Perception of the impact of COVID-19.

To assess the perception of the impact of COVID-19 on the participants, we employed a multivariate ordinal logistic regression. The participants were asked whether they agreed with the statement that COVID-19 had a negative impact on their well-being in the economic, work, psychological, and social conditions. In this case, the ordinal outcome Y of the model is the participants’ answer measured on the Likert scale, . Formally, considering the cumulative probability of the event , with , represents the level of disagreement with the statement, the ordinal linear regression model can be written as

(1)

where

(2)

is the vector of coefficients, and

(3)

is the vector of explanatory variables. In what follows, we will use the same explanatory variables as in Eq 3 unless specified otherwise.

Protective behavior.

Participants’ behavior toward protective measures was evaluated through questions about the number of vaccination intakes for COVID-19 and influenza, mask-wearing attitude, and adherence to social distancing. Social distancing and mask-wearing attitudes were assessed with an ordinal logistic regression model. The ordinal outcome Y was measured on a Likert scale, . Formally, considering the cumulative probability of the event , where , the ordinal linear regression model can be written as

(4)

Vaccination uptake was modeled with a logistic regression where the number of vaccination doses was mapped into a binary variable Y (yes or no) (since the focus was more on the attitude towards vaccination rather than the specific number of doses) and the model is expressed by equation

(5)

were and X are expressed by Eqs (2) and (3).

Model assumptions and robustness checks for protective behavior and perception model.

To verify the proportional odds assumption for the ordinal models (perception of impact of COVID-19 Eq (1) and protective behavior Eq (4)), we implemented the Brant test [29]. The results of the Brant test indicated no evidence against the proportional odds assumption, suggesting that the ordinal regression model was reasonable for our data. Details on how each type of outcome variable was incorporated into the models are described below. The analysis was performed using the MASS package for R [30]. To assess the robustness of these associations, Bayesian models (ordinal for NPIs, logistic for vaccination) were also implemented. These yielded results consistent with the frequentist approach, as detailed in the supporting information section (Sect 2.2 in S1 Text). We retained the frequentist approach in the main text due to this being a more common way to present the results.

Contact behavior.

For contact behaviors, we directly opted for a Bayesian approach. Modeling contact counts can involve substantial overdispersion and smaller effective sample sizes within subgroups, making Bayesian methods preferable for providing more stable estimates and richer uncertainty quantification. The dependent variable in this case was the total number of contacts Yi for each participant i and the model can be formally expressed by

(6)

and the link function

(7)

where and X are defined by Eqs 2 and 3, the mean number of contacts for each participant i, and α is the reciprocal dispersion parameter: the higher α is the closer Y_i is Poisson distributed. We considered weakly informative priors (Gaussian distributed) for the coefficients . We conducted the analysis separately for weekdays (Monday) and weekends (Sunday). This allowed us to investigate potential differences in contact patterns based on working and non-working participants. To discern the working behavior from the employment status, we integrated information on whether the participant had worked on the respective day into the employment variable. This resulted in considering three categories instead of one in the explanatory variable about the employment status: working-employed, i.e., participants who were employed and worked on the day of interest (weekday or weekend), non-working-employed, i.e., employed participants who did not work on the day of interest, and unemployed participants. Furthermore, we built separate models for each specific contact location, such as essential and leisure activities, transport, or health. By selecting only the contacts occurring in a particular setting, we could explore the differences in the determinants of contact behavior within that specific location and understand whether heterogeneity in contact patterns varies across the different settings.

The analysis was performed with Bambi (BAyesian Model Building Interface) [31], a Python library for Bayesian modeling. To improve the reliability of the contact data analysis, we applied a two-step filtering procedure designed to exclude participants who reported implausibly high numbers of contacts. The procedure is detailed in the supporting information section (Sect 3 in S1 Text).

Impact on work.

The perception of the negative impact of the COVID-19 pandemic on work varied significantly by age, geographical location, and employment status (Fig 1). Compared to the reference age group (35-44), participants aged 25-34 were significantly more likely to agree with the negative impact statement (OR=1.45, 95% CI=1.02–1.07). Participants living in northwestern regions were less likely to agree compared to those in the Center region (OR=0.75, 95% CI=0.56–0.99). Compared to employed participants, unemployed individuals were significantly more likely to agree (OR=1.49, calculated from Employed OR=0.67, 95% CI=0.53–0.85). Education level and gender were not significantly associated with perceptions of work impact in this model. Older adults and those living in northern regions were more likely to disagree with the perception of the negative impact of the pandemic, while unemployed individuals appeared to have relatively more negative perceptions of the work-related effects of the COVID-19 crisis.

