Evaluating the association between COVID-19 transmission and mobility in Omicron outbreaks in China

doi:10.21203/rs.3.rs-3459613/v1

Download PDF

Article

Evaluating the association between COVID-19 transmission and mobility in Omicron outbreaks in China

https://doi.org/10.21203/rs.3.rs-3459613/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Prior research has suggested a positive correlation between human mobility and COVID-19 transmission at national or provincial levels, assuming constant correlations during outbreaks. However, the correlation strength at finer scales and potential changes in relationships during outbreaks have been scarcely investigated.

Methods

We gathered case and mobility data (within-city movement, inter-city inflow, and inter-city outflow) at the city level from Omicron outbreaks in mainland China between February and November 2022. For each outbreak, we calculated the time-varying effective reproduction number (R_t). Subsequently, we estimated the cross-correlation and rolling correlation between R_t and the mobility index, comparing them and identifying potential factors affecting these correlations.

Results

We identified 57 outbreaks during Omicron wave 1 (February to June) and 171 outbreaks during Omicron wave 2 (July to December). Cross-correlation estimates varied between waves, with values ranging from 0.64 to 0.71 in wave 1 and 0.45 to 0.46 in wave 2. Oscillation models best fit the rolling correlation for almost all outbreaks, and there were significant differences between extreme values of rolling correlation and cross-correlation. Additionally, we estimated a positive relationship between the GRI and rolling correlation during the pre-peak stage, turning negative during the post-peak stage.

Conclusions

Our findings suggest a positive relationship between Omicron transmission and mobility at the city level. However, significant fluctuations in their relationship, as demonstrated by rolling correlation, indicate that assuming a constant correlation between transmission and mobility may lead to inaccurate predictions or decisions when using mobility as a proxy for transmission intensity.

Health sciences/Medical research/Epidemiology

Health sciences/Health care/Public health/Epidemiology

Health sciences/Diseases/Infectious diseases/Viral infection

Reproduction number

COVID-19

SARS-CoV-2

mobility

Although it is generally believed that more mobility leads to higher transmission of COVID-19, the underlying relationship can be more complex. To understand this relationship better, we analyzed data from various cities in mainland China that experienced Omicron outbreaks in 2022. Our findings showed a generally positive correlation between population mobility and transmission. However, this relationship can change over time and vary between different outbreaks. The level of government response also impacted this relationship. Our findings can help in predicting and managing future outbreaks by using mobility as a predictor for nowcasting and forecasting epidemics.

Monitoring real-time transmissibility of infectious pathogens is critical for guiding control policies. Transmissibility is typically monitored by estimating the time-varying effective reproduction number (R_t) ^1,2, a measure of the average number of secondary cases caused by each infected individual. Given that COVID-19 transmission occurs through close contact, population mobility is expected to strongly correlated with real-time transmissibility ^3,4. To capture this correlation, various types of mobility data have been utilized to generate different mobility index ⁵. These include data from cell phones and other electronic devices with location-tracking capabilities ^6,7, GPS data from smartphones and mobile apps ⁸, as well as bus traffic and air traffic data ^9–11.

Most studies examining the relationship between SARS-CoV-2 transmission and mobility have been conducted at the country or regional level ⁵ due to limited availability of frequently used datasets such as Google mobility ¹² and Apple mobility ¹³. However, it has been observed that the correlation between COVID-19 transmission (measured by the number of COVID-19 cases) and community mobility was specific to individual cities ¹⁴.

Moreover, the correlation patterns between R_t and mobility have shown inconsistency across different studies ⁶. One previous study analyzed the relationships between R_t and residential mobility in 125 countries, revealing a mix of positive (39), negative (58), and inconclusive (28) correlations among countries ¹⁵. Several studies suggested that the relationship between R_t and mobility may be non-linear ¹⁵ or varies across different stages of the epidemic ^16,17. Additionally, the correlation between R_t and the reduction in population mobility might vary depending on the specific variants of the SARS-CoV-2 virus that are circulating ¹⁸. On the other hand, the relationship between mobility and SARS-CoV-2 transmissibility were likely impacted by control measures, as some control measures were aimed at reducing mobility and hence transmissions in the population.

