Longitudinal assessment of depression during the first year after stroke: Dimensionality and measurement invariance

Abstract

Objective Repeated measurement of depression using summative scores of commonly used depression scales assumes that the construct is unidimensional and invariant across time, but little is known about the validity of these assumptions in stroke patients. The study examined the assumptions during the first year after mild to moderate stroke using the 8-item Patient Health Questionnaire (PHQ-8). Methods Data were obtained from the 2014–2016 Brain Attack Surveillance in Corpus Christi project, a population-based stroke surveillance study with longitudinal follow-up in Texas, United States. Depressive symptoms were assessed with the PHQ-8 at 3, 6, and 12 months post-stroke (n = 648, 542, 533, respectively). Dimensionality of the PHQ-8 was examined using exploratory bifactor analysis and confirmatory factor analysis (CFA). Based on a one-factor CFA model, longitudinal measurement invariance was examined by fitting configural, weak, strong and strict invariance models sequentially. Results Mean age was 65.6 (SD = 11.2), half were women, and 56.7% were Mexican American. Reliability measures suggested a generally unidimensional construct (Omega Hierarchical: 0.79–0.86), good internal consistency (Cronbach's alpha: 0.84–0.88) and good temporal stability (intraclass correlation coefficient: 0.838). Configural invariance for equality of item-factor structure (RMSEA = 0.038, CFI = 0.982, WRMR = 0.970) was supported. Partial weak invariance for equality of factor loadings (p = 0.11) was supported by chi-square difference tests, whereas full longitudinal invariance was supported by changes in comparative fit index and root mean square error of approximation. Conclusions The findings suggest the PHQ-8 may be a reasonable tool for repeated measurement of depression in stroke patients, and more research on its longitudinal psychometric properties is needed.