The geography of COVID-19 spread in Italy and implications for the relaxation of confinement measures ===================================================================================================== * Enrico Bertuzzo * Lorenzo Mari * Damiano Pasetto * Stefano Miccoli * Renato Casagrandi * Marino Gatto * Andrea Rinaldo ## Abstract We examine the spread of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) in Italy, to address the appropriate methodological choices for the design of selective relaxations of the current containment measures. Pressing relevance stems from the need to restart the economy dramatically affected by the lockdown. We employ a spatially explicit, data-intensive model of the patterns of disease spread in Italy, with the goal of providing tools to: estimate the baseline trajectory, i.e. the expected unfolding of the outbreak if the current containment measures were kept in place indefinitely; assess possible deviations from the baseline, should relaxations of the current lockdown result in increased disease transmission; and estimate the isolation effort required to prevent a resurgence of the outbreak. For instance, a 50% increase in effective transmission as a result of the loosening of confinement measures, to be instated on May 4, yields an epidemic curve that shows a major rebound larger than the previous peaks in most regions. A control effort, capable of isolating a daily percentage of approximately 7% of the individuals in the exposed and pre-symptomatic stages, proves necessary to counterbalance such an increase, and maintain the epidemic curve onto the decreasing baseline trajectory. We explore several scenarios, provide the basic data to design the related control strategies and discuss their feasibility. Should suitable control via tracing and testing prove unfeasible, stop-and-go enforcement or delay of the lockdown relaxations would be necessary to reduce the isolation effort required to maintain the epidemic trajectory under control. ## Introduction While the pandemic caused by SARS-CoV-2 is still ravaging most countries of the world and containment measures are implemented worldwide, a debate is emerging on whether these measures might be partially alleviated — and in case how and when.1,2 This discussion requires appropriate models that guide decision-makers through alternative actions via scenarios of the related trajectories of the epidemic. The setup of country-wise epidemiological models is particularly challenging for SARS-CoV-2 owing to submerged infections3 and to the marked spatial heterogeneity of the epidemic spread. For example, in Italy, where the (largely underestimated) reported infections and deaths are respectively 203 591 and 27 682 as of April 29, the latitudinal characters of the spread of infections showed marked delays in the beginning of the local outbreaks.4 Health-policy and science underpin the design of suitable containment strategies, which include individual and collective (local and medium- to long-distance) mobility limitations, provision of personal protective equipment (PPE), massive, possibly targeted identification of infectious cases, and the setup of layers of administrative and environmental engineering controls.5 These strategies must consider the level of effective connectivity between communities and the different epidemiological parameters that characterize them. We elaborate our analysis based on a recently published, spatially explicit model of the COVID-19 spread in Italy, inclusive of mobility among communities, the timing of infection seeding, mobility restrictions and social distancing.4 The model is a spatial system that solves in time and for each of the 107 Italian provinces the balance of, and the coupled fluxes among, several epidemiological compartments in which the total population of a community is subdivided, namely individuals who are susceptible, exposed, pre-symptomatic infectious, asymptomatic/mildly symptomatic, infected with heavy symptoms, and recovered.4 Local communities are each described by a compartmental model, and are connected via mobility fluxes of individuals from the mobile epidemiological compartments (susceptible, exposed, pre-symptomatic, asymptomatic/mildly symptomatic and recovered individuals). Thus, the force of infection of each community depends not only on the local epidemiological variables, but also on the epidemiological variables of the connected communities. Infections, therefore, not only do occur within each community, but can also be imported or exported. In addition, the model accounts for infections occurring because individuals of different communities meet in a third location, because e.g. they work or study in the same place. The relative balance of the fluxes among the various compartments is regulated by process parameters that are estimated in a Bayesian framework using reconstructed data about