In summary, previous studies have highlighted the importance of considering local determinants and social behaviors when examining the correlation between R_t and mobility, and this correlation pattern is expected to be dynamic in nature. However, we conducted a systematic review (Appendix Section 1) that yielded 26 relevant studies, with 24 of them assuming a constant relationship between mobility and transmission. Two studies adopted rolling correlation approach, but analyzing data at the province-level ¹⁹ or focused exclusively on one selected city ²⁰. We found no study explored the relationship between transmission and mobility comprehensively at the city level in a country during the period dominated by Omicron variants.

In this study, we aimed to analyze the relationship between mobility and COVID-19 transmissibility at the city level using data from the Omicron outbreaks in mainland China in 2022. Mobility was measured using the city-level Baidu Mobility index. For each outbreak, we estimated time-varying effective reproduction number (R_t) and examined the cross-correlation and rolling correlations between R_t and mobility index. Furthermore, we compared the two types of correlations. Lastly, we attempted to identify potential factors that could influence the relationship between R_t and the mobility index.

Case data

Case data from 1 January 2022 to 27 November 2022 were obtained from the daily notification of COVID-19 on the National Health Commission of the People’s Republic of China website ²¹. In China, cases were classified as either local (domestic) cases or imported cases. Local cases, including both symptomatic and asymptomatic cases, were used in our study. Cities were coded according to Baidu cityCode ²². Case data for cities in Yunnan and Xinjiang province were missing and therefore not included in our analysis. In total, data from 336 cities were included in our study.

Mobility index data

The daily mobility index data were derived from Baidu mobility big data (http://qianxi.baidu.com/). Baidu mobility data of 366 prefecture-level cities, three mobility indices including within-city movement, inter-city inflow, and inter-city outflow, were collected. The mobility index can be compared across cities. To avoid weekly fluctuations induced by the work-leisure shift, the daily mobility index was smoothed using a moving average over a 7-day window.

Government Response Index (GRI)

The daily Government Response Index was obtained from the publicly available Oxford COVID-19 Government Response Tracker (OxCGRT) ²³. The GRI was comprised 13 indicators, including containment and closure indicators, economic response indicators, and health systems indicators (Appendix Section 2). The GRI was constructed at the province level.

To construct the GRI at city-level, we first screened the textual notes about implementation and cessation of various public health measures at the city level, to ensure that those cities were included in the consideration of GRI. Then, on the days when the daily case count for a city accounted for 80% or more of the total cases in the corresponding province, the GRIs at the province level were considered as the GRI for that city. We conducted sensitivity analysis with different thresholds, including 70%, 90%, and 100%.

Definition of outbreaks

An outbreak was defined as 20 or more cases occurring in a single day ²⁴. The start date of an outbreak was defined as the date on which the first case (symptomatic or asymptomatic) occurs, going backwards from the date on which there were more than 20 cases in a single day. The end of an outbreak was defined as the day with no new cases for 7 consecutive days after the peak of the outbreak. Outbreaks with a duration longer than 14 days were included in the study.

Estimation of time-varying effective reproduction number (R_t)

Since there was pre-symptomatic transmission for SARS-CoV-2 ²⁵, reconstructing the epidemic curve by the date of infection could provide a more accurate estimation of R_t. Therefore, we first reconstructed the epidemic curve by infection date based on the epidemic curve by the report date using a deconvolution approach ²⁶, with the distribution of the delay from infection to report (Appendix section 3). Then, we estimated the R_t based on the Poisson framework in Cori et al. ¹

Relationship between transmission and mobility

We computed the cross-correlation between R_t and mobility indices for each identified outbreak using the Pearson correlation and selecting the optimal lag day based on the highest correlation. We combined correlation coefficients from each city and weighted the standard error of estimates to generate a weighted average (Appendix section 4).

Furthermore, we adopted rolling correlation between R_t and three mobility indices to measure and visualize short-term but potentially time-varying correlations ^19,27. We performed biweekly rolling correlation analysis, where the correlation on day t was estimated based on the two time series data from day t -13 to t, covering a period of 14 days. We only included city outbreaks with duration at least 42 days to ensure sufficient amount of data for estimation. We also conducted triweekly rolling correlation analyses to further explore the sensitivity of our results.

To determine if using rolling correlation was necessary, i.e. the magnitude is changing during outbreaks and not constant, we employed the non-linear least squares (NLS) method to fit the rolling correlation for each outbreak. We fitted five models (constant, linear, quadratic, sine, and cosine), and chose the optimal one based on the smallest Akaike Information Criterion (AIC) value. We also compared the cross-correlation coefficient with the minimum and maximum values of the rolling correlation coefficient.

