The Role of Mobility in the Dynamics of the COVID-19 Epidemic in Andalusia.

Metapopulation models have been a popular tool for the study of epidemic spread over a network of highly populated nodes (cities, provinces, countries) and have been extensively used in the context of the ongoing COVID-19 pandemic. In the present work, we revisit such a model, bearing a particular case example in mind, namely that of the region of Andalusia in Spain during the period of the summer-fall of 2020 (i.e., between the first and second pandemic waves). Our aim is to consider the possibility of incorporation of mobility across the province nodes focusing on mobile-phone time-dependent data, but also discussing the comparison for our case example with a gravity model, as well as with the dynamics in the absence of mobility. Our main finding is that mobility is key toward a quantitative understanding of the emergence of the second wave of the pandemic and that the most accurate way to capture it involves dynamic (rather than static) inclusion of time-dependent mobility matrices based on cell-phone data. Alternatives bearing no mobility are unable to capture the trends revealed by the data in the context of the metapopulation model considered herein.

MepoGNN: Metapopulation Epidemic Forecasting with Graph Neural Networks

Epidemic prediction is a fundamental task for epidemic control and prevention. Many mechanistic models and deep learning models are built for this task. However, most mechanistic models have difficulty estimating the time/region-varying epidemiological parameters, while most deep learning models lack the guidance of epidemiological domain knowledge and interpretability of prediction results. In this study, we propose a novel hybrid model called MepoGNN for multi-step multi-region epidemic forecasting by incorporating Graph Neural Networks (GNNs) and graph learning mechanisms into Metapopulation SIR model. Our model can not only predict the number of confirmed cases but also explicitly learn the epidemiological parameters and the underlying epidemic propagation graph from heterogeneous data in an end-to-end manner. Experiment results demonstrate our model outperforms the existing mechanistic models and deep learning models by a large margin. Furthermore, the analysis on the learned parameters demonstrates the high reliability and interpretability of our model and helps better understanding of epidemic spread. Our model and data have already been public on GitHub https://github.com/deepkashiwa20/MepoGNN.git. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Analysis of serological surveys of antibodies to SARS-CoV-2 in the United States to estimate parameters needed for transmission modeling and to evaluate and improve the accuracy of predictions.

Seroprevalence studies can estimate proportions of the population that have been infected or vaccinated, including infections that were not reported because of the lack of symptoms or testing. Based on information from studies in the United States from mid-summer 2020 through the end of 2021, we describe proportions of the population with antibodies to SARS-CoV-2 as functions of age and time. Slices through these surfaces at arbitrary times provide initial and target conditions for simulation modeling. They also provide the information needed to calculate age-specific forces of infection, attack rates, and - together with contact rates - age-specific probabilities of infection on contact between susceptible and infectious people. We modified the familiar Susceptible-Exposed-Infectious-Removed (SEIR) model to include features of the biology of COVID-19 that might affect transmission of SARS-CoV-2 and stratified by age and location. We consulted the primary literature or subject matter experts for contact rates and other parameter values. Using time-varying Oxford COVID-19 Government Response Tracker assessments of US state and DC efforts to mitigate the pandemic and compliance with non-pharmaceutical interventions (NPIs) from a YouGov survey fielded in the US during 2020, we estimate that the efficacy of social-distancing when possible and mask-wearing otherwise at reducing susceptibility or infectiousness was 31% during the fall of 2020. Initialized from seroprevalence among people having commercial laboratory tests for purposes other than SARS-CoV-2 infection assessments on 7 September 2020, our age- and location-stratified SEIR population model reproduces seroprevalence among members of the same population on 25 December 2020 quite well. Introducing vaccination mid-December 2020, first of healthcare and other essential workers, followed by older adults, people who were otherwise immunocompromised, and then progressively younger people, our metapopulation model reproduces seroprevalence among blood donors on 4 April 2021 less well, but we believe that the discrepancy is due to vaccinations being under-reported or blood donors being disproportionately vaccinated, if not both. As experimenting with reliable transmission models is the best way to assess the indirect effects of mitigation measures, we determined the impact of vaccination, conditional on NPIs. Results indicate that, during this period, vaccination substantially reduced infections, hospitalizations and deaths. This manuscript was submitted as part of a theme issue on "Modelling COVID-19 and Preparedness for Future Pandemics."

TransCode: Uncovering COVID-19 transmission patterns via deep learning.

