[Violence clusters in Pernambuco, Brazil].

Objectives: To analyze the spatial distribution of homicide mortality rates among males 15 to 49 years old in the state of Pernambuco, Brazil, for the periods of 1980 to 1984 and 1995 to 1998, and to identify violence clusters.
Methods: Mortality data were obtained from the Brazilian Ministry of Health's Mortality Information System. The mean homicide mortality rate was estimated for each municipality in the state for the two periods. The Moran coefficient was calculated to determine spatial autocorrelation. (The Moran coefficient ranges from -1 to +1, with a positive coefficient indicating a cluster of similar values, and a negative coefficient indicating adjacent dissimilar values.) To identify clusters of municipalities with either high or low homicide mortality rates, the local indicator of spatial association (LISA) was used. Finally, a Moran map was constructed to identify municipalities with statistically significant LISA values and to identify clusters of municipalities with either high or low homicide mortality rates.
Results: The Moran coefficient for 1980-1984 was 0.392, and for 1995-1998 it was 0.291 (P < 0.001). In the 1980-1984 period, one cluster of high homicide mortality rates was found in the Mata Sul region of the state, close to the metropolitan region of the state capital, Recife. In the 1995-1998 period, two violence clusters were identified: a predominantly urban one in the Recife metropolitan region, and the other in the state interior, in an area known as the "Marijuana Polygon" (Polígono da Maconha).
Conclusion: This study suggests that the violence clusters are not the result of the socioeconomic conditions per se, but rather the consequence of the interaction between poor economic conditions and drug trafficking.