Spatial epidemiology is aimed at identifying patterns in the geographical distribution of health data. It may detect irregularities such as spatial clusters of a particular disease [1, 2], for example, where a specific disease has significant high or low prevalence . Methods for the study of spatial clusters include global spatial autocorrelation, Local Indicators of Spatial Association (LISA), spatial regression, spatial scan statistics and Bayesian inference .
There are numerous examples of spatial data analysis performed on health variables, such as prevalence, incidence and mortality . In mental health, for example, Bayesian models have been used to study the relationship between poverty and social isolation, and psychiatric admission rates in acute hospitals in small urban areas of London and New York ; the variation in the incidence of psychotic disorders in urban areas in Southeast London ; the relationship between depression and schizophrenia admission rates and socioeconomic characteristics in the counties of 14 States in the USA [8, 9]; and the study of the correlation between mental retardation and clusters of developmental delay . Spatial scan statistics have been used to detect clusters of mental disorders due to psychoactive substance use, and neurotic, stress-related, and somatoform disorders, and their relationship to poverty and neighbourhood social disorganization in Malmö (Sweden) . LISA were applied to analyze spatial patterns of mental health in the slums of Dhaka (Bangladesh) . In addition, a spatial regression model has been used to analyze spatial allocation in mental health expenditure in England .
However, the studies on spatial analysis show significant problems with respect to comparability, reproducibility and generalization since different methods and techniques produce different results [14, 15]. We previously developed and tested  a Multi-Objective Evolutionary Algorithm (MOEA) that hybridised three LISA methods (Moran’s I, Geary’s C and Getis and Ord’s G) and Bayesian inference to detect schizophrenia hot spots (geographical clusters of spatial units –municipalities- with significantly high rates of selected indicators of a given disease) in Andalusia (Spain). Although this hybrid technique proved to be highly effective for this aim, there were problems when trying to precisely identify the location, shapes and boundaries of the spots, as also commonly occurs with other methods of spatial analysis [16, 17].
This study has incorporated the identification of cold spots (geographical clusters of spatial units –municipalities- with significantly low rates of treated prevalence of a given disease) into the spatial analysis of the regional mental health system in Spain. The presence of both spatial clusters were analysed using the outpatient mental health database in Catalonia (Spain).
This paper aims to obtain a precise identification and geographical location of hot and cold spots of treated prevalence of depression and check if they have any spatial relationship with the administrative (catchment areas) divisions of mental health care in Catalonia in order to facilitate evidence to enable well-informed policy decisions. The related descriptive hypotheses are: 1) spatial clusters of treated prevalence of depression (hot and cold spots) exist and, 2) these clusters are related to the administrative divisions of mental health care (catchment areas) in Catalonia.