Abstract
Background
Public health and social measures (PHSMs) are key interventions during pandemics. This study quantifies temporal changes in daily interpersonal contacts associated with PHSM implementation and their potential impact on disease transmission across different phases of the COVID-19 pandemic in Hong Kong from 2021 to 2023.
Methods
We recruited 7,053 individuals from the general population and 3,462 participants from selected high-contact groups between September 2021 and December 2023. The study period spanned the infection-naïve phase before 2022 through the city’s fifth wave of COVID-19 (January–April 2022) driven by the Omicron BA.2 variant. Reported cumulative incidence was < 0.2% of the population before the fifth wave and rose to 16% afterward. The onset of the fifth wave prompted the government to impose unprecedentedly stringent PHSMs. Participants reported their daily number of contacts via contact surveys, from which we constructed social contact matrices. We then compared the dominant eigenvalues of these contact matrices to estimate the potential association of PHSMs with changes in the basic reproduction number (
).
Results
Among the general population, the average number of daily contacts dropped significantly from 6.2 (mean; 95% confidence interval [CI]: 5.9–6.5) during pre-fifth wave to 5.1 (5.0–5.3) during the fifth wave. After the fifth wave, the average number of contacts remained low at 5.1 (5.0–5.3) when some PHSMs such as mask-wearing policies were still in place. Once all PHSMs were lifted after March 2023, the average number of contacts increased to 8.4 (8.3–8.5). Compared to the general population, participants from the selected high-contact groups reported more contacts by 23% (95% CI: 21%–26%) on average. Compared to pre-fifth wave levels, PHSMs imposed during the fifth wave were associated with a 20% reduction (median; 95% percentile interval: 1%–40%) in 
. Relaxation of PHSMs in the post-pandemic period suggested an increase in 
by 109% (61%–151%) relative to pre-fifth wave levels.
Conclusions
During the COVID-19 pandemic in Hong Kong, PHSMs were temporally associated with reduced social contacts and, potentially, disease transmission. These quantitative estimates may indicate the potential value of PHSMs for informing future pandemic preparedness and response.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12916-026-05041-1.
Keywords: Social contact pattern, Public health and social measure (PHSM), COVID-19, Age-assortative mixing, Contact matrix
Background
Understanding social contact patterns, including age-dependence, frequency, location, and type of contacts, is crucial to projecting the spread of infectious diseases and developing effective public health and social measures (PHSMs) such as school and workplace closures, stay-at-home orders, and mass gathering restrictions [1, 2]. Contact patterns changed dramatically in many countries as the COVID-19 pandemic unfolded and evolved [3, 4]. Extensive social contact studies, such as POLYMOD and CoMix, have been conducted in Western Europe and North America before and during the pandemic, providing insights to explore the corresponding changes [5–9]. For example, a survey study in the United Kingdom reported that the average number of daily contacts dropped to 2.8 in the first four days when lockdown was implemented in March 2020, compared to an estimate of 10.8 from POLYMOD in 2005–2006 [5]. In a population-based Canadian study, the number of contacts decreased from 7.8 before the pandemic in 2018–2019 to 3.1 during the lockdown in spring 2020 [9]. In Asia, a Chinese study documented a decrease in contacts in Wuhan and Shanghai from 14.6 to 18.8 in 2017–2019 to just 2.0–2.3 per day in February 2020, when the number of reported cases of the first wave of the pandemic peaked across the Chinese Mainland and stringent interventions (e.g., social distancing policies) were implemented [10]. The sizeable reductions in social contacts underscored the global impact of PHSMs on contact patterns and, potentially, transmission dynamics [10].
Hong Kong offers a unique setting for studying social contact patterns during the COVID-19 pandemic. To control transmission, the government implemented a broad suite of PHSMs, including restrictions on group gatherings (e.g., not more than 2/4/8 people) and on inbound travel for more than three years [11]. Unlike many other regions (e.g., those in Europe), Hong Kong never enacted a full citywide lockdown. Despite these prolonged PHSMs, the city experienced five waves of COVID-19 between 2020 and 2022. The first four together resulted in a cumulative incidence of less than 0.2% of the population, markedly lower than that in many other parts of the world [12, 13]. The fifth wave, occurring in January–April 2022 and driven predominantly by the highly transmissible SARS-CoV-2 Omicron BA.2 variants, led to a sharp surge in cases with a peak daily incidence of nearly 80,000 cases and a cumulative incidence of 1.2 million cases (16% of the population) [12, 14]. Sero-epidemiological estimates indicated that 45% (posterior median; 95% credible interval: 41–48%) of the population were infected during the fifth wave, underscoring significant cases under ascertainment [15].
The fifth wave was accompanied by the most stringent PHSMs implemented in the city, including school closures and work-from-home arrangements, which substantially curtailed social interactions [16, 17]. After the fifth wave, some PHSMs were gradually relaxed, although mask mandates, restrictions on group gatherings, and the half-day requirement for face-to-face schooling remained in effect [11]. By March 2023, all PHSMs were lifted, and travel between Hong Kong and the Chinese Mainland resumed after nearly three years of interruption.
In this study, we examined age- and setting-specific contact patterns during the COVID-19 pandemic and quantified how these patterns changed over time.
Methods
Participants and the survey
We conducted 12 rounds of surveys from September 2021 to December 2023. We aimed to recruit 10,000 participants, comprising individuals from the general population reached by random digit dialling and members of eight selected high-contact groups (e.g., catering and retail staff, school teachers) given their comparatively high contact rates. Each participant reported contacts for two days (one weekday and one weekend day), and no participants contributed data to multiple survey rounds.
The survey design was modelled on the POLYMOD and CoMix studies conducted in Europe [5, 18–20], adapted to the Hong Kong context and aligned with our previous study conducted in 2015–2016 [21]. Participants reported their social contacts from the preceding day via online or paper questionnaires. To minimize recall bias, automated reminders were sent the day before each assigned reporting day to support prospective recording. Details of recruitment and implementation are provided in Additional file 1: Supplementary Information (pages 1–3) [5–7, 21].
