key: cord-0724085-wm3btd0a
authors: Wambua, James; Hermans, Lisa; Coletti, Pietro; Verelst, Frederik; Willem, Lander; Jarvis, Christopher I.; Gimma, Amy; Wong, Kerry L. M.; Lajot, Adrien; Demarest, Stefaan; Edmunds, W. John; Faes, Christel; Beutels, Philippe; Hens, Niel
title: The influence of risk perceptions on close contact frequency during the SARS-CoV-2 pandemic
date: 2022-03-25
journal: Sci Rep
DOI: 10.1038/s41598-022-09037-8
sha: 92890b7da17565fa67081460e44ed0b752b43cdf
doc_id: 724085
cord_uid: wm3btd0a

Human behaviour is known to be crucial in the propagation of infectious diseases through respiratory or close-contact routes like the current SARS-CoV-2 virus. Intervention measures implemented to curb the spread of the virus mainly aim at limiting the number of close contacts, until vaccine roll-out is complete. Our main objective was to assess the relationships between SARS-CoV-2 perceptions and social contact behaviour in Belgium. Understanding these relationships is crucial to maximize interventions’ effectiveness, e.g. by tailoring public health communication campaigns. In this study, we surveyed a representative sample of adults in Belgium in two longitudinal surveys (survey 1 in April 2020 to August 2020, and survey 2 in November 2020 to April 2021). Generalized linear mixed effects models were used to analyse the two surveys. Participants with low and neutral perceptions on perceived severity made a significantly higher number of social contacts as compared to participants with high levels of perceived severity after controlling for other variables. Our results highlight the key role of perceived severity on social contact behaviour during a pandemic. Nevertheless, additional research is required to investigate the impact of public health communication on severity of COVID-19 in terms of changes in social contact behaviour.

Sample composition and GLMM summary results. Summary of the sample composition in the first and second analysis are contained in Supplementary Tables S1 and S2, respectively. The average number of contacts in the first and second analysis are contained in Table 2 in Coletti et al. 30 and Supplementary Table S3 , respectively. The average number of waves in which participants participated in the study in the first and second analysis was 4.97 and 4.87, respectively. Results from the GLMM indicated that the overdispersion parameter ranged between 2.47 (95% CI 2.29-2.67) and 2.54 (95% CI 2. 35-2.74 ) in the first analysis, whilst in the second analysis, this parameter ranged between 1.82 (95% CI 1.71-1.92) and 1.85 (95% CI 1.74-1.96) indicating substantial heterogeneity. The variance of the random effect ranged between 0.438 and 0.447 in the first analysis from the different individual perception models while in the second analysis, it ranged between 0.663 and 0.700. In both analyses, the random effect was statistically significant (p-value < 0.001 ), further indicating underlying heterogeneity in the social contact behaviour among individuals.

Perceived severity. Results from the first analysis in the GLMM model indicated that in the perceived severity model, participants with low and neutral level of perceived severity made 1.70 (95% CI 1.18-2.45) and 1.56 (95% CI 1.14-2.14) times more contacts than the participants with high levels of perceived severity, respectively (Supplementary Table S4 ). The predicted number of contacts for the high level of perceived severity was 1.41 (95% CI 1.10-1.79), while for the low and neutral levels was 2.38 (95% CI 1.69-3.36) and 2.20 (95% CI 1.61-3.00), respectively. In the second analysis, participants with low and neutral levels of perceived severity made 1.62 (95% CI 1.14-2.31) and 1.76 (95% CI 1.31-2.38) times more contacts than the participants with high levels of perceived severity, respectively (Supplementary Table S5 ). The predicted number of contacts for the high level was 1.42 (95% CI 1.05-1.91) whilst for the low and neutral levels was 2.31 (95% CI 1.67-3.18) and 2.51 (95% CI 1.89-3.33), respectively.

