Development of a new spatial analysis tool in mental health: identification of highly autocorrelated areas (hot-spots) of schizophrenia using a Multiobjective Evolutionary Algorithm model (MOEA/HS)

• Published in: Epidemiologia e psichiatria sociale Vol. 19; no. 4; p. 302
• Main Authors: García-Alonso, Carlos R, Salvador-Carulla, Luis, Negrín-Hernández, Miguel A, Moreno-Küstner, Berta
• Format: Journal Article
• Language: English
• Published: Italy 01-12-2010
• Subjects: Algorithms; Models, Statistical; Schizophrenia - epidemiology; Spain - epidemiology; Spain
• ISSN: 1121-189X
• DOI: 10.1017/S1121189X00000646

This study had two objectives: (1) to design and develop a computer-based tool, called Multi-Objective Evolutionary Algorithm/Hot-Spots (MOEA/HS), to identify and geographically locate highly autocorrelated zones or hot-spots and which merges different methods, and (2) to carry out a demonstration study in a geographical area where previous information about the distribution of schizophrenia prevalence is available and which can therefore be compared.
Local Indicators of Spatial Aggregation (LISA) models as well as the Bayesian Conditional Autoregressive Model (CAR) were used as objectives in a multicriteria framework when highly autocorrelated zones (hot-spots) need to be identified and geographically located. A Multi-Objective Evolutionary Algorithm (MOEA) model was designed and used to identify highly autocorrelated areas of the prevalence of schizophrenia in Andalusia. Hot-spots were statistically identified using exponential-based QQ-Plots (statistics of extremes).
Efficient solutions (Pareto set) from MOEA/HS were analysed statistically and one main hot-spot was identified and spatially located. Our model can be used to identify and locate geographical hot-spots of schizophrenia prevalence in a large and complicated region.
MOEA/HS enables a compromise to be achieved between different econometric methods by highlighting very special zones in complex areas where schizophrenia shows a high autocorrelation.