Document resource
Introduction To compare the influence of household and neighbourhood factors on the risk of four definitions of multimorbidity (MM), namely (1) patients with two or more long-term conditions (LTCs) (2+MM), (2) patients with three or more LTCs and (3) patients with three or more LTCs from three or more International Classification of Diseases (10th Revision) body systems and (4) 2+MM patients with at least one mental LTC and one physical LTC.Methods We used individual-level census data of Wales and health episode records from the Secured Anonymised Information Linkage databank. The cohort was all living people registered with a Welsh general practice on the 2011 census date (27 March) (n=1 676 432). Three-level multilevel logistic regression models with individuals in households in neighbourhoods were applied to each MM definition.Results Markers of socioeconomic disadvantage at the household (eg, tenure of dwelling) and neighbourhood (eg, neighbourhood deprivation) levels were associated with higher risks of MM across all definitions. Relative to urban cities and towns, rural village settings were associated with lower odds of all MM definitions in the fully adjusted models. Our partitioning of the variance in MM risk into level-wise components revealed more variance in households (between 23% and 12%) than in neighbourhoods (approximately 3%) and a sensitivity of the results to the definition of MM used. Household attributes accounted for a greater proportion of R 2 in mental–physical MM (23%) than in the other definitions (8.5% to 9.5%), which is in line with research that has demonstrated a ‘social contagion’ of mental health conditions within households.Conclusions Pertinent household and neighbourhood factors should be prioritised for targeted public interventions, and optimal strategies might vary according to definitions of MM employed, with mental-physical MM showing a different pattern from the other definitions, which have a similar pattern.