the number of daily hospitalized cases in all Italian provinces from February 24 to April 24. All relevant technical details are provided as Supplementary Material (SM). ## Results and Discussion If current lockdown measures are maintained indefinitely, the epidemic curve would continue to decrease in all Italian regions (figure 1), although at different rates. We report for convenience daily hospitalization counts aggregated for administrative regions, even though the model accounts for finer spatial granularity. ![Figure 1:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F1.medium.gif) [Figure 1:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F1) Figure 1: Clockwise from top: Daily numbers of newly hospitalized cases respectively for Italy and its hardest hit regions (Lombardia, Piemonte, Marche, Veneto and Emilia Romagna). The remaining regions are shown as SM. Red empty dots represent data. The blue solid line represents the baseline scenario, i.e. the median of the computed results with current restriction measures maintained indefinitely beyond May 3, 2020. The green and red solid lines represent the scenarios corresponding to a release of the containment measures determining an effective increase in the overall transmission rates of respectively 10 % and 20 %. The 95 % confidence intervals are color-coded in analogy to their median scenarios. Parameter values are reported as SM. The lockdown in Italy might be relaxed on May 4. How would the modes of relaxation of current confinement measures affect residual epidemic trajectories? For example, one must account for the fact that only some activities are announced to be allowed. In addition, acquired awareness may have different lasting effects on social behaviour regardless of imposed measures, and the use of PPE may change through time. Here, we propose to assess the actual increase in overall transmission by tracking the departure of the epidemic curve from the one projected by using the transmission rate achieved during the lockdown (the baseline scenario, blue curve in figure 1). We estimate that an effective increase of 10% of transmission, stemming from different combinations of the above factors, is expected to yield a less pronounced decline of the daily new hospitalization cases in most regions (figure 1). A 20% increase would instead determine a rebound of the epidemic in all regions but Lombardia. This regional differential response depends on the current level of transmission and, marginally, on the differential prevalence of the susceptible population, which is lower in the regions that were more severely hit by the first phase of the outbreak. A deviation of incoming data from the baseline in figure 1 towards an unacceptable epidemic trajectory (say, the red curve in figure 1) should raise an alert and call for control action. Matching the right scenario in real time may be achieved by data assimilation and ensemble Kalman filtering.6 Control may consist in either re-tightening of the containment measures, possibly of the stop-and-go type,2 or alternative interventions. While the strategy adopted during the first phase of the outbreak mostly relied on the isolation and treatment of symptomatic cases, a different mix of interventions is possible and desirable. A keystone of such a mix should be an increased isolation effort through tracing, and possibly testing,7 individuals who have been in close contact with a known infection,8,9 possibly with the help of technological advances like tracing apps.6 Clinical and epidemiological evidence suggests that viral shedding peaks at the end of the latent period, and that shedding rapidly declines after the symptoms’ onset or the evolution towards an asymptomatic case.10 Moreover, viral shedding is similar regardless of the emergence of symptoms in the disease course of a patient.3 This evidence suggests that isolation is much more effective if targeted at individuals in a latent stage (exposed, *E*, and pre-symptomatic, *P*, compartments of the model,4 see SM). We therefore focus on these individuals and estimate the percentage and the corresponding number of individuals that should be isolated daily (figure 2), to counterbalance the increase in transmission due to loosening the containment measures, thus maintaining the epidemic curve in the decreasing trajectory achieved during the lockdown (blue lines in figure 1). ![Figure 2:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F2.medium.gif) [Figure 2:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F2) Figure 2: The plots indicate the daily number (left scale) and the daily percentage (right scale) of latent individuals (exposed *E* and pre-symptomatic *P* compartments) to be isolated to maintain the epidemic trajectory onto the blue curve in figure 1 (corresponding to the current lockdown) despite the possible increase in transmission induced by the actual