Impact of human mobility and networking on spread of COVID-19 at the time of the 1st and 2nd epidemic waves in Japan: An effective distance approach.

BACKGROUND: The influence of human mobility to the domestic spread of COVID-19 in Japan using the approach of effective distance has not yet been assessed.
METHODS: We calculated the effective distance between prefectures using the data on laboratory-confirmed cases of COVID-19 from January 16 to August 23, 2020, that were times in the 1st and the 2nd epidemic waves in Japan. We also used the aggregated data on passenger volume by transportation mode for the 47 prefectures, as well as those in the private railway, bus, ship, and aviation categories. The starting location (prefecture) was defined as Kanagawa and as Tokyo for the 1st and the 2nd waves, respectively. The accuracy of the spread models was evaluated using the correlation between time of arrival and effective distance, calculated according to the different starting locations.
RESULTS: The number of cases in the analysis was 16,226 and 50,539 in the 1st and 2nd epidemic waves, respectively. The relationship between arrival time and geographical distance shows that the coefficient of determination was R2 = 0.0523 if geographical distance Dgeo and time of arrival Ta set to zero at Kanagawa and was R2 = 0.0109 if Dgeo and Ta set to zero at Tokyo. The relationship between arrival time and effective distance shows that the coefficient of determination was R2 = 0.3227 if effective distance Deff and Ta set to zero at Kanagawa and was R2 = 0.415 if Deff and time of arrival Ta set to zero at Tokyo. In other words, the effective distance taking into account the mobility network shows the spatiotemporal characteristics of the spread of infection better than geographical distance. The correlation of arrival time to effective distance showed the possibility of spreading from multiple areas in the 1st epidemic wave. On the other hand, the correlation of arrival time to effective distance showed the possibility of spreading from a specific area in the 2nd epidemic wave.
CONCLUSIONS: The spread of COVID-19 in Japan was affected by the mobility network and the 2nd epidemic wave is more affected than those of the 1st epidemic. The effective distance approach has the impact to estimate the domestic spreading COVID-19.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was first identified in December 2019 in Wuhan city, Hubei province, China, and has since become a worldwide pandemic [1, 2]. In Japan, the first confirmed case of coronavirus disease (COVID-19) was reported in January 15, 2020, in Kanagawa prefecture [3] and was a traveler returning from Wuhan [3]. Subsequently, laboratory-confirmed cases of COVID-19 were reported sporadically, and then increased gradually from the middle of February. The number of cases rapidly increased from approximately the end of March, and it leaded the first epidemic wave of COVID-19 in Japan. The Japanese government declared a state of emergency on April 7, 2020 [4] and many restrictions to reduce COVID-19 transmission have been taken, including the avoidance of unnecessary travel and limits on traveling across regional borders between prefectures. By the time the state of emergency was cancelled on March 25, 2020, a total of 16,581 COVID-19 cases (and 830 deaths) had been reported [5]. After the cancellation, restrictions were gradually relaxed and movements gradually increased, although people continued to wear masks and take the socially distance. For example, interprefecture travel increased and trains and railway stations became busier, particularly in urban areas. At approximately the end of June, the number of daily cases increased and a second epidemic wave was identified and was characterized by a higher number of cases than the first epidemic wave. The cumulative number of confirmed COVID-19 cases was 67,264 as of August 30, 2020 [6].

Japan has well-developed public transportation systems, such as trains, buses, and airlines, and people can easily move from place to place for business, pleasure, and other activities. COVID-19 transmission routes identified so far include sustained human-to-human transmission [7-9], transmission from an asymptomatic patient [10], and transmission from a pre-symptomatic patient [11]. It is likely that human mobility has had a strong effect on the spread of COVID-19 in Japan.

The concept of effective distance was introduced by Brockman and Helbing in 2013, using data from the influenza H1N1pdm09 pandemic and the outbreak of severe acute respiratory syndrome (SARS) in 2002–2003 SARS outbreak [12]. The concept is based on the idea that places with a dense flow of traffic between them should be effectively closer in a plausible map layout, and places with little traffic between them should be further apart [12]. Studies that have estimated the effective distance indicate that travel restrictions and international airline suspensions have contributed to the spread of COVID-19 [13, 14]. In addition, currently, there are some reports that confronted the effect of different mobility data, including people flow statistics, on the spatiotemporal distributions of SARS-CoV2 at sub-national level [15-17]. Thus, we hypothesized that the number of people traveling between Japanese cities may also affect the domestic spread of COVID-19. The effective distance may need to incorporate an estimate of the spread of COVID-19 to measure further domestic outbreaks.

The aim of the present study was to elucidate how the mobility of people in Japan affects the spread of COVID-19, and the impact of the approach of effective distance to estimate the spread of outbreaks of COVID-19 in Japan. The findings may contribute to further targeted control measures for COVID-19.

Materials and methods

Data

Japan is divided into 47 prefectures, with a population of approximately 125 million people [18] and 378 square kilometers of land [19].