Impact on social life.

Perceptions of the impact of the pandemic on social life varied significantly only by city size (Fig 1). Compared to participants living in cities with a population between 10k–100k, those living in cities with over 100,000 inhabitants were less likely to agree that the pandemic had a negative impact on their social life (OR=0.75, 95% CI=0.57–0.99). Youngest participants (age 18-24), were more likely to perceive a negative impact compared to the reference group (OR=1.53 95% CI=1.01–2.34). Geographic area, education, employment status, and gender did not show significant associations. Overall, younger age groups were significantly more likely to agree with perceptions of the negative impact of the pandemic on social life, while other factors like geography and employment status did not have a significant effect.

Impact on psychological well-being.

Perceptions of the negative impact of the pandemic on psychological well-being varied significantly by city size, geographic area, employment status, and gender (Fig 1). Compared to the reference city size (10k–100k), those in cities with over 100,000 inhabitants were less likely to agree (OR=0.69, 95% CI=0.52–0.92). Participants in the Northeast region were less likely to agree compared to those in the South (OR=0.72, 95% CI=0.53–0.97). Compared to employed individuals, unemployed participants were more likely to agree (OR=1.39, calculated from Employed OR=0.72, 95% CI=0.57–.91). Male participants were less likely to agree compared to females (OR=0.77, 95% CI=0.62–0.96). Age was a weakly significant factor in this model, with younger participants (18-24) perceiving a higher burden from the COVID-19 pandemic (OR=1.51, 95% CI=0.99–2.32). Perceptions of the pandemic’s impact on psychological well-being were influenced by age, gender, city size, and employment status, with younger, female, and unemployed participants living in smaller cities being more affected.

Impact on economic well-being.

Perceptions of the impact of the pandemic on economic status varied significantly by city size, geographic area, and employment status (Fig 1). Compared to the reference city size (10k–100k), participants in cities with over 100,000 inhabitants were less likely to agree with a negative impact (OR=0.73, 95% CI=0.55–0.96). Participants in the Northeast (OR=0.69, 95% CI=0.51–0.93) and Northwest (OR=0.72, 95% CI=0.55–0.95) were less likely to agree compared to those in the South regions. Compared to employed participants, unemployed individuals were more likely to agree (OR=1.27, calculated from Employed OR=0.79, 95% CI=0.62–0.99). Age, education, and gender were not significantly associated with perceptions of economic impact. Perceptions of the pandemic’s economic impact differed by city size and employment status, with participants living in more densely populated areas feeling less affected by the pandemic. However, unemployed individuals perceived a stronger economic impact of the pandemic than employed participants.

Socioeconomic and demographic effects on the perception of the COVID-19 impact.

Demographic and socioeconomic factors largely influenced the perception of the negative impact of the COVID-19 pandemic. Age emerged as a significant factor, with younger participants reporting experiencing the pandemic more intensely. The perceived impact appeared to diminish with increasing age, suggesting that older individuals may have been better equipped to cope with the challenges posed by the crisis [32]. Employment status also played a significant role, with unemployed individuals consistently reporting a stronger negative impact on their work, psychological, and economic well-being. Geographical differences were also observed, with participants in the northern regions perceiving a less pronounced impact on their work compared to those in the southern regions. These results may reflect the intrinsic regional differences in the economic and social stratification that might have been exacerbated by the pandemic. Gender also emerged as a factor influencing the perception of the psychological burden, with female participants reporting a higher level of agreement on the pandemic’s detrimental impact on their well-being. This suggests that women may have experienced greater mental health challenges during the crisis, confirming previous results during pandemic times [33,34]. Furthermore, the perception of the economic burden was influenced by city size, with those residing in larger urban centers reporting a lesser impact of the pandemic. This may be attributed to the greater economic resilience and resources available in larger metropolitan areas [35]. Collectively, these findings highlight the significant socioeconomic disparities in how individuals perceive the impacts of the COVID-19 pandemic, adding to the already observed disparities in the epidemic outcomes. This shows the importance of understanding socioeconomic differences even before a potential pandemic. Specifically, it is essential to investigate variations in protective and contact behaviors across socioeconomic statuses (SES). Such an understanding can help elucidate whether the observed disparities in pandemic outcomes are attributable to differences in these behaviors by SES.