Statistical analysis
We estimated the mean number of daily contacts and its 95% confidence interval (CI), with subgroup analyses stratified by age, sex, vaccination status, and contact setting. While we included survey data collected before the fifth wave (from September 2021), all data were collected during the pandemic. As such, we compared these estimates with those from our 2015–2016 pre-pandemic contact survey, which used similar methods [21]. We used negative binomial regression models to examine associations between daily contact counts and covariates including sex and vaccination status.
When analysing the social contact matrices, we first divided participants into 16 age groups (0–5, 6–10, 11–15, …, 71–75, and 76 or above). We then constructed contact matrices with entries 
across time, where 
denotes the mean number of daily contacts that individuals in age group j make with individuals in age group i. The contact matrices were presented in a form 
that was adjusted for the age distribution in the study population and reciprocity of contacts [5, 22]. Specifically, we calculated 
, where 
represents the number of individuals in the age group i in the population, such that 
. Similar to other CoMix studies, such as Jarvis et al. and Veneti et al. [5, 23], we used the contact matrices to estimate the relative changes in basic reproduction number (
). Specifically, for each contact matrix, we computed its dominant eigenvalue which, under the next-generation matrix (NGM) approach, is proportional to 
[24]. This proportional relationship between the dominant eigenvalue calculated from the contact matrix and 
is based on the assumptions that the population is fully susceptible to the circulating variants and that the transmission parameters remain constant [5, 23]. We then examined temporal changes in this metric to assess the potential impact of PHSMs on 
by comparing the calculated dominant eigenvalues over the study period with its corresponding value at the beginning of the study, a standard approach in studies on the association between social contact patterns and disease transmission [5]. We applied bootstrap resampling at the level of individual participants, with replacement, to generate 1000 replicated datasets. Using the replicated datasets, we estimated the 90% CIs of the relative changes in 
. Moreover, some earlier studies have suggested that a duration of 15 min or more (in a single contact or cumulatively) may be associated with an increased risk of SARS-CoV-2 transmission, although this threshold (15 min) may vary across different variants due to differences in transmissibility, viral load, and minimal infectious dose (e.g., the current Omicron family variants may be more transmissible) [25–28]. We performed a sensitivity analysis to estimate the changes in 
by reconstructing the contact matrices considering contacts lasting at least 15 min. This sensitivity analysis was intended to assess the robustness of the findings to a more restrictive contact definition.
Throughout the pandemic, multiple PHSMs were often implemented concurrently [11], with potential additive or synergistic effects. We divided the study period into four phases anchored around the fifth wave and associated shifts in PHSMs in Hong Kong:
Pre-fifth wave (17-Sep-2021 to 6-Jan-2022): PHSMs mainly focused on restricting inbound travel;
Fifth wave (7-Jan-2022 to 20-Apr-2022): the strictest PHSMs were in force, including work-from-home arrangements and school closures;
Post-fifth wave (21-Apr-2022 to 28-Feb-2023): some PHSMs were relaxed but mask mandate, group gathering restrictions, and half-day requirement for face-to-face schooling remained; and
Post-pandemic (1-Mar-2023 onwards): all PHSMs were lifted and schools resumed whole-day classes.
Results
A total of 10,515 (51% male) participants enrolled in the study. Over 97% completed an online questionnaire by themselves. Overall, 3,462 (33%) participants belonged to one of the selected high-contact groups: among them, 22% worked in retail, 21% in construction, and 17% in catering. Detailed demographic data are provided in Additional file 1: Table S1.
Contact patterns
Figure 1 presents temporal changes in the average number of daily contacts among respondents from the general population, stratified by age, and location and type of contacts. The average number of daily contacts varied significantly throughout the study period. During the fifth wave, when PHSMs were most stringent (from January 7 to April 20, 2022), participants reported fewer contacts than in the pre-fifth wave (Table 1 and Additional file 1: Table S3). Specifically, in the pre-fifth wave period, participants from the general population reported an average of 6.2 (95% CI: (5.9, 6.5)) daily contacts, which dropped to 5.1 (5.0, 5.3) during the fifth wave. After the fifth wave, particularly from August 2022 to January 2023, the reported number of daily contacts (5.1 (5.0, 5.3)) remained similar to the levels during the fifth wave despite relaxation of some PHSMs, e.g., school closures and work-from-home arrangements. When all PHSMs were lifted from March 2023 onward in the post-pandemic period, the average number of daily contacts increased substantially to 8.4 (8.3, 8.5).
Fig. 1.
Average number of daily contacts per individual and changes in the dominant eigenvalue of contact matrices over the study period. Note: Panel (a) presents the number of cases in the population that tested positive by polymerase chain reaction (PCR) nucleic acid tests or rapid antigen tests (RATs) across the study period [12]. The number of cases was reported up to 29-May-2023. Panel (b) shows the timeline of the implementation of different PHSMs during the study period [11]. Panel (c) presents the cumulative proportion of individuals who received at least one dose of the COVID-19 vaccine in the general population of Hong Kong by age group [12]. Panel (d) presents the average number of daily contacts among all participants stratified by the general population and the high-contact groups over the study period. Panels (e–g) depict the average number of daily contacts among the general population stratified by (e) participants’ age group, (f) location of contacts, and (g) type of physical or non-physical contacts. Regarding contact locations, the category “Other places” includes public transport, places of worship, sports facilities (e.g., gyms and pools), outdoor environments (e.g., streets and country parks), entertainment and retail venues (e.g., cinemas, supermarkets, and malls), and food and beverage venues (e.g., restaurants and bars). For panels (d–g), the markers denote the observed mean of daily contacts in each month and the error bars denote the corresponding 90% confidence intervals. An arrow sign denotes that the lower/upper of the confidence interval exceeds the limit of the y-axis. The lines present the fitted mean of contact and the shaded regions present the corresponding 95% confidence bands, both estimated by the generalized additive models (GAMs). Panel (h) presents the ratio of the dominant eigenvalues of contact matrices for all contacts, physical contacts and non-physical contacts made by participants from the general population. The data were divided into intervals, each of which contained approximately every 1,000 completed surveys. The ratios were calculated as the dominant eigenvalue of the contact matrices over time divided by that in the first interval at the beginning of the study which served as a reference baseline. A ratio over 1 represents a higher transmission potential (in terms of a higher basic reproduction number 
) compared to the baseline based on the reported contact matrix. The markers denote the median of the ratios and the error bars denote the corresponding 90% percentile intervals, based on 1,000 bootstrapping replications. The lines present the fitted mean of the ratios and the shaded regions present the corresponding 95% confidence bands, both estimated by the generalized additive models (GAMs). For GAMs (panels d–h), (1) the time of the questionnaire completed was treated as the smoothing term, with a thin plate regression spline for the penalized smoothing, and (2) the generalized cross-validation (GCV) and Mallow’s Cp were used for smoothing parameter estimation
Table 1.