Daily data on laboratory-confirmed cases of COVID-19 from January 16 to August 30, 2020, were obtained from publicly available situation reports on prefecture websites (S1 Table).

For mobility network data, we used the aggregated data on passenger volume by transportation mode for the 47 prefectures contained in the freight/passenger area flow survey conducted by the Ministry of Land, Infrastructure, Transport and Tourism of Japan issued in 2016 [20]. The data include the volume of passengers in the private railway, bus, ship, and aviation categories. The dataset counts all transport personnel across other prefectures in one year. It is the OD amount of passenger transport personnel between regions. For example, if the departure point is Tokyo and the arrival point is Osaka, all boarding / alighting personnel by prefecture on the route are counted. These data set was obtained from the “Kokudo Suchi,” the geographic information systems (GIS) data service of the Ministry of Land, Infrastructure, Transport and Tourism of Japan [21]. The mobility network diagram of this data consists of 47 nodes (i.e., prefectures) and 1,907 edges connecting prefectures. The value of transport volume excluding the edges is zero. The weight of each edge represents passenger volume between two nodes on all types of transportation (S1 Fig). Although the mobility network data were not generated during the COVID-19 epidemic, the probability of occurrence of movement between prefectures was assumed to be constant.

Effective distance

To assess the probability of COVID-19 spread within Japan, we calculated the effective distance between prefectures using the mobility network data. Previous studies have shown that effective distance, rather than geographical distance, can predict the arrival time of a virus. This metric was therefore used to identify the starting point of the virus spatial diffusion process.

The basic principle is that despite the structural complexity of the underlying network and the multiplicity of paths, the dynamic process is dominated by a set of most probable paths that can be derived from the connectivity matrix P, weighted by passenger volume. The effective distance d between the ith prefecture and the jth prefecture, which are directly connected, is defined as d = 1 –ln (P). P is the transition probability between prefectures. Moreover, the effective distance between an arbitrary reference prefecture and another prefecture in the network is calculated from the minimum of all possible paths.

Definition of dates and areas for the 1st and 2nd epidemic waves in Japan

During the observational period, the cutoff date between the 1st and the 2nd epidemic peaks was set at May 25, 2020, which was the date the state of emergency was cancelled. This date also presented the lowest number of daily cases between the 1st and 2nd epidemic peaks. Therefore, in this study, the 1st and 2nd epidemic peaks in Japan were defined as from January 16, 2020, to May 24, 2020, and from May 25 to August 30, 2020, respectively.

The arrival time of infectious disease was set according to the following original criteria. The time of arrival of the 1st epidemic wave was simply set as the day when the prefecture first issued a report of an infected person (S2 Table). The arrival time of the 2nd epidemic wave was simply set as the first day of the period in which the number of infected persons continued to rise for 1 week or more. The duration criterion of 1 week or more was chosen so that outbreaks with some cohesion (i.e., an epidemic wave) were selected rather than sporadic outbreaks. In the 2nd epidemic wave, it was confirmed that the number of cases continued to rise for more than 1 week in 39 prefectures. The other eight prefectures were not included in the analysis, as they did not show a 2nd epidemic wave.

The starting location (prefecture) of the 1st epidemic wave was defined as Kanagawa prefecture, where the first laboratory-confirmed case of COVID-19 in Japan was identified [3]. The starting location of the 2nd epidemic wave was defined as Tokyo, which had one of the highest prevalence of COVID-19 cases at the cutoff week between the 1st and the 2nd epidemic waves (and also in the following weeks).

Evaluation of the accuracy of the spread models

The accuracy of the spread models was evaluated using the correlation between time of arrival and effective distance, calculated according to the different starting locations all prefectures. Hokkaido, Chiba, Tokyo, Kanagawa, Aichi, Osaka, Fukuoka, and Okinawa that have the major international airports serving international flights to more than fifteen destination cities. The other seven locations are prefectures in which the major international airports are located, but the number of destination cities were less than eight. The first laboratory-confirmed case of COVID-19 in Japan was identified in Kanagawa [3].

Geographical distance

We calculated the geographical distance between prefectures to confirm the effectiveness of the effective distance measure. Geographical distance was measured using a geographic information system (ArcMap 10.7.1, ESRI Japan, Tokyo) that calculated the linear distance between the prefectural capitals.

Results

Number of cases and arrival time for each epidemic wave

The number of cases in the analysis data was 16,226 in the 1st epidemic wave and 50,539 in the 2nd epidemic wave. The 1st epidemic wave of infection was conspicuous in large cities in Hokkaido, Tokyo, Aichi, Osaka, and neighboring prefectures (Fig 1A). Furthermore, infection spread was observed in the local cities of Ishikawa and Toyama. In the 2nd epidemic wave (Fig 1B), as in the 1st epidemic wave, the number of cases increased in large cities and neighboring prefectures. Additionally, the infection spread to some different locations (e.g., Miyazaki, Kagoshima, and Okinawa) compared with the 1st epidemic wave.