Social distancing.

Older participants were more likely to always adhere to social distancing guidelines. In particular, those aged 45-54 (OR=1.43, 95% CI=1.03–.98) and 55-65 (OR=1.80, 95% CI=1.29–2.50) were significantly more likely to consistently comply [38], compared to younger participants (age group 35-44). Compared to employed individuals (reference), unemployed participants were more likely to comply (OR=1.47, calculated from Employed OR=0.68, 95% CI=0.53–0.87). Compared to the reference city size (10k–100k), those in cities >100k were less likely to comply (OR=0.68, 95% CI=0.51–0.91). Participants in the Northwest (OR=0.60, 95% CI=0.45–0.79) and Northeast (OR=0.67, 95% CI=0.49–0.91) were less likely to comply compared to those in the Southern regions.

Face mask.

Consistent face mask use was associated with employment status, education, and city size (Fig 2). Unemployed participants were more likely to consistently wear masks (OR=1.37, calculated from Employed OR=0.73, 95% CI=0.57–0.93). Participants with a university degree were more likely to wear masks compared to those without a degree (OR=1.32, 95% CI=0.99–1.75). Compared to the reference city size (10k–100k), those in cities >100k were less likely to consistently wear masks (OR=0.73, 95% CI=0.54–0.98). Younger participants (aged 18–24) were less likely to wear face masks than those in the 34-45 age group, controlling for all other factors.

Vaccine uptake.

COVID-19 vaccine uptake (Fig 3) varied significantly only by employment status. Unemployed individuals were significantly less likely to be vaccinated (OR=0.44, calculated from Employed OR=2.29, 95% CI=1.49–3.52). No other factors showed significant association in this model.

Influenza vaccine uptake (Fig 3) was associated with age, education, city size, and geographic area. Compared to the 35-44 age group (reference), participants aged 45-54 were less likely (OR=0.70, 95% CI=0.47–1.04, p=0.08) while those aged 55-65 were more likely (OR=1.51, 95% CI=1.04–2.20) to be vaccinated. Participants with a university degree were more likely to be vaccinated than those without a degree (OR for No Degree = 1/1.90 = 0.53, calculated from Degree OR=1.90, 95% CI=1.40–2.59). Those living in highly populated areas (>100k) were more likely to be vaccinated (OR=1.53, 95% CI=1.11–2.11). Participants in the Northeast (OR=0.59, 95% CI=0.41–0.86) and Northwest (OR=0.68, 95% CI=0.49–0.95) were less likely to be vaccinated compared to those in the Center region. It is important to note that the COVID-19 vaccine was mandatory for healthcare workers until November 2022, while the Green Pass, which restricted access to public enclosed spaces for unvaccinated individuals, was in effect for the rest of the Italian population until January 2023. We did not control for the occupation type of the participant, so the effect of being a healthcare worker cannot be captured by the analysis.

Socioeconomic and demographic factors of protective behavior.

Overall, participants’ ages were indicative of adherence to non-pharmaceutical interventions. Older participants were more likely to consistently adopt social distancing measures compared to younger participants. On the other hand, the analysis suggests that age did not play a strong role in the vaccine uptake, with only older participants (aged 55-65) more likely to be vaccinated against influenza compared to the reference group. Employment status also played a crucial role in adopting NPIs and vaccination. Unemployed participants were more likely to adopt NPIs consistently, such as face masks and social distancing, but less likely to be vaccinated against COVID-19. The geographic region also influenced vaccination rates. Participants from northeastern regions were less likely to consistently adopt social distancing compared to southern regions. Education level was another significant factor influencing vaccination decisions. Participants with a degree were more likely to be vaccinated against influenza. The results demonstrate that demographic, socioeconomic, and geographic characteristics show a strong association with how individuals adopt protective behaviors four years after the pandemic. This highlights the need to account for these factors when developing epidemiological models in order to more accurately capture disease transmission patterns and dynamics.

Contact patterns.