The interaction term between perceived severity and survey wave of the data collection was significantly associated with the number of contacts in both analyses from the Type III Wald tests (p-values = 0.002 and < 0.001, respectively). Participants with high levels of perceived severity had fewer social contacts as compared to those with low or neutral perceived severity (Fig. 3) . Marginal effects of the interaction terms between perceived severity and age group depicted similar patterns in both analyses (Fig. 4) . The interaction term between perceived severity and participants' age group was significantly associated with the number of contacts in the second analysis only (11 waves) (p-value = 0.039). For more information on the significance of the variables in both analyses for perceived severity, see Supplementary Table S6 . The interaction effects between the perceived severity and age in the first analysis were more pronounced in the youngest age group [18, 30) where large differences in the predicted number of contacts were observed between participants with high level of perceived severity and those with low or neutral perceptions. Small differences in predicted number of contacts were observed in age groups [30, 40) and 70+ among the different levels of perceived severity. In the second analysis, the interaction effects similarly showed more pronounced differences in the predicted number of contacts among the different perceived severity levels in the age group [18, 30) www.nature.com/scientificreports/ Perceived susceptibility. The GLMM results from the perceived susceptibility model indicated that the interaction terms between perceived susceptibility and age group and between perceived susceptibility and wave of data collection were significantly associated with the number of contacts in both analyses (p-values 0.025; 0.038 and <0.001; <0.001, respectively). Furthermore, the GLMM results from the first analysis indicated significant interaction effects between perceived susceptibility and face mask wearing and between perceived susceptibility and household size (p-values = 0.05 and 0.014, respectively) see (Supplementary Table S7 ). In the first analysis, the number of contacts was found to be higher for participants with low levels of perceived susceptibility in younger age groups ( [18, 30) and [30, 40) ), and generally higher for those with high levels of perceived susceptibility in the other age groups. Whilst in the second analysis, participants with high levels of perceived susceptibility generally made a higher number of contacts than those with low and neutral perceptions (Supplementary Fig. S1 ). Similarly, plots of the predicted number of contacts from the marginal effects of the interaction terms between perceived susceptibility and wave of data collection in both the first and second analyses were also slightly different. In the first analysis, participants with low levels of perceived susceptibility had a higher number of contacts. While this was not the case in the second analysis where participants with neutral perceptions of susceptibility had a higher number of contacts ( Supplementary Fig. S2 ). Plots of the predicted number of contacts from the marginal effects of the interaction terms between perceived susceptibility and face mask wearing indicated that participants who had low levels of perceived susceptibility made more number of contacts than those who had neutral or high levels. Furthermore, we observed that participants who indicated to have used a face mask made more number of contacts ( Supplementary Fig. S3 ). The interaction effects between perceived susceptibility and household size indicated that an increase in household size coincided with an increasing trend in the predicted number of contacts for all the levels of perceived susceptibility ( Supplementary Fig. S3 ).

Perceived benefit to vulnerable. In both analyses, the interaction term between perceived benefit to vulnerable and wave of data collection was significant (all p-values <0.05), see Supplementary Table S8 . However, we did not observe any distinct pattern in terms of the number of contacts among the different perception levels ( Supplementary Fig. S4 ).

Perceived effectiveness of measures. The perceived effectiveness of measures was only considered in the first analysis. We found a significant interaction effect between the perceived effectiveness of measures and wave of data collection (p-value = 0.013) in the perceived effectiveness model in the first analysis. Plots of the predicted number of contacts from the marginal effect of the interaction term between perceived effectiveness and wave of data collection showed that participants with high levels of perceived effectiveness generally made fewer contacts than those with low levels of perceived effectiveness ( Supplementary Fig. S5 ). However, the observed differences were small.

Perceived adherence to measures. Similarly, the perceived adherence to measures was considered in the first analysis only. We observed a significant interaction effect between perceived adherence to measures and participant's age group (p-value = 0.015). The plots of the predicted number of contacts from the marginal effects of the interaction term between perceived adherence to measures and participant's age group showed that in general, participants with high levels of perceived adherence made fewer contacts than those with low levels of perceived adherence to measures ( Supplementary Fig. S6 ). Similarly, the observed differences in the number of contacts were small. Descriptive plots for the aforementioned COVID-19 related perceptions and the average number of contacts in both the first and second survey rounds showed changes over time and slight variations over age groups (Supplementary Figs. S7-S11). Furthermore, the dynamics of the perceived severity and average number of reported hospitalizations tracked each other relatively well, implying the perception on severity changed with actual risk proxied by the number of new hospital admissions ( Supplementary Fig. S12 ). In all the models considered in both the first and second analyses, we observed that household size, wave of data collection, face mask wearing and participants occupation were significant predictors of the number of social contacts in addition to the different perception variables. For more information, see (Supplementary Tables S6-S9 ). The GLMM results indicated that participants who reported to have used a face mask made significantly more number of contacts than those who had not used. Further, the results indicated that a one unit increase in household size coincided with 1.25 (95% CI 1.21-1.13) and 1.45 (95% CI 1.38-1.55) increase in the number of contacts from the first and second analyses, respectively (Supplementary Tables S4-S5 ).