Fig 1

No. of cases in 1st epidemic wave and 2nd epidemic wave.

(a) No. of cases at 1st epidemic wave (January 16 to May 24). (b) No. of cases at 2nd epidemic wave (May 25 to August 30). Total no. of cases in each prefecture is shown in S1 Table.

Time of arrival of each epidemic wave is shown in (Fig 2A and 2B). In the 1st epidemic wave, the infection began to spread sporadically in geographically distant areas such as around Tokyo, around Osaka, and Hokkaido (Fig 2A). In contrast, in the 2nd epidemic wave, the starting point was limited to Tokyo, Kanagawa, and Hokkaido (Fig 2B). In either case, it is unlikely that the infection spread based on geographical proximity.

Fig 2

Time of arrival of 1st epidemic wave and 2nd epidemic wave.

(a) No. of cases at 1st epidemic wave (January 16 to May 24). (b) No. of cases at 2nd epidemic wave (May 25 to August 30). Time of arrival of the virus in each prefecture is shown in S2 Table.

Effective distance shows the spatiotemporal characteristics of infection spread

The relationship between arrival time and geographical distance is shown in (Fig 3A and 3B). This is based on the prefecture where the first case was confirmed in the 1st and 2nd epidemic waves. Fig 3A shows geographical distance Dgeo and time of arrival Ta set to zero at Kanagawa; the coefficient of determination was R2 = 0.0523 (p-value > 0.05). Fig 3B shows geographical distance Dgeo and time of arrival Ta set to zero at Tokyo; the coefficient of determination was R2 = 0.0109 (p-value > 0.05). The relationship between arrival time and effective distance is shown in (Fig 3C and 3D). Fig 3C shows effective distance Deff and time of arrival Ta set to zero at Kanagawa, where reports of infected people were first issued in Japan. The coefficient of determination was R2 = 0.3227 (p-value < 0.05). Fig 3D shows effective distance Deff and time of arrival Ta set to zero at Tokyo, which had the highest number of infected people as of May 25. The coefficient of determination was R2 = 0.415 (p-value < 0.05). The model for the 2nd epidemic wave fits the linear model better than the model for the 1st epidemic wave. Compared with the model for arrival time against effective distance in Fig 2, the arrival time and geographical distance were independent of each other. In other words, the effective distance taking into account the mobility network is a better representation of the spatiotemporal characteristics of the spread of infection.

Fig 3

Linear relationship between time of arrival Ta and effective distance Deff.

(a and b) Linear relationship between time of arrival Ta and geographical distance Dgeo at 1st epidemic wave from Kanagawa (a) and at 2nd epidemic wave from Tokyo (b). (c and d) Linear relationship between time of arrival Ta and effective distance Deff at 1st epidemic wave from Kanagawa (c) and at 2nd epidemic wave from Tokyo (d). HKD: Hokkaido, CBA: Chiba, TKY: Tokyo, KGW: Kanagawa, ACH: Aichi, OSK: Osaka, FKK: Fukuoka, OKW: Okinawa, STM: Saitama. The size of the bubble represents the sum of the number of cases in each prefecture.

Correlation between time of arrival and effective distance

Table 1 shows the Spearman correlation coefficients for multiple starting points of effective distances. In the 1st epidemic wave, the correlation was strongest when starting from Mie (r = 0.661), not Kanagawa, where the first case was confirmed in Japan. The prefecture with the second strongest correlation is Aichi (r = 0.571), and the third is Kanagawa (r = 0.568). Mie is adjacent to Aichi, where the major international airports are located, and these areas may have become seeds for transmission routes. On the other hand, Kanagawa, where the first case was confirmed in Japan, may also be the seed of the infection route. The two regions are geographically separated and it is difficult to estimate the starting point of the spread of the infection. It is also possible that the infection propagated from two geographically independent regions as a starting point. In the 2nd epidemic wave, the correlation was strongest when starting from Tokyo (r = 0.644). The prefecture with the second strongest correlation is Kanagawa (r = 0.629), and the third is Chiba (r = 0.602), which are adjacent to each other across Tokyo. In other words, it is possible that these areas were used as seeds to spread the infection. What is important here is that while it is difficult to estimate the origin of the spread of infection in the 1st epidemic wave, the 2nd epidemic wave can be estimated to have propagated based on the effective distance considering the mobility network with a specific area as the origin.

Table 1

Correlation between time of arrival and effective distance.