Reported number of daily contacts per participant in the current study and those from our previous Hong Kong-based study conducted in 2015–2016
| Variables | Pre-fifth wave 2021-09-17 to 2022-01-06 |
Fifth wave 2022-01-07 to 2022-04-20 |
Post-fifth wave 2022-04-21 to 2023-02-28 |
Post-pandemic 2023-03-01 to 2023-12-31 |
2015–2016 surveya | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| n b | Reported contacts, mean (95% CI) | n b | Reported contacts, mean (95% CI)c | n b | Reported contacts, mean (95% CI)c | n b | Reported contacts, mean (95% CI)c | n a | Reported contacts, mean (95% CI)a | |
| Among participants from the general population | ||||||||||
| Overall | 1,308 | 6.2 (5.9, 6.5) | 2,936 | 5.1 (5.0, 5.3)*** | 2,631 | 5.1 (5.0, 5.3)*** | 6,650 | 8.4 (8.3, 8.5) *** | 1,149 | 8.1 (8.0, 8.2) |
| Sex | ||||||||||
| Female | 883 | 6.1 (5.7, 6.4) | 1,420 | 5.1 (4.9, 5.3)*** | 1,438 | 5.5 (5.2, 5.7)* | 3,199 | 8.5 (8.3, 8.7)*** | 557 | 7.4 (7.3, 7.6) |
| Male | 425 | 6.5 (5.9, 7.1) | 1,516 | 5.2 (5.0, 5.4) | 1,193 | 4.8 (4.6, 4.9)** | 3,218 | 8.4 (8.2, 8.6)*** | 592 | 8.8 (8.6, 9.0) |
| Age | ||||||||||
| 0–10 | 185 | 7.4 (6.5, 8.3) | 535 | 5.2 (4.7, 5.6)*** | 220 | 5.4 (4.8, 5.9)*** | 1,154 | 10.1 (9.7, 10.5)*** | 156 | 8.7 (8.4, 8.9) |
| 11–20 | 66 | 5.9 (4.4, 7.4) | 387 | 4.7 (4.4, 5.1) | 387 | 5.0 (4.8, 5.3) | 1,144 | 8.4 (8.1, 8.7)*** | 196 | 9.4 (9.1, 9.7) |
| 21–40 | 713 | 5.8 (5.4, 6.2) | 580 | 5.2 (4.8, 5.6)* | 839 | 5.8 (5.4, 6.1) | 1,492 | 7.8 (7.5, 8.1)*** | 334 | 8.3 (8.1, 8.5) |
| 41–65 | 334 | 6.4 (5.8, 7.0) | 923 | 5.5 (5.2, 5.7) | 819 | 4.8 (4.5, 5.0)*** | 2,086 | 8.2 (8.0, 8.4)*** | 388 | 8.0 (7.8, 8.1) |
| 66+ | 10 | 4.0 (1.3, 6.7) | 511 | 4.8 (4.6, 5.0) | 366 | 4.6 (4.4, 4.8) | 774 | 7.6 (7.4, 7.9)** | 75 | 3.5 (3.3, 3.7) |
| Day of the week | ||||||||||
| Weekday | 790 | 6.8 (6.4, 7.2) | 1,532 | 5.3 (5.1, 5.5)*** | 1,121 | 5.6 (5.3, 5.8)*** | 3298 | 8.9 (8.7, 9.1)*** | 847 | 8.1 (8.0, 8.3) |
| Weekend | 518 | 5.3 (4.9, 5.7) | 1,404 | 5.0 (4.8, 5.2) | 1,510 | 4.8 (4.7, 5.0) | 3352 | 7.9 (7.8, 8.1)*** | 302 | 8.0 (7.7, 8.2) |
| Household size | ||||||||||
| 1 | 51 | 6.0 (4.4, 7.7) | 280 | 4.5 (4.2, 4.8) | 232 | 3.9 (3.6, 4.2) | 416 | 6.7 (6.3, 7.1)* | 51 | 5.4 (4.9, 5.8) |
| 2 | 212 | 5.3 (4.5, 6.0) | 591 | 4.7 (4.4, 5.0) | 577 | 4.3 (4.1, 4.6) | 1,388 | 7.3 (7.0, 7.5)*** | 178 | 6.5 (6.2, 6.8) |
| 3 | 362 | 5.8 (5.2, 6.3) | 803 | 4.8 (4.5, 5.1) | 800 | 4.9 (4.7, 5.2) | 2,058 | 7.9 (7.7, 8.2)*** | 307 | 7.7 (7.5, 7.9) |
| 4 | 389 | 6.0 (5.6, 6.5) | 742 | 5.2 (4.9, 5.5)*** | 635 | 5.5 (5.2, 5.8) | 1,752 | 9.3 (9.0, 9.5)*** | 380 | 8.6 (8.4, 8.8) |
| 5 or above | 294 | 7.7 (6.9, 8.4) | 520 | 6.4 (6.0, 6.9)** | 387 | 6.9 (6.4, 7.4) | 1,036 | 10.1 (9.7, 10.5)*** | 233 | 9.7 (9.5, 10.0) |
| Monthly household income (HK$) | ||||||||||
| 0–9,999 | 33 | 6.5 (3.9, 9.1) | 485 | 4.6 (4.3, 4.8) | 327 | 4.3 (4.1, 4.6) | 455 | 7.3 (6.9, 7.8)* | 187 | 6.5 (6.2, 6.8) |
| 10,000–19,999 | 102 | 5.3 (4.4, 6.2) | 464 | 4.8 (4.5, 5.1) | 498 | 4.9 (4.6, 5.2) | 814 | 7.7 (7.4, 8.1)*** | 223 | 8.5 (8.3, 8.8) |
| 20,000–39,999 | 443 | 6.7 (6.2, 7.3) | 916 | 5.3 (5.0, 5.6)*** | 959 | 5.0 (4.8, 5.2)*** | 2,472 | 8.3 (8.1, 8.5)*** | 225 | 8.9 (8.7, 9.2) |
| ≥40,000 | 656 | 6.1 (5.7, 6.5) | 779 | 5.6 (5.3, 6.0)* | 649 | 6.0 (5.6, 6.3) | 2,201 | 9.1 (8.8, 9.3)*** | 186 | 9.8 (9.6, 10.1) |
| Vaccinated against COVID-19 | ||||||||||
| Yes | 748 | 6.3 (5.9, 6.7) | 2,051 | 5.3 (5.1, 5.5)** | 2,268 | 5.2 (5.0, 5.3)*** | 5,138 | 8.1 (8.0, 8.3)*** | – | |
| No | 560 | 6.0 (5.6, 6.5) | 885 | 4.8 (4.5, 5.1)*** | 363 | 5.0 (4.6, 5.3) | 1,512 | 9.4 (9.0, 9.8)*** | – | |
