Article Text

Download PDFPDF
PP65 Simplifying the measurement of sleep quality: latent variable analysis of seven conceptual sleep criteria
  1. AA Alghamdi,
  2. EM Scott,
  3. GR Law,
  4. GTH Ellison
  1. Epidemiology and Biostatistic, University of Leeds, Leeds, UK


Background In epidemiological surveys, sleep quality is assessed using seven distinct criteria namely: sleep duration; latency; disturbance; coughing/snoring; perceived quality; medication and daytime sleepiness. Due to the complex functional, conceptual and mathematical interrelationships between these seven criteria epidemiologists face substantial challenges assessing which and whether/how best to combine two or more criteria. The aim of the present analyses was therefore to assess whether it is possible to reduce the number of sleep criteria used to provide a simpler indication of ‘overall sleep quality’.

Methods Self-reported sleep data,based on responses to seven items in the first wave adult self-completion questionnaire of the UK’s Understanding Society (Usoc) longitudinal study, were analysed using exploratory factor analysis (EFA) which was built based on polychromic correlation matrix of the seven sleep criteria. A value of 0.4 was chosen as the minimum acceptable loading factor to include in the ‘overall sleep quality’ model. The model was tested using confirmatory factor analysis (CFA) with an asymptomatically distribution-free estimation (ADF) as the chosen method of estimation.

Results When ‘overall sleep quality’ was considered as one continuous variable, most of its variance was explained by just four of sleep criteria: duration, latency, disturbance and perceived quality, for which the EFA loading factors were: 0.430; 0.685; 0.702; and 0.890, respectively. The associated CFA regression standardised beta coefficients were; -0.237 (SE=0.005; p < 0.001) for duration; 0.791 (SE=0.0195; p < 0.001) for latency and 0.760 (SE=0.195; p < 0.001) for perceived quality. The model fitting indices being: root mean squared error of approximation < 0.001; comparative fit index = 1.00; and Tucker-Lewis index = 1.00.

Conclusion Amongst adult Usoc participants (a sample that is broadly representative of the UK population), variation in ‘overall sleep quality’ is principally captured by just four of sleep criteria measured in its first wave questionnaire: sleep latency; sleep disturbance; and perceived sleep quality, while sleep duration plays an important role in predicting participants’ individual subjective assessments of sleep quality. These four of Pittsburgh Sleep Quality Index’s seven key subscores, are likely to provide an adequate marker of ‘overall sleep quality’ for use in epidemiological analyses, particularly those in which sleep is considered rather than an exposure or outcome of interest.

  • sleep latent variable

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.