Background/Purpose: Loneliness is a pervasive experience that has been associated with poorer health-related quality of life, and remains understudied in patients with systemic sclerosis (SSc). This study examined the psychometric properties of scores from the UCLA Loneliness Scale-6 (ULS-6; Neto, 1992), a self-report questionnaire not previously evaluated for SSc, and evaluated the equivalence of English and French versions.

Methods: This study used baseline data from 775 adults with SSc (89.9% female, 52.9% limited SSc, 66% English-speaking) enrolled in the Scleroderma Patient-Centered Intervention Network (SPIN) Cohort. Confirmatory factor analysis (CFA) was used to evaluate the single-factor structure of the ULS-6. Internal consistency reliability was assessed via McDonald’s omega and Cronbach’s coefficient alpha, and convergent validity was examined via Pearson product-moment correlations with measures of depression symptoms (PHQ-8) and social support (OSSS-3), number of people in the household, and social interaction frequency. Multiple group CFA (MGCFA) was used to determine whether the single-factor structure was supported in English and French-speaking subgroups. The MIMIC model was used to examine differential item functioning (DIF) for English versus French versions.

Results: The mean score on the ULS-6 was 7.00 (SD = 4.76, range 0-18), with a higher score representing greater loneliness. The CFA supported the expected single-factor structure (CFI = .96, SRMR = .03), although the RMSEA = .11 suggested less than adequate model fit. All standardized factor loadings for items were large and statistically significant (.62 to .86, all ps < .001). Omega and alpha were both 0.87, indicating high reliability. The ULS-6 total score correlated significantly (all ps < .001) and in expected directions with the total score for the PHQ-8 (r = .56), the total score on the OSSS (r = -.53), and frequency of social interactions with one person (r = -.162) and multiple people (r = -.24), but did not correlate significantly with number of people in the household (r = -.058, p = .11). The MGCFA indicated that the configural invariance model best fit the data (Δχ²[8] = 56.14, p < .001), suggesting non-equivalent factor loadings between language subgroups. However, unstandardized (.72 to 1.39, all ps < .001) and standardized (.48 to .90, all ps < .001) factor loadings were positive and statistically significant. The overall MIMIC model fit well and suggested cross-language metric equivalence (CFI = .939, SRMR = .044, RMSEA = .11). Statistically significant DIF were found for three items across language, although standardized differences were small (β(item 2) = .14, p < .001, β(item 4) = -.14, p = .005, β(item 6) = .13, p < .001).

Conclusion: Analyses demonstrated acceptable reliability and validity of the ULS-6 scores in English- and French-speaking adults with SSc, supporting its use to examine comparative experiences of loneliness. Although half of the items showed DIF across linguistic groups and the MGCFA did not support factor loading invariance, there were few practical implications for the total scores derived. Future studies of the ULS-6 across linguistic groups are needed to establish invariance for use with diverse SSc populations.

« Back to ACR Convergence 2021

ACR Meeting Abstracts - https://acrabstracts.org/abstract/psychometric-evaluation-and-measurement-invariance-of-english-and-french-versions-of-the-ucla-loneliness-scale-in-systemic-sclerosis-a-scleroderma-patient-centered-intervention-network-cohort-study/