Objective  To understand the changing trajectory of the incidence of COVID-19 over time in the early phase of COVID-19 epidemic in 31 provincial level administrative divisions in China, except Hong Kong, Macao, and Taiwan, and explore the development pattern and possible influencing factors of the epidemic.

  Methods  The incidence data of COVID-19 reported in China as of 8 April 2020 were obtained from the Infectious Disease Information System of Chinese Center for Disease Control and Prevention. The trajectory models of change of case numbers of COVID-19 over time at the four administrative levels were fitted by using software SAS 9.4, and the areas at different administrative levels were divided into different groups according to the trajectory characteristics, and the heterogeneity of different subgroups was analyzed to explore the potential difference.

  Results  At the level of 31 provincial administrative regions in China, the trajectory of the case numbers of COVID-19 over time could be divided into two groups. Group 1 and group 2 had a cubic incidence curve, showing an “upward-down-upward” trend. Hubei province belonged to group 2, and other provincial administrative regions belonged to group 1 with the peak value of the trajectory curve much lower than that of group 2. At the level of 30 provincial administrative divisions in China except Hubei, two groups were divided, with a cubic incidence curve. Several provinces adjacent to Hubei, parts of northeastern China and several coastal provinces in eastern China belonged to group 2, which had a higher peak incidence, and the rest belonged to group 1. At the prefecture-level in Hubei province, two groups were divided, which had quadratic or quartic were curves of incidences, respectively. Wuhan belonged to group 2, and the other areas belonged to group 1. At the level of districts/counties in Wuhan, two groups were divided, which had quadratic and quartic curves of incidence, respectively.

  Conclusion  The time-varying trajectories of the number of COVID-19 cases in different regions of China had heterogeneity, with a large number of cases, a wide range of transmission, and a fast transmission rate in the area affected by epidemic firstly. In addition, the spread of COVID-19 was closely related to geographical location, population density, travel intensity, and climatic environment. Using trajectory model to study the epidemic development trend is helpful to formulate targeted COVID-19 prevention and control strategies and provide reference for the prevention and control of emerging respiratory infectious diseases.