Who you live with and where you live: setting the context for health using multiple membership multilevel models.

STUDY OBJECTIVE Previous studies into the effect of area of residence on individuals' health have not accounted for changing residency over time, although few people remain resident in the same area throughout their life. Furthermore, few studies of area effects on health have accounted for the clustering of health at the household level. These methodological problems may have led previous studies to under estimate or over estimate the size of area level effects. This study uses multiple membership multilevel models to investigate whether longitudinal analyses of area effects on health need to take account of clustering at the household level. SETTING AND PARTICIPANTS A longitudinal survey (1991-1999) of a nationally representative sample of British households (5511 households with 10 264 adult members). DESIGN Two level (individuals within households or areas) and three level (individuals within households within areas) multiple membership models of SF-36 physical and mental health functioning scores at wave nine were analysed adjusting for age, gender, education, marital, employment, and smoking status from previous waves. RESULTS Physical and mental health functioning seem to cluster within households. Accounting for changes in household membership over time increases estimates of the clustering in functioning at the household level. The clustering of functioning within area wards is reduced when the clustering within households and risk factors for functioning are taken into account. CONCLUSIONS Clustered sampling units within study designs should be taken account of in individual level analyses. Changes in these units over time should be accounted for in longitudinal analysis.