https://orcid.org/0000-0002-2623-2075
Figures
Abstract
Despite implementing no lockdowns and having a large elderly population, Japan had a low mortality rate due to COVID-19 compared to Europe and North America. The extent to which policies impacted person-to-person contact remains unclear. In this study, we examined changes in contact patterns and their association with behaviors and governmental recommendations in Japan during the pandemic. Ten social contact surveys were conducted between 2021 and 2023 reaching over 1500 participants per survey in Osaka and Fukuoka prefectures where governmental recommendations were first implemented due to high COVID-19 incidence. Their contact patterns were assessed through their demographic characteristics, COVID-19 vaccination status, and individual disease mitigation measures. Generalized linear models were used to identify factors associated with increased contacts. The mean number of contacts during the pandemic declined by at least 49.8% (8.2 weekday contacts and 6.0 weekend contacts per individual, adjusted by age and sex) compared to a study conducted prior to 2020. Weekdays, occupation, larger household sizes, and mask wearing were associated with a higher number of contacts. The frequency and duration of contacts were negatively associated with the issuance of COVID-19 governmental measures, yet the relative change in contacts was not as prominent as pre- and post-lockdown situations in the United Kingdom. There was a gradual increase in contacts with time and less strict public health recommendations. Yet, contacts that did not increase with uptake of COVID-19 vaccination and continuous mask wearing depict cautious behavior across the survey population during the pandemic and into 2023. These results are in contrast with European countries where contacts largely increased among vaccinated individuals compared to the non-vaccinated. Social contacts are country and context specific, highlighting the need for data collection across different communities.
Citation: Nakamura T, Kinoshita R, Endo A, Atkins KE, Oshitani H, Ibuka Y, et al. (2025) Continuing to be cautious: Japanese contact patterns during the COVID-19 pandemic and their association with public health recommendations. PLOS Glob Public Health 5(9): e0004600. https://doi.org/10.1371/journal.pgph.0004600
Editor: Sheikh Taslim Ali,, University of Hong Kong Li Ka Shing Faculty of Medicine, HONG KONG
Received: April 15, 2025; Accepted: August 17, 2025; Published: September 24, 2025
Copyright: © 2025 Nakamura et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All accompanying code and data to generate all the figures in the main manuscript is available on GitHub: https://github.com/tomokanakamura/jp_contactsurvey.git.
Funding: The study was supported by Nagasaki University “Doctoral Program for World-leading Innovative and Smart Education” for Global Health, “Global Health Elite Programme for Building a Healthier World” (WISE) (TN). The social contact surveys were supported by WISE and the Ministry of Health, Labour and Welfare of Japan (grant numbers 20HA2007 and 23HA2005). PhD research stipend was supported by Nagasaki University “Doctoral Program for World-leading Innovative and Smart Education” for Global Health, KYOIKU KENKYU SHIEN KEIHI (TN). RK was supported by the Japan Society for the Promotion of Science (JSPS) (21K17307) and the Japan Agency for Medical Research and Development (JP23fk0108685). AE was supported by the Japan Science and Technology Agency (JST) (JPMJPR22R3), the Japan Society for the Promotion of Science (JSPS) (JP22K17329) and Japan Agency for Medical Research and Development (JP223fa627004). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Background
The reported cases and deaths due to coronavirus disease 2019 (COVID-19) in Japan were much lower compared to Western countries within Europe and North America. For comparison, as of 28 June 2023, there were 74,694 cumulative deaths in Japan in a total population of 125.5 million (59.5 deaths per 100,000) while there was a total of 227,524 deaths in the United Kingdom with a population approximately half of Japan (337.9 deaths per 100,000) [1]. Japan is also unique having the second largest aging population in the world, after Monaco, (29.9% of the total population who are 65 years and above) [2] who is at a higher risk of COVID-19 mortality. Throughout the pandemic from 2020 to early 2022, cumulative confirmed COVID-19 cases (adjusted by population size) were reported 10–15 times higher in the United States and United Kingdom compared to Japan [1]. Several factors, such as timing of government regulations and behavioral change interventions, have been discussed as potential determinants of a successful epidemic control [3].
Apart from strict international border control policies, Japan had relatively less strict rules compared to many other countries. Though public health emergency declarations (EDs) with different levels of strictness were issued, neither lockdowns nor curfews were implemented. National school closure due to COVID-19 regulation spanned for at least three weeks in March 2020, but most schools reopened between April and May 2020 [4]. Yet, there were key messages that continued to be addressed to the public, such as avoiding the “3Cs” which stands for settings that are closed, crowded and close contact [5]. Reducing person-to-person contact has been one of the key tactics in epidemic control as it directly shapes the risk of transmission of respiratory viruses [6].
To quantify these contact patterns, mobile phone data [7] and synthetic contact matrices [8] have been used. During the COVID-19 pandemic in Japan, mobile device data was utilized to investigate how governmental interventions could have impacted mobility [9,10]. Social contact survey data is another important input for epidemiological and mathematical models of infectious diseases. It has advantages over mobile device data as it can capture changes in contact patterns with respect to age and sex of both survey participants and their contacts. Additionally, disease mitigation measures, such as vaccination and handwashing, can be linked with contact data. Social contact surveys have been well-utilized prior to the pandemic, such as POLYMOD in Europe [11] and in Japan [12,13]. They showed that contact patterns are highly dependent on age, gender, household size and day of the week [14]. These studies provide a baseline of contact patterns prior to any physical distancing or public health related measures.