In the face of the COVID-19 pandemic that has led to unprecedented negative health outcomes and social economic burden 2,3 , it is important to understand factors that influence individual behaviour. Our study explored the relationship between 5 specific perception variables related to COVID-19 and behavioral response in terms of the reported number of social contacts. We used a generalized linear mixed effects model in order to take into account both the within-participant and between participant variability from the two longitudinal datasets.

The results indicated that individuals who perceived themselves to experience severe illness if they contract a COVID-19 infection tended to make significantly fewer contacts as compared to those who had low or neutral perceptions. The observed relationship between the perceived severity and social contact behaviour was consistent in both analyses (i.e, analyses involving survey data from the first 8 waves of data collection, and also from the subsequent 11 waves). It is important to note that these two longitudinal surveys queried respondents' behavior in two different COVID-19 pandemic waves in Belgium, with the first survey coinciding with the first COVID-19 wave, and the second survey with the second wave. Hence the similarity between the observed patterns of www.nature.com/scientificreports/ associations is suggestive of the crucial role perceived severity has on social contact behaviour. Our findings were echoed greatly by results from a study utilizing CoMix data from the United Kingdom (UK) 32 . This study found that individuals aged between 18 and 59 years who perceived high levels of seriousness if infected by the SARS-CoV-2 virus had lower mean number of contacts than those who perceived low levels of seriousness. Several studies examined the role of risk perceptions on adoption of recommended preventive measures during the COVID-19 pandemic 7, 12, [15] [16] [17] [18] [19] [20] [21] [22] . These studies have found that perceived severity was associated with the adoption of the protective behaviours, in line with the Health Belief Model. More specifically, people with higher perceived severity of the disease were found to be more likely to adopt the recommended precautionary measures. However, it is important to mention that the response variable of interest differed between studies. Whilst the response variable in our study was the number of social contacts, other studies considered indicators of avoidance of behaviour or adoption of the recommended measures as their outcome. Nonetheless, the results all point towards the critical role of perceived severity on individual's response behaviour. Furthermore, the differences in the number of contacts for individuals with high perceived severity versus individuals with low or neutral levels of severity was around one contact in our study. The evaluation of the implications of such differences on the transmission dynamics of SARS-CoV-2 are a topic for future research. In addition, with respect to other response variables, the number of contacts can more easily and more consistently be included in mathematical models of infectious diseases 33 , making the analysis presented in this work crucial for future modelling endeavours of COVID-19.

In our study, the relationship between perceived susceptibility and the number of social contacts did not yield consistent relationships. These ambiguities may have resulted from a variety of factors including, but not limited to: firstly, there could be the presence of optimism bias, a phenomenon where individuals tend to underestimate their likelihood of experiencing a negative event or overestimate the likelihood of positive events 34 . In the context of the COVID-19 pandemic, this refers to individuals underestimating their perceived risk of getting infected. Several studies have indicated the presence of optimism bias during the COVID-19 pandemic 12, 14, 35 . Secondly, individuals having a higher number of social contacts might perceive themselves more likely to get infected as a result of their behaviour and vice-versa. Results from the aforementioned study in UK 32 found that in general, participants who indicated to be likely to get infected by the SARS-CoV-2 virus had higher mean number of contacts than those who indicated to be unlikely to get the virus. And thirdly, this could be due to individuals' perception on their inherent vulnerability to infection. Thus based on our results, the relationship between perceived susceptibility and social contact behaviour remains inconclusive and thus warrants more research.

Similarly, the relationship between perceived benefit to vulnerable and number of social contacts yielded inconsistent results. There were no significant differences in social contact behaviour between individuals who had high, neutral or low perceptions in terms of protecting the vulnerable individuals in the population. This could be due to either participants responding to the questionnaire item based on the frequency of contacts with vulnerable individuals within their close social circle or occupation (i.e, health care workers in elderly homes). In addition, it might be that participants who are vulnerable (mainly elderly people with underlying comorbidities) perceive no major benefit to other vulnerable individuals as they generally make fewer social contacts. As such, more research is required in this perspective as deliberate efforts in the realm of public health messaging and communication has emphasized on adhering to recommended measures to protect others 36 .