Prefecture name	The Spearman’s correlation coefficients at 1st epidemic wave	The Spearman’s correlation coefficients at 2nd epidemic wave
Hokkaido*	0.482	0.578
Aomori	0.334	0.419
Iwate	0.348	0.369
Miyagi	0.315	0.328
Akita	0.334	0.271
Yamagata	0.275	0.308
Fukushima	0.290	0.245
Ibaraki	0.387	0.449
Tochigi	0.356	0.378
Gunma	0.370	0.251
Saitama	0.386	0.442
Chiba*	0.495	0.602
Tokyo*	0.488	0.644
Kanagawa*	0.568	0.629
Niigata	0.352	0.252
Toyama	0.263	0.191
Ishikawa	0.339	0.150
Fukui	0.279	0.097
Yamanashi	0.438	0.297
Nagano	0.345	0.215
Gifu	0.401	0.170
Shizuoka	0.502	0.437
Aichi*	0.571	0.347
Mie	0.661	0.303
Shiga	0.358	0.235
Kyoto	0.477	0.321
Osaka*	0.515	0.401
Hyogo	0.354	0.460
Nara	0.566	0.341
Wakayama	0.550	0.312
Tottori	-0.007	0.365
Shimane	0.010	0.410
Okayama	0.094	0.397
Hiroshima	0.163	0.386
Yamaguchi	0.064	0.229
Tokushima	0.269	0.468
Kagawa	0.177	0.307
Ehime	0.260	0.347
Kochi	0.258	0.422
Fukuoka*	0.243	0.370
Saga	0.211	0.347
Nagasaki	0.200	0.341
Kumamoto	0.250	0.312
Oita	0.239	0.251
Miyazaki	0.329	0.416
Kagoshima	0.216	0.393
Okinawa*	0.539	0.567

*; Hokkaido, Chiba, Tokyo, Kanagawa, Aichi, Osaka, Fukuoka, and Okinawa that have the major international airports serving international flights to more than fifteen destination cities.

Discussion

The effective distance taking into account the mobility network shows the spatiotemporal characteristics of the spread of infection better than geographical distance. The correlation of arrival time to effective distance showed the possibility of spreading from multiple areas in the 1st epidemic wave. On the other hand, the correlation of arrival time to effective distance showed the possibility of spreading from a specific area in the 2nd epidemic wave. In the 1st epidemic wave, the possibility of multiple linear models starting from multiple regions and it was difficult to estimate the starting point of the infection spread. The infection may have spread sporadically owing to the effect of overseas travelers. In contrast, in the 2nd epidemic wave, the effect of overseas travelers was not taken into consideration under border control, and the influence of domestic movement may have been noticeable.

In Japan, control measures are generally adopted on a prefecture-by-prefecture basis during an infectious disease epidemic. Therefore, the timing and period of control measures such as movement restrictions and lockdown differ from prefecture to prefecture. However, in prefectures such as Tokyo, Saitama, Chiba, and Kanagawa, which are geographically close to each other and have a short effective distance, integrated measures are necessary

Additionally, as noted in previous research [14], the assessment of regional vulnerabilities by combining effective distance and analysis of medical resources, knowing in advance which regions may be most affected, might have allowed authorities to opt for preemptive differential investments in these regions. However, it should be noted that in some areas, such as Hokkaido, the infection spread independently. Such unique areas should consider control measures in smaller units.

Our study had some limitations. We used 2016 data on mobility networks and passenger volume, and the transition probability between prefectures was assumed to be constant. It is unclear how domestic movements have changed as a result of COVID-19, and it is necessary to improve the accuracy of the transition. We therefore propose further research to calculate the effective distance based on real-time people-flow statistical data from such sources as mobile devices and Wi-Fi. The results for the relationship between effective distance and arrival time changed according to the setting of arrival time for each epidemic wave. There are no clear criteria for the arrival time of infectious disease in each region. In particular, it is difficult to determine the arrival time of the 2nd epidemic wave. There are various possible measures of arrival time, such as the date of the report of the first infected person, a report of continual infections over a specific time period, and the point at which the number of infected persons reaches a certain number. Although there is a model that estimates the period from occurrence of the 1st epidemic wave to occurrence of the 2nd epidemic wave [22], no previous studies seem to discuss the starting point of the 2nd epidemic wave. Similarly, to the best of our knowledge, previous studies using the theory of effective distance focus on the 1st epidemic [12-14]. In this study, one simple method was used to set the criteria. Future studies should address various potential criteria.

Complete mobility network diagram in Japan in 2016.

(TIF)

Click here for additional data file.

Total number of cases in each prefecture.

(DOCX)

Click here for additional data file.

Infection arrival time for each prefecture.

(DOCX)

Click here for additional data file.

8 Apr 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Chiara Poletto

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. PLOS requires an ORCID iD for the corresponding author in Editorial Manager on papers submitted after December 6th, 2016. Please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field. This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager. Please see the following video for instructions on linking an ORCID iD to your Editorial Manager account: https://www.youtube.com/watch?v=_xcclfuvtxQ

3. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors provide an application of the popular effective distance metric from Brockmann and Helbing (2013) to study the arrival times of SARS-CoV2 in Japan at prefectures level during the 1st and 2nd wave of infections. Their results suggest that Tokyo and Kanagawa prefectures were the starting geographical regions for the spreading of Covid-19 through the country.

- a detailed description of mobility data is missing, to what year to the mobility data refer to? What do they represent? Is their representativity similar? the authors mix together different types of mobility data, but it is not clear if their representativity is the same, if they cover the same number of days. To mix different sources of mobility data is a sensible process, which can determine an overrepresentation of some modes of transportation over the others.

