key: cord-1000728-zmx4hh1n
authors: KATO, H.; Takizawa, A.
title: Time series cross-correlation between home range and number of infected people during the medium term of COVID-19 Pandemic in a suburban city
date: 2022-04-10
journal: nan
DOI: 10.1101/2022.04.07.22273581
sha: 6106c2ca013bebf3368a1fd8f7f2f5f1be44a5f7
doc_id: 1000728
cord_uid: zmx4hh1n

Control of human mobility is among the most effective measures to prevent the spread of coronavirus disease 2019 (COVID-19). This study aims to clarify the correlation between home range and the number of people infected with SARS-CoV-2 during the medium-term of the COVID-19 pandemic in Ibaraki City. Home ranges are analyzed by the Minimum Convex Polygon method using mobile phone GPS location history data. We analyzed the time series cross-correlation between home range lengths and the number of infected people. Results reveal a slight positive correlation between home range and the number of infected people after one week during the medium-term of the COVID-19 pandemic. Regarding home range length, the cross-correlation coefficient is 0.4030 even at a lag level of six weeks, which has the most significant coefficient. Thus, a decrease in home range is only one of the indirect factors contributing toward a reduction in the number of infected people. This study makes a significant contribution to the literature by evaluating key public health challenges from the perspective of controliing the spread of the COVID-19 infectuion. Its findings has implications for policy makers, practitioners, and urban scientists seeking to promote urban sustainability.

The coronavirus disease 2019 pandemic has drastically changed our daily lives. The rapid increase in 29 the number of infected people risks causing a breakdown of the medical system. Control of human mobility is considered 30 one of the most effective measures to prevent the rapid spread of COVID- 19 [1] . For example, in the Osaka metropolitan 31 area, states of emergency have been declared four times since January 2020 [2] , with more substantial restrictions imposed 32 on the activities of people living in areas closer to the city center [3] . The Subcommittee on Novel Coronavirus Disease 33

Control in Japan requested citizens to reduce human mobility by 50% during this time [4] . However, the imposition of 34 emergency restrictions had significant negative impacts on the daily lives of citizens. For example, excessive restrictions 35 caused a deterioration of mental health [5] . As vaccination progresses, we need to consider more effective measures to 36 control the spread of the infection. 37

The research question of this study is as follows: Does the control of home range affect a reduction in the number of 38 infected people during the medium-term of the COVID-19 pandemic? In other words, this study verifies the possibility of 39 predicting the number of infected people based on the control of human mobility. In particular, it is difficult to predict the 40 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Purpose 48

This study aims to clarify the time series cross-correlation between home range and the number of people infected 49 with SARS-CoV-2 during the medium-term of the COVID-19 pandemic in a suburban city. Home ranges were analyzed 50 by the Minimum Convex Polygon (MCP) method using mobile phone GPS location history (LH) data. LH data includes 51 location history data collected from individual devices, unlike area-based data such as Google mobility reports [8] . A 52 timeseries analysis was conducted using panel data of every Wednesday from April 2020 to July 2021, a time frame where 53 four waves of the pandemic were witnessed in Japan. This study analyzed the time series cross-correlation between the 54 home range lengths and the number of infected people. 55

The case study research was conducted in Ibaraki City, which is a typical suburban city in the Osaka metropolitan 56 area. Fig 1 shows the location of Ibaraki City, which has a population of approximately 280,000, and an area of 10 km east-57 west and 17 km north-south [9] . Due to the city's extensive train network, residents can commute in about 30 min to Osaka 58 City or Kyoto City. Fig 1 shows Many studies on human mobility use mobile phone data to estimate the number of people infected with SARS-CoV-73 2. In particular, many studies focused on Wuhan, where the first infected person was identified. In the early stages of the 74 pandemic, the spatial distribution of the number of infected people was found to explain the movement of the population 75 from Wuhan between January 1 and January 24, 2020 [13] . After January 24, 2020, findings suggested that the lockdown 76 reduced the number of COVID-19 cases in Wuhan by limiting human mobility within the city [14] . Based on these results, 77 simulations to determine the feasibility of measures for the successful control and containment of the COVID-19 pandemic 78

showed the necessity of restricting human mobility by 20%-40%, using the case of Shenzhen in China [15] . In the United 79