Perceived effectiveness of measures and perceived adherence to measures were both inversely associated with the number of contacts. Participants with high levels of perceived effectiveness of measures made lower number of contacts than those with low levels. Similarly, participants with high levels of perceived adherence to measures made fewer contacts than those low levels. However, the observed differences were generally small. According to the theory of Protection Motivation and Self-efficacy, persons' belief in effectiveness of an intervention measure, and their confidence to adhere to the measure predicts the likelihood of engaging in the preventive behaviour 24 . Previous studies conducted under this theoretical framework-that explore the relationships between perceived effectiveness of measures and perceived adherence to measures with the recommended health behaviour-do not explicitly use the number of social contacts as a proxy of the recommended health behaviour. Instead, they use indicators of avoidance of behaviours or adoption of recommended measures as above-mentioned. However, our results are consistent with results from previous studies [4] [5] [6] 15, 25, 26, 37 despite the outcome variables being slightly different. It is worth mentioning that the number of social contacts is a proxy of contact events responsible for disease transmission and is influenced by underlying determinants such as household size, day of the week (weekday versus weekend), age, among others as indicated in our study as well as in previous studies 27, 38 . Thus, more studies utilizing the number of social contacts as a proxy of the adoption of recommended measures will be pertinent to shed more light on the influence of perceptions on contact behaviour, while controlling for possible confounders. Furthermore, data on perceived effectiveness of measures and perceived adherence to measures was only collected in the first 8 waves (i.e, the first wave of COVID-19 pandemic), and thus continued data collection on these contextual factors could be of great importance to gain additional insights in the observed relationships. It is worth mentioning that both the perceptions and number of social contacts changed over time with slight differences observed by age groups. Furthermore, the wave of data collection which coincided with changing regimes of intervention measures and also changing landscape of the pandemic, was an important factor in the interaction effects of the perception variables, further highlighting that perceptions and social contact behaviour were dynamic in time. This is consistent with results from 2 studies that found evolution of both perceptions and protective behaviours during the influenza A(H1N1)v2009 pandemic 9,10 , and a recent study from UK conducted during the COVID-19 pandemic 22 .

Our findings highlight the importance of aligning the public's COVID-19 related perceptions with reality. That is, people who perceive COVID-19 to be more severe, will be more inclined to engage in preventive behaviours (here measured as the number of social contacts). Based on our results, we can suggest that public health communication and targeted messaging could yield more impact if tailored to messages emphasizing the severity www.nature.com/scientificreports/ of COVID-19. Thus, it is important to stress the severity of COVID-19-e.g in terms of excess mortality 39 or long-term effects post COVID-19 infection 40 . Furthermore, we found significant interaction effects between age and both perceived severity and perceived susceptibility, hence age-adjusted campaigns with respect to disease severity and susceptibility are required to enhance social distancing measures. A collaborative multidisciplinary approach by scientists, policymakers and communication experts is pivotal to formulate an effective and contextualized strategy that could optimise the impact of public health messaging 41 . Our study has several limitations. The associations between the perception variables and number of social contacts could have been affected by the level of stringency of the intervention measures that were being implemented. For example, during a lockdown, participants may not be able to contact people outside their household, even if they wanted to. However, this effect should be minimal as we controlled for the survey wave of data collection where different intervention measures were put in place. Information on COVID-19 vaccination was only partly available during the second survey. The percentage of the vaccinated individuals ranged from 0.5% in wave 12 to 14.8% in wave 19. Hence, due to the small sample of the vaccinated respondents, the vaccination status was not included in the analyses. However, a descriptive analysis (Supplementary Fig. S13 ) revealed no apparent differences in risk perceptions in the vaccinated individuals (before and after vaccination), and also in social contact behaviour between the vaccinated and not vaccinated (Supplementary Fig. S14 ). Although the panel of participants was representative by gender, age and region of residence in each survey wave, the voluntary opt-in of participants in each subsequent survey wave could be subject to self-selection bias where individuals more concerned about the pandemic in general would be more likely to participate. However, the participation rate was relatively high with 67.5% having participated in 3 or more waves in the first 8 survey waves and 63.19% in the subsequent 11 survey waves. Based on their importance in the context of social contact behaviour-we made sure the sampling design ensured representativeness in terms of age, gender and region of residence 27, 38 . However, other potential factors such as race, urban/rural dwelling, income and education were not considered. Future studies of social contact patterns could take the latter factors into account to obtain an even more representative sample and to assess the impact of these factors on social contact patterns. In the process of model building for the different perception variables which entailed numerous hypothesis testing, our models could have missed potential significant interaction effects other than the ones we mainly focused on in our exploratory modeling. The latter were selected due to their epidemiological relevance in the SARS-CoV-2 pandemic. This could have marginally affected the significance of the terms in the final models in our analyses. While potentially having an impact on respondents' risk perception, we did not collect information about their COVID-19 infection history. However, given the study was conducted in the early phases of the COVID-19 pandemic, where the percentages of the already infected in the population ranged between 0.04% and 0.62% in the first survey, and 4.05% and 7.72% in the second survey, the population-level effect is expected to be minimal. Our study findings could be subject to reverse causality since we assumed that perceptions precede the social contact behaviour, which might not necessarily be true. Our study could also suffer from social desirability bias, despite that anonymity of responses was assured. The results of our study apply to the Belgian population and caution is required when extrapolating these to other populations.