| Among participants from high-contact groups | – | |||||||||
| Overall | 500 | 8.8 (8.0, 9.6) | 1,023 | 7.8 (7.3, 8.2) | 2,214 | 6.7 (6.5, 6.9)** | 3,094 | 9.3 (9.1, 9.6)*** | – | |
| Selected groups | ||||||||||
| Residents from residential care homes d | 2 | 20.5 (18.2, 22.8) | 19 | 7.2 (5.6, 8.8)* | 253 | 7.8 (7.4, 8.3)* | 103 | 9.1 (8.5, 9.8)* | – | |
| Staff from catering businesses | 101 | 7.5 (6.1, 8.9) | 189 | 8.2 (7.0, 9.4) | 306 | 5.4 (4.9, 6.0)* | 536 | 9.5 (8.8, 10.2)** | – | |
| Domestic helpers | 2 | 4.0 (4, 4) | 4 | 4.8 (1, 8.2) | 250 | 5.4 (5.1, 5.7) | 296 | 8.3 (7.7, 8.9) | – | |
| School teachers | 69 | 11.8 (8.2, 15.3) | 91 | 5.3 (4.4, 6.2)*** | 206 | 8.6 (7.4, 9.8) | 311 | 9.1 (8.3, 9.9) | – | |
| Staff from residential care homes d | 10 | 16.8 (3.7, 29.9) | 98 | 10.0 (8.0, 12.0) | 212 | 8.8 (8.0, 9.6) | 270 | 10.8 (9.7, 11.9) | – | |
| Staff from retail businesses | 200 | 8.7 (7.7, 9.7) | 345 | 7.1 (6.5, 7.8) | 286 | 5.2 (4.6, 5.8)*** | 629 | 9.9 (9.3, 10.4)*** | – | |
| Delivery staff | 63 | 7.6 (6.0, 9.1) | 137 | 9.6 (8.2, 11.0) | 148 | 9.3 (7.9, 10.7) | 285 | 8.7 (7.9, 9.5) | – | |
| Construction workers | 57 | 8.0 (6.2, 9.8) | 153 | 7.1 (6.2, 8.0) | 554 | 6.0 (5.6, 6.4) | 686 | 9.1 (8.7, 9.5)* | – | |
| Sex | – | |||||||||
| Female | 284 | 9.9 (8.7, 11.1) | 505 | 8.7 (8.0, 9.4) | 1,074 | 7.2 (6.8, 7.6)** | 1,626 | 9.9 (9.5, 10.2)* | – | |
| Male | 216 | 7.4 (6.5, 8.4) | 518 | 6.8 (6.3, 7.4) | 1,140 | 6.2 (5.9, 6.5) | 1,468 | 8.7 (8.4, 9.0)*** | – | |
| Age | ||||||||||
| 21–40 | 376 | 8.9 (8.0, 9.9) | 537 | 9.0 (8.3, 9.7) | 1,228 | 7.3 (7.0, 7.7) | 1,913 | 9.1 (8.8, 9.4)*** | – | |
| 41–65 | 108 | 9.1 (7.4, 10.9) | 379 | 6.7 (6.1, 7.3)*** | 828 | 6.0 (5.7, 6.3)*** | 1,061 | 9.7 (9.3, 10.1)* | – | |
| Day of the week | – | |||||||||
| Weekday | 288 | 9.9 (8.8, 11.1) | 502 | 8.2 (7.6, 8.9)* | 1,094 | 7.2 (6.9, 7.6)** | 1,524 | 9.7 (9.4, 10.1)** | – | |
| Weekend | 212 | 7.3 (6.4, 8.3) | 521 | 7.3 (6.7, 7.9) | 1,120 | 6.2 (5.9, 6.5) | 1,570 | 8.9 (8.6, 9.2)*** | – | |
| Vaccinated against COVID-19 | – | |||||||||
| Yes | 364 | 9.2 (8.2, 10.2) | 891 | 7.8 (7.3, 8.3) | 2,140 | 6.7 (6.5, 6.9)** | 2,843 | 9.4 (9.2, 9.7)*** | – | |
| No | 136 | 7.8 (6.8, 8.9) | 132 | 7.5 (6.3, 8.7) | 74 | 6.1 (4.8, 7.4) | 251 | 8.2 (7.3, 9.2) | – | |
Abbreviations: CI confidence interval
aThe results of daily contacts in the 2015–2016 study were based on Leung et al. 2017 [21]. “n” refers to the number of participants recruited in the 2015–16 survey. The average number of daily contacts was weighted by inverse probability of treatment weighting (IPTW) using the propensity score for the mode of questionnaires
b“n” refers to the number of completed surveys
cComparisons to before the fifth wave were conducted using the Mann-Whitney U test, with *, **, and *** present p-values <0.05, <0.01, and <0.001, respectively
dResidential care homes included residential care homes for elderly persons (RCHE) and residential care homes for persons with disabilities (RCHD)
Location-specific trends suggested the impact of PHSMs on social behaviours. The average number of daily contacts made at home, at workplaces, and at school, and those made physically demonstrated a substantial drop during the fifth wave (Fig. 1f and g). The number of contacts made in public venues such as venues for entertainment and sports, restaurants, and public transport (“Other places” in Fig. 1f) and the number of non-physical contacts (Fig. 1g) increased greatly after March 2023 when all PHSMs were lifted, concurrent with the rise in the overall number of contacts in the post-pandemic period.