To elucidate the changes in contact patterns relevant in the transmission of SARS-CoV-2 in Japan, we conducted 10 repeated cross-sectional surveys from 2021 to 2023 (Fig 1A, 1B). In this paper, we aim to describe the changes in social contacts and other behaviors, such as hand hygiene and mask wearing, during the pandemic and into 2023. As different levels of EDs were issued and lifted with time, we evaluated the association of these governmental measures on social contacts. We also present a statistical model that explores individual characteristics and behavior that impact the frequency of contacts.
Methods
Ethics statement
Participation in the survey was voluntary and all analyses were performed on anonymized data. The study design including the survey questions and informed consent were approved by the ethics committee of Nagasaki University School of Tropical Medicine and Global Health (approval number: NU_TMGH_2022_162_4). Informed written consent was obtained from all participants who provided data from the surveys.
Survey design
Ten contact surveys, in which the participants were asked to report their individual characteristics and contact patterns, were conducted in Osaka and Fukuoka prefectures, Japan in 2021–2023 (Fig 1). Osaka prefecture is located on the west of mainland Japan and has the third highest population next to Tokyo and Kanagawa prefectures [15]. Fukuoka prefecture is on Kyushu Island, which is south of mainland Japan, and has the ninth highest population [15]. In 2021, three EDs were issued across Japan including Osaka and Fukuoka during which two surveys (surveys 1 and 3, Fig 1) were conducted. During the height of the BA.1 (Omicron) transmission in 2022, a semi-ED was issued, which consisted of a less strict recommendation than an ED (Table A in S1 Text) [16]. Two surveys (surveys 7 and 8, Fig 1) were conducted during this time. Six surveys (surveys 2, 4, 5, 6, 9, and 10, Fig 1) were conducted when ED was absent with the last survey in February 2023. The timing of the surveys was semi-strategic; they were planned during expected changes in contact patterns and funding availability.
Survey participants were recruited by a Japanese online survey company (F-press). Participants included anyone at least 18 years old residing in either Osaka or Fukuoka prefecture who agreed to participate through written informed consent. Children under 18 years old participated with the consent of their guardians/parents who recorded their information on their behalf. Every individual in each household can fill out the survey as a participant, and all these individuals were linked by household. The number of participants per survey was a minimum of 1,500, powered to detect a difference in contact numbers of 2.5 between pairs of observations with a 90% power and 5% Type I error and 20% loss to follow-up. The survey participants were compensated 300 Japanese yen per survey and were able to participate in as many surveys as they opted for between 10 February 2021 and 12 February 2023.
The survey was adapted from the UK CoMix study [17–19] in Japanese to capture the daily frequency and type of contacts. In each survey, participants recorded their contacts during one weekday and one of the days during the weekend. A “contact” was defined as physical or non-physical: physical contact included handshakes, hugging, kissing, and playing a contact sport while non-physical contact was defined as facing another individual (with or without mask wearing) and exchanging at least three Japanese sentences to each other in a conversation. The participants were asked to record their contacts in a diary format with instructions a week prior to the day when the survey was implemented. They reported their contacts including their age, sex, type of contact (non-physical and/or physical), location of contact, approximate time of contact, and whether the contact was made indoors and/or outdoors (Table A in S1 Text). These specified contacts were for the first 10 contacts. Those who reported more than 10 contacts were asked to approximate the number of contacts, location of contacts, and their age categories.
Participants were asked to report their individual characteristics as well as their individual preventive behavior such as frequency of mask wearing, handwashing, and teleworking (Table A in S1 Text). They were asked COVID-19 related questions including their history of having tested PCR positive for COVID-19, their concern towards getting infected with COVID-19 (rank from one to five), and their vaccination status.
Statistical analysis
For each survey, we calculated the mean and median of contacts during the weekday and weekend, mean and median age, COVID-19 vaccine coverage, frequency of mask wearing, and the proportion of survey participants who tested COVID-19 positive. The population estimates of the mean contacts and vaccination coverage were adjusted by age and sex based on the 2021 October census [15].
We compared our data to contact patterns analyzed by Ibuka et al. [12] as baseline data prior to the COVID-19 pandemic. Their survey methodology was similar to ours; the data included all reported contacts per individual during the week as well as the duration and location of contacts. Their contact surveys were conducted with an objective to understand influenza-related behaviors which is relevant in the context of respiratory disease transmission. For our study, contact patterns were stratified by weekday and weekend and compared across eight age groups (0–9, 10–19, 20–29, 30–39, 40–49, 50–59, 60–69, and 70+) and across the study period. These age groups were utilized to enable direct comparison with baseline contacts, and the survey was designed in a way that allowed ease of reporting for the participants. If they did not know the exact age of their contacts, it would be easier to distinguish age by decade. For the multivariable regression model, the exact ages of all participants were reported, and hence, it was possible to distinguish school-aged children by analyzing in seven age groups (0–5, 6–17, 18–29, 30–39, 40–49, 50–59, 60+).
