The use of remotely sensed environmental parameters for spatial and temporal schistosomiasis prediction across climate zones in Ghana.

Schistosomiasis control in sub-Saharan Africa is enacted primarily through preventive chemotherapy. Predictive models can play an important role in filling knowledge gaps in the distribution of the disease and help guide the allocation of limited resources. Previous modeling approaches have used localized cross-sectional survey data and environmental data typically collected at a discrete point in time. In this analysis, 8 years (2008-2015) of monthly schistosomiasis cases reported into Ghana's national surveillance system were used to assess temporal and spatial relationships between disease rates and three remotely sensed environmental variables: land surface temperature (LST), normalized difference vegetation index (NDVI), and accumulated precipitation (AP). Furthermore, the analysis was stratified by three major and nine minor climate zones, defined using a new climate classification method. Results showed a downward trend in reported disease rates (~ 1% per month) for all climate zones. Seasonality was present in the north with two peaks (March and September), and in the middle of the country with a single peak (July). Lowest disease rates were observed in December/January across climate zones. Seasonal patterns in the environmental variables and their associations with reported schistosomiasis infection rates varied across climate zones. Precipitation consistently demonstrated a positive association with disease outcome, with a 1-cm increase in rainfall contributing a 0.3-1.6% increase in monthly reported schistosomiasis infection rates. Generally, surveillance of neglected tropical diseases (NTDs) in low-income countries continues to suffer from data quality issues. However, with systematic improvements, our approach demonstrates a way for health departments to use routine surveillance data in combination with publicly available remote sensing data to analyze disease patterns with wide geographic coverage and varying levels of spatial and temporal aggregation.