OBJECTIVES: The pharmacological prescription can present geographical variation due to multiple factors. The aim of this communication is to introduce the spatial variability in the evaluation of different individual outcomes. METHODS: Pharmacological prescription data of 10,410 persons with diagnosis of osteoarthrosis were analyzed. The prescriptions were fulfilled during the year 2006 in the health department 11 of the Comunitat Valenciana (Spain). The number of recipes, the number of prescribed principles and the pharmaceutical expense for every patient were evaluated. The relationship with gender and age was explored by means of generalized linear models. The geographical variation between different administrative units is introduced by means of heterogeneity and structured random effects. The proposed hierarchical models were analyzed using Bayesian methods with MCMC procedures. RESULTS: Models with random spatial effects explain an important part of the variability and fit better. As a measure of quality of the fitted model we have chosen the Deviance Information Criteria (DIC). The improvement in the deviance of the model for the number of principles per patient is of 1790 and in DIC of 1760. The maps reveal geographical inequalities and zones with greater average values. CONCLUSION: The Bayesian hierarchical models have been a very useful tool for incorporating geographical information in the analysis of pharmacological prescription data. They allow mapping spatial components expressing the trend of that geographical variation.