In this section, we describe the age-stratified contact patterns represented by the contact matrices. We consider the direct contact patterns of the participants on weekdays and weekends, first without differentiating with respect to the contact location, and subsequently, considering different contact locations. In Fig 4, we show the contact matrices for the two days of the surveys. Overall contact patterns show similar trends on both days, with more contacts on average during the weekend for younger participants. Younger participants have more interactions overall, in particular among the same age group, but also with the oldest age group. Considering different contact locations, Fig F in S1 Text and Fig G in S1 Text, overall, younger participants had more contacts than older ones for leisure activities and transportation, both on weekdays and weekends. Contacts on the weekdays during essential activities are mostly heterogeneous across the age groups. Contacts in health locations were mostly with older contacts for all the age groups, both on weekdays and weekends. Expectedly, homophilic contact patterns were more prevalent for leisure activities on both days, while contact patterns were more heterogeneous on public transport, especially on weekdays.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Contact matrices on weekday and weekend.

The matrix element (m_ij) is the median of 10000 bootstrap realizations of the average number of contacts of age group i with participants of age group j.

https://doi.org/10.1371/journal.pcbi.1013262.g004

Furthermore, to provide context for post-pandemic social mixing, we examine the contact patterns from our study alongside those from the pre-pandemic POLYMOD [22] and pandemic-period COMIX [18] surveys in Italy. Fig D in S1 Text displays the age-stratified contact matrices from these three periods. A common feature observed across all three datasets is the pattern of age homophily, where individuals tend to have more contacts with people in their own or adjacent age groups. This tendency remains visible in our post-pandemic survey data. Additionally, examining our current study’s matrix (Fig D in S1 Text, right panel), we observe relatively heterogeneous mixing patterns, particularly involving older participant age groups (45-54 and 55-65), who report non-negligible contacts with various other age groups beyond their immediate diagonal neighbors. It is important to note that, differently from POLYMOD and COMIX, our survey does not cover participants older than 65 or younger than 18 years old, and the survey methodology was CAWI (Computer Assisted Web Interviewing) [27]. The methodological differences in the design of the three studies allow only a qualitative comparison of the contact patterns within each period. Further details on the methods employed can be found in the Supplementary Information (Sect 3.1 in S1 Text), with Fig D in S1 Text also reporting the explicit median and interquartile range values for each element.

Association of socioeconomic and demographic factors with contact patterns.

To analyze the distribution of contacts across the two survey days, we applied the negative binomial regression, both for indirect and for direct contacts. The results of the indirect contacts model are in the supporting information section (Sect 4.1 in S1 Text). Figs 5 to 9 show the posterior distributions of the model coefficients (β), which represent changes in the expected number of contacts on a logarithmic scale. This additive representation is particularly useful for visualization, as it enables clearer comparisons across categories.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Highest Density Interval (HDI) of the posterior distribution for the coefficients (β) from the Bayesian negative binomial model for the number of direct contacts per participant on weekend and weekday.

The y-axis shows the names of each category of the explanatory variables, excluding the reference categories used for the intercept: age_group[35-44], education[Nodegree], gender[Female], city_size[10000-100000], country_area[South], employment[Unemployed]. The x-axis represents the estimated coefficient value (log-scale effect) relative to the reference category. A 97,5% HDI excluding zero suggests a statistically significant difference from the reference group.

https://doi.org/10.1371/journal.pcbi.1013262.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Highest Density Interval (HDI) of the posterior distribution for the coefficients (β) from the Bayesian negative binomial model for the number of direct contacts during essential activities, on weekend and weekday.

The y-axis shows the names of each category of the explanatory variables, excluding the reference categories used for the intercept: age_group[35-44], education[Nodegree], gender[Female], city_size[10000-100000], country_area[South], employment[Unemployed]. The x-axis represents the estimated coefficient value (log-scale effect) relative to the reference category. A 97,5% HDI excluding zero suggests a statistically significant difference from the reference group.

https://doi.org/10.1371/journal.pcbi.1013262.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Highest Density Interval (HDI) of the posterior distribution for the coefficients (β) from the Bayesian negative binomial model for the number of direct contacts during leisure activities, on weekend and weekday.

