COVID-19 report on a meta-population model for Belgium: a first

status report

Pietro Coletti, Pieter Libin, Oana Petrof, Steven Abrams, Christel Faes,Elise Kuylen, James Wambua, Lander Willem, the SIMID COVID-19 team,

Philippe Beutels, Niel Hens

May 18, 2020

This report is a preprint and has not been certified by peer review. The copyright holder for thispreprint is the author. All rights are reserved and no reuse is allowed without permission.

Main messages

• Intervention measures were effective, reducing reproductive number from a median value of3.18 to a median value of 0.70

• Care should be taken in relieving lockdown.

• Leisure activities have the largest potential impact on epidemic resurgence.

• Contact tracing and isolation of suspected cases are crucial to limit epidemic resurgence.

Assumptions and limitations

Below we report the main assumptions and related limitations of our model:

• We consider two age classes (0-18 and 19+). Therefore there is limited heterogeneity withage.

• We use pre-pandemic data for social mixing and commuting, assessing their reduction duringlockdown from a fit to data on new hospitalizations. Information on how people change theircontact and travel behavior in response to interventions will become available in the nextweeks. This will allow us to lift some of the assumptions made.

• We define the impact of distancing measures and exit scenarios in terms of location-specific“contact reductions”, which are used as proxy for changes in the force of infection.

• We present projections up to August 31st.

1

Introduction

The COVID-19 pandemic has put an incredible pressure on modern society. While the global deathtoll of the virus has risen above 280,000 deaths on the 11th of May [1], several countries are lookingat strategies to cope with the virus on the medium to long term. As specific therapeutics aremissing and a vaccine will not be available in the short term, non-pharmaceutical interventionshave been proven effective in reducing the pressure on healthcare systems [2, 3, 4, 5, 6]. After amassive implementation of lockdown measures, affecting as much as one third of the global worldpopulation [7], governments have now started to relieve some of the social distancing measures.Belgium has started to do so on the 4th of May, with a plan for a gradual reopening over thenext month. The absence of substantial population immunity after this first wave of COVID-19in Belgium stresses the risk of subsequent waves when interventions are relaxed, requiring realtime monitoring and early warning signals to be in place. Here, we assess the impact of differentimplementations of re-openings, quantifying their effect on the epidemic in Belgium.

Model details

We adapted a previously developed meta-population model, that was validated for seasonal influenza[8], to study the COVID-19 outbreak in Belgium. The model reproduces the demography of children(0-18 years) and adults (19 years and above) in the different municipalities of Belgium. In oursimulation setting, people commute every day to their work/school municipality and interact withpeople therein, therefore acting as potential carriers of the infection. At the end of the day, they goback to their home municipality, possibly having been infected. In what follows we illustrate themain model features, and refer to [8] for more details.Data from a social contact survey conducted in Flanders, Belgium anno 2010-2011 [9] is usedto inform mixing patterns of the population. We account for intervention measures by reducingcontacts at specific locations i.e. home, work, school, transportation, leisure activity and other [10].Mobility data from the 2001 census is used to inform mobility fluxes due to school attendanceand work. When individuals commute, we assume that they establish half of their contacts in thepatch they commute to. In this regard, our model accounts for commuting mobility and not forshort-term trips. We use a stochastic SEIpIaIR model (Figure 1) in which we distinguish pre-symptomatic (Ip), asymptomatic (Ia), and symptomatic (Ims and Iss) transmission by assumingdifferent transmission rates, governed by different contact patterns. In particular, we assume thatsymptomatic individuals (both mildly symptomatic Ims and severely symptomatic Iss) reduce theirnumber of contacts (with values in line with the ones measured during the 2009 Influenza pandemic[11]) and their mobility (avoid commuting to work/school). A fraction of symptomatic individuals,mostly adults, can show severe symptoms and therefore require hospitalization, either in a hospitalward (W) or in intensive care unit (ICU). Once this happens, we assume that they cannot furtherinfect other people due to isolation measures. We assume that children have a lower susceptibility(50%) with respect to adults [4, 12]. Table 1 shows the model parameters for the disease progression.

2

Table 1: Model parameters.