Measures of uncertainty in age-specific contact numbers and duration were obtained using the bootstrap; the mean and 95% confidence intervals (CI) were obtained by sampling with replacement for 1000 times. Consistent with previous contact survey studies, we truncated the total number of contacts to avoid a few observations with hyperinflated contact numbers affecting age-specific summary statistics [11,19]. We selected a truncation cutoff of 250 through a visual check by plotting the mean number of contacts with a range of cutoff points using a Weibull distribution. (Fig A in S1 Text).
To investigate factors associated with the reported number of contacts (February 2023 survey only), we used a multivariable regression assuming a Weibull distribution with log link function. This distribution was selected to address the coefficient of variation in reported contacts and fit the right skewed distribution of contacts better than a negative binomial distribution. The dependent variable was the reported contacts, and 15 variables were selected from a list of variables (Table A in S1 Text) as the model covariates based on a hypothesis-driven approach. The model was referenced on a 40–49 year old individual who lives with two others, works as a company employee at their workplace, and fully vaccinated against COVID-19 with at least 3 or more doses. The participants’ age was retained in the model as it could be a potential confounder. All other covariates that were selected in the multivariable model were associated (p < 0.05) with the number of contacts in an univariable model. Multicollinearity of the covariates was checked using the Variance Inflation Factor (VIF). The multivariable model was developed using both forward and backward stepwise selections until the Akaike Information Criterion (AIC) had no further improvement. The incidence rate ratios calculated from the regression model are referred here as contact rate ratios (CRR) where the relative mean number of contacts per day is compared to the reference of each covariate. Model fit was examined with residual plots and comparing predicted vs. observed data. When analyzing summary statistics of contacts across the various time points and age categories, the Kruskal-Wallis test was used. When comparing two nested generalized linear models to test the significance of a variable, likelihood ratio test was used. Statistical results were reported with p-values (significance level 0.05). Extremely small p-values under 1x10-5 were simply reported as p < 1x10-5.
All statistical analyses were done in R version 4.0.3, using the packages “MASS”, “survival”, “tidyverse”, “dplyr”, “reshape2”, “scales”, “rstatix”, “ggplot2”, “olsrr,” and “cowplot.”
Results
Comparison of contact survey participants with national data
A range of 1513–1721 participants recorded their contact patterns between February 2021 and February 2023 in Fukuoka and Osaka prefectures, Japan (Table 1). Depending on the survey, the mean age of the participants was between 45 and 48 years old. Between 52–54% of the participants were female. As of the 2021 national census, the mean age of the Japanese population was 48.4 and 51.3% being female [15]. COVID-19 vaccine coverage among the survey participants increased with time. In Japan, healthcare workers were first vaccinated in February 2021, followed by the older population ≥ 65 years) from March 2021, and the rest of the population from June 2021 [22]. Vaccine coverage reported in February 2023 was slightly higher among the survey participants compared to the national vaccine coverage as of February 2023 [21]. Compared to national vaccine coverage reported in January 2023, vaccine coverage across all age groups were slightly higher among the survey participants in February 2023 (Fig B in S1 Text). The percent of individuals and/or household members having tested COVID-19 positive increased from 0.3% (5/1721) in February 2021 to 8.7% (136/1569) in February 2023. A serosurvey (N = 5627) conducted between 3rd February and 4th March 2023 showed 35.8% of Osaka residents and 31.3% of Fukuoka residents tested positive with infection-induced antibodies [23]. Our survey appropriately approximated the national statistics in terms of demographics but with a slightly higher vaccination coverage and lower percentage of people who tested COVID-19 positive.
The mean number of contacts during the pandemic reduced by 49.8% during the weekday and by 53.3% during the weekend when compared to pre-pandemic times [12]; we report an average of 8.18 weekday contacts and 5.98 weekend contacts per individual. The sample distribution of contacts was right-skewed (Fig 2) where 60.2% of the participants contacted less than five individuals per day during the weekday (and 76.5% during the weekend) in February 2023. The sample distribution of contacts from the earlier surveys showed a similar distribution (Fig C in S1 Text). Prior to the pandemic, the distribution of daily reported contacts (N = 3146 participants) was also right-skewed and less than 2% of contacts reported zero contacts [12].
The blue bars show the observed contacts reported from the contact survey. The pink bars show the predicted contacts based on the multivariable regression model using a Weibull distribution. Sixty participants (3.82%) recorded over 50 contacts.
Change in contact patterns across a typical week
To test whether there were substantial differences in frequency of contacts across a typical week, two generalized linear models with and without survey time were compared using a likelihood ratio test. Survey time point significantly explained the variability in the mean of contacts during the weekday (chi-squared = 29.33, p-value = 0.00057) (Fig 3A) and during the weekend (chi-squared = 98.79, p < 1x10-5) (Fig 3B). Particularly among the 10’s and 40’s, the mean number of contacts increased in calendar time during the weekday and weekend.
The mean and 95% confidence intervals were obtained by bootstrapping.