BACKGROUND: The heterogeneity of COVID-19 spread dynamics is determined by complex spatiotemporal transmission patterns at a fine scale, especially in densely populated regions. In this study, we aim to discover such fine-scale transmission patterns via deep learning. METHODS: We introduce the notion of TransCode to characterize fine-scale spatiotemporal transmission patterns of COVID-19 caused by metapopulation mobility and contact behaviors. First, in Hong Kong, China, we construct the mobility trajectories of confirmed cases using their visiting records. Then we estimate the transmissibility of individual cases in different locations based on their temporal infectiousness distribution. Integrating the spatial and temporal information, we represent the TransCode via spatiotemporal transmission networks. Further, we propose a deep transfer learning model to adapt the TransCode of Hong Kong, China to achieve fine-scale transmission characterization and risk prediction in six densely populated metropolises: New York City, San Francisco, Toronto, London, Berlin, and Tokyo, where fine-scale data are limited. All the data used in this study are publicly available. RESULTS: The TransCode of Hong Kong, China derived from the spatial transmission information and temporal infectiousness distribution of individual cases reveals the transmission patterns (e.g., the imported and exported transmission intensities) at the district and constituency levels during different COVID-19 outbreaks waves. By adapting the TransCode of Hong Kong, China to other data-limited densely populated metropolises, the proposed method outperforms other representative methods by more than 10% in terms of the prediction accuracy of the disease dynamics (i.e., the trend of case numbers), and the fine-scale spatiotemporal transmission patterns in these metropolises could also be well captured due to some shared intrinsically common patterns of human mobility and contact behaviors at the metapopulation level. CONCLUSIONS: The fine-scale transmission patterns due to the metapopulation level mobility (e.g., travel across different districts) and contact behaviors (e.g., gathering in social-economic centers) are one of the main contributors to the rapid spread of the virus. Characterization of the fine-scale transmission patterns using the TransCode will facilitate the development of tailor-made intervention strategies to effectively contain disease transmission in the targeted regions.

Pathogen diversity in meta-population networks

The pathogen diversity means that multiple strains coexist, and widely exist in the biology systems. The new mutation of SARS-CoV-2 leading to worldwide pathogen diversity is a typical example. What are the main factors of inducing the pathogen diversity? Previous studies indicated the pathogen mutation is the most important reason for inducing the pathogen diversity. The traffic network and gene network are crucial in shaping the dynamics of pathogen contagion, while their roles for the pathogen diversity still lacking a theoretical study. To this end, we propose a reaction–diffusion process of pathogens with mutations on meta-population networks, which includes population movement and strain mutation. We extend the Microscopic Markov Chain Approach (MMCA) to describe the model. Traffic networks make pathogen diversity more likely to occur in cities with lower infection densities. The likelihood of pathogen diversity is low in cities with short effective distances in the traffic network. Star-type gene network is more likely to lead to pathogen diversity than lattice-type and chain-type gene networks. When pathogen localization is present, infection is localized to strains that are at the endpoints of the gene network. Both the increased probability of movement and mutation promote pathogen diversity. The results also show that the population tends to move to cities with short effective distances, resulting in the infection density is high. © 2022 Elsevier Ltd

Modeling social interaction and metapopulation mobility of the COVID-19 pandemic in main cities of highly populated Java Island, Indonesia: An agent-based modeling approach

IntroductionCoronavirus transmission is strongly influenced by human mobilities and interactions within and between different geographical regions. Human mobility within and between cities is motivated by several factors, including employment, cultural-driven, holidays, and daily routines. MethodWe developed a sustained metapopulation (SAMPAN) model, an agent-based model (ABM) for simulating the effect of individual mobility and interaction behavior on the spreading of COVID-19 viruses across main cities on Java Island, Indonesia. The model considers social classes and social mixing affecting the mobility and interaction behavior within a sub-population of a city in the early pandemic. Travelers' behavior represents the mobility among cities from central cities to other cities and commuting behavior from the surrounding area of each city. ResultsLocal sensitivity analysis using one factor at a time was performed to test the SAMPAN model, and we have identified critical parameters for the model. While validation was carried out for the Jakarta area, we are confident in implementing the model for a larger area with the concept of metapopulation dynamics. We included the area of Bogor, Depok, Bekasi, Bandung, Semarang, Surakarta, Yogyakarta, Surabaya, and Malang cities which have important roles in the COVID-19 pandemic spreading on this island. DiscussionOur SAMPAN model can simulate various waves during the first year of the pandemic caused by various phenomena of large social mobilities and interactions, particularly during religious occasions and long holidays.