Quantity Median (95% CI) Distribution Source

Latent period (ε) 1.4 days ([0:7] days) Exponential [13, 14]Pre-symptomatic period (θ) 2.4 days ([0:13] days) Exponential [13, 14]Children infectivity (wrt adults) 0.5 — [4, 15, 12, 16]Proportion asymptomatic (pa) 0.45 — [15, 16, 17, 18]Proportion mild symptoms (pm) 0.5/0.45 (children/adults) — [15, 12, 16, 17]Proportion severe symptoms (ps) 0.05/0.1 (children/adults) — [12, 16]Symptomatic/asymptomatic period (µ) 2.4 days ([0:13] days) Exponential [19]Symptom onset to hospitalization (σ) 4.7 days ([0:17] days) Chi-square [19]

Figure 1: Schematic representation of the compartmental model. Individuals start as sus-ceptible (S) and can become exposed to the disease (E) when interacting with infected individuals(Ip, Ia, Ims and Iss). After a latent period,exposed individuals enter a pre-symptomatic phase (Ip),after which they can either acquire symptoms (Ims and Iss) or not (Ia). Symptomatic individuals canhave mild symptoms (Ims) or severe symptoms (Iss). In the latter case, they are hospitalized, eitherin a ward (W ) or they are admitted to ICU (ICU). The final outcome of infected individuals is eitherrecovery (R) or death (D).

Interventions on 14/03/2020

To calibrate our model through the estimation of transmission parameters for the Belgian setting,we used national data on daily hospital admissions [20]. The calibration is done in two differentsteps:

• Pre-intervention phase: we used data on contact patterns and mobility from pre-pandemictime to fit the per-contact transmission probability to hospitalization data up to the 18th ofMarch (Fig. 2,a).

• Intervention phase: keeping the same epidemiological parameters of the pre-intervention pe-riod, we modulated population mixing and mobility to fit hospitalization data up to the 1stof May (Fig. 2,b).

Based on reported hospital admissions in Belgium during the lockdown period, we estimated a 100%reduction in school contacts; 90% contact reduction at work and during transport, leisure activitiesand in other locations and, on top of these reductions, a further reduction of 35% of contacts betweenadults. The overall social contact reduction, relative to the pre-lockdown behavior, is estimated tobe 80%.

3

Figure 2: Parameter estimation and model fitting. Panel a: calibration of the pre-interventionphase in terms of per-contact transmission probability. Data on hospital admissions is shown in com-parison with the best-fit model. Black points are fitted to calibrate the model in the pre-interventionphase. Panel b: calibration for the intervention phase. Data on hospital admissions is shown incomparison with the best-fit model.

Lifting lockdown

The Belgian government lifted the lockdown gradually from the 4th of May. Table 2 shows asummary of the different phases and their implementation. Changes with respect to the previousphase (i.e. the previous row) are shown in bold. In our scenario analysis we considered three phases:

• Phase 1: from the 4th of May, increasing the contacts made at work and during commuting byadults, to account for the increase of people going back to work. Mobility of adults increasesaccordingly. We also consider a small increase in contacts during leisure and in other locations(from 10% of the soft lockdown up to 20%) to account for increased general mixing of thepopulation.

• Phase 2: from the 18th of May, increasing contacts made at school and during commuting bychildren to account for school re-opening. Mobility of children increases accordingly.

• Phase 3: from the 8th of June, increasing contacts made during leisure and in other locations,to assess the impact of a possible re-opening of leisure activities.

Table 2: Timing and concepts of lockdown relief. Each phase is implemented incrementallyon the previous ones. Bold values highlight the changes with respect to the previous phase (i.e. theprevious row). Intermediate parameter values are reported, with the full range in bracket.

Timing Work School Mobility Mobility Leisurestart/end & transportation contacts (%) adults (%) children(%) & other

contacts (%) contacts (%)

Lockdown 14-03/04-05 10 0 10 0 10Phase 1 (work) 04-05/17-05 40 [20-60] 0 40 [20-60] 0 10 [10-20]Phase 2 (school) 18-05/07-06 40 [20-60] 40 [20-60] 40 [20-60] 40 [20-60] 10 [10-20]Phase 3 (leisure) 08-06/30-06 40 [20-60] 40 [20-60] 40 [20-60] 40 [20-60] 30 [20-40]Summer holidays 01-07/31-08 40 [20-60] 0 40 [20-60] 0 30 [20-40]

4

Figure 3 shows three examples of the simulated scenarios, corresponding to the two extreme casesof lowest/highest reduction of epidemic spread and a intermediate scenario, corresponding to pa-rameter values that are chosen as the intermediate value of each parameter interval. The outermostscenarios show that, depending on the exit strategy implemented, the number of new hospitaliza-tions on the 31st of August can vary between a few hundreds and more than 7000 (not shown inthe Figure, beyond y-axis value).