On the contrary, the older population (70+) consistently had low contacts during the weekday and weekend. Particularly during the weekdays, the mean number of contacts across all survey time points (February 2021-February 2023) was significantly lower amongst the 70 + year old compared to the 40’s (chi-squared = 112.65, p < 1x10-5). There were clear temporal changes in the duration of contacts in the 40’s during the weekday (chi-squared = 49.47, p < 1x10-5) (Fig 4A) and during the weekend (chi-squared = 54.04, p < 1x10-5) (Fig 4B). For example, the mean duration of weekday contacts for the 40’s increased from 6.10 hours (95% CI: 5.29-6.85) in February 2021 to 8.63 hours (95% CI: 7.75-9.48) in February 2023. There were similar temporal changes in the duration of contacts amongst the 70 + year old (chi-squared = 21.04, p-value = 0.012 for weekday; chi-squared = 25.51, p-value = 0.0025 for weekend).
The mean and 95% confidence intervals were obtained by bootstrapping.
Association of Emergency Declarations (ED) and Semi-ED
We explored whether EDs (observed during surveys 1 and 3) were associated with reported contacts and their duration, and whether the strength of ED provided further granularity. After adjusting for age and sex in a generalized linear model, the issuance of an ED was negatively associated with the mean number of contacts during weekends (Adjusted CRR: 0.84 (95% CI: 0.79-0.88) compared with periods without any ED, yet there was no association during weekdays (Adjusted CRR: 0.96 (95% CI: 0.91-1.01). On the other hand, there was a slightly positive association between the issuance of a semi-ED (surveys 7 and 8) with contacts during weekends (Adjusted CRR: 1.06 (95% CI: 1.01-1.12)) and weekdays (Adjusted CRR: 1.05 (95% CI: 1.00-1.11)) compared with periods without any ED.
For example, a woman in her 40’s would contact an average of 3.32 individuals (95% CI: 3.09-3.56) during weekends when an ED was issued, but she would contact an average of 3.97 individuals (95% CI: 3.74-4.20) during periods when any level of ED was absent (Table 2). Since surveys 9 and 10 were conducted two years after the initial survey and could have significantly different contact patterns due to “pandemic fatigue,” the same analysis was conducted by excluding these two surveys, but the negative association between ED and contacts remained the same.
When surveys 1 and 3 (ED issued) were compared with survey 2 (ED absent) after adjusting for age and sex, there was no evidence of an association between the frequency of contacts and the issuance of an ED during the weekday (Adjusted CRR: 1.03 (95% CI: 0.95-1.12) or the weekend (Adjusted CRR: 0.95 (95% CI: 0.88-1.04).
The issuance of an ED was negatively associated with the duration of contacts during both weekdays (Adjusted CRR: 0.90 (95% CI: 0.86-0.95)) and weekends (Adjusted CRR: 0.87 (95% CI: 0.82-0.91)) after adjusting for age and sex. Following the same example given previously, a woman in her 40’s would have an average contact duration of 5.09 hours (95% CI: 4.76-5.43) with other individuals during weekends when an ED was issued and 5.86 hours (95% CI: 5.55-6.18 during periods without any ED.
Because one of the key restrictions during ED was either complete closure or shortening of restaurant/bar hours with restricted hours of serving alcohol (Table B in S1 Text), we investigated the number of individuals who reported contacts at restaurants/bars. For this analysis, the contact location of restaurants and bars were combined which included izakaya—informal Japanese bars where alcohol and food are served—karaoke, and movie theatres. After adjusting for age and sex, these contacts were negatively associated with the issuance of ED during weekdays (Adjusted CRR: 0.77 (95% CI: 0.72-0.83)) and weekends (Adjusted CRR: 0.68 (95% CI: 0.63-0.74)) compared to periods when ED was absent.
Factors associated with contact patterns
Based on our multivariable regression model, an individual with the reference characteristics had an expected 2.35 contacts (95% CI: 1.70-3.23) during the weekday. Each variable was compared to this reference, and we reported on characteristics that were statistically different to this value (Table 3). To check the model fit, the predicted values of weekday contacts were plotted against the observed values reported from the February 2023 contact survey (Fig 2). Residual plots were also evaluated (Fig D in S1 Text). Heteroscedasticity was evident (Fig D.1–2 in S1 Text) which was expected due to the sample distribution showing overdispersion. Model fit was good as the error terms of each predictor in the model showed approximate linearity (Fig D.3–6 in S1 Text). Collinearity was observed with some covariates that had VIFs higher than 2 (Table C in S1 Text), such as school-aged children and those who reported contacts at schools.
Contact patterns varied by location (Table 3). Participants who reported contacts at home, school, restaurants, and bar settings during the weekday had higher contacts than those who did not report any contacts at these locations. Those who lived in a household of four or five people had higher contacts than those who lived in a household of three people. The association between contacts and occupation differed depending on the occupation type where healthcare professionals and government employees, including public school teachers, had significantly higher contacts.
Different work conditions were associated with contact patterns. While those who teleworked had lower contacts, those who moved to a different prefecture at least six times in the past month for work/school purposes had higher contacts. As of February 2023, among those who reported as employed or attending school, 73.2% (818/1118) reported they have never teleworked from home.
Disease mitigation measures, such as mask wearing frequency increased significantly with higher number of contacts, but there was no increasing trend of handwashing frequency with more contacts. There was no evidence of association between contacts and vaccination status and their level of concern in getting infected with COVID-19. Note that among those who reported their frequency of teleworking, there were 451 people (28.71%) who reported not employed and there were 15 people (0.96%) who reported NA on their vaccination status (not shown on Table 3).
