National Natural Science Foundation of China
The severe outbreak of coronavirus disease 2019 (COVID-19) demonstrates the importance of disease risk assessment. The existing risk assessment methods are limited by the real time and accuracy of data. Most of them take the administrative statistical unit as the analysis scale, which has modifiable areal unit problem (MAUP). First, based on a random forest method, we integrated COVID-19 transmission data at community scale and multisource geospatial data to map COVID-19 disease outbreak risks at fine scale. The experimental results (overall accuracy=0.85, Kappa=0.70) indicated the feasibility of the model. Second, we built a spatial variable-infection risk model at community and place scale to assess the risk degree of epidemic spread in different places and facilities. Last, we analyzed the possibly spatial drivers of disease transmission. The results show that (1) the central area of Wuhan city has the highest risk of infection and the risk map presents a trend of decreasing from the center to the periphery; (2) The top five facilities with the highest risk of COVID-19 infection are shopping, medical, financial, transportation and public facilities; (3) The transmission risk of the epidemic is low in primary and middle schools, but high in colleges and universities; (4) The model determines the degree of epidemic risk at the community scale and predicts that shopping and traffic places are two most significant driving factors with the epidemic outbreak. In conclusion, this study suggests a new method of disease risk assessment based on a fine scale, which can pave the way for future disease risk assessment.