Participants in high-contact groups consistently reported more daily contacts than the general population (Fig. 1d), with a relative increase by 23% (21%, 26%) after adjustment for participants’ age, gender, and vaccination status (Table 2). Specifically, the average daily number of contacts made by participants in the high-contact groups was 8.8 (8.0, 9.6) before the fifth wave, decreasing to 7.8 (7.3, 8.2) during the fifth wave and 6.7 (6.5, 6.9) in post-fifth wave (Table 1). In the post-pandemic period, they reported 9.3 (9.1, 9.6) daily contacts on average.
Table 2.
Relative number of reported daily contacts made by an individual
| Factors | Relative number of reported daily contacts | 95% confidence interval | p-value |
|---|---|---|---|
| Age of participants | |||
| 0–20 years | 1.02 | (0.99, 1.06) | 0.23 |
| 21–65 years | 1 | Reference | |
| 66+ years | 0.95 | (0.92, 0.99) | 0.015 |
| Gender | |||
| Male | 1 | Reference | |
| Female | 1.06 | (1.04, 1.08) | < 0.001 |
| Household size | |||
| 1–2 | 0.87 | (0.85, 0.89) | < 0.001 |
| 3–4 | 1 | Reference | |
| 5 or above | 1.19 | (1.16, 1.23) | < 0.001 |
| Day of contact | |||
| Weekday | 1 | Reference | |
| Weekend | 0.89 | (0.87, 0.90) | < 0.001 |
| Vaccinated against COVID-19a | |||
| Yes | 1.06 | (1.02, 1.09) | < 0.001 |
| No | 1 | Reference | |
| Category of participants | |||
| General population | 1 | Reference | |
| Selected groups | 1.23 | (1.21, 1.26) | < 0.001 |
| Study periodb | |||
| Pre-fifth wave | 1 | Reference | |
| Fifth wave | 0.88 | (0.84, 0.91) | < 0.001 |
| Post-fifth wave | 0.85 | (0.81, 0.89) | < 0.001 |
| Post-pandemic | 1.31 | (1.26, 1.37) | < 0.001 |
| Vaccination coverage in the populationc | 0.9995 | (0.9986, 1.0004) | 0.26 |
Note: a“Vaccinated against COVID-19” refers to participants’ self-reported status of receiving at least one dose of COVID-19 vaccine by the time of completing the survey
bPre-fifth wave, fifth wave, post-fifth wave, and post-pandemic refer to contacts documented between 17-Sep-2021 and 6-Jan-2022, between 7-Jan-2022 and 20-Apr-2022, between 21-Apr-2022 and 28-Feb-2023, and from 1-Mar-2023 onwards, respectively
cVaccination coverage refer to the proportions of individuals in the general population who have received at least one dose of COVID-19 vaccines by the time the survey was completed. Vaccination coverage specific to the age group and gender of the respondents was considered [12]
Compared to our pre-pandemic survey in 2015–2016, which recorded an average of 8.1 (8.0, 8.2) daily contacts in the general population, respondents in this study generally reported fewer daily contacts in the pre-fifth wave period (6.2 (5.9, 6.5)) and during the fifth wave (5.1 (5.0, 5.3)) (Table 1) [21]. In the post-pandemic period, when all PHSMs were lifted and whole-day face-to-face schooling resumed, the average number of daily contacts rebounded to 8.4 (8.3, 8.5) which was similar to pre-pandemic levels.
Contact matrices
The contact matrices showed strong age-assortative mixing along the main diagonal for contacts made at home and at school (the second and third rows of Fig. 2). Off-diagonal bands in the contact matrices reflect intergenerational interactions typically in multi-generational households (home) and the student-teacher interactions (school). In contrast, contacts made at workplaces and other places (the fourth and the last rows of Fig. 2), as well as non-physical contacts (third row of Additional file 1: Figure S7), exhibited broader, cross-age mixing.
Fig. 2.
Social contact matrices by the location of contact (rows) and time of responses (columns) among participants from the general population. Note: Social contact matrices present the mean number of daily reported contacts made by participants, with results shown for all contacts reported and at the specific locations by row. The x-axis of the matrices represents the age of participants, and the y-axis represents the age of the contacts. A darker colour represents a higher frequency of contact. The social contact matrices were adjusted for the age distribution in the study population and the reciprocity of contacts. The contact matrices for the 2015–2016 survey were based on data from Leung et al. 2017 [21]. Regarding contact locations, the category “Other places” includes public transport, places of worship, sports facilities (e.g., gyms and pools), outdoor environments (e.g., streets and country parks), entertainment and retail venues (e.g., cinemas, supermarkets, and malls), and food and beverage venues (e.g., restaurants and bars)
The contact matrices mirrored the potential effect of PHSMs across the study period. The marked reduction in school contacts during the fifth wave compared to the pre-fifth wave period (Fig. 2k–l) reflects complete school closures. From April 2022 onwards, schools reopened progressively, initially with half-day classes, and later with whole-day schooling permitted contingent on vaccine uptake among students and teachers [11]. Across different phases of school resumption, contacts among school-aged children gradually rebounded (Fig. 2m–n). Particularly, after March 2023, when all schools resumed whole-day face-to-face classes without additional requirements (Fig. 2n), school contacts increased sharply, underscoring the central role of educational institutions in shaping age-specific mixing. For example, the average number of daily intra-age-group contacts made at school among 6–10-year-olds dropped from 2.1 before the fifth wave to 0.2 during the fifth wave, and rebounded to 2.2 when whole-day schooling resumed in the post-pandemic period.