Discussion
Our study is one of the few that have characterized social contact patterns in Japan during the COVID-19 pandemic through repeated cross-sectional surveys. Contact surveys were conducted in Japan during and after the Tokyo Olympics in August 2021 that resulted in having a median of three contacts per day (mean of 8.92) [24], similar to our results from the March 2022 survey. The social contact patterns of individuals during illness and after recovery was the subject of another set of contact surveys in Japan from 2022 to 2024 [25]. The authors identified a gradual rise in the overall frequency of contacts compared to 2021. Further, the factors associated with increased contacts, such as school-aged children, increased number of household members, and mask wearing, were similar to those we found in the general population in Fukuoka and Osaka.
Though our surveys did not capture contact patterns in 2020, a study using mobility data showed that there was a 70% reduction in daily total contacts in April 2020 after there were government recommendations to telework and close the public schools [26]. This is comparable to countries such as Norway (67–73% decrease) [27], France (70% decrease) [28], Germany (73% decrease) [29], United Kingdom (74% decrease) [17], Belgium (80% decrease) [30] and United States (82% decrease) [31] where strict lockdowns were implemented. Many contact patterns remained subdued even after post-lockdowns (e.g., China, 21 European countries) [32,33].
Although Japan never implemented lockdowns, our results suggest that the objective of ED issuance was met as it was associated with changes in frequency and duration of contacts. Particularly during the weekend, there were fewer contacts and shorter duration of contacts. This suggests that individuals may have changed their behavior during the days when they have more control over compared to the weekdays when they either commute to school or work. Even as of February 2023, two years after the second ED, the majority (73.2%) of those who were employed or attending school had never worked from home. This corresponds to a survey that showed among approximately 20,000 employed individuals, 70.3% were not teleworking during the first ED in April 2020 when stay-at-home and teleworking recommendations were issued for the first time during the pandemic [34]. Prior to the pandemic, teleworking was a practice rarely implemented in Japanese society; a national survey in September 2019 showed 20.2% of 2,118 companies had implemented teleworking strategies for their employees [35].
The duration of contacts increased with time particularly among children, teenagers, and adults in their 40s. This is correlated with governmental restrictions that were gradually lifted, such as having less stringent measures during semi-EDs compared to EDs. However, the mean number of contacts remained subdued across the first three surveys in 2021, including March when ED was absent, showing that the lifting of ED was not what may have influenced the increase in contact patterns per se. These can be signs of pandemic fatigue [36]—a decline in adhering to public health measures due to the prolonged health crisis—that can lead to a gradual shift towards behavior more like pre-pandemic times. It is also worth noting that there were other factors that also increased with time, such as vaccination and circulation of Omicron, a less virulent strain of SARS-CoV-2. It is, however, important to note that even after ED was lifted in March 2021 before mass vaccination started, the frequency and duration of contacts did not change. This contrasts with periods when the first and second lockdowns were lifted in the UK, also prior to vaccine rollout to the entire public, adults (18 years and above) increased their contacts from 29 to 59% compared to during the first lockdown [19]. These empirical data illustrate how contact patterns can vary from country to country during a pandemic. Understanding how individuals behave, especially during periods prior to vaccine rollout, may provide hints on the level of necessity, timing, and severity of governmental recommendations in future epidemics.
The Japanese population that had consistently low number of contacts was those over 70, which consist of 22.6% of the national population [15]. They not only met with fewer than five contacts during the week, but each contact remained short (average of one hour of contact per day) with very limited physical contact throughout the pandemic (Fig E in S1 Text). Their contact patterns also did not change after mass vaccination started from May 2021 for 65 years and above. With the combination of shielding the older populations in long-term care facilities [37], this may have helped in preventing a surge of COVID-19 outbreaks and deaths amongst this population during the beginning of the pandemic—an issue that was apparent in the US [38] and the UK [38,39].
“Prosocial behavior,” such as physical distancing, mask wearing, and getting tested as described by Sachs JD et al. [3], is critical in pandemic control, and countries in WHO’s Western Pacific region were quick in encouraging it as part of their “suppression strategy” [3]. Our results in Japan reflect this. In addition to EDs, the 3C policy was implemented from early 2020 which was especially relevant in the Japanese context as our results showed that indoor contacts were more frequent than outdoor contacts (Fig F in S1 Text). Individuals who went to restaurants and bars had higher contacts compared to those who did not. The close contact that is likely in these settings highlight the importance of mitigation approaches such as improved ventilation for disease control [40].
In our surveys, there were increased hours of mask wearing with higher number of contacts, demonstrating prosocial behavior. Frequency of mask wearing continued to be stable throughout 2021–2023 which contrasts with the UK where mask wearing was strongly associated with changes in government policy [19]. On the other hand, handwashing was not associated with the frequency of contacts. This may be due to the public message of 3C where avoiding crowded conditions and close-contact settings were highlighted in Japan more than handwashing and disinfection—another key difference compared to other countries [5]. Contacts were not associated with vaccination status, opposite from the European countries where vaccinated individuals reported higher contacts [41]. This could be due to Japan not implementing any restrictions in accessing public or indoor areas due to vaccination status, unlike in Europe where vaccine certificates were issued. Such differences suggest behavioral patterns that are triggered by risk perception vs. governmental intervention. The number of deaths increased after the introduction of the omicron variant, but the suppression strategy in Japan from 2020-2021 – a combination of reduced contacts, prosocial behavior, and high vaccination coverage – may have helped in keeping a low cumulative mortality like other Western Pacific countries where the suppression strategy was implemented early [3].
