Introduction: In Brazil, deaths by external causes rank first in the mortality statistics. Nevertheless, studies which investigate the relationship between mortality by external causes and race/skin color are scarce.
Objectives: To evaluate the relative contribution of race/skin color to the spatial distribution of mortality by homicide in Salvador, state of Bahia, Brazil, in the period 1998 - 2003.
Material and methods: This is a spatial aggregate study including secondary data on 5,250 subjects, using a unit of analysis called the "weighting area" (WA). Annual average death rates by homicide were estimated. The Global and Local Moran Index were used to evaluate the presence of spatial autocorrelation and the Conditional Auto Regressive (CAR) model was employed to evaluate the referred effect, using the R statistical package.
Results: Global and Local Moran's I tests were significant. CAR regression showed that the predicted mortality rate increases when there is a growth in the proportion of black males aged between 15 and 49 years. Geometrically weighted regression (GWR) showed a very small variation of the local coefficients for all predictors.
Conclusion: We demonstrated that the interrelation between race, violence and space is a phenomenon which results from a long process of social inequality. Understanding these interactions requires interdisciplinary efforts that contribute to advancement of knowledge that leads to more specific Public Health interventions.