Factors affecting cross-correlation and rolling correlation

We aimed to investigate what factors may impact the cross-correlation and rolling correlation. Regarding cross-correlation, we conducted Pearson correlation tests on several potential factors, including outbreak duration, peak value of R_t, and GRI.

Regarding rolling correlation, we investigated if rolling correlations were different by stages of outbreaks, and level of GRI. Outbreaks were divided into two stages, namely pre-peak and post-peak stage, based on the peak value of R_t (Appendix section 5). We employed a mixed-effect regression to assess the impact of different stages, and GRI, on the rolling correlation between R_t and the mobility index. Analysis stratified by outbreak stage was also conducted. The mixed-effect regression model included a random intercept term to account for variations among different outbreak cities. The rolling correlation for each outbreak was used as the outcome variable. The city-level GRI and different stages served as the predictor variable, respectively. We applied a Fisher transformation to the rolling correlation before fitting the models.

The significance level for all tests was set at p < 0.05. Processing Python 3.8.6 (Python Software Foundation) and related libraries were utilized to capture the required data. All analyses were performed using R software version 3.6.3 (R Foundation for Statistical Computing, Austria).

Informed consent was not required for this study since the data used was obtained from publicly available data sources.

Identifications of outbreaks

In 2022, mainland China experienced two major waves caused by the Omicron variants. Wave 1 (February to June) was caused by the Omicron BA.2 variant and resulted in a total of 766, 000 cases. We identified 57 outbreaks in 57 cities across 23 provinces during this period. Wave 2 (July to December) was caused by the more transmissible Omicron BA.5 variants. However, it is worth noting that the surveillance was gradually discontinued in different regions between November 27 and December 7, 2022 ²⁸, due to the cessation of the zero-COVID policy in mainland China ²⁹. As of November 27, there were a total of 554,000 cases, and we identified 171 outbreaks in 171 cities across 29 provinces during wave 2.

For each outbreak in each city, we first estimated time-varying effective reproduction number (R_t). Then we computed cross-correlation and rolling-correlation, and tested models with different assumptions on variation of correlation during outbreaks. We used the outbreak in Shanghai during wave 1 as an example (Fig. 2). Details of each outbreak were available in Figure S1 and S2.

Cross-correlation

We found that a lag of zero had the highest correlation between R_t and mobility indices for all outbreaks in wave 1 and 2 (Figure S5). Overall, we observed positive cross-correlations between R_t and three mobility indices (Fig. 3), with weighted averages ranging from 0.64–0.71 in wave 1 and 0.45–0.46 in wave 2. However, 12%-23% of wave 1 outbreaks and 22%-26% of wave 2 outbreaks showed uncorrelated R_t and mobility indices, while 5%-9% of wave 1 outbreaks and 19%-24% of wave 2 outbreaks showed negative correlations. Among the 40 cities with outbreaks in both wave 1 and wave 2, we found no statistically significant correlation between the cross-correlation in two waves from the same cities (correlation: 0.08, p = 0.61).

Rolling correlation

The rolling correlation among 27 outbreaks in wave 1 and 64 outbreaks in wave 2 demonstrated frequent and substantial fluctuations during the outbreaks lasted at least 42 days, particularly those with prolonged outbreaks. The rolling correlation of using three different types of mobility indices for an outbreak showed a similar changing pattern (Fig. 4, Figure S1, and Figure S2). In particular, we observed an oscillating pattern between R_t and mobility indices.

Regression analysis showed that the constant model performed significantly worse than non-constant models in all outbreaks (Fig. 4D, 4E, and 4F). Specifically, among 91 outbreaks (27 in wave 1 and 64 in wave 2), the sine/cosine models yielded the best fit in 68–75 (75%-82%) outbreaks for within-city movement, inter-city inflow, and inter-city outflow, respectively. In the remaining outbreaks, the best fit was achieved using either a linear or quadratic model. sensitivity analysis using triweekly rolling correlation instead of biweekly rolling correlation showed smoother patterns (Figure S3 and S4), but the constant model still performed significantly worse than all other models (Figure S6).

We also tested using GAM model to fit the rolling correlation and the results strongly support non-linearity (Figure S7). We also tested the relationship between the rolling correlations of wave 1 and wave 2 for the same city (Appendix section 6.4), but found no statistically significant correlation between the rolling correlations in the two waves among 15 cities with outbreaks in both wave 1 and wave 2.