Figure 1h presents the relative changes in the dominant eigenvalue of the contact matrix relative to the pre-fifth wave baseline over time. This metric can be considered to reflect relative changes in the 
of the Omicron BA.2 variant attributable solely to changes in contact patterns associated with PHSMs (i.e., excluding the effect of susceptible depletion and the emergence of new variants). Under this interpretation, accounting for all contacts (the blue line in the plot), 
declined throughout the fifth wave and the subsequent months, reaching a nadir of 20% (median; 95% percentile interval: 1% to 40%) below the pre-fifth wave level. Thereafter, 
increased steadily. By December 2023, nearly a year after all PHSMs had been lifted, 
was 2.1 (1.6, 2.5) times the pre-fifth wave level. In the sensitivity analysis that reconstructed the contact matrices with contacts lasting at least 15 min only, the estimated 
(Additional file 1: Figure S9) showed a similar pattern to the original analysis based on all contacts regardless of contact duration (Fig. 1h). Specifically, the estimated 
dropped during the fifth wave, remained low in the subsequent months, and reached a nadir of 26% (11%, 42%) below the pre-fifth wave level. The estimated 
then increased gradually when PHSMs were lifted after March 2023, and was approximately 2.1 (1.6, 2.6) times the pre-fifth wave level by December 2023.
Discussion
The COVID-19 pandemic substantially altered how people interact. Implementation of PHSMs, availability of pharmaceutical interventions such as vaccines and antiviral medications, and behavioural responses to fear of contagion were among the factors that reshaped contact patterns in populations [29–31]. Assessing changes in social contact patterns across phases of PHSM implementation is critical for understanding the effectiveness of these interventions in reducing disease transmission [3]. We examined this issue in Hong Kong, which implemented prolonged PHSMs, and observed temporal changes in contact patterns that differed from those reported in other countries and regions.
Throughout the entire acute phase of the COVID-19 pandemic of over three years in Hong Kong, the stringency of PHSMs varied during different waves, leading to significant changes in transmission opportunities, as reflected by the estimated effective reproduction numbers [14]. To examine how the contact patterns changed across the pandemic and with PHSMs implementation, we surveyed over 10,000 participants in more than 28 months, which covered the fifth wave with the highest incidence in the city during the period from January to April 2022 [11]. Among the general population, compared to our pre-pandemic survey in 2015–2016 (8.1), we observed a reduction of 37% in the number of daily contacts during the fifth wave (5.1 on average) [21]. In the post-pandemic period, the mean number of daily contacts rebounded to 8.4, a level similar to the pre-pandemic estimate. The relative changes in the number of daily contacts in our study were less pronounced than those reported in European regions and the Chinese Mainland [21]. For example, the CoMix Europe Working Group estimated that the average number of daily contacts during the pandemic between March 2020 and March 2022 dropped by 66–74% in Belgium, Germany, the Netherlands, and the United Kingdom, when compared to the POLYMOD study in 2005–2006 [6, 29, 30]. In China, the number of daily contacts decreased by 70–88% in Wuhan, Shanghai, Shenzhen, and Changsha during the first-wave outbreak period in January–April 2020, compared to their respective baseline estimates [10, 32].
Moreover, the number of contacts in many countries/regions after the pandemic remained considerably lower than the pre-pandemic levels. Compared to the POLYMOD study, the number of contacts in the post-pandemic period (November–December 2022) was lowered by 30–45% in the United Kingdom, Belgium, and the Netherlands [33]. Another study in the Netherlands reported a drop of 13% in the number of reported daily contacts after the pandemic (May 2023) compared to the pre-pandemic estimate in 2016–2017 [34].
Many European countries had lockdowns that mandated citizens to stay at home and greatly reduced contact between people [35]. The work-from-home practice may also introduce a new norm in contact patterns, particularly after the pandemic [36]. In China, stringent social distancing policies such as containment measures and movement restrictions were implemented during the first wave of the outbreak in 2020 [37]. However, there was no lockdown in Hong Kong although the government and many companies recommended staff work from home. The government also adjusted the stringency of PHSMs according to the severity of infections in the community. This may partly explain the less pronounced drop during the pandemic in our survey when compared to the CoMix European project and the Chinese case.
Furthermore, while the stringency level of PHSMs impacted social contacts, contact patterns also evolved throughout the epidemic due to changes in health risk perception and PHSM fatigue. A local study reported that people may perceive a lower severity of the disease and show a lower intention to avoid going out as time passed during the pandemic [14]. Consequently, timely data on social contacts and adherence to PHSMs are essential to assess the effectiveness of PHSMs and enhance public health messaging. Such information could offer actionable insights for tailoring public health strategies in future pandemics. Notably, over 97% of our responses were from self-completed online surveys, demonstrating its feasibility for the prompt assessment of PHSM effectiveness. With clear instructions, self-reported online contact surveys can effectively track changes in contact patterns. This can help policymakers evaluate and adjust PHSMs as needed.
Based on the changes in the dominant eigenvalues of contact matrices, we estimated that 
in late 2023 (about one year after all PHSMs were lifted) was approximately 2.1 times the pre-fifth wave level in late 2021. The increase was smaller when 
was based on physical contacts only than when non-physical contacts were considered. This may reflect the resumption of routine social activities while some caution regarding close physical contact remained (such as maintaining social distancing) [38]. This may also indicate that physical contacts, often concentrated within households or close networks, changed less than non-physical contacts in public settings [39]. These findings, however, should be interpreted cautiously, because this approach of 
estimation assumes a fully susceptible population and constant transmission parameters, assumptions that did not hold throughout the pandemic [40, 41]. For example, the more pronounced increase in 
based on non-physical contacts likely indicated greater recovery of broader social mixing, rather than a proportional increase in actual transmission risk through non-physical contacts.
Our study has several limitations. First, respondents were asked to record their contact patterns for the day before they were contacted for the survey. We sent out automated text reminders to the mobile phones of participants on the day before the proposed contact recording day [42, 43]. While this approach may have helped participants remember their contacts, recall bias regarding contact details (e.g., the age of the contacted persons and respective contact duration) may remain, particularly for participants who had a lot of contacts (e.g., participants from the high-contact groups). Second, social desirability bias may have led to underreported contacts even though the research data did not include personal identifiers. Third, our study started recruitment one and a half years after the first emergence of the pandemic. The changes in contact patterns and the correlation with PHSMs in the early phase of the pandemic were not captured in our study. Fourth, our previous pre-pandemic study conducted in 2015–2016 utilized both online and paper-based questionnaires, among which the average number of contacts based on the online questionnaire (5.1) was notably lower than that reported using the paper-based format (9.9) [21]. In the current study, the average number of contacts when all PHSMs were lifted (8.4) was similar to the estimate from the previous study (8.1, after the adjustment to the mode of questionnaire). Participants in 2015–2016 may have been less familiar with completing online questionnaires compared to participants in the current study (conducted in 2021–2023). The reported number of contacts from the online survey in the previous study may have been underestimated, and thus comparisons between the current study and the pre-pandemic survey should be interpreted with caution. Nevertheless, comparisons within the current study, such as those before and after the fifth wave, were not affected by this issue. Fifth, we estimated the changes in 
by referring to the changes in the dominant eigenvalues calculated from social contact matrices. The proportional relationship between 
and the calculated dominant eigenvalues is based on the assumptions that the population is fully susceptible to the circulating variants and the transmission parameters are constant across the time horizon. However, during the study period, changes in vaccination coverage in the population, levels of natural infection, and the circulating variants invalidated these assumptions. As such, the corresponding findings should be interpreted carefully. For example, the estimated relative changes in 
based on the changes in contact patterns may overestimate the actual impact on observable transmission if the probability of disease transmission per contact had increased substantially over time or if the circulating variants were more transmissible than the variant circulating when the baseline dominant eigenvalue was estimated [44, 45].
