Objective: To analyse the multi-dimension nature of severe acute respiratory syndrome (SARS) transmission.
Methods: Based on the data of SARS in 2003 and the geographic information system of Beijing, as well as under the broad range of the theorems and techniques of data-driven and model-driven knowledge mining, hierarchical techniques were used to test the hot spots. Wavelet technique was also used to decompose Moran's I frequency to survey the spatial clustering process of SARS. For factors analysis, BW test was used to distinguish factors which influencing SARS process. In temporal aspects, susceptive-infective-removal model (SIR) without Taylor expansion was solved by a genetic-simulated annealing algorithm, that directly provided a new approach to obtain epidemic parameters from the SIR model.
Results: Different order of spatial hot spots were noticed and the clustering were relevant with the means of transportation. Diffusion dynamics were changed along with the temporal process of SARS. Regarding factor analysis, geographic relationship, population density, the amount of doctors and hospitals appeared to be the key elements influencing the transmission of SARS. The predictable number of SARS cases evolving with time were also calculated.
Conclusions: Cluster detection of close contacts of SARS infective in Beijing revealed the spatial characters of urban population flow and having important implications in the prevention and control of this communicable diseases. Some human and physical environment factors played statistical significant roles in different periods during SARS epidemics. An efficient algorithm was developed to solve SIR model directly, enabling the estimation of epidemic parameters from SIR and early forecast.