This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.
Mobility restriction was one of the primary measures used to restrain the spread of COVID-19 globally. Governments implemented and relaxed various mobility restriction measures in the absence of evidence for almost 3 years, which caused severe adverse outcomes in terms of health, society, and economy.
This study aimed to quantify the impact of mobility reduction on COVID-19 transmission according to mobility distance, location, and demographic factors in order to identify hotspots of transmission and guide public health policies.
Large volumes of anonymized aggregated mobile phone position data between January 1 and February 24, 2020, were collected for 9 megacities in the Greater Bay Area, China. A generalized linear model (GLM) was established to test the association between mobility volume (number of trips) and COVID-19 transmission. Subgroup analysis was also performed for sex, age, travel location, and travel distance. Statistical interaction terms were included in a variety of models that express different relations between involved variables.
The GLM analysis demonstrated a significant association between the COVID-19 growth rate ratio (GR) and mobility volume. A stratification analysis revealed a higher effect of mobility volume on the COVID-19 GR among people aged 50-59 years (GR decrease of 13.17% per 10% reduction in mobility volume;
The association between mobility reduction and COVID-19 transmission significantly varied according to mobility distance, location, and age. The substantially higher impact of mobility volume on COVID-19 transmission for longer travel distance, certain age groups, and specific travel locations highlights the potential to optimize the effectiveness of mobility restriction strategies. The results from our study demonstrate the power of having a mobility network using mobile phone data for surveillance that can monitor movement at a detailed level to measure the potential impacts of future pandemics.
The COVID-19 pandemic had led to over 630 million infections and 6 million deaths worldwide by November 2022 [
In almost 3 years of the COVID-19 pandemic, unprecedented efforts have been made to explore the association between human mobility and COVID-19 transmission [
In this study, we integrated anonymized geolocalized mobile phone data with census and demographic data in the Greater Bay Area of China. We aimed to analyze the impact of mobility reduction on COVID-19 transmission according to travel distance (long and short), location (workplaces, schools, recreation areas, shopping areas, transit stations, and other areas), age (≤18, 19-29, 30-39, 40-49, 50-59, and ≥60 years), and sex. The coverage rate of mobile phone use among the population aged 15-65 years was almost 100% [
Large volumes of anonymized aggregated mobile phone position data between January 1 and February 24, 2020, were collected for the 9 megacities of Guangzhou, Shenzhen, Foshan, Huizhou, Dongguan, Zhongshan, Zhaoqing, Zhuhai, and Jiangmen in Guangdong-Hong Kong-Macao Greater Bay Area, China. The Greater Bay Area is the most populated and largest urban area, and is 1 of the 4 largest bay areas in the world. Mobile phone data were provided by 1 of the 3 leading mobile phone service providers. Origin-destination matrices were constructed by computing the number of people that move between different locations on an hourly basis, as done previously [
Daily incidences of COVID-19 were obtained from official governmental reports in the Greater Bay Area, China [
We used anonymized and aggregated mobile phone data at the population level without individual travel patterns for strict protection of personal privacy. All data were obtained in an anonymous format without personal identifying information. This study was approved by the Institutional Review Board of Shenzhen University, China (review number PN-202300030).
The growth rate ratio (GR) of COVID-19 was computed as the average number of new cases per day over the previous 3 days to that over the previous 7 days. The static correlation and dynamic correlation between the GR and mobility volume were determined by Pearson correlation and rolling correlation. We tested the different day lags between the GR and mobility in the correlation analysis, as there may be a time lag between reported cases and true community infections. A generalized linear model (GLM) was established to test the association between mobility volume and COVID-19 transmission. Subgroup analysis was also performed for sex, age, travel location, and travel distance. The statistical interaction terms were included in a variety of models that express different relations between the involved variables. The Rt in Shenzhen by time for real transmission was estimated according to the likelihood-based estimation method [
An interaction analysis was performed by including the interaction terms of mobility volume ratio (VR) and distance ratio (DR) in the GLM analysis to determine whether the impact of mobility volume on COVID-19 transmission differs by mobility distance in the Greater Bay Area, China. The VR and DR were defined for each day (t), which quantified the change in mobility patterns and were similar to previous studies [
where
The DR represents the change in the distance of individual trips made to each area per day, relative to ordinary behavioral patterns (ie, before COVID-19), which was calculated as follows:
where
The public health interventions in the Greater Bay Area, China during the study period are illustrated in
Relationship between mobility volume and the GR in the Greater Bay Area, China. (A) Average mobility volume based on the mobile phone position data from January to February 2020 in the Greater Bay Area, China. (B) The change in mobility volume for 9 cities from January to February 2020 (the dots represent the observed data, and the plotted lines are smoothed by a generalized additive model). (C) The GR of COVID-19 from January to February 2020 (the line represents GLM fit to the data, and the shadow represents the 95% CI). (D) Correlation between mobility volume and the GR. (E) Association between mobility volume and the GR (the line represents GLM fit to the data, and the shadow represents the 95% CI). The dashed lines represent the main public health interventions. GLM: generalized linear model; GR: growth rate ratio.
