Social contact patterns during the early COVID-19 pandemic in Norway: insights from a panel study, April to September 2020

Background During the COVID-19 pandemic, many countries adopted social distance measures and lockdowns of varying strictness. Social contact patterns are essential in driving the spread of respiratory infections, and country-specific measurements are needed. This study aimed to gain insights into changes in social contacts and behaviour during the early pandemic phase in Norway. Methods We conducted an online panel study among a nationally representative sample of Norwegian adults by age and gender. The panel study included six data collections waves between April and September 2020, and 2017 survey data from a random sample of the Norwegian population (including children < 18 years old) were used as baseline. The market research company Ipsos was responsible for carrying out the 2020 surveys. We calculated mean daily contacts, and estimated age-stratified contact matrices during the study period employing imputation of child-to-child contacts. We used the next-generation method to assess the relative reduction of R0 and compared the results to reproduction numbers estimated for Norway during the 2020 study period. Results Over the six waves in 2020, 5 938 observations/responses were registered from 1 718 individuals who reported data on 22 074 contacts. The mean daily number of contacts among adults varied between 3.2 (95%CI 3.0-3.4) to 3.9 (95%CI 3.6–4.2) across the data collection waves, representing a 67–73% decline compared to pre-pandemic levels (baseline). Fewer contacts in the community setting largely drove the reduction; the drop was most prominent among younger adults. Despite gradual easing of social distance measures during the survey period, the estimated population contact matrices remained relatively stable and displayed more inter-age group mixing than at baseline. Contacts within households and the community outside schools and workplaces contributed most to social encounters. Using the next-generation method R0 was found to be roughly 25% of pre-pandemic levels during the study period, suggesting controlled transmission. Conclusion Social contacts declined significantly in the months following the March 2020 lockdown, aligning with implementation of stringent social distancing measures. These findings contribute valuable empirical information into the social behaviour in Norway during the early pandemic, which can be used to enhance policy-relevant models for addressing future crises when mitigation measures might be implemented. Supplementary Information The online version contains supplementary material available at 10.1186/s12889-024-18853-8.


Group contacts
From CoMix wave 2 to CoMix wave 6, 453 participants reported group contacts (449 of those also reported individual contacts); ranging from 11% to 19% of the individuals with contact data per data collection wave.We noticed that some participants reported high numbers with few of them reporting extreme values, such as in CoMix wave 2 where there were two participants that reported 4 847 and 7 021 contacts each.As mentioned in the main article the group contacts were excluded in the main analysis.
As an additional analysis to assess the impact of group contacts in CoMix wave 2 to 6, we calculated the mean number of contacts after fixing the maximum number of group contacts at 50, thereby reducing the influence of participants with exceptionally high contact numbers to the population mean.This affected 12 individuals in wave 2, 16 in wave 3, 8 in wave 4, 16 in wave 5 and 12 in wave 6, all of whom had reported more than 50 group contacts.We present the results from this analysis in Table S1.When comparing these figures with the mean contact numbers reported in Table 2, we observed an increase of 1,2 to 3,1 on average throughout CoMix waves 2 to 6. Similar as in Table 2, the lowest number of contacts was reported in wave 4 and the highest in wave 3.

Representativeness of study population
Our participants were sampled to be representative by age, gender, and region of residence.The dropouts and new recruitment of participants impacted a little in the composition of the sample with only small changes in the representativeness (Table S2).
We observed that the representativeness by gender and county was quite good.Regarding the age distribution, the 18-29 age group were undersampled ranging from 7 to 16% across the CoMix waves compared to the 20% representation in the general population.Conversely, the 50-69 age group showed a slight oversampling, ranging from 35 to 40% through the waves in contrast to the 30% representation in the general population.In addition, participants residing in Oslo County were marginally overrepresented throughout the waves, constituting 17-18% of the sample compared to the 13% presentation in the general population.While these variances were considered minor for the purpose of our study, we decided to conduct weighted analysis for the contact pattern data to account for these differences.3. Control measures overview during the study period In Norway, the stringency of control measures implemented since the beginning of the pandemic varied depending on the evolving epidemiological situation.In Table 3, we present some of the social distance measures implemented by the Norwegian government during the data collection period of the CoMix study (1,2).We should note that the social distance measures included in Table S3 are referring to the national guidelines, and there may have been variations in local measures may varied in some instances.
In Norway, children's daycare and schools were closed on 13 March 2020, and reopened under strict IPC measures from 20 April 2020.From end of May, a traffic light model was developed to guide administrators on IPC strategies in primary schools (grade 1-7, children 6-13 years of age) (3,4).This three-tiered system, with non-pharmaceutical measures, depended on local incidence and infection pressure.The guidelines advised the establishment of cohorts consisting of small permanent groups of children and staff with limited interaction between cohorts, alongside timely testing and isolation of symptomatic cases, and tracing and quarantine of their contacts.During the CoMix study period, the schools for children below 18 years old were all closed in wave 1 (lockdown) and 4 (summer holidays).Primary schools had strict IPC measures in wave 2 and the 'traffic model' restrictions in wave 3, 5 and 6.All other schools had some IPC measures in place even though they were open for the rest of the waves.Moreover, individual companies may have implemented their own distinct measures.**Note: From 1 st of June, cafes/restaurants/bars could reopen as long as they could provide a distance of one meter between guests and seating for all guests.All places had to close by 24.00..  1 042 30% 21% a: A contact could have been made in more than one place at the same day.Therefore, the proportions were calculated using the total of contacts for all places reported.b: A contact could have been made in more than one place at the same day (outdoors and indoors).Therefore, the proportions were calculated using the total of contact for both places.*Note: A contact could have been made in more than one place at the same day.Therefore, the proportions here were calculated using the total of contacts for all places reported.