The y-axis shows the names of each category of the explanatory variables, excluding the reference categories used for the intercept: age_group[35-44], education[Nodegree], gender[Female], city_size[10000-100000], country_area[South], employment[Unemployed]. The x-axis represents the estimated coefficient value (log-scale effect) relative to the reference category. A 97,5% HDI excluding zero suggests a statistically significant difference from the reference group.

https://doi.org/10.1371/journal.pcbi.1013262.g007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Highest Density Interval (HDI) of the posterior distribution for the coefficients (β) from the Bayesian negative binomial model for the number of direct contacts during transport, on weekend and weekday.

The y-axis shows the names of each category of the explanatory variables, excluding the reference categories used for the intercept: age_group[35-44], education[Nodegree], gender[Female], city_size[10000-100000], country_area[South], employment[Unemployed]. The x-axis represents the estimated coefficient value (log-scale effect) relative to the reference category. A 97,5% HDI excluding zero suggests a statistically significant difference from the reference group.

https://doi.org/10.1371/journal.pcbi.1013262.g008

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Highest Density Interval (HDI) of the posterior distribution for the coefficients (β) from the Bayesian negative binomial model for the number of direct contacts in health locations, on weekend and weekday.

The y-axis shows the names of each category of the explanatory variables, excluding the reference categories used for the intercept: age_group[35-44], education[Nodegree], gender[Female], city_size[10000-100000], country_area[South], employment[Unemployed]. The x-axis represents the estimated coefficient value (log-scale effect) relative to the reference category. A 97,5% HDI excluding zero suggests a statistically significant difference from the reference group.

https://doi.org/10.1371/journal.pcbi.1013262.g009

In contrast, in the main text we report Incidence Rate Ratios (IRRs) [39], defined as , which describe the multiplicative change in the expected number of contacts relative to a reference category. An IRR with a Highest Density Interval (HDI) that does not include 1 indicates a statistically significant difference from the reference group at the level. IRRs are used in the resullt’s description because they more intuitively convey the magnitude of differences in contact numbers.

Results are reported for total contacts on weekdays (Table D in S1 Text) and weekends (Table F in S1 Text), as well as for contacts occurring in specific contexts: essential activities (Table H in S1 Text, Table J in S1 Text), leisure activities (Table L in S1 Text, Table N in S1 Text), transport (Table P in S1 Text, Table R in S1 Text), and health-related settings (Table T in S1 Text, Table V in S1 Text).

The reference group used in all models is: Age 35–44, Female, No degree, living in a city with 10,000–100,000 inhabitants, located in the South region, and currently unemployed.

Overall contacts.

Age showed a strong association with contact numbers (Table D in S1 Text, Table F in S1 Text). Compared to the reference group [35-44], participants aged [18-24] had significantly more contacts on weekdays (IRR = ; 97,5% HDI = [1.23–2.39]) and weekends (IRR = ; 97,5% HDI = [1.01–1.86]). This suggests about 67% more contacts on weekdays and 36% more on weekends for the youngest group. Older participants aged [55-65] had contact numbers similar to the reference group on weekdays, but potentially fewer contacts on weekends (IRR = ; 97,5% HDI = [0.66–1.03], not significant as HDI includes 1). Employment status was a key determinant. Compared to unemployed participants (reference), employed individuals who worked on the survey day had significantly more contacts on weekdays (IRR = ; 97,5% HDI = [2.61–3.82]) and also on weekends (IRR = ; 97,5% HDI = [1.73–2.89]). Employed individuals not working on the survey day had fewer contact with respect to the unemployed on weekdays (IRR = ; 97,5% HDI = [0.59–1.03]) and rather similar during weekends (IRR = ; 97,5% HDI = [0.97–1.36]). Gender (Male vs Female reference) and Education (Degree vs Nodegree reference) did not show statistically significant associations with the overall number of contacts on either weekdays or weekends (Table D in S1 Text, Table F in S1 Text), as the 95% HDIs for their IRRs included 1.

Geographic differences were limited. Compared to the South (reference), only participants in the North-East showed a tendency towards fewer contacts, primarily on weekends (IRR = ; 97,5% HDI = [0.71–1.08], not significant) (Table F in S1 Text).

City size showed a significant association only on weekends for the largest cities. Compared to the reference size (10k-100k), those in cities >100k reported significantly fewer contacts (IRR = ; 97,5% HDI = [0.58–0.85]) (Table F in S1 Text).

Essential activities.

