Abstract
The Pittsburgh Sleep Quality Index (PSQI) is a widely used measure for assessing sleep impairment. Although it was developed as a unidimensional instrument, there is much debate that it contains multidimensional latent constructs. This study aims to investigate the dimensionality of the underlying factor structure of the PSQI in a multi-ethnic working population in Singapore. The PSQI was administered on three occasions (baseline, 3 months and 12 months) to full-time employees participating in a workplace cohort study. Exploratory factor analysis (EFA) investigated the latent factor structure of the scale at each timepoint. Confirmatory factor analysis (CFA) evaluated the model identified by EFA, and additionally evaluated it against a single factor and a three-factor model. The EFA identified a two-factor model with similar internal consistency and goodness-of-fit across each timepoint. In the CFA, the two- and three-factor models were both superior to the unidimensional model. The two- and three-factor models of the PSQI were reliable, consistent and provided similar goodness-of-fit over time, and both models were superior to the unidimensional measure. We recommend using the two-factor model to assess sleep characteristics in working populations in Singapore, given that it performs as well as the three-factor model and is simpler compared to the latter.