Objective: to assess whether compositional (education, income index, number of family members) and contextual (area socioeconomic index) risk factors independently predict all cause and specific mortality.
Design: a multilevel (hierarchic) logistic regression model was applied to the individual data of a cohort followed up from 01.01.2002 till 31.12.2007.
Setting and participants: the study evaluated 40-79 years old people resident in Palermo at 01.01.2002, for whom it was possible to match register office and census data (220,723 people, 74.8% of the same age group total population).
Main outcomes measures: odds ratios for specific risk factors.
Results: mortality was generally lower in people with better socioeconomic conditions and living in more affluent neighbourhoods. Individual risk factors odds ratios do not vary in models with and without area related risk factor. Variance partition component and other between area and total variability ratio index show small values.
Conclusion: compositional and contextual socioeconomic factors are independent predictors of mortality; area related variability is only a small fraction of total variability.