Figure 3: Predicted hospital admissions per day for three scenarios. In the highest/lowestparameter values scenarios, parameters are chosen to give the maximum/minimum reduction in numberof contacts for each phase of the exit strategy. In the intermediate parameters scenario parameters arechosen to be in the intermediate of the parameter range. Table 3 reports the corresponding parametervalues.

Figure 4 shows the impact of the different phases of the exit strategy. In Figure (4,a), on the 4thof May, contacts at work and on transportation are increased, ranging from 20% to 60% of pre-pandemic values. Phases 2 and 3 are implemented identically. As expected, there is a delay betweenthe implementation of the phase 1 of the exit strategy and its effect on the number of hospitaladmissions: after 2 weeks the number of hospital admissions stops to decrease as in the interventionphase. One further week is needed to see differences between the three implementations of phase 1.In Figure (4,b) we show the impact of phase 2 (school re-opening) once phase 1 is implemented forthe intermediate value of contacts at work/transportation (40%). The percentage of school contactsranges from 20% to 60%, while phase 3 is implemented identically in all three scenarios. In thiscase, the different curves start to diverge 5 weeks after the re-opening of schools. Summer schoolholidays, starting on the 1st of July have a considerable effect on the number of hospital admissionsonly in the 60% school contacts scenario. In Figure (4,c) we show the impact of phase 3, oncephase 1 and 2 are implemented with intermediate values of the considered parameters. Differentimplementations of the phase 3 give different results already after two weeks. This means that phase3 is the one with the largest potential impact, at least for the parameter values explored here.

Table 3: Parameters for the exit scenario analysis presented in Figure 3.

Scenario Work School Mobility Mobility Leisure& transportation contacts (%) adults (%) children(%) & othercontacts (%) contacts (%)

Highest param. values 60 60 60 60 40Intermediate param. values 40 40 40 60 30Lowest param. values 20 20 20 20 20

5

Figure 4: Exit scenarios using different timings and location-specific reductions. Panel a:work re-opening (phase 1); Panel b: school re-opening (phase 2); Panel c: leisure re-opening(phase 3). Only parameters for the considered phase differ while the others remain fixed to 40%of work/transportation contacts, 40% school contacts, 10% leisure/other contacts.

Case isolation

When extensive contact tracing and testing is available, a viable option for disease mitigation isto isolate (suspected) cases. By case isolation, we mean all the measures taken to trace infectedindividuals’ contacts, test them and isolate them to reduce further spread. Since all contacts ofinfected individuals are tested, case isolation is assumed to affect also asymptomatic individualsthat would otherwise slip out of the usual surveillance methods. We present our results in term ofa synthetic quantity, α, that is the percentage of individuals entering the symptomatic and asymp-tomatic classes (Ia, Ims and Iss) that are effectively isolated. We assume that these quarantinedindividuals reduce their contacts by a factor of ten. We do not cover here how to link the targetα to an optimal strategy for contact tracing and testing that takes into account test specifics andthe time of the testing [21, 22, 23]. We also assume that no isolation of pre-symptomatic people isimplemented (Ip). We considered that case isolation can start at the start of phase 2 (i.e. on the18th of May) or at the start of phase 3 (i.e. on the 8th of June), to assess the impact of delay inimplementation. Figure 5 shows the impact of case isolation for the scenario marked as “intermedi-ate” in Figure 3. The ability to isolate newly infected individuals has a considerable impact on thenumber of hospital admissions. The isolation of 25% of new cases is able to reduce the expectednumber of hospital admission at the end of August from 2500 to less than 1000. The isolationof twice as many cases (50% instead of 25%) would stop the growth of new hospital admissions.Starting case isolation 3 weeks after (phase 3 vs phase 2), in the 25% isolation scenario, would leadby the end of August to 250 more daily hospital admissions.