Fig. S1 :
Fig. S1: Mean number of daily contacts reported,with 95% confidence intervals, that occurred A) outdoors and B) indoors, stratified by participant age group and data collection period (CoMix wave).The starting date of each CoMix data collection wave is indicated in the brackets.These results are weighted by gender.A)

Fig. S2 :
Fig. S2: Percentage (crude) of daily contacts reported by location* of participants in each CoMix wave.The starting date of each CoMix data collection wave is indicated in the brackets.

Fig S3 :
Fig S3: Boxplots of the estimated ratio of the maximum eigenvalues (CoMix/Baseline) of the adult-toadult contract matrices for each CoMix wave by location and type of contact.The ratios of dominant eigenvalues represent bootstrap sample pairs (N=10 000) that were used in the imputation process to scale contacts of children in the baseline contact matrices.

Fig S4 :
Fig S4: Boxplots of the estimated ratio of the maximum eigenvalues (CoMix/Baseline) of the full population matrices for each CoMix wave by location and type of contact.The ratios of dominant eigenvalues represent bootstrap sample pairs (N=10 000).

Fig S5 :
Fig S5: Imputed social contact matrices showing the mean number of daily physical contacts in the six CoMix waves; the corresponding matrix from the 2017 baseline survey is shown below as a reference.The matrices report bootstrap mean values from N=10 000 samples.Data were weighted on age and adjusted for reciprocity of contacts.

Fig S6 :
Fig S6:Social contact matrices of the six CoMix waves showing the mean number of daily contacts reported; below is shown the corresponding matrix of contacts reported by adults in the 2017 baseline survey as a reference.The figures represent bootstrap mean values from N=10 000 samples.Data were weighted on age.Note: Data on children contacts were not available (noted as NA: not available) for the CoMix waves and were available for the 2017 survey but were not included (noted as NS: not shown) in these matrices).

Fig S7 :
Fig S7:Social contact matrices of the six CoMix waves showing the mean number of daily physical contacts reported; below is shown the corresponding matrix of physical contacts reported by adults in the 2017 baseline survey as a reference.The matrices represent bootstrap mean values from N=10 000 samples.Data were weighted on age.Note: Data on children contacts were not available (noted as NA: not available) for the CoMix waves and were available for the 2017 survey but were not included (noted as NS: not shown) in these matrices).

Fig S8 :
Fig S8: Imputed, setting-specific social contact matrices showing the mean number of daily physical contact for the six CoMix waves.Locations include all contacts made in the home, at schools, at workplaces and other community contacts (transport, sport activities etc.)The figures represent bootstrap mean values of N=10 000.Data were weighted on age and adjusted for reciprocity of contacts.

Fig S9 :
Fig S9: Comparison of CoMix and baseline dominant eigenvectors and eigenvalues in four different assumption scenarios regarding the magnitude of reduction among children contacts; children-to-children contacts reduced by 25%, 50%, 65% and 80%: (A) Dominant normalised eigenvectors from contact matrices of all contacts by age group, (B) Dominant normalised eigenvectors from contact matrices of physical contacts by age group.

Fig S10 :
Fig S10: Boxplots showing the estimated basic reproduction numbers, R0, derived from the six CoMix wave surveys in four different assumption scenarios regarding the magnitude of reduction among children contacts; children-to-children contacts reduced by 25%, 50%, 65% and 80%.The R0 values were calculated based on contact matrices that consider all contacts (left panel) and exclusively physical contacts (right panel).The estimation involves multiplying the maximum eigenvalue ratios (Imp CoMix/baseline) by an initial R0-value established for Norway before the March 2020 lockdown.

Table S1 :
Summary of daily individual and group contacts reported in each CoMix wave in 2020.The crude mean of contacts (not weighted) per wave is presented in the last column.
*N.A.: Not available.In wave 1, participants were not able to report group contacts.**Total contacts= Sum of group and individual contacts.

Table S2 :
Sample characteristics in the baseline survey and of each CoMix wave (starting date of each CoMix data collection wave is indicated in brackets) compared to the proportions of the 2020 Norwegian adult population.

Table S4 :
Percentage (crude) of contacts that all participants reported by place* and type of contact in each CoMix wave.The starting date of each CoMix data collection wave is indicated in the brackets.