- how are the trips considered in the dataset? If a person travels from Tokyo to Osaka while stopping through all the prefectures in the middle, are all the steps considered or you only get the number of people travelling directly from Tokyo to Osaka? For example, is any stop time considered in order to separate trips? Or the dataset only accounts for the origin and destination showed on the purchased tickets?

- The authors say “The accuracy of the spread models was evaluated using the correlation between time of arrival and effective distance, calculated according to the different starting locations: Hokkaido, Chiba, Tokyo, Kanagawa, Aichi, Osaka, Fukuoka, and Okinawa that have the major international airports serving international flights to more than fifteen destination cities. The other seven locations are prefectures in which the major international airports are located, but the number of destination cities were less than eight. “

The authors said that they had 47 prefectures in the dataset, but the models are tested taking as seeding locations only 8 prefectures, whereas later they say “the other 7 locations are prefectures in which the major international airports are located”. I am confused at this point on how many administrative areas are they considering in their model. The model is tested only on a subset of the total dataset and the destinations seem reduced to 15 out of 47. I do not understand the logic of this choice. Why not considering all the prefectures are starting locations and destination locations? The ones that you listed are surely those with the biggest airports, but you said that the mobility data includes all modes of transportations, such as trains, bus, etcetera. Hence finally all prefectures could be tested as starting (seeding) points, possibly showing similar correlation scores if they are strongly connected.

- the authors may want to discuss the role of cases underdetection and the different testing capacities implemented in the different prefectures, which may have affected the detection of the first infected person in the area, and hence the time of arrival of the virus in the geographical area. Figure 2 shows arrival times that differ by few days between all the prefectures taken into account, small differences in cases underdetection would strongly affect the correlation, and indeed points in Fig.2a look very dispersed. The p-value for these correlation should be reported in the picture.

-Why not considering the date of reaching a certain amount of daily case incidence instead of using such a sensible measure like the arrival time of the first case (see for example Ref.[1] in assessing the spreaders role)? The role of asymptomatic infections has been already widely studied in the case of SARS-CoV2, single cases do not necessarily trigger an outbreak, e.g. many reported individual and isolated cases in Europe were reported in January 2020 without starting any local epidemic. Moreover, long times of disease incubation hinders the detection of the first infected patients in their actual areas of arrival from international travel. So what is the logic of relying on such a definition from a public health point of view for SARS-CoV2, when the detection of a single case are not necessarily representative for the triggering of an outbreak in the area?

- As a more general reasoning, I would like to know what is the authors’ thought on the following problem. The model is tested only on few starting prefectures, which surely are those who are the most connected ones in terms of international flights, however they also correspond to the most populated areas in which testing capacity may be higher than other peripheral prefectures, and hence outbreak detection is more efficient. What would be the probability of correctly detecting the arrival time of an infected patient in a peripheral area that has a poor testing capacity? In brief, would we be able through the effective distance model to correctly assess the start of an outbreak from a peripheral area, given these circumstances? Or would we always detect the earliest arrival time in the most tested prefectures and mobility hubs as the result of a demographic and testing capacity bias?

- In this sense, how can be we sure that Tokyo and Kanagawa are effectively the starting prefectures of the 1st and 2nd waves, only from the correlation of effective distance and arrival times, given the possible confounder effects represented by population density? Infrastructures are planned on the basis of gravity (or radiation) models that take into account origin and destination populations, so effective distance from mobility hubs at prefectures level would reflect the population hierarchy. Given these confounding factors, how can we be sure that this model provides evidence to say that Tokyo and Kanagawa were the seeding prefectures in Japan?

- the introduction on previous works and state of the art on mobility and epidemics is insufficient. There are many many works that lately confronted the effect of different mobility data on the spatiotemporal invasion of SARS-CoV2 at sub-national level that need to be correctly referenced.

See for example these three papers and their references to build a more general overview in the introduction:

[1] Mazzoli, Mattia, et al. "Interplay between mobility, multi-seeding and lockdowns shapes COVID-19 local impact." PLoS computational biology 17.10 (2021): e1009326.

[2] Kraemer, Moritz UG, et al. "Spatiotemporal invasion dynamics of SARS-CoV-2 lineage B. 1.1. 7 emergence." Science 373.6557 (2021): 889-895.

[3] Kraemer, Moritz UG, et al. "The effect of human mobility and control measures on the COVID-19 epidemic in China." Science 368.6490 (2020): 493-497.

Reviewer #2: This work analyses how the human mobility in Japan may have affected the spreading of SARS-CoV-2 in the first two waves. It does so by computing the correlation between the effective distance of a prefecture from the starting point of the epidemic in Japan and the time of first-arrival of the epidemic in that prefecture.

This idea to use the concept of effective distance to gain insights on the epidemics is not new, but to the best of my knowledge this is the first time that it has been used for the public transportation network of Japan, making the paper original worth publishing.