This study assessed the relationship between COVID-19 perceptions and social contact behaviour using two longitudinal surveys from a panel of individuals between April and August 2020, and November 2020 and April 2021 in Belgium. We found that individuals who perceived COVID-19 to be a serious illness for them made a significantly lower number of contacts as compared to those who had low or neutral perceptions. Similarly, individuals with high levels of perceived effectiveness and perceived adherence to measures made fewer contacts as compared to those with low levels. Given the importance of human behaviour in the transmission dynamics of SARS-CoV-2 virus, tailored communication strategies by public health officials about the severity of COVID-19 is crucial.

Ethics statement. Ethics approval for the study was given by the ethics committee of Antwerp University Hospital (reference number EC UZA 20/13/147). Participants aged 18 years and older opted to voluntarily participate in the study. In the second longitudinal survey, in addition, parents were invited to provide data on behalf of their child (a randomly selected child if a parent had several children). Informed consent was obtained from all subjects involved in the study. All methods were performed in accordance with the relevant guidelines and regulations. Survey methodology. The survey methodology and the sample characteristics for the first survey have been described by Coletti et al. 30 . Briefly, the first survey involved 8 waves of data collection in a representative panel of adults in terms of age, gender, and region of residence. More specifically, the panels of adults in the waves of data collection aimed for the same individuals filling out multiple surveys. Thus, a large proportion of responses were from the same participants. However, due to respondents' drop-out in subsequent waves of data collection, the panel was topped up with additional individuals to match the target quota. For more information on the representativeness of the sample in comparison with the Belgian population, see Supplementary  Table S1 that provides more details on the composition of the sample in each survey wave of data collection for the above-mentioned factors. These 8 waves coincided with different regimes of intervention measures that were implemented (Fig. 1) . Although much of the survey methodology is outlined in 30 , below we highlight some factors that are deemed relevant for the current study. These include participants' age, gender, information on socioeconomic status, high-risk status, number of social contacts, perception related to risk of infection, perception towards the effectiveness of the imposed intervention measures and confidence to adherence to the intervention measures. The high-risk status corresponds to one or more of the following health conditions; chronic respira- www.nature.com/scientificreports/ tory disease, chronic kidney disease, chronic heart disease, chronic neurological disease, chronic liver disease, immunosuppression, diabetes (all types), asplenia or dysfunction of the spleen, class III obesity, and pregnancy. The number of social contacts were measured between 5 am the day preceding the survey and 5 am of the survey day and represented all contacts that the participant made. The second survey involved 11 waves of data collection and corresponds largely to the second wave of the COVID-19 pandemic in Belgium (Fig. 2) . The survey also ensured representativeness in terms of age, gender and region of residence. More information is contained in the Supplementary Table S2 . The questionnaire used in the first survey was modified to collect data on behalf of children. The questionnaire items of the risk related perceptions remained the same in both surveys. However, questions related to perception on effectiveness of intervention measures and confidence to adhere to the measures were not collected in the second survey due to limitations on survey length. Similarly during data collection in subsequent waves, as the original panel loses some participants in specific waves, the cohort is replenished to fulfil sample size requirements. We utilize these two unique longitudinal surveys collected during the first and second wave of COVID-19 in Belgium to explore how perceptions relate with changes in social contacts behaviour. Thus our study provides a good basis to understand the link between perceptions and number of social contacts during the current pandemic.