Conclusions
During the COVID-19 pandemic in Hong Kong, our data and analyses suggest that PHSMs (including school closures, work-from-home arrangements, and restrictions on group gatherings and inbound travel) were temporally associated with reduced social contacts and hence may potentially affect disease transmission. These quantitative estimates may indicate the potential value of PHSMs for informing future pandemic preparedness and response.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary Material 1: Additional file 1: Supplementary Information. Tables S1-S3, Figures S1-S9.
Acknowledgements
This work was supported by a grant from the Health and Medical Research Fund (COVID19F05) of the Government of the Hong Kong Special Administrative Region, China. The study was also supported by the Hong Kong Jockey Club Global Health Institute (HKJCGHI), Hong Kong Special Administrative Region, China, and AIR@InnoHK administered by the Innovation and Technology Commission, Hong Kong Special Administrative Region, China.
Abbreviations
- CI
Confidence interval
- PHSM
Public health and social measure
Author contributions
EL, GML, MJ, KL, and JTW designed and conceptualized the study, and acquired research funding. HC and SK contributed to project management and administration. HC, KP, SK, and KL performed data curation and validation, and conducted formal data analysis. HC, KP, MJ, KL, and JTW drafted and reviewed the manuscript. All authors read and approved the final manuscript.
Authors’ social media handles
Kathy Leung – X: @kathysleung.
Joseph T Wu – X: @jtkwu.
Funding
Health and Medical Research Fund, and AIR@InnoHK.
Data availability
The datasets generated and analyzed during the current study are available in the GitHub repository, https://github.com/hcchoihk/Social_contact_survey_HK [46].
Declarations
Ethics approval and consent to participate
The study was approved by the Institutional Review Boards of HKU/Hospital Authority Hong Kong West Cluster (HKU/HA HKW IRB) (reference number: UW21-195) and the observational study ethics committee of the London School of Hygiene & Tropical Medicine (reference number: 29947).
Consent for publication
Not applicable.
Competing interests
Prof. Mark Jit is a member of the BMC Medicine Editorial Board. The authors declare no other potential competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Horace Choi and Kiesha Prem contributed equally to this work.
Contributor Information
Kathy Leung, Email: ksmleung@ust.hk.
Joseph T Wu, Email: joewu@hku.hk.
References
- 1.Mousa A, Winskill P, Watson OJ, Ratmann O, Monod M, Ajelli M, et al. Social contact patterns and implications for infectious disease transmission - a systematic review and meta-analysis of contact surveys. eLife. 2021;10:e70294. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Mistry D, Litvinova M, Pastore YPA, Chinazzi M, Fumanelli L, Gomes MFC, et al. Inferring high-resolution human mixing patterns for disease modeling. Nat Commun. 2021;12(1):323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Liu Y, Diamond C, Abbott S, Wong K, Schmidt T, Edmunds WJ, et al. The impact of public health and social measures (PHSMs) on SARS-CoV-2 transmission in the WHO European Region (2020–2022). Influenza Other Respir Viruses. 2024;18(12):e70036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Leung K, Wu JT, Leung GM. Effects of adjusting public health, travel, and social measures during the roll-out of COVID-19 vaccination: a modelling study. Lancet Public Health. 2021;6(9):e674–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jarvis CI, Van Zandvoort K, Gimma A, Prem K, Auzenbergs M, O’Reilly K, et al. Quantifying the impact of physical distance measures on the transmission of COVID-19 in the UK. BMC Med. 2020;18(1):124. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Verelst F, Hermans L, Vercruysse S, Gimma A, Coletti P, Backer JA, et al. SOCRATES-CoMix: a platform for timely and open-source contact mixing data during and in between COVID-19 surges and interventions in over 20 European countries. BMC Med. 2021;19(1):254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Feehan DM, Mahmud AS. Quantifying population contact patterns in the United States during the COVID-19 pandemic. Nat Commun. 2021;12(1):893. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Drolet M, Godbout A, Mondor M, Beraud G, Drolet-Roy L, Lemieux-Mellouki P, et al. Time trends in social contacts before and during the COVID-19 pandemic: the CONNECT study. BMC Public Health. 2022;22(1):1032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Zhang J, Litvinova M, Liang Y, Wang Y, Wang W, Zhao S, et al. Changes in contact patterns shape the dynamics of the COVID-19 outbreak in China. Science. 2020;368(6498):1481–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.The Government of the Hong Kong Special Administrative Region. COVID-19: Press release. 2023. https://www.coronavirus.gov.hk/eng/press_release.html. Accessed 24 June 2024.
- 12.DATA.GOV.HK. Data in Coronavirus Disease (COVID-19). https://data.gov.hk/en-data/dataset/hk-dh-chpsebcddr-novel-infectious-agent. Accessed.
