Mapping of homicide risk with binomial cokriging and simulation: a case study in Sao Paulo, Brazil

The impact of violence on the population's health profile in large Brazilian cities is a real concern and has been incorporated into the country's public health agenda. Although violent crimes affect the entire population, they are distributed unequally in the city. In this context, analytical tools to evaluate the risk and spatial distribution of homicide are important for surveillance and prevention. The current article is a contribution in this direction. A geostatistical methodology is used to estimate and map homicide risk, and a new semivariogram estimator is presented and its application is evaluated in the city of Sao Paulo, Brazil, for the period 2002 to 2004. Homicide risk scenarios are generated using binomial cokriging and stochastic simulation procedures, allowing the demarcation of areas with greater or lesser risk that can be used to study spatial determinants of violence and improve the understanding of problems through analysis of differences across urban territories.