I only have one doubt: throughout the paper it is mentioned various times that an higher Spearman coefficient indicates a better fit with the linear model. However the Spearman coefficient measures how closely the data follow a monotonic function, which in general may not be linear. So in my opinion the paper would gain in rigorousness if this problem was addressed or if instead of “linear model” the authors used a more generic term like “positive correlation”.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

24 Jun 2022

Point-by-point response to reviewers’ comments

We would like to express our sincere appreciation to the editor and reviewers for their kind comments and suggestions regarding our manuscript. We have included our responses to the reviewers’ comments herein, for ease of reference. For clarity, the comments are shown in blue and bold font and our responses are shown in black and regular font.

Reviewer reports:

Reviewer 1:

- The authors provide an application of the popular effective distance metric from Brockmann and Helbing (2013) to study the arrival times of SARS-CoV2 in Japan at prefectures level during the 1st and 2nd wave of infections. Their results suggest that Tokyo and Kanagawa prefectures were the starting geographical regions for the spreading of Covid-19 through the country.

- a detailed description of mobility data is missing, to what year to the mobility data refer to? What do they represent? Is their representativity similar? the authors mix together different types of mobility data, but it is not clear if their representativity is the same, if they cover the same number of days. To mix different sources of mobility data is a sensible process, which can determine an overrepresentation of some modes of transportation over the others.

Thank you very much for taking your time for reviewing our manuscript and providing valuable advice and suggestions.

Regarding the mobility data, the issued year and some additional descriptions have been included in the method section. We have rewrite them for more clearly understanding.

Also, we mentioned about using the different sources of mobility data in 2016 as the study limitation in the discussion section.

P. 6-7; Lines 52-60.

For mobility network data, we used the aggregated data on passenger volume by transportation mode for the 47 prefectures contained in the freight/passenger area flow survey conducted by the Ministry of Land, Infrastructure, Transport and Tourism of Japan issued in 2016 [20]. The data include the volume of passengers in the private railway, bus, ship, and aviation categories. The dataset counts all transport personnel across other prefectures in one year. It is the OD amount of passenger transport personnel between regions. For example, if the departure point is Tokyo and the arrival point is Osaka, all boarding / alighting personnel by prefecture on the route are counted. These data set was obtained from the “Kokudo Suchi,” the geographic information systems (GIS) data service of the Ministry of Land, Infrastructure, Transport and Tourism of Japan [21].

P. 17-18; Lines 228-230.

We used 2016 data on mobility networks and passenger volume, and the transition probability between prefectures was assumed to be constant.

- how are the trips considered in the dataset? If a person travels from Tokyo to Osaka while stopping through all the prefectures in the middle, are all the steps considered or you only get the number of people travelling directly from Tokyo to Osaka? For example, is any stop time considered in order to separate trips? Or the dataset only accounts for the origin and destination showed on the purchased tickets?

The dataset counts all transport personnel across other prefectures in one year. It is the OD amount of passenger transport personnel between regions. For example, if the departure point is Tokyo and the arrival point is Osaka, all boarding / alighting personnel by prefecture on the route are counted. We added these detailed explanation in the materials and methods section in the text.

P. 6-7; Lines 55-59.

The data include the volume of passengers in the private railway, bus, ship, and aviation categories. The dataset counts all transport personnel across other prefectures in one year. It is the OD amount of passenger transport personnel between regions. For example, if the departure point is Tokyo and the arrival point is Osaka, all boarding / alighting personnel by prefecture on the route are counted.

- The authors say “The accuracy of the spread models was evaluated using the correlation between time of arrival and effective distance, calculated according to the different starting locations: Hokkaido, Chiba, Tokyo, Kanagawa, Aichi, Osaka, Fukuoka, and Okinawa that have the major international airports serving international flights to more than fifteen destination cities. The other seven locations are prefectures in which the major international airports are located, but the number of destination cities were less than eight. “

The authors said that they had 47 prefectures in the dataset, but the models are tested taking as seeding locations only 8 prefectures, whereas later they say “the other 7 locations are prefectures in which the major international airports are located”. I am confused at this point on how many administrative areas are they considering in their model. The model is tested only on a subset of the total dataset and the destinations seem reduced to 15 out of 47. I do not understand the logic of this choice. Why not considering all the prefectures are starting locations and destination locations? The ones that you listed are surely those with the biggest airports, but you said that the mobility data includes all modes of transportations, such as trains, bus, etcetera. Hence finally all prefectures could be tested as starting (seeding) points, possibly showing similar correlation scores if they are strongly connected.

In this article, considering the transmission route from abroad, we focused only on the eight prefectures where the major international airports are located. As you point out, we need to show the correlation between arrival dates and effective distances in 47 prefectures. Therefore, Table 1 was revised to show the correlation between 47 prefectures (P. 14-16; Lines 198-201). In response to this revision, the text has been revised as follows.

P. 10; Lines 109-111.

The accuracy of the spread models was evaluated using the correlation between time of arrival and effective distance, calculated according to the different starting locations all prefectures.