There are limitations of our study. Our surveys were limited to Osaka and Fukuoka prefectures, where incidence was typically higher than the remaining 45 prefectures in Japan. Extrapolating our findings to other prefectures in Japan should be done with caution, as contact rates may be lower. However, Japanese contact survey studies prior to the COVID-19 pandemic by Ibuka et al. [12] and Munasinghe et al. [13] did not report any differences in contact rates across prefectures, while Tsuzuki et al. [24] examined contact rates between rural and urban areas and did not find any meaningful differences. We can therefore have moderate confidence in the relevance of our findings for the rest of the country. Selection bias could have been introduced as the surveys captured a sample population that had a slightly higher vaccination coverage than the national average. Recording contacts retrospectively may result in recall bias, leading to an underestimation of contacts.
Lastly, there are limitations of the multivariable regression model. The Weibull distribution that was selected for the model matches the high contact numbers better but yields a lesser fit for the lowest contact numbers. This reflects the Weibull distribution’s emphasis on the heavy tail, and the model may better explain the factors associated with individuals who reported higher contacts. The drop seen in the observed contacts after 10 contacts may also be due to the design of the survey where participants were asked to enter additional information on over 10 contacts. Caution is needed when interpreting the coefficients of covariates that showed correlation particularly the school-aged children. Since we hypothesized that they play an important role in contact dynamics, we retained these factors in our model. Due to the limited number of questions that could be asked in each survey, our multivariable regression model could have included covariates that were not included in the model, leading to residual bias. Some of these variables could have impacted contact patterns with time. Thus, future work is planned to investigate factors that could have influenced the change in contacts during the pandemic by analyzing the same individuals or households across multiple time points. A mediation analysis can be another approach to investigate causal mechanisms of the changes in contact patterns.
Compared to other countries, Japan has been unique in tackling the COVID-19 pandemic without implementing lockdowns and long school closures. It mostly relied on prefecture-specific recommendations to the population and some limitations in the business hours and operation of the restauration industry. These measures are difficult to capture with tools such as the Oxford Stringency Index, which emphasize strict interventions, such as work and school closures, implemented in countries including the UK and the US [42]. This underscores the importance of contextualizing different public health measures in each country or cultural setting.
Prosocial behavior, such as limiting contacts especially during the weekend in 2021 and continued mask wearing until 2023 were evident. Although issuance of EDs occurred as the pandemic progressed in waves, contacts remained subdued compared to pre-pandemic times. Contact surveys conducted in late 2022 in western Europe showed post-pandemic behavior with a gradual increase in contacts that followed after most of the COVID-19 restrictions were lifted [43]. However, during this time, Japan had reached its second highest number of COVID-19 cases, which could explain the continued observation of prosocial behavior in 2023. Behavior can be difficult to predict and contact patterns may differ in a future pandemic. The age-specific contacts from our study can be utilized in mathematical modelling, for example to characterize infectious disease spread and support pandemic planning. Most importantly, identifying these country-specific factors that influence human behavior provides further support for policies in controlling disease transmission that are context specific, not only for COVID-19 but also for other infectious diseases that may emerge in the future.
Conclusions
We conducted repeated surveys in Japan to understand how the population modified its behavior and contact patterns during the pandemic. Our results showed that daily contacts dropped by approximately 50% during the pandemic, rebounding only slowly after the government recommendations were relaxed. People proceeded with caution as they wore masks longer if they had more contacts and consistently wore masks until 2023, even after a full year since the last governmental recommendation was lifted. The frequency of contacts did not increase after individuals received the COVID-19 vaccine which is in contrast with European countries where the opposite trend was reported. Our findings provide evidence on the importance of reduced contacts, careful behavior, and high vaccination coverage that potentially limited disease transmission and mortality in Japan.
Supporting information
S1 Text. Supporting Information for “Continuing to be Cautious: Japanese Contact Patterns during the COVID-19 Pandemic and their Association with Public Health Recommendations.”
https://doi.org/10.1371/journal.pgph.0004600.s001
(PDF)
Acknowledgments
We would like to acknowledge F-press that implemented the social contact surveys online and the study population in Osaka and Fukuoka prefectures who kindly participated in our surveys. We thank John Edmunds and our colleagues within the Centre for Mathematical Modelling of Infectious Diseases at LSHTM, the National Institute for Infectious Diseases in Japan, and Nagasaki University’s School of Tropical Medicine and Global Health and the Department of Clinical Medicine for their inputs during the development of this study.
References
- 1.
World Health Organization. WHO Coronavirus (COVID-19) Dashboard | WHO Coronavirus (COVID-19) Dashboard With Vaccination Data. https://covid19.who.int/. 2023. 2023 April 5.
- 2.
The World Bank. World Development Indicators | DataBank. https://databank.worldbank.org/reports.aspx?source=2&series=SP.POP.65UP.TO.ZS&country=. 2023. 2023 July 5.
- 3. Sachs JD, Karim SSA, Aknin L, Allen J, Brosbøl K, Colombo F, et al. The Lancet Commission on lessons for the future from the COVID-19 pandemic. The Lancet. 2022;400(10359):1224–80.
