The COVID-19 pandemic, originating in Wuhan, China in late 2019, rapidly spread within the country. This coincided with the Lunar New Year (LNY) holiday period, Chunyun, the world's largest annual human movement. The mass travel during Chunyun, involving billions of journeys in 2019, raised concerns about its potential role in accelerating COVID-19 propagation within China and internationally. On January 23, 2020, a cordon sanitaire was implemented in Wuhan, restricting all non-essential movement. While several studies have examined the effectiveness of Wuhan's cordon sanitaire and other travel restrictions, a comprehensive understanding of human mobility patterns during the initial phase of the pandemic in China remains crucial for informing global strategies. This study utilizes Baidu Huiyan's location-based service (LBS) data to analyze human movement between Chinese prefectures, examining patterns from Wuhan and their impact on the overall Chinese travel network. The research also explores the relationship between travel patterns and regional healthcare capacity to assess the impact of human movement on healthcare pressure during the epidemic's spread.

Previous research has focused on assessing the effectiveness of the Wuhan cordon sanitaire and other domestic travel restrictions in China in the context of COVID-19 control. Studies have used Baidu movement data in transmission models and examined changes in patterns around Wuhan. However, a key unknown was the extent to which observed travel patterns were part of regular seasonal movements or responses to the emerging epidemic or interventions. This study aims to address this gap by providing a detailed examination of travel patterns using various data science techniques and a broader geographic scope.

The study used daily prefecture-level movement data from Baidu Huiyan, encompassing overall migration indices and the percentage of travelers arriving at or leaving specific locations. The data were processed to create symmetrical matrices of daily travel between all Chinese prefectures. Population sizes (2018) were obtained from the China Statistics Yearbook, and healthcare capacity was measured by the number of Grade II and III hospitals per 100,000 residents. COVID-19 case data were sourced from DXY.cn. Time series analysis, k-means clustering, and network analysis (Leiden algorithm for community structure) were employed. Two metrics were used to assess outbound travel surges from prefectures: a proportion-based metric reflecting the relative between-year difference and an anomaly-based metric capturing deviation from 2019. Healthcare pressure was calculated by dividing weekly confirmed COVID-19 cases by healthcare capacity. Non-parametric Mann–Whitney U tests were used to compare healthcare pressure between regions with low and high healthcare capacity. The study also analyzed hierarchical patterns of movement between prefectures of different population sizes. Sensitivity analyses were performed, and the limitations of using Baidu Huiyan data, including its potential bias towards Baidu users, were acknowledged.

Analysis of outbound travel from Wuhan revealed an early-January peak followed by a sharper peak before the LNY holidays, exceeding 2019 levels. K-means clustering identified four temporal patterns of outbound travel from Wuhan, with cluster membership associated with COVID-19 detection timing and prefecture population sizes. Similar patterns were observed in other large Chinese cities. While there was evidence of an increased outbound travel surge from Wuhan before the cordon sanitaire, similar increases were observed in many other prefectures. Analysis of movement across China showed consistent trends in inbound and outbound travel volume across population quartiles, with increased flows before LNY and a sharp drop after the Wuhan cordon sanitaire. However, the composition of these flows differed substantially, revealing a hierarchical divergence of movement. Larger prefectures were more connected to other large prefectures, and smaller prefectures were more connected to other small prefectures. Before LNY, movement was away from large population centers and high healthcare capacity areas. After LNY, a gradual return to high healthcare capacity areas occurred, but the distribution hadn't fully recovered by March 1st. This movement was associated with higher COVID-19-related healthcare pressure in low-capacity areas. Network analysis showed that the overall transportation network did not undergo significant structural reorganization due to Wuhan's cordon sanitaire and other regional restrictions. The modularity of the network showed a temporary increase immediately following the implementation of restrictions in Wuhan, potentially reflecting the movement of resources to the city. Overall connectivity decreased across China after the cordon sanitaire, consistent with nationwide restricted movement policies.

The study's findings suggest that the LNY holiday travel played a more significant role in mobility changes than the impending travel restrictions. The limited relationship between spatial proximity and epidemic spread highlights the importance of considering functional connectivity measures, such as human mobility patterns, when planning travel restrictions. The hierarchical divergence in movement between prefectures of different sizes suggests that medium-sized prefectures could play a key role in limiting the spread to smaller, less populated areas. The shift in healthcare pressure to areas with lower capacity emphasizes the need for resource mobilization in such regions. The lack of significant structural reorganisation in the transportation network indicates that short-term travel restrictions may have limited long-term impacts on mobility.

This study provides valuable insights into the complex interplay between human mobility, holiday travel, and the spread of COVID-19 in China. The findings highlight the importance of considering not only spatial proximity but also functional connectivity and healthcare capacity when implementing travel restrictions. Future research could benefit from higher-resolution mobility data, including traveller characteristics, to further refine risk assessments and intervention strategies. Longitudinal studies are also needed to understand the long-term impacts of travel restrictions on mobility networks.

The study relied on Baidu Huiyan data, which may not represent the entire population of China and may have limitations in the temporal resolution of data. The analysis was limited to the period between January 1st and March 1st 2020, preventing the examination of longer-term impacts of travel restrictions. Furthermore, the study did not explicitly account for the potential influence of other non-pharmaceutical interventions.