P. 14; Lines 181-196.

In the 1st epidemic wave, the correlation was strongest when starting from Mie (r =0.661), not Kanagawa, where the first case was confirmed in Japan. The prefecture with the second strongest correlation is Aichi (r = 0.571), and the third is Kanagawa (r = 0.568). Mie is adjacent to Aichi, where the major international airports are located, and these areas may have become seeds for transmission routes. On the other hand, Kanagawa, where the first case was confirmed in Japan, may also be the seed of the infection route. The two regions are geographically separated and it is difficult to estimate the starting point of the spread of the infection. It is also possible that the infection propagated from two geographically independent regions as a starting point. In the 2nd epidemic wave, the correlation was strongest when starting from Tokyo (r = 0.644). The prefecture with the second strongest correlation is Kanagawa (r = 0.629), and the third is Chiba (r = 0.602), which are adjacent to each other across Tokyo. In other words, it is possible that these areas were used as seeds to spread the infection. What is important here is that while it is difficult to estimate the origin of the spread of infection in the 1st epidemic wave, the 2nd epidemic wave can be estimated to have propagated based on the effective distance considering the mobility network with a specific area as the origin.

P. 16; Lines 207-212.

The correlation of arrival time to effective distance showed the possibility of spreading from multiple areas in the 1st epidemic wave and the possibility of spreading from a specific area in the 2nd epidemic wave. In the 1st epidemic wave, the possibility of multiple linear models starting from multiple regions and it was difficult to estimate the starting point of the infection spread.

- the authors may want to discuss the role of cases underdetection and the different testing capacities implemented in the different prefectures, which may have affected the detection of the first infected person in the area, and hence the time of arrival of the virus in the geographical area. Figure 2 shows arrival times that differ by few days between all the prefectures taken into account, small differences in cases underdetection would strongly affect the correlation, and indeed points in Fig.2a look very dispersed. The p-value for these correlation should be reported in the picture.

The significance level based on the calculation of the p-value was added to Fig.3 and text.

P. 12-13; Lines 153-163

Fig 3(a) shows geographical distance Dgeo and time of arrival Ta set to zero at Kanagawa; the coefficient of determination was R2 = 0.0523 (p-value > 0.05). Fig 3(b) shows geographical distance Dgeo and time of arrival Ta set to zero at Tokyo; the coefficient of determination was R2 = 0.0109 (p-value > 0.05). The relationship between arrival time and effective distance is shown in Fig 3(c and d). Fig 3(c) shows effective distance Deff and time of arrival Ta set to zero at Kanagawa, where reports of infected people were first issued in Japan. The coefficient of determination was R2 = 0.3227 (p-value < 0.05). Fig 3(d) shows effective distance Deff and time of arrival Ta set to zero at Tokyo, which had the highest number of infected people as of May 25. The coefficient of determination was R2 = 0.415 (p-value < 0.05).

-Why not considering the date of reaching a certain amount of daily case incidence instead of using such a sensible measure like the arrival time of the first case (see for example Ref.[1] in assessing the spreaders role)? The role of asymptomatic infections has been already widely studied in the case of SARS-CoV2, single cases do not necessarily trigger an outbreak, e.g. many reported individual and isolated cases in Europe were reported in January 2020 without starting any local epidemic. Moreover, long times of disease incubation hinders the detection of the first infected patients in their actual areas of arrival from international travel. So what is the logic of relying on such a definition from a public health point of view for SARS-CoV2, when the detection of a single case are not necessarily representative for the triggering of an outbreak in the area?

As you point out, there is a time lag in inducing infection, and it is not possible to determine the signs of virus arrival from a single case that reaches the area. Since the purpose of this study is to discuss how the infection was reached in 47 prefectures, the time when the infection was first announced in that prefecture is used. In fact, there is no clear standard for arrival time (listed at the end of the sentence).

When performing a detailed infection spread simulation considering the incubation period in a specific area, it is necessary to calculate by focusing on the case incidence rate you propose. The three documents you presented [1-3] performed simulations based on real-time datasets, and provided excellent suggestions for considering the arrival time and the impact of mobility networks. These references were referenced in the Limitation section. After that, I would like to challenge the real data set and complicated condition setting to improve the accuracy.

P. 18; Lines 231-233.

We therefore propose further research to calculate the effective distance based on real-time people-flow statistical data from such sources as mobile devices and Wi-Fi.

- As a more general reasoning, I would like to know what is the authors’ thought on the following problem. The model is tested only on few starting prefectures, which surely are those who are the most connected ones in terms of international flights, however they also correspond to the most populated areas in which testing capacity may be higher than other peripheral prefectures, and hence outbreak detection is more efficient. What would be the probability of correctly detecting the arrival time of an infected patient in a peripheral area that has a poor testing capacity? In brief, would we be able through the effective distance model to correctly assess the start of an outbreak from a peripheral area, given these circumstances? Or would we always detect the earliest arrival time in the most tested prefectures and mobility hubs as the result of a demographic and testing capacity bias?