- 4.
Ministry of Education Culture Sports Science and Technology. Nationwide simultaneous temporary closure of schools (from 28 February to before spring break). 全国一斉臨時休業関係(2/28～春季休業前まで): 文部科学省. https://www.mext.go.jp/a_menu/coronavirus/mext_00006.html. 2020. 2023 April 4.
- 5. Oshitani H. COVID lessons from Japan: the right messaging empowers citizens. Nature. 2022;605(7911):589. pmid:35606523
- 6. Wallinga J, Teunis P, Kretzschmar M. Using data on social contacts to estimate age-specific transmission parameters for respiratory-spread infectious agents. Am J Epidemiol. 2006;164(10):936–44. pmid:16968863
- 7. Leung K, Wu JT, Leung GM. Real-time tracking and prediction of COVID-19 infection using digital proxies of population mobility and mixing. Nat Commun. 2021;12(1):1501. pmid:33686075
- 8. Prem K, Zandvoort K van, Klepac P, Eggo RM, Davies NG, Centre for the Mathematical Modelling of Infectious Diseases COVID-19 Working Group, et al. Projecting contact matrices in 177 geographical regions: An update and comparison with empirical data for the COVID-19 era. PLoS Comput Biol. 2021;17(7):e1009098. pmid:34310590
- 9. Nomura S, Tanoue Y, Yoneoka D, Gilmour S, Kawashima T, Eguchi A, et al. Mobility Patterns in Different Age Groups in Japan during the COVID-19 Pandemic: a Small Area Time Series Analysis through March 2021. J Urban Health. 2021;98(5):635–41. pmid:34379269
- 10. Nagata S, Nakaya T, Adachi Y, Inamori T, Nakamura K, Arima D, et al. Mobility change and COVID-19 in Japan: mobile data analysis of locations of infection. Journal of Epidemiology. 2021;31(6):387.
- 11. Mossong J, Hens N, Jit M, Beutels P, Auranen K, Mikolajczyk R, et al. Social contacts and mixing patterns relevant to the spread of infectious diseases. PLoS Med. 2008;5(3):e74. pmid:18366252
- 12. Ibuka Y, Ohkusa Y, Sugawara T, Chapman GB, Yamin D, Atkins KE, et al. Social contacts, vaccination decisions and influenza in Japan. J Epidemiol Community Health. 2016;70(2):162–7. pmid:26424846
- 13. Munasinghe L, Asai Y, Nishiura H. Quantifying heterogeneous contact patterns in Japan: a social contact survey. Theor Biol Med Model. 2019;16(1):6. pmid:30890153
- 14. Liu CY, Berlin J, Kiti MC, Del Fava E, Grow A, Zagheni E, et al. Rapid Review of Social Contact Patterns During the COVID-19 Pandemic. Epidemiology. 2021;32(6):781–91. pmid:34392254
- 15.
National Statistics Center. National population estimates as of 1 October 2021. Portal Site of Official Statistics of Japan. https://www.e-stat.go.jp/stat-search/files?page=1&layout=datalist&toukei=00200524&tstat=000000090001&cycle=7&year=20210&month=0&tclass1=000001011679&tclass2val=0. 2021. 2023 April 17.
- 16.
Government of Japan. 基本的対処方針に基づく対応｜内閣官房新型コロナウイルス等感染症対策推進室. 2023 https://corona.go.jp/emergency/
- 17. Jarvis CI, Van Zandvoort K, Gimma A, Prem K, CMMID COVID-19 working group, Klepac P, et al. Quantifying the impact of physical distance measures on the transmission of COVID-19 in the UK. BMC Med. 2020;18(1):124. pmid:32375776
- 18. Jarvis CI, Gimma A, van Zandvoort K, Wong KLM, CMMID COVID-19 working group, Edmunds WJ. The impact of local and national restrictions in response to COVID-19 on social contacts in England: a longitudinal natural experiment. BMC Med. 2021;19(1):52. pmid:33602244
- 19. Gimma A, Munday JD, Wong KLM, Coletti P, van Zandvoort K, Prem K, et al. Changes in social contacts in England during the COVID-19 pandemic between March 2020 and March 2021 as measured by the CoMix survey: A repeated cross-sectional study. PLoS Med. 2022;19(3):e1003907. pmid:35231023
- 20.
United Nations. World Population Prospects - Population Division - United Nations. https://population.un.org/wpp/Download/Standard/MostUsed/. 2022. 2023 July 24.
- 21.
Digital Agency of Japan. COVID-19 Vaccination Coverage Status of Japan 新型コロナワクチンの接種状況 | デジタル庁. https://info.vrs.digital.go.jp/dashboard. 2023. 2023 August 24.
- 22.
National Institute of Infectious Diseases Japan. Update on the COVID-19 Vaccine. https://www.niid.go.jp/niid/ja/diseases/ka/corona-virus/2019-ncov/2484-idsc/10569-covid19-53.html. 2021. 2023 April 13.
- 23.
