Efficiency of Geographically Weighted Regression in Modeling Human Leptospirosis Based on Environmental Factors in Gilan Province, Iran

Abstract

It is of little debate that Leptospirosis is verified as the most important zoonosis disease in tropical and humid regions. In North of Iran, maximum reports have been dedicated to Gilan province and it is considered as an endemic problem there. Therefore, modeling or researching about different aspects of it seems indispensable. Hence, this paper investigated various models of Geographically Weighted Regression (GWR) approach and impacts of seven environmental variables on modelling leptospirosis in Gilan. Accordingly, counts of patients were considered as dependent variable during 2009–2011 at village level and environmental variables were utilized as independent variables in the modelling. In addition, performance of two Kernels (Fixed and Adaptive), two Weighting Functions (Bisquare and Gaussian) and three Bandwidth Selection Criteria (AIC (Akaike Information Criterion), CV (Cross Validation) and BIC (Bayesian information criterion)) were compared and assessed in GWR models. Results illustrated: (1) Leptospirosis and effective variables vary locally across the study area (positive and negative); (2) Adaptive kernel in comparison to Fixed kernel, Bisquare weighting function to Gaussian, and also AIC to CV and BIC (due to R2 and Mean Square Error (MSE) validation criteria); (3) Temperature and humidity were founded as impressive factors (include higher values of coefficients); Finally, contain more reliable results consecutively. However, the provided distribution maps asserted that central villages of Gilan not only are more predisposed to leptospirosis prevalence, but also prevention programs should focus on these regions more than others.

Keywords

• infectious disease
• leptospirosis
• Geospatial Information System
• spatial analysis
• GWR