- 13.COVID-19 National Preparedness Collaborators. Pandemic preparedness and COVID-19: an exploratory analysis of infection and fatality rates, and contextual factors associated with preparedness in 177 countries, from Jan 1, 2020, to Sept 30, 2021. Lancet. 2022;399(10334):1489–512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Yang B, Lin Y, Xiong W, Liu C, Gao H, Ho F, et al. Comparison of control and transmission of COVID-19 across epidemic waves in Hong Kong: an observational study. Lancet Reg Health West Pac. 2024;43:100969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lau JJ, Cheng SMS, Leung K, Lee CK, Hachim A, Tsang LCH, et al. Real-world COVID-19 vaccine effectiveness against the Omicron BA.2 variant in a SARS-CoV-2 infection-naive population. Nat Med. 2023;29(2):348–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cowling BJ, Ali ST, Ng TWY, Tsang TK, Li JCM, Fong MW, et al. Impact assessment of non-pharmaceutical interventions against coronavirus disease 2019 and influenza in Hong Kong: an observational study. Lancet Public Health. 2020;5(5):e279–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.World Health Organization. Public health and social measures in response to COVID-19. 2024. https://phsm.euro.who.int/covid-19. Accessed 24 June 2024.
- 18.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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Latsuzbaia A, Herold M, Bertemes J-P, Mossong J. Evolving social contact patterns during the COVID-19 crisis in Luxembourg. PLoS ONE. 2020;15(8):e0237128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Backer JA, Mollema L, Vos ER, Klinkenberg D, van der Klis FR, de Melker HE, et al. Impact of physical distancing measures against COVID-19 on contacts and mixing patterns: repeated cross-sectional surveys, the Netherlands, 2016–17, April 2020 and June 2020. Eurosurveillance. 2021;26(8):2000994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Leung K, Jit M, Lau EHY, Wu JT. Social contact patterns relevant to the spread of respiratory infectious diseases in Hong Kong. Sci Rep. 2017;7(1):7974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.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. [DOI] [PubMed] [Google Scholar]
- 23.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. BMC Public Health. 2024;24(1):1438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Vynnycky E, White R. An introduction to infectious disease modelling. Oxford University Press; 2010.
- 25.Pringle JC, Leikauskas J, Ransom-Kelley S, Webster B, Santos S, Fox H, et al. COVID-19 in a correctional facility employee following multiple brief exposures to persons with COVID-19 - Vermont, July-August 2020. MMWR Morb Mortal Wkly Rep. 2020;69(43):1569–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.World Health Organization. Public health surveillance for COVID-19: interim guidance. 2022. https://apps.who.int/iris/bitstream/handle/10665/360580/WHO-2019-nCoV-SurveillanceGuidance-2022.2-eng.pdf. Accessed 10 April 2026.
- 27.Oliva C, Favato G. From 15 minutes to 15 seconds: How the Delta variant changed the risk of exposure to COVID-19. A comparative epidemiological investigation using community mobility data from the metropolitan area of Genoa, Italy. Front Public Health. 2022;10:872698. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Allen H, Tessier E, Turner C, Anderson C, Blomquist P, Simons D, et al. Comparative transmission of SARS-CoV-2 Omicron (B.1.1.529) and Delta (B.1.617.2) variants and the impact of vaccination: national cohort study, England. Epidemiol Infect. 2023;151:e58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wong KLM, Gimma A, Coletti P, CoMix Europe Working G, 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Henwood A, Rinck M, Krpan D. Pandemic related changes in social interaction are associated with changes in automatic approach-avoidance behaviour. Sci Rep. 2023;13(1):4637. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Backer JA, Vos ERA, den Hartog G, van Hagen CCE, de Melker HE, van der Klis FRM, et al. Contact behaviour before, during and after the COVID-19 pandemic in the Netherlands: evidence from contact surveys, 2016 to 2017 and 2020 to 2023. Eurosurveillance. 2024;29(43):2400143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Al-Salem W, Moraga P, Ghazi H, Madad S, Hotez PJ. The emergence and transmission of COVID-19 in European countries, 2019–2020: a comprehensive review of timelines, cases and containment. Int Health. 2021;13(5):383–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Adjodah D, Dinakar K, Chinazzi M, Fraiberger SP, Pentland A, Bates S, et al. Association between COVID-19 outcomes and mask mandates, adherence, and attitudes. PLoS ONE. 2021;16(6):e0252315. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Li Z, Chen Q, Feng L, Rodewald L, Xia Y, Yu H, et al. Active case finding with case management: the key to tackling the COVID-19 pandemic. Lancet. 2020;396(10243):63–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Crawford FW, Jones SA, Cartter M, Dean SG, Warren JL, Li ZR, et al. Impact of close interpersonal contact on COVID-19 incidence: Evidence from 1 year of mobile device data. Sci Adv. 2022;8(1):eabi5499. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Backer JA, Bogaardt L, Beutels P, Coletti P, Edmunds WJ, Gimma A, et al. Dynamics of non-household contacts during the COVID-19 pandemic in 2020 and 2021 in the Netherlands. Sci Rep. 2023;13(1):5166. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Diekmann O, Heesterbeek JA, Roberts MG. The construction of next-generation matrices for compartmental epidemic models. J R Soc Interface. 2010;7(47):873–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.van den Driessche P, Watmough J. Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission. Math Biosci. 2002;180:29–48. [DOI] [PubMed] [Google Scholar]
- 42.Hoang T, Coletti P, Melegaro A, Wallinga J, Grijalva CG, Edmunds JW, et al. A systematic review of social contact surveys to inform transmission models of close-contact infections. Epidemiology. 2019;30(5):723–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Mikolajczyk RT, Kretzschmar M. Collecting social contact data in the context of disease transmission: Prospective and retrospective study designs. Social Networks. 2008;30(2):127–35. [Google Scholar]
- 44.Reichmuth ML, Heron L, Beutels P, Hens N, Low N, Althaus CL. Social contacts in Switzerland during the COVID-19 pandemic: Insights from the CoMix study. Epidemics. 2024;47:100771. [DOI] [PubMed] [Google Scholar]
- 45.Munday JD, Jarvis CI, Gimma A, Wong KLM, van Zandvoort K, Group CC-W, et al. Estimating the impact of reopening schools on the reproduction number of SARS-CoV-2 in England, using weekly contact survey data. BMC Med. 2021;19(1):233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.GitHub. Social_contact_survey_HK. https://github.com/hcchoihk/Social_contact_survey_HK. Accessed 25 November 2025.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Material 1: Additional file 1: Supplementary Information. Tables S1-S3, Figures S1-S9.
Data Availability Statement
The datasets generated and analyzed during the current study are available in the GitHub repository, https://github.com/hcchoihk/Social_contact_survey_HK [46].