Objectives  To describe the spatiotemporal diffusion of the severe acute respiratory syndrome (SARS) epidemic in mainland China, and to analyse the spatial pattern of SARS transmission from the Beijing epicentre to its neighbouring areas.

Methods  Probable SARS cases occurring between November 2002 and May 2003 in mainland China were compiled from different sources and geo‐coded into a geographical information database based on onset location. Spatial analyses including kernel density estimation, and spatial statistical and tracking analyses were performed to characterise the spatiotemporal distribution of SARS cases based on onset location/date. SARS cases that got infected in Beijing but were reported in three provinces surrounding Beijing were mapped, and logistic regression using a ‘case–control’ design at the county level was performed to analyse the impact of travel‐related risk factors in the diffusion pattern.

Results  The SARS epidemic in mainland China spanned a large geographical extent but clustered in two areas: first in Guangdong Province, and about 3 months later in Beijing with its surrounding areas in Shanxi Province, Inner Mongolia Autonomic Region, Hebei Province and Tianjin. Counties in the neighbourhood of Beijing that were crossed by a national highway or inter‐provincial freeway showed the highest risk of acquiring SARS infections, even after correction for population density and medical staff density. Being intersected by a railway did not significantly associate with risk of SARS.

Conclusions  This study provides the first complete documentation of the spatial and temporal characteristics of the SARS epidemic in mainland China. Our analyses confirmed that SARS had benefited from national highways and inter‐provincial freeways for its spread from epicentres to neighbouring areas, whereas trains showed no significant association. This knowledge may be important for the control of re‐emerging SARS, or other future emerging human‐to‐human transmittable infections.