Discussion

We used a stochastic, discrete time, data-driven meta-population model to predict the impact oflifting the lockdown in phases. The model includes data on pre-pandemic mobility and mixing, andis calibrated on hospital admissions. The initial phase of COVID-19 spread in Belgium is charac-terized by a fast spread of the disease, with a doubling time of 3.09 days (95% CI [3.05 : 3.14]), inline with values from other European countries [5, 2, 24] and worldwide [25, 26]. Combined withour parameter choices, this results in an R0 = 3.18 (95% CI [3.13 : 3.22]), which lies within theinterval estimated in a recent meta-analysis (mean=2.6, standard deviation=0.54 [27]). Our model

6

Figure 5: Effect of case isolation. The number of hospital admissions is shown for three differentvalues of case isolation. The baseline scenario (no case isolation) is the intermediate parameter valuesscenario of Figure 4, which is detailed in table 3.

properly describes hospital admissions during the lockdown period if a strong reduction (80%) inthe number of contacts is established. In this situation the number of hospital admissions starts todecrease after 3 weeks from the start of the lockdown allowing the health system to cope with ICUdemands and the reproduction number is estimated to be 0.70 (95% CI [0.68 : 0.74]). Such a strongreduction in the average number of contacts marks the disruption that a lockdown has on everydaylife. Studies in Wuhan and Shanghai [4] found an even stronger reduction in the number of contactsduring lockdown, while a recent survey in the UK [27] measured a reduction of 75%. Preliminaryanalysis of social contact data collected in Belgium [28] after the lockdown shows similar resultscompared to [4] and [27], in line with our model results. Adherence to country-specific contact datais paramount, as intervention measures can vary substantially between countries, both in terms ofimplementation and in terms of compliance. Collecting country specific contact data during thedifferent stages of the epidemic (i.e. before, during and after intervention) is therefore of crucialimportance to adequately assess the impact of social distancing. However, our knowledge of contactpatterns before the COVID-19 crisis can be used to explore synthetic scenarios and to identify therelative impact of introducing social distancing in different locations. In the current analysis thisapproach was taken, whilst considering a plausible range of reductions in social contacts in differentcircumstances. According to our model, leisure activities have the largest potential impact on theepidemic profile: allowing leisure activity contacts to exceed 10% of pre-pandemic values duringphase 3, would lead to an excessive projected number of hospital admissions by mid July. This isconsistent with leisure/other contacts accounting for 25% to 40% of the total contacts people make,according to representative surveys [29, 30]. We observed less impact of school closure on hospitaladmissions in contrast to social mixing at work and during transport or leisure activities. First,as expected, school closure leads to observable effects only in those scenarios in which a consistentfraction (i.e. 40% or more) of school contacts are established in the population. If smaller valuesare considered, the relative contribution of school contacts to the overall spread of the disease isnegligible, and therefore no strong transition from school-term to holiday-term is observed. Second,as children have a much lower probability of being infected, symptomatic (and as such hospitalized)with respect to adults [26], increased diffusion among children increases the observed hospital admis-sions mostly indirectly, through the increase of infected adults. We tested in a sensitivity analysisa scenario in which children have the same susceptibility to the disease: in this case school closure