We next investigated the impact of a reduction in mobility volume on COVID-19 transmission in various subgroups of sex and age. Although the mobility volume for females was less than that for males before, during, and after the COVID-19 pandemic (
Among different age groups, those aged 50-59 and ≥60 years exhibited the lowest levels of reduction in mobility volume of 72.9% and 67.0%, respectively, during the pandemic. The percentage reductions in mobility volume were 78.1%, 81.5%, 81.4%, and 77.7% for those aged ≤18, 19-29, 30-39, and 40-49 years, respectively (
Relationship between mobility volume and the GR by sex and age. (A) Time series of the daily average mobility volume and (B) the sex-specific GR for males and females. The dots represent the raw data, while the plotted lines are smoothed by a generalized additive model. (C) Relationship between the sex-specific GR and sex-specific mobility volume. The line represents GLM fit to the data, and the shadow represents the 95% CI. (D) Time series of the daily average mobility volume and (E) the age-specific GR for various age groups. The dots represent the raw data, while the plotted lines are smoothed by a generalized additive model. (F) Relationship between the age-specific GR and age-specific mobility volume. The line represents GLM fit to the data, and the shadow represents the 95% CI. The dashed lines represent the main public health interventions. (G) The change in the demographic-specific GR per 10% reduction in mobility volume. GLM: generalized linear model; GR: growth rate ratio.
The distribution of mobility volume at 6 types of locations at different time periods in the city of Shenzhen (one of the Greater Bay Area cities) is illustrated in
Relationship between mobility volume and COVID-19 transmission by various destinations. (A) The distribution of mobility volume for various destinations in the city of Shenzhen at different time periods in the Greater Bay Area, China (the peak of 3D bars represents the mobility volume in the given period), and (B) the time series of mobility volume at these destinations. (C) Time series of the instantaneous reproduction number (Rt) in Shenzhen. The shadow represents the 95% CI. (D) The change in the Rt per 10% reduction in mobility volume for a certain destination. Adjusted R2 represents the goodness of fit for the models.
The change in mobility volume occurred earlier and was greater than the change in distance during the pandemic in the Greater Bay Area, China (
ANOVA showed that removing the interaction did significantly affect the fit of the model (
Relationship between mobility volume and COVID-19 transmission by mobility distance. (A) Time series of the VR and DR in the Greater Bay Area, China. The dots represent the observed data, and the plotted lines are smoothed by a generalized additive model. (B) Time series of the instantaneous reproduction number (Rt) in the Greater Bay Area, China. The shadow represents the 95% CI. (C) Relationship between the VR and Rt ratio by different DRs in the Greater Bay Area, China. The line represents GLM fit to the data, and the shadow represents the 95% CI. DR: distance ratio; GLM: generalized linear model; VR: volume ratio.
In this study, we found that the impact of reductions in human mobility on COVID-19 transmission significantly varied by travel distance, location, and age. There was a significant positive interaction between mobility volume and mobility distance regarding COVID-19 transmission, with steeper slopes for the association (larger coefficient in the regression analysis) between mobility volume and COVID-19 transmission with increasing mobility distance. We found a significantly steeper slope for the association between the reduction in mobility volume and COVID-19 transmission among persons aged 50-59 years than among other age groups. Furthermore, the slope for the association was steeper for the locations of transit stations and shopping areas, compared with workplaces, schools, recreation areas, and other locations.