States, human mobility was severely restricted during the early stages of the pandemic [16] . Using mobile phone data 80 across the United States from January 1 to April 20, 2020, it was found that COVID-19 transmission correlated strongly 81 with mobility patterns [17] . In addition, based on mobile phone data from March 1 to June 9, 2020, a positive correlation 82 was found between the number of infected people and mobility inflow at the country level in the United States [18] . Further, 83

in the European context, the spread of COVID-19 was positively correlated with the number of people staying in each area 84 and with human mobility between March 1 to June 6, 2020, at lag levels of one, two, and three weeks [19] . This could be 85 attributed to the fact that in Europe, lockdowns generally affect long-distance travel behavior [19] . In addition, between 86 January 1 and April 15, 2020, it was found that the estimated adequate reproduction number of COVID-19 correlated 87 strongly with human mobility (or social contact) in Tokyo, Japan [20] . The changes in human mobility pertaining to 88 nightlife spaces were more significantly associated with the number of COVID-19 cases [21] . These studies of the early 89 stages of the pandemic indicate that the number of infected people correlates with human mobility. 90

The novelty of this study is to analyze the medium-term relationship between the number of infected people and home 91 range. In Japan, the state of emergency was called a "soft lockdown" [22] because the Japanese government did not restrict 92 the activities of individuals [23] . Therefore, most citizens could at least go out in a limited capacity even under the state of 93 emergency. Thus, while multiple emergency declarations were in effect, the home range did not change significantly, but 94 the number of infected people decreased steadily. This suggests that factors other than home range might influence the 95 spread of the infection. This study will allow policymakers to develop policies for controlling the infection during the 96 COVID-19 pandemic in the medium to long term. 97 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint 5 98

Location History Data

The study used LH data collected by Agoop Corporation. Agoop Corporation collected the LH data by obtaining 101 consent from users, who contracted with a specific mobile phone carrier company or installed specific applications [24] . 102

All participants were provided with information regarding the type of data collected, purpose of use and provision to third 103 parties, and a privacy policy [25] . Agoop Corporation provides anonymized LH data for research purposes. Due to the 104 availability of high-quality data in a Japanese context, many studies utilized LH data relating to the spread of COVID-19 105

[10-12]. This research protocol was approved by the ethics committee of the Graduate School of Life Science, Osaka City 106

University . Additionally, all methods were carried out in accordance with "Guidelines for the Use of Device 107

Location Data," a common regulation for location data analysis in Japan [26] . The guideline prohibits using GPS data for 108 any purpose that involves identifying individual users to protect the privacy of users' GPS location history. 109

Depending on the mobile phone type, the LH data were collected in the form of logs approximately every 15 min, 110 and the LH data were obtained from mobile phones with users' consent. In Ibaraki City, the number of logs was 111 approximately 1,600,000 per day, and the number of users was approximately 12,000, indicating that 5% of the residents 112 of Ibaraki City was adequate for analyzing the home range of residents. 113

The variables in the LH data used in this study were user ID, year, month, day, hour, minutes, and latitude and 114 longitude. The user IDs are anonymized 96-digit alphanumeric codes, the permanent ID assigned to each device, and enable 115 panel data analysis. 116

This study used a time series analysis method to analyze the cross-correlation between home range length and the 118 number of infected people. The time-series cross-correlation allows us to understand the similarity of data in a time series 119 and the lag of the period. In this study, the lag was set to eight weeks, considering that the duration between the date of 120 exposure and onset of symptoms is usually a few weeks [27] . 121

The analysis period was from April 2020 to July 2021, during which Japan experienced four waves, and a state of 122 emergency was imposed four times in Ibaraki City. On April 7, 2020, when a state of emergency was declared in all 123 prefectures, the Japanese government requested people to stay at home. Following the end of this state of emergency, the 124 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint Japanese government attempted to recover the economy through various measures such as the "Go-To Travel" campaign 125 for the hotel and restaurant industries [28] . However, by the winter of 2020, the number of infected people had gradually 126 increased [29] . A second state of emergency was subsequently declared from January 14 to February 28, 2021. The Japanese 127 government also developed priority preventive measures before the next emergency declaration Distance) allows us to estimate the total travel distance. The HR-length (Total Travel Distance) was used as an indicator in 144 a previous study [20] . 145