Comparison of cross-correlation and rolling correlation

No correlation was found between the magnitude of the cross-correlation and the extreme values of the rolling correlation (Fig. 5). There were negative values of minimum of rolling correlation among 86%-90% of outbreaks (Fig. 5A, 5C, and 5E). Moreover, the sign of cross-correlation and minimun of rolling correlation was opposite among 68–77% of outbreaks. On the other hand, the maximum of rolling correlation was almost 1 in all outbreaks, but the cross-correlation is negative in 13%-18% of outbreaks (Fig. 5B, 5D, and 5F).

Factors affecting cross-correlation and rolling correlation

We found negative correlations between the duration of outbreaks and cross-correlations between R_t and three mobility indices during both wave 1 and wave 2, ranging from − 0.37 to -0.21 (Table 1). In wave 2, the peak value of R_t was positively correlated with the cross-correlation between R_t and within-city movement and inter-city inflow, but not the inter-city outflow.

Table 1

Factors affecting cross-correlations and rolling correlation between R_t and mobity index.
Correlation	Wave	Factor	Within-city movement	Inter-city inflow	Inter-city outflow
Cross-correlation	Wave 1	Peak value of R_t	-0.07 (-0.32, 0.20)	-0.09 (-0.34, 0.18)	-0.05 (-0.31, 0.21)
		Outbreak Duration	-0.35 (-0.56, -0.10)	-0.37 (-0.58, -0.12)	-0.35 (-0.56, -0.10)
		Maximum value of GRI	0.19 (-0.07, 0.43)	0.20 (-0.06, 0.44)	0.27 (0.01, 0.49)
	Wave 2	Peak value of R_t	0.31 (0.15, 0.45)	0.24 (0.07, 0.39)	0.12 (-0.04, 0.28)
		Outbreak Duration	-0.22 (-0.37, -0.05)	-0.25 (-0.40, -0.09)	-0.21 (-0.36, -0.04)
		Maximum value of GRI	-0.03 (-0.20, 0.13)	-0.04 (-0.21, 0.12)	-0.05 (-0.22, 0.12)
Rolling correlation (Fisher transformation)	Wave 1: Pre-peak stage	GRI	0.13 (0.07, 0.20)	0.06 (0.00, 0.11)	0.09 (0.03, 0.15)
	Wave 1: Post-peak stage	GRI	-0.05 (-0.07, -0.03)	-0.05 (-0.06, -0.03)	-0.02 (-0.03, 0.00)
	Wave 1	Pre-peak	ref	ref	ref
	Wave 1	Post-peak	0.05 (-0.12, 0.23)	0.33 (0.17, 0.50)	0.25 (0.09, 0.42)
GRI, Government Response Index.
For cross-correlation, Pearson correlation coefficients and 95% confidence intervals (CI) were displayed.
For rolling correlation, fixed effect results, including coefficients and 95% CI, were shown.

Based on the 27 completed outbreaks in wave 1, mixed-effect regression analysis revealed that the rolling correlation for inter-city inflow and inter-city outflow was significantly positively correlated with the post-peak stage, relative to the pre-peak stage (Table 1). For inter-city inflow and inter-city outflow, the rolling correlation increased from 0.50 (95% CI: 0.26, 0.68) to 0.71 (95% CI: 0.57, 0.81) (p < 0.01), and from 0.50 (95% CI: 0.29, 0.66) to 0.66 (95% CI: 0.52, 0.77) (p < 0.01) respectively, when the stage changed from pre-peak to post-peak. However, there was no significant change for within-city movement (Fig. 6A).

We explored the relationship between rolling correlation and GRI for 12 outbreaks in wave 1 due to availability of GRI data. We estimated a positive correlation between rolling correlation and GRI during the pre-peak stage, but a negative correlation during the post-peak stage (Fig. 6B and 6C). Specifically, during the pre-peak stage, when the GRI increased from 55 to 75, the rolling correlation was estimated to be increased from 0.04 to 0.99 for within-city movement (p < 0.01), and from 0.31 to 0.97 for inter-city outflow (p < 0.01), respectively, but no significant change for inter-city inflow (Fig. 6B). In contrast, during the post-peak stage, when the GRI increased from 55 to 75, the rolling correlation was estimated to be decreased from 0.86 to 0.33 for within-city movement (p < 0.01), from 0.89 to 0.45 for inter-city inflow (p < 0.01), and from 0.78 to 0.57 for inter-city outflow (p = 0.01) (Fig. 6C).

