The importance of household composition in epidemiological analyses of sleep: Evidence from the Understanding Society longitudinal panel survey


Aims: To establish the relationship between household composition and sleep, we: 1) used latent class analysis (LCA) to classify households; 2) examined the reliability and stability of household composition classes over time; 3) conducted multinomial logistic regression analyses to determine the relationship between household class and the self-reported sleep duration and quality of adults. Methods: Data were sourced from Waves 1 and 2 of the United Kingdom “Understanding Society” (USoc) longitudinal panel survey. LCA was used to classify household composition as a categorical latent construct using data on the number and ages of household occupants and the number of rooms used for sleeping. The Bayesian Information Criterion assessed model fit and identified the optimum number of latent classes. Multi-nomial logistic regression was used to investigate cross-sectional relationships between the household classes and self-reported sleep duration and quality amongst adults, after adjustment for confounders. Results: Household composition was best defined by 7 latent classes in data from Wave 1 of USoc. This finding was confirmed in Wave 2. Compared to the reference class (households with no children and no overcrowding), there was a higher risk of short sleep (≤5 hours) versus 7-8 hours sleep for latent household composition classes that included children (RR: 1.56; 95% CI: 1.29-1.89) and for those with both children and overcrowding (RR: 1.57; 95% CI: 1.31-1.88). Similarly the risk of “very bad” versus “fairly good” quality sleep was significantly higher in those household classes with overcrowding, particularly those with extended (RR: 1.75; 95% CI: 1.34-2.29) and large (RR: 1.51; 95% CI: 1.21-1.87) households. Conclusion: These analyses of a recent, nationally representative cohort from the UK, demonstrated that latent household composition classes are reliable over time; and that these latent household composition classes are important correlates of self-reported sleep amongst adult occupants. We showed that household composition is an important contextual variable to consider in most epidemiological studies of sleep.

Share and Cite:

Fowler, H. , T. H. Ellison, G. , M. Scott, E. and R. Law, G. (2014) The importance of household composition in epidemiological analyses of sleep: Evidence from the Understanding Society longitudinal panel survey. Open Journal of Epidemiology, 4, 46-55. doi: 10.4236/ojepi.2014.41009.

Conflicts of Interest

The authors declare no conflicts of interest.


[1] Krueger, P.M. and Friedman, E.M. (2009) Sleep duration in the United States: A cross-sectional population-based study. American Journal of Epidemiology, 169, 1052-1063.
[2] Chapman, D.P., Wheaton, A.G., Perry, G.S., Sturgis, S.L., Strine, T.W. and Croft, J.B. (2012) Household demo-graphics and perceived insufficient sleep among US adults. Journal of Community Health, 37, 344-349.
[3] Groeger, J.A., Zijlstra, F.R.H. and Dijk, D.J. (2004) Sleep quantity, sleep difficulties and their perceived consequences in a representative sample of some 2000 British adults. Journal of Sleep Research, 13, 359-371.
[4] Burgard, S.A. and Ailshire, J.A. (2013) Gender and time for sleep among US adults. American Sociological Review, 78, 51-69.
[5] University of Essex, Institute for Social and Economic Research and National Centre for Social Research (2012) Understanding society: Waves 1-2, 2009-2011 (UKDA-6614-stata12). 4th Edition, UK Data Archive SN, Colchester, 6614.
[6] Department for Communities and Local Government (2003) Housing (overcrowding) bill. HMSO, London.
[7] Law, G.R., Head, R.F. and Ellison, G.T.H. (2012) Confounding and causal path diagrams. In: Tu, Y.K. and Greenwood, D., Eds., Modern Methods for Epidemiology, Springer, London, Chapter 1, 1-14.
[8] Textor, J., Hardt, J. and Knuppel, S. (2011) DAGitty: A graphical tool for analysing causal diagrams. Epidemiology, 22, 745.
[9] Greenland, S., Pearl, J. and Robins, J.M. (1999) Causal diagrams for epidemiologic research. Epidemiology, 10, 37-48.
[10] Pearl, J. (1993) Comment: graphical models, causality and intervention. Statistical Science, 8, 266-2669.
[11] Tu, Y.K., West, R., Ellison, G.T.H. and Gilthorpe, M.S. (2005) Why evidence for the fetal origins of adult disease might be a statistical artefact: The “reversal paradox” for the relation between birth weight and blood pressure in later life. American Journal of Epidemiology, 161, 27-32.
[12] Law, G.R. and Pascoe, S.W. (2013) Statistical Epidemiology. CABI, Oxfordshire.
[13] Statistical Innovations Inc. (2003) Latent gold software: Release 4.5. Belmont.
[14] Vermut, J.K. and Magidson, J. (2005) Latent GOLD 4.0 user’s guide. Statistical Innovations Inc., Belmont.
[15] Schwartz, G. (1978) Estimating the dimenstions of a model. Annals of Statistics, 6, 461-464.
[16] Vermut, J.K. and Magidson, J. (2005b) Technical guide for latent GOLD 4.0. Statistical Innovations Inc., Belmont.
[17] STATACORP (2012) Stata statistical software: Release 12.1. Stata Press, College Station.

Copyright © 2023 by authors and Scientific Research Publishing Inc.

Creative Commons License

This work and the related PDF file are licensed under a Creative Commons Attribution 4.0 International License.