7

would have a larger impact on the number of infections, especially in the children age class. Therole of children is still unclear [31, 32] and, although their secondary attack rate in household [33],this is difficult to reconcile with the absence of a documented case of infection caused by a childbelow 10 in literature [15] nor to the low number of confirmed cases in children [12].In our results, isolation of newly infected individuals has a tremendous impact on epidemic mit-igation, allowing to re-establish social interactions while still ensuring epidemic containment. Inour 25% case isolation scenario, 25% of newly infected individuals reduce their spreading potentialby 90%; this already results in a massive reduction in the number of new hospitalizations. Westress here that although we quantified the reduction of spreading potential in terms of number ofcontacts, this may also come as a combination of different effects, like the use of antivirals in theearly phase of the infection [34]. Whether this 25% is achievable in practice, will depend on theinfrastructure put in place to identify infected individuals, trace their contacts and isolate them.Also, a fast setup is crucial: a 3 weeks delay in implementing case isolation leads to a considerableimpact on the number of new hospital admissions. As a fast and reliable contact tracing is of fore-most importance, several digital solutions have been proposed [35] to match the need for personalinformation with privacy concerns.Other models have been applied to the emergence of COVID-19 in Belgium, either specifically [36,37, 19, 38] or in multi-country applications [39]. Using different model paradigms allows to focus ondistinct aspects of the outbreak, like delay distributions of the clinical history of patients [19], in-cluding serology in the model input [37] or exploring individual-specific contact tracing options [36].When evaluating intervention strategies with profound societal impact, ideally different modelsshould be compared [40].In this work we impose a set of assumptions. As we consider just two age classes (children andadults) we can not fully capture age-specific effects of the clinical history of individuals. Althoughclinical severity is known to vary with age [41, 5], we opted for two age classes as age-specific dataon hospitalizations was not available when this model was developed. Age-specific hospitalizationcan be addressed in a future extension of this work. Also, we fitted the reduction of social inter-actions during the lockdown to data: even though this leads to reasonable results, it reduces theinterpretability of forecasted scenarios, as all situations have to be expressed in terms of the pre-pandemic situation. How contacts re-shape during social distancing depends on how interventionsare implemented and on population compliance, as well as on cultural habits. To overcome this lim-itation, social contact data collected during the lockdown in Belgium would be needed [28]. Lastly,we did not consider the impact of increased hygiene, nor the use of personal protective equipmentsuch as face masks [42] on the spreading potential of COVID-19. However, in our model this wouldtranslate into an overall reduction of the force of infection, that can be accounted for by a re-scalingof the percentages of contact reduction. While we consider (e.g.) a scenario with 75% reduction inthe number of contacts (i.e. a fourfold reduction), a model with 50% reduction in the number ofcontacts and 50% reduction in the transmissibility per contact would give similar results.In conclusion, our results show the predicted impact of a phase-based relieve of lockdown. Whileeconomical and societal needs urge government to relieve strict distancing measures and mobilityrestrictions, caution should be taken when evaluating different scenarios and re-assessing the situa-tion on a regular basis. Social distancing will still play an important role in the months to come andinfrastructure for extensive testing and contact tracing will increase the effectiveness of mitigationstrategies.

Websites with further information

www.simid.bewww.socialcontactdata.orgwww.uhasselt/dsi-covid19

8

Acknowledgements

This work received funding from the European Research Council (ERC) under the European Union’sHorizon 2020 research and innovation program (grant agreement 682540 TransMID). This projecthas received funding from the European Union’s Horizon 2020 research and innovation program- project EpiPose (No 101003688). LW received funding from the Research Foundation Flanders(1234620N). We gratefully acknowledge the Flemish Supercomputer Centre (VSC) and its staff. Wethank Giulia Pullano, Laura Di Domenico and Vittoria Colizza for useful discussions.

References

[1] Dong, E., Du, H., Gardner, L.: An interactive web-based dashboard to track COVID-19 in realtime. The Lancet Infectious Diseases (5), 533–534. doi:10.1016/S1473-3099(20)30120-1

[2] Di Domenico, L., Pullano, G., Sabbatini, C.E., Boelle, P.-Y., Colizza, V.: Ex-pected impact of lockdown in ıle-de-france and possible exit strategies (2020).doi:10.1101/2020.04.13.20063933

[3] Prem, K., Liu, Y., Russell, T.W., Kucharski, A.J., Eggo, R.M., Davies, N., Flasche, S., Clifford,S., Pearson, C.A.B., Munday, J.D., Abbott, S., Gibbs, H., Rosello, A., Quilty, B.J., Jombart,T., Sun, F., Diamond, C., Gimma, A., van Zandvoort, K., Funk, S., Jarvis, C.I., Edmunds,W.J., Bosse, N.I., Hellewell, J., Jit, M., Klepac, P.: The effect of control strategies to reducesocial mixing on outcomes of the COVID-19 epidemic in Wuhan, China: a modelling study.The Lancet Public Health, 261–270 (2020). doi:10.1016/S2468-2667(20)30073-

[4] Zhang, J., Litvinova, M., Liang, Y., Wang, Y., Wang, W., Zhao, S., Wu, Q., Merler, S., Viboud,C., Vespignani, A., Ajelli, M., Yu, H.: Changes in contact patterns shape the dynamics of thecovid-19 outbreak in china (2020). doi:10.1126/science.abb8001