The analysis process is summarized in Fig 2. Changes in HR-length (Farthest Distance) and HR-length (Total Travel 146 Distance) during the COVID-19 pandemic were analyzed. In addition, these two HR-lengths were analyzed the cross-147 correlation with the number of people infected. 148

The study analyzed LH data in Ibaraki City. However, this data included the logs of people who only passed through 149

Ibaraki City, such as people commuting from Tokyo to Osaka by express. Therefore, to isolate the data of people living in 150

Ibaraki City, the study extracted user IDs of the first log located in Ibaraki City after 0:00 h of every day. The study then 151 analyzed the user ID data that appeared for more than two days in the analyzed period. The intent of the study was to 152 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint analyze the user IDs of individuals living in Ibaraki City, not those just passing through it. With respect to the LH data of 153 the user ID, this study analyzed the home range from 0:00 to 23:59 h. 154

This study used LH data from every Wednesday and analyzed the change in home range based on this data from every 155 Wednesday, considering the home range changes that occurred between weekdays and holidays [10]. On weekdays, people 156 are involved in steady-state activities such as work, which is suitable for analyzing the impact of the COVID-19 pandemic. 157

Wednesday was chosen as the most appropriate day of the week to analyze the impact of the COVID-19 pandemic for a 158 number of reasons. Wednesday is the weekday with the fewest holidays from April 2020 to July 2020: only May 6, 2020, 159

and May 5, 2021, were holidays. Furthermore, December 30, 2020, marked the beginning of the year-end vacation. 160

Moreover, on December 30, many people tend to be off work and school during the year-end vacations. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint 8 (Farthest Distance) and HR-length (Total Travel Distance) from April 2020 to July 2021. In both figures, weekly and 180 monthly data followed similar trends. In particular, the home range showed a significant decrease every week from April 181 to May 2020, when the first state of emergency was declared. This result was verified by previous research [10] [11] [12] . 182

However, since June 2020, the home range has gradually increased. It was found that the HR-length decreased significantly 183 only from April to May 2020, when the first emergency declaration was issued, but thereafter, travel to Ibaraki City and 184

Osaka City did not change significantly. 185 the number of infected people. The first wave lasted from April to May 2020, the second from July to September 2020, the 208 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint third from December 2020 to February 2021, and the fourth from March to June 2021. The fifth wave began in July 2021. 209

Each wave lasted progressively longer than the preceding one. 210

States of emergency were declared for the first, third, and fourth waves. Following each declaration of emergency, 211 the number of infected people decreased significantly. In particular, the number of infected people decreased to zero from 212

May 17-23, 2020, at which time the first emergency declaration was lifted; to five people from February 21-27, 2021, 213 when the second emergency declaration was lifted; and to 13 people from June 13-19, 2021, when the third emergency 214 declaration was lifted. This suggests that the declaration of a state of emergency effectively led to a decrease in the number 215 of infected people. To summarize these results, home range was found to be slightly positively correlated with the number of infected 233 people after six weeks, with the highest correlation coefficient being 0.40. This means that changing the home range 234 contributes to a decrease in the number infected people, but it is not a strong effect. 235 236 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 previous studies [18, 19] . In addition, prior research that analyzed the short-term period clarified a strong correlation [17, 20] . 244

In contrast to these studies, the novelty of the present study is that it clarifies the slight positive correlation between home 245 range and the number of infected people in the medium-term period following the onset of the COVID-19 pandemic. 246

It was found that the number of infected people was slightly more correlated with HR-length (Total Travel Distance) 247 than HR-length (Farthest Distance). Controlling travel distance is more effective than controlling the tendency to go out 248 farther. With regard to HR-length (Total Travel Distance), the CCC was found to be 0.4030 even at a lag level of six weeks, 249 which had the most significant coefficient. However, it was inexplicable at a lag level of six weeks. The reason for this is 250 the length of the incubation period of SARS-CoV-2; it takes approximately one-two weeks from the date of infection for 251 the first symptoms to appear [27] . Specifically, after the emergency declaration was issued, the number of infected people 252 decreased significantly in approximately one-two weeks. However, the home range had decreased even before the 253 emergency declaration was issued, and the decrease was not significant. This might be the reason for the low correlation 254 coefficient. It may be concluded that a decrease in home range is only one of the indirect factors contributing toward a 255 reduction in the number of infected people. 256