Brazilian Collaborative Network For COVID-19 Modeling: Successful Experience Of Using Real-Time Science To Support Evidence-Based Decision-Making

Introduction. Modeling is important for guiding policy during epidemics. The objective of this work was to describe the experience of structuring a multidisciplinary collaborative network in Brazil for modeling coronavirus disease 2019 (COVID-19) to support decisionmaking throughout the pandemic. Methods. Responding to a national call in June 2020 for proposals on COVID-19 mitigation projects, we established a team of investigators from public universities located in various regions throughout Brazil. The team's main objective was to model severe acute respiratory syndrome coronavirus 2 transmission dynamics in various demographic and epidemiologic settings in Brazil using different types of models and mitigation interventions. The modeling results aimed to provide information to support policy making. This descriptive study outlines the processes, products, challenges, and lessons learned from this innovative experience. Results. The network included 18 researchers (epidemiologists, infectious diseases experts, statisticians, and modelers) from various backgrounds, including ecology, geography, physics, and mathematics. The criteria for joining the network were having a communication channel with public health decision-makers and being involved in generating evidence for public policy. During a 24-month period, the following sub-projects were established: (i) development of a susceptible-exposed-infected-recovered-like, individual-based metapopulation and Markov chain model;(ii) projection of COVID-19 transmission and impact over time with respect to cases, hospitalizations, and deaths;(iii) assessment of the impact of non-pharmacological interventions for COVID-19;(iv) evaluation of the impact of reopening schools;and (v) determining optimal strategies for COVID-19 vaccination. In addition, we mapped existing COVID-19 modeling groups nationwide and conducted a systematic review of relevant published research literature from Brazil. Conclusions. Infectious disease modeling for guiding public health policy requires interaction between epidemiologists, public health specialists, and modelers. Communicating modeling results in a nonacademic format is an additional challenge, so close interaction with policy makers is essential to ensure that the information is useful. Establishing a network of modeling groups will be useful for future disease outbreaks.

Pathogen diversity in meta-population networks

The pathogen diversity means that multiple strains coexist, and widely exist in the biology systems. The new mutation of SARS-CoV-2 leading to worldwide pathogen diversity is a typical example. What are the main factors of inducing the pathogen diversity? Previous studies indicated the pathogen mutation is the most important reason for inducing the pathogen diversity. The traffic network and gene network are crucial in shaping the dynamics of pathogen contagion, while their roles for the pathogen diversity still lacking a theoretical study. To this end, we propose a reaction–diffusion process of pathogens with mutations on meta-population networks, which includes population movement and strain mutation. We extend the Microscopic Markov Chain Approach (MMCA) to describe the model. Traffic networks make pathogen diversity more likely to occur in cities with lower infection densities. The likelihood of pathogen diversity is low in cities with short effective distances in the traffic network. Star-type gene network is more likely to lead to pathogen diversity than lattice-type and chain-type gene networks. When pathogen localization is present, infection is localized to strains that are at the endpoints of the gene network. Both the increased probability of movement and mutation promote pathogen diversity. The results also show that the population tends to move to cities with short effective distances, resulting in the infection density is high.

Estimating the effective reproduction number for heterogeneous models using incidence data.

The effective reproduction number, R ( t ) , plays a key role in the study of infectious diseases, indicating the current average number of new infections caused by an infected individual in an epidemic process. Estimation methods for the time evolution of R ( t ) , using incidence data, rely on the generation interval distribution, g(τ), which is usually obtained from empirical data or theoretical studies using simple epidemic models. However, for systems that present heterogeneity, either on the host population or in the expression of the disease, there is a lack of data and of a suitable general methodology to obtain g(τ). In this work, we use mathematical models to bridge this gap. We present a general methodology for obtaining explicit expressions of the reproduction numbers and the generation interval distributions, within and between model sub-compartments provided by an arbitrary compartmental model. Additionally, we present the appropriate expressions to evaluate those reproduction numbers using incidence data. To highlight the relevance of such methodology, we apply it to the spread of COVID-19 in municipalities of the state of Rio de Janeiro, Brazil. Using two meta-population models, we estimate the reproduction numbers and the contributions of each municipality in the generation of cases in all others.

Investigating time, strength, and duration of measures in controlling the spread of COVID-19 using a networked meta-population model.