[5] Riccardo, F., Ajelli, M., Andrianou, X., Bella, A., Del Manso, M., Fabiani, M., Bellino, S.,Boros, S., Mateo Urdiales, A., Marziano, V., Rota, M.C., Filia, A., D’Ancona, F.P., Siddu, A.,Punzo, O., Trentini, F., Guzzetta, G., Poletti, P., Stefanelli, P., Castrucci, M.R., Ciervo, A.,Di Benedetto, C., Tallon, M., Piccioli, A., Brusaferro, S., Rezza, G., Merler, S., Pezzotti, P.,: Epidemiological characteristics of covid-19 cases in italy and estimates of the reproductivenumbers one month into the epidemic (2020). doi:10.1101/2020.04.08.20056861

[6] Ferguson, N.M., Laydon, D., Nedjati-Gilani, G.e.a.: Impact of non-pharmaceutical interven-tions (npis) to reduce covid-19 mortality and healthcare demand (2020). doi:10.25561/77482.https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/

gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf

[7] Coronavirus: The world in lockdown in maps and charts. https://www.bbc.com/news/

world-52103747

[8] Luca, G.D., Kerckhove, K.V., Coletti, P., Poletto, C., Bossuyt, N., Hens, N., Colizza, V.: Theimpact of regular school closure on seasonal influenza epidemics: a data-driven spatial trans-mission model for belgium. BMC Infectious Diseases 18(1), 1–16 (2018). doi:10.1186/s12879-017-2934-3

[9] Willem, L., van Kerckhove, K., Chao, D.L., Hens, N., Beutels, P.: A Nice Day for an Infection?Weather Conditions and Social Contact Patterns Relevant to Influenza Transmission. PLoSONE 7(11) (2012). doi:10.1371/journal.pone.0048695

9

[10] Willem, L., Hoang, T.V., Funk, S., Coletti, P., Beutels, P., Hens, N.: Socrates: An online toolleveraging a social contact data sharing initiative to assess mitigation strategies for covid-19(2020). doi:10.1101/2020.03.03.20030627

[11] Van Kerckhove, K., Hens, N., Edmunds, W.J., Eames, K.T.D.: The Impact of Illness onSocial Networks: Implications for Transmission and Control of Influenza. American Journal ofEpidemiology 178(11), 1655–1662 (2013). doi:10.1093/aje/kwt196

[12] Ludvigsson, J.F.: Systematic review of covid-19 in children shows milder cases and a betterprognosis than adults. Acta Paediatrica 109(6), 1088–1095. doi:10.1111/apa.15270

[13] Lauer, S.A., Grantz, K.H., Bi, Q., Jones, F.K., Zheng, Q., Meredith, H.R., Azman, A.S.,Reich, N.G., Lessler, J.: The Incubation Period of Coronavirus Disease 2019 (COVID-19)From Publicly Reported Confirmed Cases: Estimation and Application. Annals of InternalMedicine (2020). doi:10.7326/M20-0504

[14] Guan, W.-j., Ni, Z.-y., Hu, Y., Liang, W.-h., Ou, C.-q., He, J.-x., Liu, L., Shan, H., Lei, C.-l.,Hui, D.S.C., Du, B., Li, L.-j., Zeng, G., Yuen, K.-Y., Chen, R.-c., Tang, C.-l., Wang, T., Chen,P.-y., Xiang, J., Li, S.-y., Wang, J.-l., Liang, Z.-j., Peng, Y.-x., Wei, L., Liu, Y., Hu, Y.-h.,Peng, P., Wang, J.-m., Liu, J.-y., Chen, Z., Li, G., Zheng, Z.-j., Qiu, S.-q., Luo, J., Ye, C.-j., Zhu, S.-y., Zhong, N.-s.: Clinical characteristics of coronavirus disease 2019 in china. NewEngland Journal of Medicine 382(18), 1708–1720 (2020). doi:10.1056/NEJMoa2002032

[15] Boast, A., Munro, G.H. A and: An evidence summary of paediatric covid-19 literature (2020). doi:10.31440/DFTB.24063. https://dontforgetthebubbles.com/

evidence-summary-paediatric-covid-19-literature/

[16] WHO question & answer. https://www.who.int/news-room/q-a-detail/

q-a-similarities-and-differences-covid-19-and-influenza

[17] Day, M.: Covid-19: identifying and isolating asymptomatic people helped eliminate virus initalian village 368 (2020). doi:10.1136/bmj.m1165