In Fig 6 we show the results of the negative binomial regression for contacts during essential activities, details are shown in the supporting information section (Sect 4.2 in S1 Text). Younger individuals ([18-24]) tended to have more contacts than the reference [35-44] group, though the effect was not statistically significant. Employed working participants had significantly more contacts than unemployed on weekdays (IRR ; 97,5% HDI = [1.02–2.05]). Those in the largest cities (>100k) had significantly fewer contacts than the reference size on weekends (IRR = ; 97,5% HDI = [0.45–0.93]).

Leisure activities.

In Fig 7 we show the results for contacts that took place during leisure activities. Age effects were present, with the [18-24] group having more contacts than the [35-44] reference on weekdays, though not always statistically significant (Weekday IRR = ; 97,5% HDI = [0.71–3.76]). On weekdays, employed participants who were working reported significantly more contacts than unemployed individuals (IRR = ; HDI = [1.27–3.46]). A similar trend was observed on weekends, where employed individuals - regardless of whether they were currently working - also had more contacts. Although the HDI for this effect does not lie entirely above zero, the majority of the mass supports a positive association. More detailed information is reported in the supporting information section (Sect 4.3 in S1 Text).

Transport.

Fig 8 shows the results for contacts on public transport. Age effects were very strong here. Compared to the [35-44] reference group, the [18-24] group had dramatically more contacts on weekdays (IRR = ; 97,5% HDI = [2.27–34.8]) and weekends (IRR = ; 97,5% HDI = [1.03–9.10]). Employed working individuals had significantly more contacts than the unemployed on weekdays (IRR = ; 97,5% HDI = [1.43–10.47]). Residents in the North-West had significantly more transport contacts than Southerners on weekdays (IRR = ; 97,5% HDI = [1.06–7.77]).

Health locations.

In Fig 9 we show the results for contacts in health locations. Contact rates were generally lower. The [18-24] age group tended to have more contacts than the [35-44] reference, but this was not statistically significant. Employment status differences were generally not significant. Living in the largest cities (>100k) was associated with significantly fewer contacts on weekdays compared to mid-sized cities (IRR = ; 97,5% HDI = [0.23–1.01], borderline).

Comparison with indirect contacts.

We performed the same analyses for indirect contacts (Table E in S1 Text, Table G in S1 Text, Table I in S1 Text, Table K in S1 Text, Table M in S1 Text, Table O in S1 Text, Table Q in S1 Text, Table S in S1 Text, Table U in S1 Text, Table W in S1 Text). Generally, the main predictors identified for direct contacts also influenced indirect contacts in the same direction. For instance, the strong positive association of young age ([18-24]) and working employment status with weekday contacts was observed for both direct and indirect types, often with comparable effect sizes (IRRs), particularly in leisure and transport settings. Similarly, the significant negative association of large city size (>100k) with overall weekend contacts and health contacts was consistent across both direct and indirect measures. However, some subtle differences emerged, particularly on weekends, where certain factors significantly associated with direct contacts (e.g., age [18-24] for overall weekend or transport contacts) showed weaker or non-significant associations for indirect contacts. Overall, while minor variations exist, there were no fundamental differences suggesting distinct socioeconomic or demographic drivers for direct versus indirect contacts; factors influencing one type tended to influence the other similarly. Detailed results for indirect contacts are available in the Supplementary Information.

Socioeconomic determinants of contact behavior.

Relative to a baseline of [35-44] year old, unemployed females, with no degree in mid-sized Southern cities, age and employment status were the most consistent predictors of contact patterns. Younger individuals ([18-24]) reported significantly higher contact numbers overall (around 1.4-1.7 times the baseline) and dramatically higher contacts in transport settings (up to 8 times the baseline on weekdays). Employment status displayed a strong temporal pattern: employed individuals working on the survey day had substantially more contacts than unemployed individuals (around 2-3 times more, overall and in leisure/transport settings), especially on weekdays. Gender and education level showed limited association with contact frequency relative to their baselines. Geographic location and city size had context-dependent effects, with large cities sometimes associated with fewer contacts (overall weekend, essential, health) but potentially more transport contacts, and regional differences emerging primarily on weekends or for specific activities like transport. These findings underscore that contact behavior is shaped by a complex interplay of demographics (age), socioeconomic position (employment), and the specific context (day type, activity/location).