This paper estimates the technical efficiency of SUS of Brazilian states with data from 2006. The DEA model is used with nondiscricionary imputs to analyze how some demographics and socio-economics characteristics can be change technical efficiency of public health offer. In addition the paper introduces a selection variable method that reduces the original set of variable improving the discrimination power of DEA model. The results shown that the reduced model improves discrimination without loss information. Beside, the results suggest that the population is the only nondiscricionary imput that creates actually a more favorable environment to offers public health.
