Spatiotemporal pattern recognition and dynamical analysis of COVID-19 in Shanghai, China

Abstract

Shanghai suffered a large outbreak of Omicron mutant of COVID-19 at the beginning of March 2022. To figure out the spatiotemporal patterns of the epidemic, a retrospective statistical investigation, coupled with a dynamic model, is implemented in this study. The hotspots of SARS-CoV-2 transmissions are identified, and strong aggregative effects in the decay stage are found. Besides, the visualization of disease diffusion is provided to show how COVID-19 disease invades all districts of Shanghai in the early stage. Furthermore, the calculations from the dynamic model manifest the effect of detections to suppress the epidemic dissemination. These results reveal the strategies to improve the spatial control of disease.

Keywords: Autocorrelation analysis; Daily reproduction number; Dynamic model; Spatial statistics.

Fig. 1

Administrative division map of Shanghai. There are $16$ administrative districts in Shanghai and the central downtown area consists of HP, XH, CN, PT, JA, HK, YP.

Fig. 2

Illustration of the daily new increment of confirmed cases, asymptomatic cases and their summation in Shanghai. The asymptomatic cases accounts for the majority of positive cases and the confirmed cases accounts for a few. The data begins on 1$\mathrm{st}$ March and ends on 31$\mathrm{th}$ May.

Fig. 3

Illustrations of stage classification for the whole outbreak in Shanghai. Panel (a) presents the three stages including import and diffusion stage (IDs), outbreak and epidemic stage (OEs) and decay and elimination stage (DEs) of total positive cases in Shanghai. The switch times are 28$\mathrm{th}$ March and 29$\mathrm{th}$ April. Panel (b) presents the difference values of daily new increment of total cases in Shanghai.

Fig. 4

Local indicator spatial association (LISA) clustering maps for three stages of the outbreaks in Shanghai.

Fig. 5

Getis hotspot maps for three stages of the outbreaks in Shanghai.

Fig. 6

Diffusion maps of districts with positive cases in Shanghai. The color shades indicate the cumulative value of positive cases in each district.

Fig. 7

Illustrations of the fitting result of model (2) in Shanghai. Panels (a), (b) indicate the real values of daily increment of asymptomatic (triangle marker) and confirmed (square marker) cases in each district in Shanghai, and panels (c), (d) indicate the corresponding fitting values of asymptomatic and confirmed cases, respectively. Panels (e), (f) indicate the daily undetected rate functions with respect to time of each district.

Fig. 8

Heatmaps and boxcharts of daily reproduction numbers under spatiotemporal classification (5). Panels (a) and (c) are the heat maps of daily reproduction number caused by importing one asymptomatic or confirmed case, where the values are normalized values according to the stages. Panels (b) and (d) are the boxcharts of daily reproduction number caused by importing one asymptomatic or confirmed case. In panels (b) and (d), *** and ${}^{\text{\#\#\#}}$ denote $P$-values $<0.001$ comparing to $S_1$ and $S_2$.