Abstract
Few study has revealed spatial transmission characteristics of COVID-19 in Wuhan, China. We aimed to analyze the spatiotemporal spread of COVID-19 in Wuhan and its influence factors. Information of 32,682 COVID-19 cases reported through March 18 were extracted from the national infectious disease surveillance system. Geographic information system methods were applied to analysis transmission of COVID-19 and its influence factors in different periods. We found decrease in effective reproduction number (Rt) and COVID-19 related indicators through taking a series of effective public health measures including restricting traffic, centralized quarantine and strict stay-at home policy. The distribution of COVID-19 cases number in Wuhan showed obvious global aggregation and local aggregation. In addition, the analysis at streets-level suggested population density and the number of hospitals were associated with COVID-19 cases number. The epidemic situation showed obvious global and local spatial aggregations. High population density with larger number of hospitals may account for the aggregations. The epidemic in Wuhan was under control in a short time after strong quarantine measures and restrictions on movement of residents were implanted.
Introduction
Early epidemiological research into the COVID-19 outbreak in Wuhan primarily focused on temporal dynamics—tracking the speed of viral spread over time. However, granular spatial analysis at the street level is essential for understanding why specific neighborhoods became hotspots. This study addresses this gap by examining the spatiotemporal characteristics of the outbreak across Wuhan’s 179 streets, identifying urban features that drove transmission and evaluating the impact of varying intervention intensities.
Methodology
The researchers integrated three primary datasets: 32,682 confirmed cases from the national surveillance system, demographic data from the 2018 Wuhan Statistical Yearbook, and urban infrastructure locations (hospitals, shopping centers, transit stations) scraped from digital maps.
The analysis was divided into three distinct periods based on the effective reproduction number ($R_t$):
- Period 1 (Pre-cognitive): Natural spread without strong interventions.
- Period 2 (Control): Initial city lockdown and traffic suspension.
- Period 3 (Fading): Implementation of centralized quarantine and strict stay-at-home policies.
Spatial transmission was analyzed using Moran’s I for global aggregation and Local Indicators of Spatial Association (LISA) for identifying local clusters. A Spatial Lag Model (SLM) was employed to determine the influence of factors like population density and proximity to public facilities while accounting for spatial dependencies between neighboring streets.
Results
Temporal and Spatial Dynamics
The $R_t$ peaked at 3.86 on January 23, coinciding with the start of the city lockdown. Following the implementation of more stringent measures in Period 3, the $R_t$ plummeted to below 0.1 by mid-March. Correspondingly, the doubling time for cases increased from 3.6 days in the early stages to 103.9 days by the end of the study period.
Spatial analysis revealed significant global and local aggregation. The number of streets showing non-significant clustering increased from 18 to 54 as the outbreak progressed, indicating that strict movement restrictions successfully disrupted transmission chains.
Drivers of Transmission
The Spatial Lag Model identified the number of hospitals as the strongest predictor of case numbers (coefficient: 27.263, P < 0.001). This "Hospital Effect" suggests that medical facilities may have inadvertently served as super-spreader hubs due to high patient density and nosocomial infections early in the outbreak. Population density was also a significant factor (coefficient: 0.001, P < 0.001). Interestingly, the ratio of elderly residents did not significantly correlate with street-level case counts, likely due to the lower mobility and lower overall density in areas with higher elderly populations.
Discussion
The findings demonstrate that while initial lockdowns slowed the spread, centralized quarantine was necessary to dissolve spatial clusters formed by household transmission. The strong correlation between hospital proximity and case numbers highlights the need for improved infection control in medical settings during the early stages of an epidemic.
Limitations of this study include its retrospective observational nature, which precludes definitive causal claims, and the use of 2018 demographic data for a 2020 event. Future work should incorporate real-time mobility data to better capture the dynamic interactions between urban infrastructure and viral transmission.
Summary updated