than one and scree plot criteria EFA extracted three factors accounting for 65.204% of the variance. The rst factor was labelled as nancial worry (comprising items 1, 3, 4, and 5; eigenvalue = 3.188), the second factor was labelled as nancial control/satisfaction (comprising items 2, 6, and 7; eigenvalue = 2.367), and the third factor was labelled as direct and indirect costs (comprising items 9, 10, and 11; eigenvalue = 2.686). The eighth item of COST was removed due to a high cross-loading
sharif2020 - p. 5
The results of performing CFA on the factor structure obtained with EFA consisting of three dimensions and 10 items indicated a good t (χ2 (df = 32, N = 199) = 64.176, p < .01; χ2/df = 2.006, comparative t index (CFI) = .966, Tucker-Lewis index (TLI) = .953, Incremental Fit Index (IFI) = .967, Relative Fit Index (RFI) = .910, Normed Fit Index (NFI) = .936, Standardized Root Mean Square Residual (SRMR) = .051 and RMSEA = .071 (90% con dence interval = .046–.096)). All item loadings were greater than .7 (varying from .729 to .836) and statistically signi cant at .001. Both composite reliability ( nancial worry: .889, nancial control: .814, and direct and indirect costs concerns: .837) and maximal reliability ( nancial worry: .930, nancial control: .947, and direct and indirect costs concerns: .837) of the three dimensions were greater than .7 indicating good construct reliability. Also, average variance extracted for nancial worry (.661), nancial control (.595), and direct and indirect costs concerns (.630) was greater than .5 that established convergent validity of the dimensions. Moreover, average variance extracted of each dimension was greater than its maximum shared variance ( nancial worry: .212, nancial control: .216, and direct and indirect costs concerns: .216) with other dimensions ful lling the requirements of discriminant validity. Figure 1 shows the measurement model with the results of CFA.