Background SARS-CoV-2 dynamics are driven by human behaviour. Social contact data are of utmost importance in the context of transmission models of close-contact infections. Methods Using online representative panels of adults reporting on their own behaviour as well as parents reporting on the behaviour of one of their children, we collect contact mixing (CoMix) behaviour in various phases of the COVID-19 pandemic in over 20 European countries. We provide these timely, repeated observations using an online platform: SOCRATES-CoMix. In addition to providing cleaned datasets to researchers, the platform allows users to extract contact matrices that can be stratified by age, type of day, intensity of the contact and gender. These observations provide insights on the relative impact of recommended or imposed social distance measures on contacts and can inform mathematical models on epidemic spread. Conclusion These data provide essential information for policymakers to balance non-pharmaceutical interventions, economic activity, mental health and wellbeing, during vaccine rollout.

Background During the COVID-19 pandemic, the CoMix study, a longitudinal behavioral survey, was designed to monitor social contacts and public awareness in multiple countries, including Belgium. As a longitudinal survey, it is vulnerable to participants’ “survey fatigue”, which may impact inferences. Methods A negative binomial generalized additive model for location, scale, and shape (NBI GAMLSS) was adopted to estimate the number of contacts reported between age groups and to deal with under-reporting due to fatigue within the study. The dropout process was analyzed with first-order auto-regressive logistic regression to identify factors that influence dropout. Using the so-called next generation principle, we calculated the effect of under-reporting due to fatigue on estimating the reproduction number. Results Fewer contacts were reported as people participated longer in the survey, which suggests under-reporting due to survey fatigue. Participant dropout is significantly affected by household size and age categories, but not significantly affected by the number of contacts reported in any of the two latest waves. This indicates covariate-dependent missing completely at random (MCAR) in the dropout pattern, when missing at random (MAR) is the alternative. However, we cannot rule out more complex mechanisms such as missing not at random (MNAR). Moreover, under-reporting due to fatigue is found to be consistent over time and implies a 15-30% reduction in both the number of contacts and the reproduction number ($$R_0$$ R 0 ) ratio between correcting and not correcting for under-reporting. Lastly, we found that correcting for fatigue did not change the pattern of relative incidence between age groups also when considering age-specific heterogeneity in susceptibility and infectivity. Conclusions CoMix data highlights the variability of contact patterns across age groups and time, revealing the mechanisms governing the spread/transmission of COVID-19/airborne diseases in the population. Although such longitudinal contact surveys are prone to the under-reporting due to participant fatigue and drop-out, we showed that these factors can be identified and corrected using NBI GAMLSS. This information can be used to improve the design of similar, future surveys.

The analysis of hierarchical data that take the form of clusters with random size has received considerable attention. The focus here is on samples that are very large in terms of number of clusters and/or members per cluster, on the one hand, as well as on very small samples (e.g., when studying rare diseases), on the other. Whereas maximum likelihood inference is straightforward in medium to large samples, in samples of sizes considered here it may become prohibitive. We propose sample-splitting (Molenberghs et al., 2011) as a way to replace iterative optimization of a likelihood that does not admit an analytical solution, with closed-form calculations. We use pseudo-likelihood (Molenberghs et al., 2014), consisting of computing weighted averages over solutions obtained for each cluster size occurring. As a result, the statistical properties of this approach need to be investigated. This is especially important because the minimal sufficient statistics involved are incomplete. The operational characteristics are studied using a first set of simulations. Simulations are also done to compare the proposed method to existing techniques developed to circumvent difficulties with unequal cluster sizes, such as multiple imputation. It follows that the proposed non-iterative methods have a strong beneficial impact on computation time; at the same time, the method is the most precise among its competitors considered. The findings are illustrated using data from a developmental toxicity study, where clusters are formed of fetuses within litters.

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.

Background The European Centre for Disease Prevention and Control (ECDC) systematically collates information from sources to rapidly detect early public health threats. The lack of a freely available, customisable and automated early warning tool using data from Twitter prompted the ECDC to develop epitweetr, which collects, geolocates and aggregates tweets generating signals and email alerts. Aim This study aims to compare the performance of epitweetr to manually monitoring tweets for the purpose of early detecting public health threats. Methods We calculated the general and specific positive predictive value (PPV) of signals generated by epitweetr between 19 October and 30 November 2020. Sensitivity, specificity, timeliness and accuracy and performance of tweet geolocation and signal detection algorithms obtained from epitweetr and the manual monitoring of 1,200 tweets were compared. Results The epitweetr geolocation algorithm had an accuracy of 30.1% at national, and 25.9% at subnational levels. The signal detection algorithm had 3.0% general PPV and 74.6% specific PPV. Compared to manual monitoring, epitweetr had greater sensitivity (47.9% and 78.6%, respectively), and reduced PPV (97.9% and 74.6%, respectively). Median validation time difference between 16 common events detected by epitweetr and manual monitoring was -48.6 hours (IQR: −102.8 to −23.7). Conclusion Epitweetr has shown sufficient performance as an early warning tool for public health threats using Twitter data. Since epitweetr is a free, open-source tool with configurable settings and a strong automated component, it is expected to increase in usability and usefulness to public health experts.