To characterize the risk perception, participants were asked to indicate their level of agreement or disagreement on 3 items. See Table 1 for full item wordings. To characterize the perception on the effectiveness of intervention measures, participants had to indicate their perception on the level of effectiveness on each of 9 social and behavioral intervention measures. The responses were categorised into 5 categories; very effective, fairly effective, not very effective, not at all effective, and don't know (Table 1 ). In terms of the confidence to adhere to the imposed measures, participants were provided with 7 items corresponding to the intervention measures and asked to indicate their level of confidence, that they could adhere to these measures if they wanted to. The responses were grouped into five categories; very confident, fairly confident, not very confident, not at all confident, and don't know (Table 1) .

Data pre-processing. We coded the items corresponding to risk perceptions on 5-point Likert scales.

Items corresponding to the perception on effectiveness of intervention measures and confidence to adhere were converted into 4-point Likert scales (Table 1) . Since participants were also given an option don't know in the list of responses, in the process of coding these items into Likert-scales, this option was treated as a missing value. This was solely to enhance exploratory analysis for the internal consistency of the items and also for the computation of the mean changes of the perceptions over time. Hence, in the statistical analyses, it was included as a response level as explained in the next paragraph. www.nature.com/scientificreports/ We used Cronbach's alpha as the reliability measure of internal consistency of the Likert-scales 42 . We considered a threshold of 0.80 of the Cronbach's alpha reliability coefficient to group the items 42 . Items corresponding to the risk perceptions had Cronbach's alpha of 0.5 (95% confidence interval (CI) 0.48-0.52) and 0.52 (95% CI 0.50-0.53), in the first and the second survey, respectively. Thus, the three items could not be combined into one variable representing the underlying risk perception construct. Therefore, we considered the risk perception questions as three separate constructs (i.e, perceived severity for the item "coronavirus would be a serious illness for me", perceived susceptibility for the item "I am likely to catch coronavirus", and perceived benefit to the vulnerable for the item "If I don't follow the government's advice, I might spread the coronavirus to someone who is vulnerable", see Table 1 ). Motivated by exploratory analyses, we categorised the 6 response levels into three response levels: high perception ('strongly agree' and 'tend to agree'), low perception ('tend to disagree' and 'strongly disagree'), and neutral ('neither agree nor disagree' and 'don't know').

Cronbach's alpha for the 9 items of perceived effectiveness of measures was 0.86 (95% CI 0.85-0.87) and for the 7 items of perceived adherence to measures was 0.87 (95% CI 0.86-0.88), thus we obtained composite scores for the perceived effectiveness and perceived adherence to measures. More information on the survey items is contained in Table 1 . In the remainder of this paper, we use the terms perceived severity, perceived susceptibility and perceived benefit to vulnerable for the risk related perceptions. While perceived effectiveness of measures, and perceived adherence to measures will be used for the perceptions on effectiveness of measures and confidence to adhere to the measures.

We used a generalized linear mixed effects model (GLMM) to model the number of contacts in the two surveys 43 . We used random effects in the model to incorporate the correlations among observations from the same participant. A negative binomial distribution allowing for overdispersion was used to define the error distribution, while we applied zero-inflation to deal with excess zeroes in the number of social contacts. The model adjusted for the survey wave (time of data collection as categorical variable), day of the week (weekday versus weekend), the participant's household size, gender, age, face mask wearing and high risk status to control for possible confounding. The model was fitted using maximum likelihood estimation. We used R version 4.0.3 and the glmmTMB package (version 1.0.2.1) 44 for all statistical analyses.

Model building was performed for each individual perception variable. This was informed by preliminary exploratory analyses which identified significant interaction effects between the different perception variables and several control variables. Hence in total, we had 5 models for the first survey where each model represented an individual perception variable (i.e. perceived severity, perceived susceptibility, perceived benefit to vulnerable, www.nature.com/scientificreports/ III Wald tests and a significance level of 5% was considered. For convenience and ease of reporting results, the analyses involving data from the first survey (8 waves) will be termed as first analysis, whilst the one involving the second survey (11 waves) will be termed as second analysis in the rest of the work. 

The datasets utilized in this study will be made publicly available through the zenodo platform and any other additional information will be provided by the authors upon a reasonable request. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.

