Spatial epidemiological determinants of severe fever with thrombocytopenia syndrome in Miyazaki, Japan: a GWLR modeling study

Abstract

Abstract Background Cases of severe fever with thrombocytopenia syndrome (SFTS) have increasingly been observed in Miyazaki, southwest Japan. It is critical to identify and elucidate the risk factors of infection at community level. In the present study, we aimed to identify areas with a high risk of SFTS virus infection using a geospatial dataset of SFTS cases in Miyazaki. Methods Using 10 × 10-km mesh data and a geographically weighted logistic regression (GWLR) model, we examined the statistical associations between environmental variables and spatial variation in the risk of SFTS. We collected geospatial and population census data as well as forest and agriculture mesh information. Altitude and farmland were selected as two specific variables for predicting the presence of SFTS cases in a given mesh area. Results Using GWLR, the area under the receiver operating characteristic curve (AUC) was estimated at 73.9%, outperforming the classical logistic regression model (72.4%). The sensitivity and specificity of the GWLR model were estimated at 90.9 and 58.7%, respectively. We identified altitude (odds ratio (OR) = 0.996, 95% confidence interval (CI): 0.994–0.999 per one-meter elevation) and farmland (OR = 0.999, 95% CI: 0.998–1.000 per % increase) as useful negative predictors of SFTS cases in Miyazaki. Conclusions Our study findings revealed that the risk of SFTS is high in geographic areas where farmland area begins to diminish and at mid-level altitudes. Our findings can help to improve the efficiency of ecological and animal surveillance in high-risk areas. Using finer geographic resolution, such surveillance can help raise awareness among local residents in areas with a high risk of SFTS.

Keywords

• Severe fever with thrombocytopenia syndrome
• Ticks
• Statistical model
• Altitude
• Farms
• Spatial regression