The conclusions suggest that factors other than the home range might contribute to a decrease in the number of 257 infected people; for instance, restrictions imposed under the state of emergency in the Osaka Prefecture, which applied to 258 the residents of Ibaraki City during this period. Citizens were required to refrain from not only non-urgent outings but also 259 from drinking alcohol in a group on the street or in a park; the operation of restaurants after 8:00 p.m. was also suspended, 260 and large-scale events were prohibited [32] . Restrictions with regard to wearing face masks might also have had an impact 261 on the rate of infection. For example, as the home range expands, one factor influencing the spread of infection might be 262 an increase in the number of situations in which people remove their masks, such as while eating lunch and smoking [33] . 263

In addition, during the third emergency declaration, household infections had become a significant problem, which was 264 associated with sharing a bedroom and speaking with an index case individual for 30 min or longer [34] . 265 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint Based on these results, it is possible to improve current measures for an emergency declaration. For instance, instead 266 of controlling human mobility, the number of infected people could be effectively reduced by the imposition of mask 267 mandates, reducing the opening hours of restaurants, and increasing the use of hotel facilities for medical treatment. These 268 measures would also make it possible for people to fulfill their work and study commitments while taking steps to protect 269 themselves from infection. 270

The limitation of this study is that we analyzed only the indicator of home range in Ibaraki City. It is necessary to 271 analyze not only the distance traveled using LH data, but also the place of stay using area-based data. The analysis might 272 provide a higher correlation coefficient. Further, due to privacy issues, the discrepancies between samples of the number 273 of infected people and human mobility pose a research challenge. Therefore, it is also necessary to analyze the actual 274 number of infected people and the distance they travel. Moreover, we analyzed data from Ibaraki City, a suburban city, but 275

human mobility in central cities should also be considered in determining the significant factors influencing infection 276 spread. In the future, it is necessary to study different types of cities to examine the correlation between home range and 277 the number of infected people in a Japanese context, such as Osaka City, the more metropolitan capital of the Osaka 278

Prefecture. 279 280 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint

The impact of policy measures on human 282 mobility, COVID-19 cases, and mortality in the US: A spatiotemporal perspective

Japanese Cabinet. Secretariat (office for COVID-19 and other emerging infectious disease control)

Change Restrictions Lift State Emerg

Spatiotemporal patterns of human mobility and its association with land 288 use types during COVID-19 in New York City

Press Conference by the Prime Minister 291 regarding the Novel Coronavirus

Mental health problems and social media 294 exposure during COVID-19 outbreak

Territoriality and home range concepts as applied to mammals

How Does the location of urban facilities affect the forecasted population change in the Osaka 299 metropolitan fringe area? Sustainability

Ibaraki open data. Population/Households

Impact of COVID-19 pandemic on home range in a suburban city 305 in the Osaka metropolitan area

Development of a spatio-temporal analysis method to support the prevention of COVID-19 307 infection: Space-time kernel density estimation using GPS location history data BT -Urban informatics 308 and future cities

Changes in walkable streets during the COVID-19 pandemic in a suburban city 311 in the Osaka metropolitan area

Population flow drives spatio-temporal 313 distribution of COVID-19 in China

The effect of human mobility 315 and control measures on the COVID-19 epidemic in China

Effects of human mobility restrictions on the

China: a modelling study using mobile phone data

Human mobility trends during the early 321 stage of the COVID-19 pandemic in the United States

the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted

Association between mobility 324 patterns and COVID-19 transmission in the USA: A mathematical modelling study

Mobile device data reveal the dynamics in a positive 327 relationship between human mobility and COVID-19 infections

The moderating effect of solar radiation on the 330 association between human mobility and COVID-19 infection in Europe

Non-compulsory measures 333 sufficiently reduced human mobility in Tokyo during the COVID-19 epidemic

Mobile data analysis of locations of infection

The Japan times. The Coronavirus Jpn's Constitution

Outline of the act partially amending the act on special measures against 342 novel influenza, etc., and other relevant acts

Guidelines for the use of device location data

Early transmission dynamics in Wuhan

of novel coronavirus-infected pneumonia

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted April 10, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 10, 2022. ; https://doi.org/10.1101/2022.04.07.22273581 doi: medRxiv preprint