COVID-19 situation update worldwide, as of 7th of

Negative impacts of COVID-19 lockdown on mental health service access and follow-up adherence for immigrants and individuals in socio-economic difficulties

The extent of COVID-19 pandemic socio-economic impact on global poverty. a global integrative multidisciplinary review

An outbreak of the severe acute respiratory syndrome: Predictors of health behaviors and effect of community prevention measures in hong kong, china

SARS related preventive and risk behaviours practised by hong kong-mainland china cross border travellers during the outbreak of the sars epidemic in hong kong

Public perceptions, anxiety, and behaviour change in relation to the swine flu outbreak: Cross sectional telephone survey

COVID-19 is rapidly changing: Examining public perceptions and behaviors in response to this evolving pandemic

Demographic and attitudinal determinants of protective behaviours during a pandemic: A review

People at risk of influenza pandemics: The evolution of perception and behavior

The dynamics of risk perceptions and precautionary behavior in response to 2009 (H1N1) pandemic influenza

SARS risk perception, knowledge, precautions, and information sources, the Netherlands

Relationships between initial COVID-19 risk perceptions and protective health behaviors: A national survey

Government intervention, risk perception, and the adoption of protective action recommendations: Evidence from the COVID-19 prevention and control experience of China

Optimistic bias and preventive behavioral engagement in the context of COVID-19

Public perceptions, anxiety, and the perceived efficacy of health-protective behaviours to mitigate the spread of the SARS-Cov-2/COVID-19 pandemic

The impacts of knowledge, risk perception, emotion and information on citizens' protective behaviors during the outbreak of COVID-19: a cross-sectional study in china

Psychological responses, behavioral changes and public perceptions during the early phase of the COVID-19 outbreak in china: A population based cross-sectional survey

Comparing public perceptions and preventive behaviors during the early phase of the COVID-19 pandemic in hong kong and the united kingdom: Cross-sectional survey study

Risk perception regarding the covid-19 outbreak among the general population: A comparative middle east survey. Middle East Curr

The moderating effects of perceived severity on the generational gap in preventive behaviors during the COVID-19 pandemic in the US

Risk perceptions of COVID-19 around the world

Covid-19 risk perception: A longitudinal analysis of its predictors and associations with health protective behaviours in the united kingdom

Historical origins of the health belief model

Protection motivation and self-efficacy: A revised theory of fear appeals and attitude change

Psychosocial factors influencing the practice of preventive behaviors against the severe acute respiratory syndrome among older chinese in hong kong

The role of illness perceptions, coping, and self-efficacy on adherence to precautionary measures for COVID-19

Social contacts and mixing patterns relevant to the spread of infectious diseases

Changes in contact patterns shape the dynamics of the Covid-19 outbreak in China

Quantifying the impact of physical distance measures on the transmission of COVID-19 in the UK

CoMix: Comparing mixing patterns in the Belgian population during and after lockdown

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

Comix: Changes in social contacts as measured by the contact survey during the COVID-19 pandemic in england between

Design and analysis of social contact surveys relevant for the spread of infectious diseases

The optimism bias

Perceptions of the adult US population regarding the novel coronavirus outbreak

Using social and behavioural science to support COVID-19 pandemic response

Are common sense model constructs and self-efficacy simultaneously correlated with self-management behaviors and health outcomes: A systematic review

Close contact infection dynamics over time: insights from a second large-scale social contact survey in Flanders

Belgian COVID-19 mortality, excess deaths, number of deaths per million, and infection fatality rates

More than 50 long-term effects of COVID-19: A systematic review and meta-analysis

The COVID-19 pandemic: a wake-up call for better cooperation at the science-policysociety interface

Calculating, interpreting, and reporting cronbach's alpha reliability coefficient for likert-type scales. In Calculating, interpreting, and reporting Cronbach's alpha reliability coefficient for Likert-type scales

Models for discrete longitudinal data

J.W., P.C., L.W., P.B., N.H. Conceived the study. J.W., P.C., C.I.J, A.G, K.W contributed to data cleaning. J.W., P.C. performed data analysis. J.W., N.H., F.V. wrote the first draft of the manuscript. All authors contributed and reviewed the manuscript, and approved the final version for publication.

The authors declare no competing interests.

The online version contains supplementary material available at https:// doi. org/ 10. 1038/ s41598-022-09037-8.Correspondence and requests for materials should be addressed to J.W.Reprints and permissions information is available at www.nature.com/reprints.