[18] Nishiura, H., Kobayashi, T., Suzuki, A., Jung, S.M., Hayashi, K., Kinoshita, R., Yang, Y.,Yuan, B., Akhmetzhanov, A.R., Linton, N.M., Miyama, T.: Estimation of the asymptomaticratio of novel coronavirus infections (COVID-19). International journal of infectious diseases: IJID : official publication of the International Society for Infectious Diseases 94, 154–155(2020). doi:10.1016/j.ijid.2020.03.020

[19] C. Faes et al.: From symptom onset to hospitalisation to recovery or death: a statistical analysisof different time-delay distributions in belgian covid-19 patients. In preparation (2020)

[20] Sciensano, Covid-19 Belgium Epidemiological Situation. https://datastudio.google.com/embed/u/0/reporting/c14a5cfc-cab7-4812-848c-0369173148ab/page/tpRKB

[21] Hellewell, J., Abbott, S., Gimma, A., Bosse, N.I., Jarvis, C.I., Russell, T.W., Munday, J.D.,Kucharski, A.J., Edmunds, W.J., Sun, F., Flasche, S., Quilty, B.J., Davies, N., Liu, Y., Clifford,S., Klepac, P., Jit, M., Diamond, C., Gibbs, H., van Zandvoort, K., Funk, S., Eggo, R.M.:Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. The LancetGlobal Health 8(4), 488–496 (2020). doi:10.1016/S2214-109X(20)30074-7

[22] Kucirka, L.M., Lauer, S.A., Laeyendecker, O., Boon, D., Lessler, J.: Variation in false-negativerate of reverse transcriptase polymerase chain reaction–based sars-cov-2 tests by time sinceexposure. Annals of Internal Medicine. doi:10.7326/M20-1495

10

[23] Kucharski, A.J., Klepac, P., Conlan, A., Kissler, S.M., Tang, M., Fry, H., Gog, J., Edmunds,J., : Effectiveness of isolation, testing, contact tracing and physical distancing on reducingtransmission of sars-cov-2 in different settings (2020). doi:10.1101/2020.04.23.20077024

[24] Jit, M., Jombart, T., Nightingale, E.S., Endo, A., Abbott, S., for Mathematical Mod-elling of Infectious Diseases COVID-19 Working Group, L.C., Edmunds, W.J.: Estimatingnumber of cases and spread of coronavirus disease (covid-19) using critical care admissions,united kingdom, february to march 2020. Eurosurveillance 25(18) (2020). doi:10.2807/1560-7917.ES.2020.25.18.2000632

[25] Zhao, S., Lin, Q., Ran, J., Musa, S.S., Yang, G., Wang, W., Lou, Y., Gao, D., Yang, L.,He, D., Wang, M.H.: Preliminary estimation of the basic reproduction number of novelcoronavirus (2019-ncov) in china, from 2019 to 2020: A data-driven analysis in the earlyphase of the outbreak. International Journal of Infectious Diseases 92, 214–217 (2020).doi:10.1016/j.ijid.2020.01.050

[26] Liu, Y., Gayle, A.A., Wilder-Smith, A., Rocklov, J.: The reproductive number of COVID-19 is higher compared to SARS coronavirus. Journal of Travel Medicine 27(2) (2020).doi:10.1093/jtm/taaa021. taaa021

[27] Jarvis, C.I., Zandvoort, K.V., Gimma, A., Prem, K., working Group, C.C.-., Klepac, P., Rubin,G.J., Edmunds, W.J.: Quantifying the impact of physical distance measures on the transmis-sion of COVID-19 in the UK. BMC Medicine (2020). doi:10.1186/s12916-020-01597-8

[28] The SIMID COVID19 task-force: Social contact survey in belgium during covid-19 pandemic.In preparation (2020)

[29] Mossong, J., Hens, N., Jit, M., Beutels, P., Auranen, K., Mikolajczyk, R., Massari, M., Salmaso,S., Tomba, G.S., Wallinga, J., Heijne, J., Sadkowska-Todys, M., Rosinska, M., Edmunds, W.J.:Social contacts and mixing patterns relevant to the spread of infectious diseases. PLOS Medicine5(3), 1–1 (2008). doi:10.1371/journal.pmed.0050074