To assess the robustness of our findings, we conducted analyses using different GRI data extraction thresholds, and consistently observed similar results (Figure S10).

Comparison of cross-correlation and rolling correlation across three mobility indices

Inter-city inflow and inter-city outflow had high average correlations (0.88–0.96) in both waves, while within-city movement had lower correlations with inter-city inflow/outflow (0.31–0.35) (Appendix section 8.1). However, the cross-correlations between R_t and mobility of within-city movement, inter-city inflow, and inter-city outflow were highly correlated with each other (ranging from 0.63 to 0.88, Appendix section 8.2). Additionally, mixed-effect regression analysis showed no significant differences in rolling correlations for these three mobility indices (Figure S12).

Previous studies usually explored the relationship between mobility and transmisison at the province or national level. Here, we analyzed this realationship at a finer scale, namely city level, using data from Omicron outbreaks in mainland China in 2022, and three types of mobility indices (within-city movement, inter-city inflow, and inter-city outflow). Overall, we found their relationship could be different among cities and outbreaks, suggesting directly using mobility to proxy transmsision intensity may be inaccurate. In particular, we found the cross-correlation and rolling correlation between the waves for the same city could be different. Futhermore, the rolling correlations were also non-constant in all outbreaks, suggesting the relationship between transmision and mobility was time-varying during outbreaks. Despite these variations, we estimated the rolling correlation was higher after peak in an outbreaks, and intensity of control measured (measured by GRI) could modify rolling correlation.

We found that the cross-correlation between R_t and mobility indices in outbreaks in different waves in the same cities could be different, particularly the cross-correlation during wave 2 was significantly lower than that of wave 1. This was consistent with previous studies reporting lower correlation during later waves ^3,30 than earlier waves. It maybe explained by increased transmisbility of the dominated virus, in which Omicron BA.5 in wave 2 was more trasnmissible than Omicron BA.2 ³¹. Another potential explaination was the pandemic fatigue ³² that self-protection behavrior may have changed in later wave, which cannot be fully capture by mobility indices. Therefore, the previous estimated relationship between mobility and transsmission may not be generalizable to future outbreaks even in the same regions.

By using rolling correlation, we found that the relationship between mobility and transmission varied during outbreaks at the city level. Overall, mobility was positively correlated with transmisisons, but in some outbreaks, the minimum of rolling correlation could be negative. This finding was consistent with a previous study that showed models allowing for different associations between R_t and mobility in subperiods outperformed models without subperiods ³³. Furthermore, prior studies conducted at both province and city levels have demonstrated significant fluctuations in the rolling correlation between R_t and mobility indices throughout the entire epidemic period ^19,27.

Increased mobility is expected to increase transmission due to more contacts in the community, but the magnitude of this effect can vary, resulting in a time-varying association between R_t and mobility. This variability may be due to 1) factors that may affect transmission but not mobility, 2) factors that may disproportionally affect mobility and transmsision, and 3) transmissions could have impact on mobility. First, mobility in the community could not capture transmission intensity in other settings of smaller scales, such as within households or neighborhoods. Second, individual behaviors may change during outbreaks, including the degree of self-protection and contact patterns ^34,35 that may have impact on tranmission intensity, but could not be captured by mobility indices. Also higher transmissions may increase the individual self-protective behaviors, such as self-isolation or reducing going outside. Third, the intensity of self-protective behavior, and intensity and adherence to implemented containment measures may also change due to pandemic fatigue ³², which can disproportionally affect both mobility and R_t ^36,37. On the other hand, prolonged and repeated outbreaks may also cause pandemic fatigue.

We estimated that the higher GRI was associated with higher rolling correlation before epidemic peak but lower rolling correlation after peak, which may explain the time-varying relationship between mobility and transmissions. As containment measure may reduce both mobility and transmsision, causing positive and higher rolling correlatoin in early stage of outbreaks. However, in later stage of outbreaks, there could be implementations of further measures, such as facial coverings, depending on the transsmision intesnity, that could weaken the relationship between R_t and mobility ³⁸. For example, higher transmission intensity may further trigger implementation of measures that increased GRI, resulting decreased rolling correlation. Additionally, when transmission began to decline, control measures were not immediately relaxed, causing a delay in the decline of GRI. Furthermore, a decrease in risk perception and compliance towards control measures among public could lead to an increase in mobility ³⁹, resulting in a weaker relationship between R_t and mobility despite high GRI.