Ministry of Health Labor and Welfare. Report on COVID-19 serosurvey of Japan 新型コロナウイルス感染症に関する抗体保有状況調査について｜厚生労働省. 2023. https://www.mhlw.go.jp/stf/seisakunitsuite/bunya/0000121431_00132.html
- 24. Tsuzuki S, Asai Y, Ibuka Y, Nakaya T, Ohmagari N, Hens N, et al. Social contact patterns in Japan in the COVID-19 pandemic during and after the Tokyo Olympic Games. J Glob Health. 2022;12:05047. pmid:36462208
- 25. Jeong H, Park S, Chun JY, Ohmagari N, Kim Y, Tsuzuki S. Chronological trend of social contact patterns in Japan after the emergence of COVID-19. J Infect Public Health. 2025;18(2):102629. pmid:39733687
- 26. Yabe T, Tsubouchi K, Fujiwara N, Wada T, Sekimoto Y, Ukkusuri SV. Non-compulsory measures sufficiently reduced human mobility in Tokyo during the COVID-19 epidemic. Scientific Reports. 2020;10(1).
- 27.
Veneti L, Robberstad B, Steens A, Forland F, Winje BA, Vestrheim DF, et al. Social contact patterns during the early COVID-19 pandemic in Norway: insights from a panel study, April to September 2020. Cold Spring Harbor Laboratory. 2023. doi: https://doi.org/10.1101/2023.11.18.23298731
- 28. Bosetti P, Huynh BT, Abdou AY, Sanchez M, Eisenhauer C, Courtejoie N, et al. Lockdown impact on age-specific contact patterns and behaviours, France, April 2020. Eurosurveillance. 2021;26(48):1.
- 29. Tomori DV, Rübsamen N, Berger T, Scholz S, Walde J, Wittenberg I, et al. Individual social contact data and population mobility data as early markers of SARS-CoV-2 transmission dynamics during the first wave in Germany-an analysis based on the COVIMOD study. BMC Med. 2021;19(1):271. pmid:34649541
- 30. Coletti P, Wambua J, Gimma A, Willem L, Vercruysse S, Vanhoutte B, et al. CoMix: comparing mixing patterns in the Belgian population during and after lockdown. Sci Rep. 2020;10(1):21885. pmid:33318521
- 31. Feehan DM, Mahmud AS. Quantifying population contact patterns in the United States during the COVID-19 pandemic. Nat Commun. 2021;12(1):893. pmid:33563992
- 32. Zhang J, Litvinova M, Liang Y, Zheng W, Shi H, Vespignani A, et al. The impact of relaxing interventions on human contact patterns and SARS-CoV-2 transmission in China. Sci Adv. 2021;7(19):eabe2584. pmid:33962957
- 33. Wong KLM, Gimma A, Coletti P, CoMix Europe Working Group, Faes C, Beutels P, et al. Social contact patterns during the COVID-19 pandemic in 21 European countries - evidence from a two-year study. BMC Infect Dis. 2023;23(1):268. pmid:37101123
- 34.
Ministry of Internal Affairs and Communications. Part 1: Special Feature: Digitally Supporting Lifestyles and the Economy 総務省｜令和3年版 情報通信白書｜テレワークの実施状況. https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/r03/html/nd123410.html. 2021. 2023 July 6.
- 35.
Ministry of Internal Affairs and Communications. Results of the 2020 Survey on Communications Usage Trends. 総務省｜報道資料｜令和元年通信利用動向調査の結果. https://www.soumu.go.jp/menu_news/s-news/01tsushin02_02000148.html. 2020. 2023 July 6.
- 36.
World Health Organization RO. Pandemic fatigue – reinvigorating the public to prevent COVID-19: policy framework for supporting pandemic prevention and management. World Health Organization. 2020. https://iris.who.int/handle/10665/335820
- 37. Abe K, Kawachi I. Deaths in Nursing Homes During the COVID-19 Pandemic-Lessons from Japan. JAMA Health Forum. 2021;2(2):e210054. pmid:36218790
- 38. Werner RM, Hoffman AK, Coe NB. Long-Term Care Policy after Covid-19 - Solving the Nursing Home Crisis. N Engl J Med. 2020;383(10):903–5. pmid:32459918
- 39. Giddings R, Krutikov M, Palmer T, Fuller C, Azmi B, Shrotri M, et al. Changes in COVID-19 outbreak severity and duration in long-term care facilities following vaccine introduction, England, November 2020 to June 2021. Euro Surveill. 2021;26(46):2100995. pmid:34794537
- 40. Saeedi R, Ahmadi E, Hassanvand MS, Mohasel MA, Yousefzadeh S, Safari M. Implemented indoor airborne transmission mitigation strategies during COVID-19: a systematic review. J Environ Health Sci Eng. 2023;21(1):11–20. pmid:37152068
- 41. Wambua J, Loedy N, Jarvis CI, Wong KLM, Faes C, Grah R, et al. The influence of COVID-19 risk perception and vaccination status on the number of social contacts across Europe: insights from the CoMix study. BMC Public Health. 2023;23(1):1350. pmid:37442987
- 42.
Mathieu E, Ritchie H, Rodés-Guirao L, Appel C, Giattino C, Hasell J, et al. Coronavirus Pandemic (COVID-19). https://ourworldindata.org/covid-stringency-index. 2020. 2024 May 31.
- 43. Jarvis CI, Coletti P, Backer JA, Munday JD, Faes C, Beutels P, et al. Social contact patterns following the COVID-19 pandemic: a snapshot of post-pandemic behaviour from the CoMix study. Epidemics. 2024;48:100778. pmid:38964131