[30] Hoang, T., Coletti, P., Melegaro, A., Wallinga, J., Grijalva, C.G., Edmunds, J.W.,Beutels, P., Hens, N.: A Systematic Review of Social Contact Surveys to InformTransmission Models of Close-contact Infections. Epidemiology 30(5), 723–736 (2019).doi:10.1097/EDE.0000000000001047

[31] Jones, T.C., Mu hlemann, B., Veith, T., Zuchowski, M., Hofmann, J., Stein,A., Edelmann, A., Corman, V.M., Drosten, C.: An analysis of sars-cov-2 vi-ral load by patient age (2020). https://zoonosen.charite.de/fileadmin/user_

upload/microsites/m_cc05/virologie-ccm/dateien_upload/Weitere_Dateien/

analysis-of-SARS-CoV-2-viral-load-by-patient-age.pdf

[32] Held, L.: A discussion and reanalysis of the results reported in jones et al (2020): ”an analysisof sars-cov-2 viral load by patient age”. https://osf.io/bkuar/

[33] Bi, Q., Wu, Y., Mei, S., Ye, C., Zou, X., Zhang, Z., Liu, X., Wei, L., Truelove, S.A., Zhang, T.,Gao, W., Cheng, C., Tang, X., Wu, X., Wu, Y., Sun, B., Huang, S., Sun, Y., Zhang, J., Ma,T., Lessler, J., Feng, T.: Epidemiology and transmission of COVID-19 in 391 cases and 1286of their close contacts in Shenzhen, China: a retrospective cohort study. The Lancet InfectiousDiseases 3099(20), 1–9 (2020). doi:10.1016/S1473-3099(20)30287-5

[34] Torneri, A., Libin, P.J.K., Vanderlocht, J., Vandamme, A.-M., Neyts, J., Hens, N.:A prospect on the use of antiviral drugs to control local outbreaks of covid-19 (2020).doi:10.1101/2020.03.19.20038182

11

[35] Ferretti, L., Wymant, C., Kendall, M., Zhao, L., Nurtay, A., Abeler-Dorner, L., Parker, M.,Bonsall, D., Fraser, C.: Quantifying sars-cov-2 transmission suggests epidemic control withdigital contact tracing 368(6491) (2020). doi:10.1126/science.abb6936

[36] L. Willem et al.: The impact of contact tracing and household bubbles on covid-19 transmissionin belgium. In preparation (2020)

[37] S. Abrams et al. In preparation (2020)

[38] Covid-19 outbreak modelling and control (2020)

[39] Russell, T., Hellewell, J., Abbott, S., Golding, N., Gibbs, H., Jarvis, C., van Zandvoort,K., nCov working group”, C., Flasche, S., Eggo, R., JW, E., Kucharski, A.: Using a delay-adjusted case fatality ratio to estimate under-reporting (2020). https://cmmid.github.io/topics/covid19/global_cfr_estimates.html

[40] Den Boon, S., Jit, M., Brisson, M., Medley, G., Beutels, P., White, R., Flasche, S.,Hollingsworth, T.D., Garske, T., Pitzer, V.E., Hoogendoorn, M., Geffen, O., Clark, A., Kim,J., Hutubessy, R.: Guidelines for multi-model comparisons of the impact of infectious diseaseinterventions. BMC Medicine 17(1), 1–13 (2019). doi:10.1186/s12916-019-1403-9

[41] Wu, J.T., Leung, K., Bushman, M., Kishore, N., Niehus, R., de Salazar, P.M., Cowling, B.J.,Lipsitch, M., Leung, G.M.: Estimating clinical severity of COVID-19 from the transmissiondynamics in Wuhan, China. Nature Medicine 26(4), 506–510 (2020). doi:10.1038/s41591-020-0822-7

[42] Leung, N.H.L., Chu, D.K.W., Shiu, E.Y.C., Chan, K.-H., McDevitt, J.J., Hau, B.J.P., Yen,H.-L., Li, Y., Ip, D.K.M., Peiris, J.S.M., Seto, W.-H., Leung, G.M., Milton, D.K., Cowling,B.J.: Respiratory virus shedding in exhaled breath and efficacy of face masks. Nature Medicine26(May) (2020). doi:10.1038/s41591-020-0843-2

12