Latent class analysis reveals six distinct sleeping patterns that are associated with key sociodemographic and health characteristics in both Wave 1 and Wave 4 of Understanding Society.

Presenter: Amal Alghamdi, University of Leeds

Author: Amal Alghamdi

Co-author(s): Graham Law, Eleanor Scott and George Ellison

As a multifactorial trait, sleep is challenging to operationalise. We assessed whether latent class analysis (LCA) might identify distinct sleeping patterns that capture the complexity of sleep and are meaningfully associated with socio-demographic and health characteristics.

Data for Understanding Society’s seven sleep variables (duration, latency, disturbance, medication, snoring/coughing, quality and daytime sleepiness), collected from respondents with complete data in Waves 1 and 4, were subjected to LCA using Latent Gold (Statistical Innovations, MA). LCA model fit was assessed using the Log-Likelihood Bayesian/Akaike Information Criteria and classification error parameters. Latent classes were interpreted according to their association with the seven sleep variables; and with age, gender, education, employment, household composition and subjective health.

The best fitting LCA models identified six latent sleep classes in both waves, each class containing 6.5-31.6% of respondents. The distribution of the seven sleep variables across these six sleep classes suggested the latter might be described as: ‘long good sleepers’; ‘long moderate sleepers’; ‘snoring good sleepers’; ‘snoring bad sleepers’; ‘short bad sleepers’; ‘struggle to sleepers’. These classes were significantly associated with all of the socio-demographic/health variables. For example: a disproportionate number of employed, healthy, well-educated females living in a couple with children were ‘good long sleepers’; while a disproportionate number of employed, well-educated healthy males living in a couple with children were ‘good snoring sleepers’.

Latent class analysis revealed six distinct sleeping patterns that are associated with key socio-demographic and health characteristics and appear stable across Waves 1 and 4 of Understanding Society.