The oscillation observed in the rolling correlation between transmission and mobility highlighted multiple factors may make their relationship dynamic. Hence, the predicting transmsision based on mobility assuming a constant relationship between these variables could be inaccurate. In partciular, for almost all outbreaks the maximum of rolloing correlation could reach one, suggesting that using cross-correlation may underuse the mobility data. More complicated models may be needed to ultize the mobility data to proxy transmission intensity.

When comparing rolling correlation calculated by different mobiltiy indices, there were no significant differences in rolling correlation during the pre-peak stage and during the post-peak stage when using within-city movement as the mobility index. However, there were significant differences when inter-city inflow or outflow were used. Theses observations suggest that within-city movement may be a more sensitive indicator of outbreaks and capture outbreak information earlier than the other two mobility indices.

This study has several limitations. First, our analysis relied on the accuracy of the mobility data in measuring human movement within and between cities, and this may not capture all important flow changes of the public during the epidemic period ⁴⁰. Second, there were missing data, such as city-level case data for Yunnan and Xinjiang provinces. Third, extracting city-level GRI from the province-level data resulted in information loss for some outbreaks, as not all cities had city-level GRI available. Lastly, we could not validate the relationship between rolling correlation and GRI using data from outbreaks in wave 2, due to the presence of censored data. Further validation is necessary to ensure the strength and reliability of our findings.

In conclusion, using data from a finer scale (city level), our study revealed the dynamic relationship between mobility and transmission, varying across different waves, outbreaks, stages within an outbreak, and level of government response. These suggested that when employing mobility index as a proxy of real-time transmissibility, assuming a constant relationship between them throughout the entire stage may result in inaccurate evaluation of transmisison intesnity. Therefore, nowcasting and forecsating epidemic using mobility may requre further consideration of other factors and development of methodology.

Competing interests

BJC reports honoraria from AstraZeneca, Fosun Pharma, GSK, Haleon, Moderna, Pfizer, Roche and Sanofi Pasteur. All other authors reported no competing interest.

Author contributions

TKT, KECA and BJC were responsible for study design. LP and XH were responsible for data collection. LP was responsible for data analysis. LP, KECA, BJC, PW, TKT were responsible for data interpretation. LP wrote the first draft of the manuscript. All authors contributed to the final draft.

Acknowledgments

The authors thank Zirong Liu for technical assistance. This work was supported in part with federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services, under contract no. 75N93021C00016, and the Theme-based Research Scheme (Project No. T11-705/21-N) of the Research Grants Council of the Hong Kong SAR Government. BJC is supported by an RGC Senior Research Fellowship (grant number: HKU SRFS2021-7S03) and the AIR@innoHK program of the Innovation and Technology Commission of the Hong Kong SAR Government.

Data availability

The raw data analyzed in the study is sourced from publicly available data. The aggregated dataset used for the analysis in this study is provided upon request.

