Psychometric assessment of the 10-item, revised experience of close relationship (ECR-R-10) in nonclinical and clinical populations of adults and older adults in Thailand

Abstract

AbstractThe experiences of close relationships-revised (ECR-R) is a widely used 36-item self-report measurement for measuring adult attachment. However, various short versions of the ECR-R have been developed and tested psychometrically. Given the cultural impact, a short version of the Thai ECR-R should be derived from the existing Thai version of the ECR-R. This study aimed to develop a 10-item version of the ECR-R that demonstrates comparable psychometric properties to the previous Thai version and the 18-item ECR-R. This study included four studies with a total of 1,322 participants. In study 1, 434 adults in a nonclinical setting were used for the development of the 10-item Thai ECR-R and tested in an independent sample. Studies 2, 3, and 4 were conducted on 312 adults in the clinical setting, 227 older adults in the nonclinical, and 123 older adults in clinical settings. The Cronbach alphas and corrected correlations between the ECR-R-18 and the ECR-R-10 in each study were calculated. Confirmatory factor analysis of the first-order two-factor solution model with fit statistics was examined with each sample. Correlations of the ECR-R-18 and the ECR-R-10 with other measurements were presented and compared. Known-group validity and measurement invariance test were also examined. The Cronbach alphas of the ECR-R-10 among all samples were acceptable, ranging between .77 and .85 for avoidance subscales and between .82 and .86 for anxiety subscales. The corrected correlation between the ECR-R-18 and ECR-R-10 was between .61 (p < .001) and .82 (p < .001). The values of the comparative fit index and Tucker-Lewis index for the model of ECR-R-10 were between .903 and .985, whereas the root-mean-square error of approximation was between .082 and .036, indicating that the model fits were acceptable. The ECR-R-10 was related to the measurements with a similar construct; however, no difference in the magnitude of correlation was observed between ECR-R-18 and ECR-R-10. Known group validity was established. Measurement invariance was successfully established across different age and gender groups, although it was only partially achieved with respect to clinical status. The ECR-R-10 provided equal or superior psychometric properties to the ECR-R-18 across age groups and settings. As it is a briefer scale, the ECR-R-10 can be practically used in general and clinical samples to reduce the burden of assessment, especially with older adults. Further investigation is needed to test the scale's temporal stability.