In response to your suggestions, we have revised Table 1 to show the correlation for 47 prefectures. As shown in Table 1, in the 1st epidemic wave, the correlation coefficient was the highest when Mie prefecture, which has a relatively small population, was used as a seed. In other words, it indicates that the infection may have spread from areas where international flights are not connected. However, unfortunately, the probability of correctly detecting the arrival time is unclear because no clear criteria have been set for the arrival time of the infection. This is the limitation of analysis in this study.

- In this sense, how can be we sure that Tokyo and Kanagawa are effectively the starting prefectures of the 1st and 2nd waves, only from the correlation of effective distance and arrival times, given the possible confounder effects represented by population density? Infrastructures are planned on the basis of gravity (or radiation) models that take into account origin and destination populations, so effective distance from mobility hubs at prefectures level would reflect the population hierarchy. Given these confounding factors, how can we be sure that this model provides evidence to say that Tokyo and Kanagawa were the seeding prefectures in Japan?

In response to your suggestions, we have revised Table 1 to show the correlation for 47 prefectures. The text has been revised accordingly. The possibilities of the prefectures where the 1st epidemic wave started have expanded. However, unfortunately, the probability of correctly detecting the arrival time is unclear because no clear criteria have been set for the arrival time of the infection. This is the limitation of analysis in this study. Therefore, although it is impossible to determine the seeding prefecture, we were able to grasp the tendency of the spread of infection with seeds in a specific area as shown in the text.

- the introduction on previous works and state of the art on mobility and epidemics is insufficient. There are many many works that lately confronted the effect of different mobility data on the spatiotemporal invasion of SARS-CoV2 at sub-national level that need to be correctly referenced.

See for example these three papers and their references to build a more general overview in the introduction:

[1] Mazzoli, Mattia, et al. "Interplay between mobility, multi-seeding and lockdowns shapes COVID-19 local impact." PLoS computational biology 17.10 (2021): e1009326.

[2] Kraemer, Moritz UG, et al. "Spatiotemporal invasion dynamics of SARS-CoV-2 lineage B. 1.1. 7 emergence." Science 373.6557 (2021): 889-895.

[3] Kraemer, Moritz UG, et al. "The effect of human mobility and control measures on the COVID-19 epidemic in China." Science 368.6490 (2020): 493-497.

Thank you very much for your kind suggestion. Per your advice, we referred the issue about three major reports regarding the effect of different mobility data on the spatiotemporal invasion of SARS-CoV2 in the introduction section. Also, we added the comment about the related issue of theses references in the limitation section.

P. 5; Lines 34 - 36.

In addition, currently, there are some reports that confronted the effect of different mobility data, including people flow statistics, on the spatiotemporal distributions of SARS-CoV2 at sub-national level [15 - 17].

P. 18; Lines 231-233.

We therefore propose further research to calculate the effective distance based on real-time people-flow statistical data from such sources as mobile devices and Wi-Fi.

Reviewer #2:

This work analyses how the human mobility in Japan may have affected the spreading of SARS-CoV-2 in the first two waves. It does so by computing the correlation between the effective distance of a prefecture from the starting point of the epidemic in Japan and the time of first-arrival of the epidemic in that prefecture.

This idea to use the concept of effective distance to gain insights on the epidemics is not new, but to the best of my knowledge this is the first time that it has been used for the public transportation network of Japan, making the paper original worth publishing.

I only have one doubt: throughout the paper it is mentioned various times that an higher Spearman coefficient indicates a better fit with the linear model. However the Spearman coefficient measures how closely the data follow a monotonic function, which in general may not be linear. So in my opinion the paper would gain in rigorousness if this problem was addressed or if instead of “linear model” the authors used a more generic term like “positive correlation”.

As you point out, the analysis should show Spearman's correlation coefficient, not the judgment of compatibility with the linear model. I refrained from using the term linear model in the text as much as possible.

Submitted filename: Rebuttal letter_220624.docx

Click here for additional data file.

1 Aug 2022

Impact of human mobility and networking on spread of COVID-19 at the time of the 1st and 2nd epidemic waves in Japan: an effective distance approach

PONE-D-21-37364R1

Dear Dr. Nohara,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Chiara Poletto

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Dear Authors,

all my concerns have been addressed.

I consider the manuscript suitable for publication.

Reviewer #2: The authors did a good job in answering my requests. I am now satisfied with the content of the paper.

However I found that the authors say in their Data Availability Statement that the data are "available upon reasonable request". If the reason behind this choice was that the data cannot be anonymized or shared publicly that should have been explained in the Data Availability Statement. Otherwise the authors are encouraged to share the data on a public repository. Please provide an explanation and change the Statement accordingly.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

**********

3 Aug 2022

PONE-D-21-37364R1

Impact of human mobility and networking on spread of COVID-19 at the time of the 1st and 2nd epidemic waves in Japan: an effective distance approach

Dear Dr. Nohara:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Chiara Poletto

Academic Editor

PLOS ONE