Geographical clustering of mortality from systemic lupus erythematosus in the United States: contributions of poverty, Hispanic ethnicity and solar radiation.

The objective was to investigate whether spatial variation in poverty, Hispanic ethnicity, and solar radiation explains the strong pattern of geographical clustering of mortality from systemic lupus erythematosus (SLE) in the United States. SLE mortality counts for women and men of black and white race in US counties, 1979-1998, were obtained from the National Center for Health Statistics. County-level poverty rates and proportions of Hispanic residents were drawn from the 1990 national census. The annual cumulative level of ambient ultraviolet 'B' radiation (UVB) was estimated for each county according to latitude, longitude, and elevation. Maps for the full study population and for sex - and race-specific subpopulations showed that the national pattern of geographical variation in SLE mortality primarily reflected the experience of white women. Formal spatial analysis of the data for white women identified 10 statistically significant, multi-county clusters--four with elevated and six with reduced SLE mortality rates. Multivariate regression modeling established that higher levels of poverty, Hispanic ethnicity, and UVB were each associated with elevated local rates of SLE mortality among white women. Statistical adjustment via the regression model was used to remove effects of these factors on local rates. In a re-application of spatial analysis to the adjusted rates, four clusters 'disappeared'. In those clusters, poverty, Hispanic ethnicity and UVB had explained an average of 58.2% of the deviations between local and national SLE mortality rates. In six clusters (including three that disappeared with adjustment), Hispanic ethnicity explained a larger percentage of the deviations between local and national rates than either poverty or UVB. In multivariate models based on data for black women and for men of both races, poverty and UVB had similar effects on SLE mortality rates to those observed among white women. However, Hispanic ethnicity was not a significant predictor of SLE mortality in any of these other groups. The contributions of poverty, Hispanic ethnicity and solar radiation to geographical variation of SLE mortality in the United States are substantial, but these factors do not fully account for the clustering phenomenon.