Cori, A., Ferguson, N. M., Fraser, C. & Cauchemez, S. A new framework and software to estimate time-varying reproduction numbers during epidemics. Am J Epidemiol 178, 1505–1512, doi:10.1093/aje/kwt133 (2013).
Gostic, K. M. et al. Practical considerations for measuring the effective reproductive number, Rt. PLoS Comput Biol 16, e1008409, doi:10.1371/journal.pcbi.1008409 (2020).
Lison, A., Persson, J., Banholzer, N. & Feuerriegel, S. Estimating the effect of mobility on SARS-CoV-2 transmission during the first and second wave of the COVID-19 epidemic, Switzerland, March to December 2020. Euro Surveill 27, doi:10.2807/1560-7917.ES.2022.27.10.2100374 (2022).
Kajitani, Y. & Hatayama, M. Explaining the effective reproduction number of COVID-19 through mobility and enterprise statistics: Evidence from the first wave in Japan. PLoS One 16, e0247186, doi:10.1371/journal.pone.0247186 (2021).
Hu, T. et al. Human mobility data in the COVID-19 pandemic: characteristics, applications, and challenges. International Journal of Digital Earth 14, 1126–1147, doi:10.1080/17538947.2021.1952324 (2021).
Zhang, M. et al. Human mobility and COVID-19 transmission: a systematic review and future directions. Annals of GIS 28, 501–514, doi:10.1080/19475683.2022.2041725 (2022).
Tan, S. et al. Mobility in China, 2020: a tale of four phases. Natl Sci Rev 8, nwab148, doi:10.1093/nsr/nwab148 (2021).
Unacast, <https://www.unacast.com/> (2023).
EUROCONTROL. Eurocontrol Dashboard, <https://www.eurocontrol.int/ServiceUnits/Dashboard/EnRouteMainDashboard.html> (2023).
Leung, K., Lau, E. H. Y., Wong, C. K. H., Leung, G. M. & Wu, J. T. Estimating the transmission dynamics of SARS-CoV-2 Omicron BF.7 in Beijing after adjustment of the zero-COVID policy in November-December 2022. Nat Med 29, 579–582, doi:10.1038/s41591-023-02212-y (2023).
Leung, K., Wu, J. T. & Leung, G. M. Real-time tracking and prediction of COVID-19 infection using digital proxies of population mobility and mixing. Nat Commun 12, 1501, doi:10.1038/s41467-021-21776-2 (2021).
Google. COVID-19 Community Mobility Reports, <https://www.google.com/covid19/mobility/> (2023).
Apple. Mobility Trends Reports, <https://covid19.apple.com/mobility> (2022).
da Silva, T. T., Francisquini, R. & Nascimento, M. C. V. Meteorological and human mobility data on predicting COVID-19 cases by a novel hybrid decomposition method with anomaly detection analysis: A case study in the capitals of Brazil. Expert Syst Appl 182, 115190, doi:10.1016/j.eswa.2021.115190 (2021).
Setti, M. O. & Tollis, S. In-depth Correlation Analysis of SARS-CoV-2 Effective Reproduction Number and Mobility Patterns: Three Groups of Countries. J Prev Med Public Health 55, 134–143, doi:10.3961/jpmph.21.522 (2022).
Gottumukkala, R. et al. Exploring the relationship between mobility and COVID- 19 infection rates for the second peak in the United States using phase-wise association. BMC Public Health 21, 1669, doi:10.1186/s12889-021-11657-0 (2021).
Nouvellet, P. et al. Reduction in mobility and COVID-19 transmission. Nat Commun 12, 1090, doi:10.1038/s41467-021-21358-2 (2021).
Zheng, J. X. et al. The rapid and efficient strategy for SARS-CoV-2 Omicron transmission control: analysis of outbreaks at the city level. Infect Dis Poverty 11, 114, doi:10.1186/s40249-022-01043-2 (2022).
Ainslie, K. E. C. et al. Evidence of initial success for China exiting COVID-19 social distancing policy after achieving containment. Wellcome Open Res 5, 81, doi:10.12688/wellcomeopenres.15843.2 (2020).
Ji, H. et al. The effectiveness of travel restriction measures in alleviating the COVID-19 epidemic: evidence from Shenzhen, China. Environ Geochem Health 44, 3115–3132, doi:10.1007/s10653-021-00920-3 (2022).
National Health Commission of the People's Republic of China. Tracking the COVID-19 Epidemic in China < http://www.nhc.gov.cn/xcs/yqtb/list_gzbd> (2024).
Baidu. Baidu cityCode, <https://lbsyun.baidu.com/index.php?title=open/%E5%BC%80%E5%8F%91%E8%B5%84%E6%BA%90> (2023).
Blavatnik School of Government & University of Oxford. Oxford Covid-19 Government Response Tracker (OxCGRT), <https://github.com/OxCGRT/covid-policy-tracker> (2023).
Zha, H. et al. "Chinese Provincial Government Responses to COVID-19". Version 2.. Blavatnik School of Government Working Paper (2022).
He, X. et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nat Med 26, 672–675, doi:10.1038/s41591-020-0869-5 (2020).
Miller, A. C. et al. Statistical Deconvolution for Inference of Infection Time Series. Epidemiology 33, 470–479, doi:10.1097/EDE.0000000000001495 (2022).
Ji, H. et al. The effectiveness of travel restriction measures in alleviating the COVID-19 epidemic: evidence from Shenzhen, China. Environ Geochem Health, doi:10.1007/s10653-021-00920-3 (2021).
Peng, L. et al. Comparative epidemiology of outbreaks caused by SARS-CoV-2 Delta and Omicron variants in China. Epidemiol Infect 152, e43, doi:10.1017/s0950268824000360 (2024).
Goldberg, E. E., Lin, Q., Romero-Severson, E. O. & Ke, R. Swift and extensive Omicron outbreak in China after sudden exit from 'zero-COVID' policy. Nat Commun 14, 3888, doi:10.1038/s41467-023-39638-4 (2023).
Dainton, C. & Hay, A. Quantifying the relationship between lockdowns, mobility, and effective reproduction number (Rt) during the COVID-19 pandemic in the Greater Toronto Area. BMC Public Health 21, 1658, doi:10.1186/s12889-021-11684-x (2021).
Xu, A. et al. Sub-lineages of the SARS-CoV-2 Omicron variants: Characteristics and prevention. MedComm (2020) 3, e172, doi:10.1002/mco2.172 (2022).
Gao, H. et al. Pandemic fatigue and attenuated impact of avoidance behaviours against COVID-19 transmission in Hong Kong by cross-sectional telephone surveys. BMJ Open 11, e055909, doi:10.1136/bmjopen-2021-055909 (2021).
Kurita, J., Sugishita, Y., Sugawara, T. & Ohkusa, Y. Evaluating Apple Inc Mobility Trend Data Related to the COVID-19 Outbreak in Japan: Statistical Analysis. JMIR Public Health Surveill 7, e20335, doi:10.2196/20335 (2021).
Zhang, J. et al. Changes in contact patterns shape the dynamics of the COVID-19 outbreak in China. Science 368, 1481–1486, doi:10.1126/science.abb8001 (2020).
Kraemer, M. U. G. et al. The effect of human mobility and control measures on the COVID-19 epidemic in China. Science 368, 493–497, doi:10.1126/science.abb4218 (2020).
Huisman, J. S. et al. Estimation and worldwide monitoring of the effective reproductive number of SARS-CoV-2. Elife 11, doi:10.7554/eLife.71345 (2022).
Du, E., Chen, E., Liu, J. & Zheng, C. How do social media and individual behaviors affect epidemic transmission and control? Sci Total Environ 761, 144114, doi:10.1016/j.scitotenv.2020.144114 (2021).
Bergman, N. K. & Fishman, R. Correlations of mobility and Covid-19 transmission in global data. PLoS One 18, e0279484, doi:10.1371/journal.pone.0279484 (2023).
Petherick, A. et al. A worldwide assessment of changes in adherence to COVID-19 protective behaviours and hypothesized pandemic fatigue. Nat Hum Behav 5, 1145–1160, doi:10.1038/s41562-021-01181-x (2021).
Kishore, N. et al. Evaluating the reliability of mobility metrics from aggregated mobile phone data as proxies for SARS-CoV-2 transmission in the USA: a population-based study. Lancet Digit Health 4, e27-e36, doi:10.1016/s2589-7500(21)00214-4 (2022).