Policy makers around the world are facing unprecedented challenges in making decisions on when and what degrees of measures should be implemented to tackle the COVID-19 pandemic. Here, using a nationwide mobile phone dataset, we developed a networked meta-population model to simulate the impact of intervention in controlling the spread of the virus in China by varying the effectiveness of transmission reduction and the timing of intervention start and relaxation. We estimated basic reproduction number and transition probabilities between health states based on reported cases. Our model demonstrates that both the time of initiating an intervention and its effectiveness had a very large impact on controlling the epidemic, and the current Chinese intense social distancing intervention has reduced the impact substantially but would have been even more effective had it started earlier. The optimal duration of the control measures to avoid resurgence was estimated to be 2 months, although would need to be longer under less effective controls.

Reassessment of contact restrictions and testing campaigns against COVID-19 via spatio-temporal modeling.

Since the earliest outbreak of COVID-19, the disease continues to obstruct life normalcy in many parts of the world. The present work proposes a mathematical framework to improve non-pharmaceutical interventions during the new normal before vaccination settles herd immunity. The considered approach is built from the viewpoint of decision makers in developing countries where resources to tackle the disease from both a medical and an economic perspective are scarce. Spatial auto-correlation analysis via global Moran's index and Moran's scatter is presented to help modulate decisions on hierarchical-based priority for healthcare capacity and interventions (including possible vaccination), finding a route for the corresponding deployment as well as landmarks for appropriate border controls. These clustering tools are applied to sample data from Sri Lanka to classify the 26 Regional Director of Health Services (RDHS) divisions into four clusters by introducing convenient classification criteria. A metapopulation model is then used to evaluate the intra- and inter-cluster contact restrictions as well as testing campaigns under the absence of confounding factors. Furthermore, we investigate the role of the basic reproduction number to determine the long-term trend of the regressing solution around disease-free and endemic equilibria. This includes an analytical bifurcation study around the basic reproduction number using Brouwer Degree Theory and asymptotic expansions as well as related numerical investigations based on path-following techniques. We also introduce the notion of average policy effect to assess the effectivity of contact restrictions and testing campaigns based on the proposed model's transient behavior within a fixed time window of interest.

The Impact of Quarantine and Medical Resources on the Control of COVID-19 in Wuhan based on a Household Model.

In order to understand how Wuhan curbed the COVID-19 outbreak in 2020, we build a network transmission model of 123 dimensions incorporating the impact of quarantine and medical resources as well as household transmission. Using our new model, the final infection size of Wuhan is predicted to be 50,662 (95%CI: 46,234, 55,493), and the epidemic would last until April 25 (95%CI: April 23, April 29), which are consistent with the actual situation. It is shown that quarantining close contacts greatly reduces the final size and shorten the epidemic duration. The opening of Fangcang shelter hospitals reduces the final size by about 17,000. Had the number of hospital beds been sufficient when the lockdown started, the number of deaths would have been reduced by at least 54.26%. We also investigate the distribution of infectious individuals in unquarantined households of different sizes. The high-risk households are those with size from two to four before the peak time, while the households with only one member have the highest risk after the peak time. Our findings provide a reference for the prevention, mitigation and control of COVID-19 in other cities of the world.

Effect of Non-homogeneous Mixing and Asymptomatic Individuals on Final Epidemic Size and Basic Reproduction Number in a Meta-Population Model.

To uncover the effective interventions during the pandemic period, a novel mathematical model, which incorporates separate compartments for incubation and asymptomatic individuals, has been developed in this paper. On the basis of a general mixing, final size relation and next-generation matrix are derived for a meta-population model by introducing the matrix blocking. The final size ([Formula: see text]) and the basic reproduction number ([Formula: see text]) are no longer a simple monotonous relationship. The analytical results of heterogeneity illustrate that activity is more sensitive than the others. And the proportion of asymptomatic individuals is a key factor for final epidemic size compared to the regulatory factor. Furthermore, the impact of preferential contact level on [Formula: see text] and [Formula: see text] is comparatively complex. The isolation can effectively reduce the final size, which further verifies its effectiveness. When vaccination is considered, the mixing methods maybe influence the doses of vaccination used and its effective. Moreover, using the present predictive model, we can provide the valuable reference about identifying the ideal strategies to curb the pandemic disease.

