OBJECTIVE:To evaluate the impact of a homicide prevention program. METHODS:A quasi-experimental study was performed using time series analysis of homicide incidence in the Morro das Pedras area in the city of Belo Horizonte, Southeastern Brazil, from 2002 to 2006. The number of homicides occurring in this location was compared to other violent and nonviolent favelas and to other neighborhoods of the city, during each of the Program phases. To test the hypothesis that homicide reduction was caused by the actions implemented by the program, a statistical model was developed based on generalized linear models. RESULTS:In the fi rst six months a 69% reduction in the number of homicides was obtained. During the other Program periods, the effect on the reduction of homicides lessened, but the difference among coeffi cients compared to the initial period was not statistically signifi cant. Even with full Program implementation, the effect continued to be similar to the previous periods, probably because the program was implemented in other violent favelas in the city. CONCLUSIONS:The results suggest that the Staying Alive Program model can be an important alternative for the prevention of youth homicides in communities that have characteristics similar to the pilot program in Morro das Pedras.