Mortality effects of average education: a multilevel study of small neighbourhoods in rural and urban areas in Norway

BACKGROUND: The intention was to find out whether there was an association between the socio-economic resources in a small neighbourhood ("basic statistical unit"; BSU) and individual mortality, net of individual resources, and whether this association differed between municipalities including a quite large city and others. The possibility of a rural-urban difference in the health effect of community resources has not been checked earlier. METHODS: Discrete-time hazard models for mortality at age 60-89 were estimated for 1990-1992 and 2000-2002, using register data that cover the entire Norwegian population. For each person, the educational level and the municipality and BSU of residence in 1990 and 2000 were known. Average education was computed by aggregating over the individual data. In total, there were about 200000 deaths in more than 13000 BSUs during 5 million person-years of observation. RESULTS: There was a significant relationship between average education in the BSU and individual mortality, but only in the medium-sized and largest municipalities. The sharpest relationship was seen in the latter, where for example OR per year of education was 0.908 (95% CI 0.887-0.929) in the 1990-92 period. The findings were robust to various alternative specifications. CONCLUSION: These results from a large data set are consistent with the idea that neighbourhood socio-economic resources may affect individual mortality, but suggest that distinctions according to population size or density be made in future research and that one should be careful, if focusing on cities, to generalize beyond that setting. With these data, one can only speculate about the reasons for the rural-urban difference. A stronger higher-level spatial segregation in urban areas may be one explanation.