Benchmarking Optimal Control for Network Dynamic Systems with Plausible Epidemic Models

The sheer dimension of network dynamic systems adds a challenge of scale to synthesizing optimal control, which the techniques such as mean-field approximation, reinforcement learning, and graphon mean field games attempt to overcome. We propose to use compartmental metapopulation epidemic models derived from open data to benchmark these advanced approaches on an important problem with intuitive visualization options such as choropleth maps. To this end, we formalize a procedure for generating plausible instances of such models with 1–64,735 nodes based on open census data for the contiguous U.S., each with a network of daily commute and airplane travel, coupled with a formal aggregation routine enabling a view of the same geography at different resolutions, illustrated by merging the 2,072 census tracts in Oregon and Washington states, together with their travel networks, into 75 county-level nodes, 23 “airport service area” nodes, and 2 nodes for states themselves. These four cases, and ten other, are then put through 180-day “patient zero” scenarios in a Metapopulation SIR Model with per-node “lockdown level” control, with the objective of minimizing the cumulative number of infections and the lockdown level. The optimal control is derived through the Pontryagin Maximum Principle and numerically computed with the forward-backward sweep method. To ensure reproducibility, the instance generator, solver, and visualization routines are available at https://github.com/yvs314/epi-net-m. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Epidemic spreading with migration in networked metapopulation.

Migration plays a crucial role in epidemic spreading, and its dynamic can be studied by metapopulation model. Instead of the uniform mixing hypothesis, we adopt networked metapopulation to build the model of the epidemic spreading and the individuals' migration. In these populations, individuals are connected by contact network and populations are coupled by individuals migration. With the network mean-field and the gravity law of migration, we establish the N-seat intertwined SIR model and obtain its basic reproduction number ℛ 0 . Meanwhile, we devise a non-markov Node-Search algorithm for model statistical simulations. Through the static network migration ansatz and ℛ 0 formula, we discover that migration will not directly increase the epidemic replication capacity. But when ℛ 0 > 1 , the migration will make the susceptive population evolve from metastable state (disease-free equilibrium) to stable state (endemic equilibrium), and then increase the influence area of epidemic. Re-evoluting the epidemic outbreak in Wuhan, top 94 cities empirical data validate the above mechanism. In addition, we estimate that the positive anti-epidemic measures taken by the Chinese government may have reduced 4 million cases at least during the first wave of COVID-19, which means those measures, such as the epidemiological investigation, nucleic acid detection in medium-high risk areas and isolation of confirmed cases, also play a significant role in preventing epidemic spreading after travel restriction between cities.

Metapopulation epidemic models with a universal mobility pattern on interconnected networks.

The global pandemic of the coronavirus disease 2019 (COVID-19) exemplifies the influence of human mobility on epidemic spreading. A framework called the movement-interaction-return (MIR) model is a model to study the impact of human mobility on epidemic spreading. In this paper, we investigate epidemic spreading in interconnected metapopulation networks. Specifically, we incorporate the human mobility pattern called the radiation model into the MIR model. As a result, the proposed model is more realistic in comparison to the original MIR model. We use the tensorial framework to develop Markovian equations that describe the dynamics of the proposed model on interconnected metapopulation networks. Then we derive the corresponding epidemic thresholds by converting tensors into matrices. Comprehensive numerical simulations confirm our analysis.

Existing human mobility data sources poorly predicted the spatial spread of SARS-CoV-2 in Madagascar.

For emerging epidemics such as the COVID-19 pandemic, quantifying travel is a key component of developing accurate predictive models of disease spread to inform public health planning. However, in many LMICs, traditional data sets on travel such as commuting surveys as well as non-traditional sources such as mobile phone data are lacking, or, where available, have only rarely been leveraged by the public health community. Evaluating the accuracy of available data to measure transmission-relevant travel may be further hampered by limited reporting of suspected and laboratory confirmed infections. Here, we leverage case data collected as part of a COVID-19 dashboard collated via daily reports from the Malagasy authorities on reported cases of SARS-CoV-2 across the 22 regions of Madagascar. We compare the order of the timing of when cases were reported with predictions from a SARS-CoV-2 metapopulation model of Madagascar informed using various measures of connectivity including a gravity model based on different measures of distance, Internal Migration Flow data, and mobile phone data. Overall, the models based on mobile phone connectivity and the gravity-based on Euclidean distance best predicted the observed spread. The ranks of the regions most remote from the capital were more difficult to predict but interestingly, regions where the mobile phone connectivity model was more accurate differed from those where the gravity model was most accurate. This suggests that there may be additional features of mobility or connectivity that were consistently underestimated using all approaches but are epidemiologically relevant. This work highlights the importance of data availability and strengthening collaboration among different institutions with access to critical data - models are only as good as the data that they use, so building towards effective data-sharing pipelines is essential.