Our study indicated that the introduction of mobility restrictions in the Greater Bay Area, China led to a marked decrease in COVID-19 transmissibility. The time lag between mobility reduction and decline in the GR was estimated as 2 days at the very beginning of the pandemic in this area, which is shorter than the time of around 2 weeks in a similar study in the United States reported by Badr et al [
The mobility reduction was associated with a greater reduction in the GR for the age group of 50-59 years than the other age groups in our study. The slope of the association between mobility volume and the GR for those aged 50-59 years was steeper than that for the other age groups. Those aged 50-59 years showed lower mobility reduction than those aged ≤49 years, but were vulnerable to COVID-19 infection with a higher proportion of underlying chronic diseases compared with young people [
The mobility reduction for transit stations and shopping centers was associated with a greater reduction in COVID-19 transmission in the whole city compared with the findings for workplaces, schools, and recreation areas. The slopes for the association between mobility to transit stations and shopping centers and the Rt were steeper than the slopes for the association involving other locations. These high-contact environments are more crowded and therefore have a higher risk. However, the mobility reduction for transit stations was less than that for other locations. Many governments applied various policies for mobility restriction at specific locations, since there was not enough evidence available regarding which locations should be closed or which locations should remain open. Our study provides evidence that controlling mobility to a small number of locations could reduce transmission in the entire city. Therefore, location-specific mobility restrictions should be taken into consideration for precise interventions and reopening strategies with substantially lower economic costs [
Our results have public health and policy implications. First, we analyzed the relationship between mobility responses and COVID-19 transmission using mobility data of only a certain travel destination, travel distance, or demographic subgroup (age and sex groups) to gain more insightful knowledge. Our study provides evidence to identify hotspots of transmission and guide policy interventions for specific age groups or mobility patterns associated with higher risks of mobility-related COVID-19 transmission. Second, mobile phone data at fine spatial and temporal resolutions provide strong added value for explaining variations in COVID-19 transmission. The results from our study demonstrate the power of having mobility networks using mobile phone data that monitor movement at a detailed level across cities to measure the potential impacts of public health events. The network can be regularly updated and used to identify populations and travel characteristics at risk of adverse impacts during future pandemics or other crises. It is also suggested to set up a national mobility network that captures human mobility habits, which can form the basis for longitudinal studies.
It is important to note that our study has several limitations. First, we illustrated the detailed structural change in mobility patterns using mobile phone data from users in the Greater Bay Area, China, and these patterns may not be fully representative of other locations in China. However, the mobility change in the Greater Bay Area based on the mobile phone was quite similar to the change in the Baidu mobility index for the whole country, owing to nationally unified public health interventions. We believe that this analysis for the Greater Bay Area is an intuitive and representative estimate of the structural change in mobility patterns in China, but future extension of this analysis for the whole country should be further explored. Second, we focused on quantifying the relationship between mobility patterns and COVID-19 transmission; there is extensive evidence indicating that population-wide social distancing and other potential mitigating factors (eg, wearing face masks and washing hands) all contribute to achieving control of the COVID-19 pandemic [
The COVID-19 pandemic was the first time in human history that human mobility showed a large-scale decline after mobility restrictions to prevent and control the infectious disease. Our study demonstrated that the impact of reductions in human mobility on COVID-19 transmission was significantly modified by travel distance, travel location, and age. The higher impact of mobility reduction on COVID-19 transmission for longer distances, certain age groups, and specific locations highlights the potential to optimize mobility restriction policies to balance adverse health, society, and economic outcomes and the benefits of controlling the spread of COVID-19. It is of great significance to understand the impact of mobility reduction on the spread of infectious diseases in detail, and it provides evidence for the prevention and control of future pandemics.
Supplementary materials and methods.
Main public health interventions in the Greater Bay Area, China during the study period.
Correlations between mobility volume and growth rate ratio at different time lags (in days).
The magnitude of mobility volume change for various demographic groups.
The time series of mobility volume from mobile phone data in the Greater Bay Area, China and the mobility index from Baidu for the whole country of China.
distance ratio
generalized linear model
growth rate ratio
instantaneous reproduction number
volume ratio
We thank Mr Renzhe Xu for helpful suggestions, and Mr Juecheng Li and Ms Gongjian Wu for technical assistance. This work was supported by the National Key Research and Development (R&D) Program of China (grant number: 2018YFB2100704), Research on Prevention and Control of COVID-19 in the Guangdong Education Department (grant number: 2020KZDZX1171), Shenzhen Basic Research Fund (grant number: JCYJ20190808174209308), and National Natural Science Foundation of China (grant numbers: 42171400, 2021A1515011324, 82103945, and 71961137003). The funding body was not involved in the collection, analysis, or interpretation of the data, the writing of the article, or the decision to submit the manuscript for publication.
We purchased the mobile phone data (January to March 2020) from a service provider (China Unicom). Our data purchase agreement with China Unicom prohibits us from sharing the data with third parties, but interested parties can contact China Unicom to obtain the same data. We collated epidemiological data from publicly available data sources (government websites). All epidemiological information that was used has been documented in the article.
YZ ran the study, analyzed and interpreted the data, wrote the paper, and obtained funding for the study. JX extracted the data, analyzed and interpreted the data, and obtained funding for the study. QL participated in discussions and obtained funding for the study. KY, Xiling W, Xiong W, DH, YY, BJC, EC, and ZD participated in discussions on the undertaking of the study. All authors reviewed the paper for content and approved the final report.
BJC has received honoraria from AstraZeneca, Fosun Pharma, GSK, Haleon, Moderna, Roche, and Sanofi Pasteur.