Developing an Early Detection Model for improving Vector-Borne Disease Surveillance

Abstract. Given the increasing incidence of vector-borne diseases, there is a need for an effective predictive model, to support timely public health responses in urban areas. However, most of the study have been limited to district level and few climatic variables, which may not be sufficient for localized mitigation efforts. To bridge this gap an Early Detection model for dengue fever is developed, analysing key spatial-temporal variables influencing local transmission. The model integrates meteorological variables such as rainfall, temperature, humidity along with physical factors such as NDVI, land cover and population distribution. Dengue cases data was obtained form District Medical Office, Bhopal, while other independent variables were generated through Geographic Information System (GIS) operations on Earth Observation (EO) datasets such as Landsat, MODIS, etc. Multiple Linear Regression (MLR), Generalized Linear Model (GLM), and Sequential Regression Model (SRM) were employed to capture temporal and spatial dependencies. SRM emerged as the most effective model (Adjusted Pseudo-R² = .407, RMSE Test = 1.810), outperforming MLR (Adjusted-R² = . 256, RMSE Test = 1.701) and GLM (Adjusted Pseudo-R² = .555, RMSE Test = 2.623), to identify high-risk areas. Humidity, NDVI, LULC water and forest significantly influenced dengue cases, as these factors favours mosquito breeding. The study highlights the effectiveness of GIS and Machine Learning (ML) in strengthening the disease surveillance and control. Applying this approach in Indian cities such as Bhopal, Madhya Pradesh, demonstrated its potential to facilitated timely, targeted, tailored and resource efficient interventions.