release of restrictions (horizontal axes). Solid lines refer to median values, shaded areas to the 95 % (lighter blue) and 50 % confidence intervals. In each panel, median and 95 % confidence interval of *E*, *P* and new daily symptomatic cases (C) are given (estimates refer to May 4). The other Italian regions are shown as SM. ![Figure 3:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F3.medium.gif) [Figure 3:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F3) Figure 3: Epidemic scenarios for the regions not shown in figure 1. ![Figure 4:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F4.medium.gif) [Figure 4:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F4) Figure 4: Same as figure 1 but with a increase in transmission equal to 50 %. ![Figure 5:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F5.medium.gif) [Figure 5:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F5) Figure 5: Isolation target for the regions not shown in figure 2. ![Figure 6:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F6.medium.gif) [Figure 6:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F6) Figure 6: Isolation target as shown in figure 2 but for a starting date of the relaxation of the restrictive measures delayed by one month. ![Figure 7:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/05/2020.04.30.20083568/F7.medium.gif) [Figure 7:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/F7) Figure 7: Isolation target assuming a symptomatic fraction *σ* = 0.1. While the daily percentage of latent individuals that must be isolated (right axes) is similar to the case shown in the main text (figure 2), the abundance of exposed and pre-symptomatic individuals is more than two-fold, and so the isolation target in terms of number of individuals (left axes). To complement the above information, figure 2 reports the estimated abundance of exposed and pre-symptomatic individuals in the considered regions at the date of the announced new measures (May 4), along with the expected number of new daily symptomatic cases (C) predicted by the model. As an example, in Lombardia an increase in transmission of 50 % would quickly lead to a ramping epidemic curve (SM). However, daily isolation of about ~ 1800 out of ~ 26000 (~ 7 %) exposed and pre-symptomatic individuals would effectively counterbalance the increase in transmission and bring back the curve to the baseline. The estimated number of latent individuals that needs to be isolated every day is of the same order of magnitude of the daily new symptomatic cases that are likely to self-report to the health structures (~ 1031). Therefore, the isolation effort should prioritize all the close contacts that the daily new symptomatic cases have had in the previous two days (i.e. the peak of viral shedding10). As infected individuals might not immediately test positive, and because obtaining test results takes time, this strategy might imply to precautionary isolate, at least temporarily, all close contacts. The percentage of these contacts that are actually infected, and thus contribute to reaching the required isolation target, depends on the secondary attack rate. The latter needs to be carefully evaluated depending on the context where the contacts occurred, and on the new conditions upon the relaxation of the confinement measures with increased people awareness and the use of PPE. Should targeting all primary contacts of all new daily symptomatic cases prove insufficient to reach the required target number, secondary contacts would also need to be targeted. In this case, testing primary contacts would help identifying actually infected cases, thus refining the tracing of secondary contacts.7 If the isolation target proves impossible to achieve for the limits of resources and/or logistical reasons, a possible strategy may consist in delaying the relaxation of the confinement measures. To that end, we note that delaying by an additional month the release of containment measures would more than halve the abundance of exposed and pre-symptomatic individuals, thus proportionally reducing the number of individuals who need to be isolated (results shown as SM). Results shown here are derived by assuming that 75 % of infections are mildly symptomatic or asymptomatic,4 which is consistent with empirical evidence.3,6 Such assumption leads to an infection fatality ratio (IFR) of 1·5 %. We repeated the analysis (including parameter estimation) by assuming 90 % of asymptomatic/mildly symptomatic infections, which leads to an IFR of 0·5 %. These two values bracket the available estimates of IFR for western countries.11 While the daily percentage of latent individuals that must be isolated in the two cases is similar, the abundance of exposed and pre-symptomatic individuals in the latter scenario is more than twofold, and so is the isolation target in terms of the number of individuals (SM). The