Optimal allocation of resources to healthcare workers or the general populace: a modelling study.

We consider a model that distinguishes susceptible; infected, but not yet infectious; pre-symptomatic, symptomatic, asymptomatic, and hospitalized infectious; recovered and dead members of two groups: healthcare workers (HCW) and members of the community that they serve. Because of the frequency or duration of their exposures to SARS-CoV-2, a greater fraction of HCW would experience severe COVID-19 symptoms that require medical care, which reduces mortality rates, absent personal protective equipment (PPE). While N95 masks (and, possibly, other scarce medical resources) are available to members of both groups, they do not use them equally well (i.e. efficacy and compliance differ). We investigated the optimal allocation of potentially scarce medical resources between these groups to control the pandemic and reduce overall infections and mortality via derivation and analysis of expressions for the reproduction numbers and final size. We also simulated prevalence and cumulative incidence, quantities relevant to surge capacity and population immunity, respectively. We found that, under realistic conditions, the optimal allocation is virtually or entirely to HCW, but that allocation of surplus masks and other medical resources to members of the general community also reduces infections and deaths.

Dynamics of COVID-19 pandemic in India and Pakistan: A metapopulation modelling approach.

India has been the latest global epicenter for COVID-19, a novel coronavirus disease that emerged in China in late 2019. We present a base mathematical model for the transmission dynamics of COVID-19 in India and its neighbor, Pakistan. The base model was rigorously analyzed and parameterized using cumulative COVID-19 mortality data from each of the two countries. The model was used to assess the population-level impact of the control and mitigation strategies implemented in the two countries (notably non-pharmaceutical interventions). Numerical simulations of the basic model indicate that, based on the current baseline levels of the control and mitigation strategies implemented, the pandemic trajectory in India is on a downward trend. This downward trend will be reversed, and India will be recording mild outbreaks, if the control and mitigation strategies are relaxed from their current levels. By early September 2021, our simulations suggest that India could record up to 460,000 cumulative deaths under baseline levels of the implemented control strategies, while Pakistan (where the pandemic is comparatively milder) could see over 24,000 cumulative deaths at current mitigation levels. The basic model was extended to assess the impact of back-and-forth mobility between the two countries. Simulations of the resulting metapopulation model show that the burden of the COVID-19 pandemic in Pakistan increases with increasing values of the average time residents of India spend in Pakistan, with daily mortality in Pakistan peaking in mid-August to mid-September of 2021. Under the respective baseline control scenarios, our simulations show that the back-and-forth mobility between India and Pakistan could delay the time-to-elimination of the COVID-19 pandemic in India and Pakistan to November 2022 and July 2022, respectively.

A data-driven metapopulation model for the Belgian COVID-19 epidemic: assessing the impact of lockdown and exit strategies.

BACKGROUND: In response to the ongoing COVID-19 pandemic, several countries adopted measures of social distancing to a different degree. For many countries, after successfully curbing the initial wave, lockdown measures were gradually lifted. In Belgium, such relief started on May 4th with phase 1, followed by several subsequent phases over the next few weeks. METHODS: We analysed the expected impact of relaxing stringent lockdown measures taken according to the phased Belgian exit strategy. We developed a stochastic, data-informed, meta-population model that accounts for mixing and mobility of the age-structured population of Belgium. The model is calibrated to daily hospitalization data and is able to reproduce the outbreak at the national level. We consider different scenarios for relieving the lockdown, quantified in terms of relative reductions in pre-pandemic social mixing and mobility. We validate our assumptions by making comparisons with social contact data collected during and after the lockdown. RESULTS: Our model is able to successfully describe the initial wave of COVID-19 in Belgium and identifies interactions during leisure/other activities as pivotal in the exit strategy. Indeed, we find a smaller impact of school re-openings as compared to restarting leisure activities and re-openings of work places. We also assess the impact of case isolation of new (suspected) infections, and find that it allows re-establishing relatively more social interactions while still ensuring epidemic control. Scenarios predicting a second wave of hospitalizations were not observed, suggesting that the per-contact probability of infection has changed with respect to the pre-lockdown period. CONCLUSIONS: Contacts during leisure activities are found to be most influential, followed by professional contacts and school contacts, respectively, for an impending second wave of COVID-19. Regular re-assessment of social contacts in the population is therefore crucial to adjust to evolving behavioral changes that can affect epidemic diffusion.