Yes there is potential Competing Interest. BJC reports honoraria from AstraZeneca, Fosun Pharma, GSK, Haleon, Moderna, Pfizer, Roche and Sanofi Pasteur. All other authors reported no competing interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

Evaluating the association between COVID-19 transmission and mobility in Omicron outbreaks in China

Status:

Version 1

Abstract

Figures

Plain language summary

INTRODUCTION

METHODS

Case data

Mobility index data

Government Response Index (GRI)

Definition of outbreaks

Estimation of time-varying effective reproduction number (R_t)

Relationship between transmission and mobility

Factors affecting cross-correlation and rolling correlation

RESULTS

Identifications of outbreaks

Cross-correlation

Rolling correlation

Comparison of cross-correlation and rolling correlation

Factors affecting cross-correlation and rolling correlation

Comparison of cross-correlation and rolling correlation across three mobility indices

DISCUSSION

Declarations

Competing interests

Author contributions

Acknowledgments

Data availability

References

Additional Declarations

Supplementary Files

Status:

Version 1

Evaluating the association between COVID-19 transmission and mobility in Omicron outbreaks in China

Status:

Version 1

Abstract

Figures

Plain language summary

INTRODUCTION

METHODS

Case data

Mobility index data

Government Response Index (GRI)

Definition of outbreaks

Estimation of time-varying effective reproduction number (Rt)

Relationship between transmission and mobility

Factors affecting cross-correlation and rolling correlation

RESULTS

Identifications of outbreaks

Cross-correlation

Rolling correlation

Comparison of cross-correlation and rolling correlation

Factors affecting cross-correlation and rolling correlation

Comparison of cross-correlation and rolling correlation across three mobility indices

DISCUSSION

Declarations

Competing interests

Author contributions

Acknowledgments

Data availability

References

Additional Declarations

Supplementary Files

Status:

Version 1

Estimation of time-varying effective reproduction number (R_t)