trajectories shown in figure 1 are robust with respect to the assumed asymptomatic fraction, at least for the relatively short projection horizon considered here. ## Conclusions Health policy decisions concerning the relaxation of current containment measures of the COVID-19 epidemic in Italy imply decisions focused on the expected social and economic benefits. However, costs are hard to put a price tag on, because they depend on the macroscopic effects of the relaxations measures, stemming from their real-life application in terms of PPE adoption, social distancing, potential and realized mobility patterns, evolution of alertness about precautionary measures. Because proper choices of cost/benefit ratios are necessary to decide key health policy actions, epidemiological scenarios are necessary to provide at any time a reasoned estimate of the actual number of local infections based on the modelling scenario that best fits the early signs of local revamping of the epidemic. To that end, monitoring on a daily basis the number of new hospitalizations is essential to assimilate data with suitable spatial granularity. Interventions include the reinstatement of human mobility while enforcing the strict adoption of precautions and hygiene measures. Taken together, they reflect an overall exposure hardly predictable *a priori*. To bring the epidemic curve back on track (e.g. near the baseline trajectory observed prior to relaxation) health policy makers should consider a mix of interventions that may include the re-tightening of confinement measures, possibly by a stop-and-go implementation based also on seasonality effects, or evidence on limits to the acquired immunity,2 and/or the effective isolation of infectious individuals.1,6 Modeling studies can provide reasoned estimates of the minimum target to be attained, in this study through the latters. The proper strategy to achieve the isolation target is the domain of public health policy. The complementary use of testing in the control strategy is instead the domain of virology and epidemiology. To both domains, proper modelling scenarios offer information otherwise unavailable. Our results suggest that each Italian region should carefully evaluate its current tracing, testing and isolation capacity in order to plan and manage the second phase of the epidemic. ## Data Availability All data used in this manuscript are publicly available. COVID-19 epidemiological data for Italy are available at [https://github.com/pcm-dpc/COVID-19](https://github.com/pcm-dpc/COVID-19). Mobility data at municipality scale are available at [https://www.istat.it/it/archivio/139381](https://www.istat.it/it/archivio/139381). Population census data are available at [http://dati.istat.it/Index.aspx?QueryId=18460](http://dati.istat.it/Index.aspx?QueryId=18460). [https://github.com/pcm-dpc/COVID-19](https://github.com/pcm-dpc/COVID-19) [https://www.istat.it/it/archivio/139381](https://www.istat.it/it/archivio/139381) [http://dati.istat.it/Index.aspx?QueryId=18460](http://dati.istat.it/Index.aspx?QueryId=18460) ## Contributors EB, MG and AR were responsible for conceiving the work. EB, DP and LM were responsible for numerical simulations and model parameter estimation. All authors were responsible for data analysis and statistics and writing the manuscript. ## Declarations of interest We declare no competing interests. ## Supplementary Material ### Epidemiological model #### The original *SEPIA* model Here, we use the model *SEPIA*4 which includes the following compartments: Susceptible (*S*), Exposed (*E*), Pre-symptomatic (*P*), Infected with heavy symptoms (*I*), Asymptomatic/mildly symptomatic (*A*), Hospitalized (*H*), Quarantined at home (*Q*), Recovered (*R*) and Dead (*D*) individuals. The local dynamics of transmission is given by: ![Formula][1] Susceptible individuals (*S*) become exposed to the viral agent by contacting individuals who are in each of the three infectious stages: pre-symptomatic, heavily symptomatic or asymptomatic/mildly symptomatic. Frequency-dependent contact rates are assumed, so that exposure is governed by the following force of infection ![Formula][2] where *βP*, *βI*, and *βA* are the stage-dependent transmission rates. Exposed individuals (*E*) are latently infected, until they enter the pre-symptom stage (at rate *δE*). Pre-symptomatic individuals (*P*) progress (at rate *δP*) to become (with probability *σ*) either symptomatic individuals with heavy symptoms (*I*) or (with probability 1 − *σ*) asymptomatic/mildly symptomatic individuals (*A*). Heavily symptomatic infectious individuals (*I*) exit their compartment if/when (a) they are isolated from the community (at rate *η*) because they are hospitalized (a fraction 1 − *ζ*) or quarantined at home (a fraction *ζ*), (b) recover from infection (at rate *γI*), or (c) die (at rate *αI*). Asymptomatic/mildly symptomatic individuals (*A*) leave instead their compartment after having recovered from infection (at rate *γA*). Hospitalized individuals (*H*) may either recover from infection (at rate *γH*) or die (at rate *αH*). Quarantined (i.e., home-isolated) individuals (*Q*) leave their compartment upon recovery (at rate *γQ*). People recovering from infection or dying because of COVID-19 populate the classes of recovered (*R*) and dead (*D*) individuals, respectively. #### The effect of isolation Following lockdown release, the expected increase in the transmission rates can be compensated by isolation of cases. We argue in the main text that isolation is more effective if targeted to exposed and pre-symptomatic phases. Therefore, when isolation is enforced, two out-fluxes, *ρEE* and *ρPP*, must be considered from the exposed and pre-symptomatic compartments, respectively. The parameters *ρE* and *ρ*P (days−1) represent the rate at which infected individuals in the *E* and *P* classes are effectively isolated from the community. For the sake of simplicity, we assume *ρ*E = *ρ*P = *ρ*. Also, individuals isolated are simply removed from the community, without any further consideration of their clinical trajectories. #### Spatial dynamics We couple the *n* local communities via a community-dependent force of infection that incorporate human mobility at a geographically suitable resolution. The force of infection for community *i* is given by ![Formula][3] where ![Graphic][4] (with *X* ∈ {*S,E,P,I,A,R*}) is the probability ![Graphic][5] for all *i*’s and *X*’s) that individuals in epidemiological state *X* who are from community *i* enter into contact with individuals who are present at community *j* as either residents or because they are traveling there from community *k* (note that *i*, *j* and *k* may coincide). Spatial coupling is parameterized by using information from the latest nation-wide assessment of mobility fluxes, which was produced by the Italian National Institute of Statistics (ISTAT) in 2011 (data available at [https://www.istat.it/it/archivio/139381](https://www.istat.it/it/archivio/139381)). For each second-level administrative unit (province), say i, two quantities are extracted from the ISTAT data, namely the fraction *pi* of mobile people, i.e. the residents of *i* who defined themselves as commuters, and the fraction *qij* of mobile people between *i* and all other administrative units *j =* 1… *n* (including *j = i*). The contact probabilities ![Graphic][6] are then defined based on the quantities *pi* and *qij*. Specifically, we assume ![Formula][7] where the parameter *rX* (0 ≤ *rX* ≤ 1) describes the fraction of contacts occurring while individuals in epidemiological compartment *X* are traveling. In other words, for community *i*, the social contacts of non-mobile people (a fraction 1 − *pi* of the community size), those of mobile people that do not occur during travel (a fraction 1 − *rX* of total contacts for people in epidemiological compartment *X*) and those associated with mobility for people who travel within their community (a fraction *qii* of mobile people) contribute to social mixing within the community. Conversely, the contacts occurring between two different communities, say *i* and *j*, are a fraction *rX* of the total contacts of the individuals in epidemiological compartment *X*, multiplied by the probability *pi* that people from *i* travel (independently of the destination) and the probability *qij* that the travel occurs between *i* and *j*. ### Epidemiological data For the calibration of the model we consider the epidemiological data collected by the Dipartimento della Protezione Civile (data available at [https://github.com/pcm-dpc/COVID-19](https://github.com/pcm-dpc/COVID-19)), which are released daily and comprehend: at the regional level, the cumulative numbers of positive, dead and recovered individuals, together with the actual number of positive individuals that are severe symptomatic (hospitalized) or are under quarantine at home; at the province level, the cumulative numbers of positive cases. Due to the strong space-time variations in the number of tests performed, the most trustworthy variable to monitor the outbreak is the daily number of hospitalizations individuals, in the following indicated with *H*in. This quantity corresponds to the flux (1 − *ζ*)*ηI* in the *H* compartment of our model, and grants a straightforward link between data and model variables. However, *H*in is not directly provided in the online data, and we thus adopt a stochastic approach to derive *H*in combining data regarding the number of hospitalized individuals and deaths, and estimated distribution of delays between hospitalization and death or discharge. At any given day *k*, ![Graphic][8] is obtained by the observed variations in the daily number of hospitalized individuals, *Hk* − *Hk*−1, plus the daily deaths ![Graphic][9], and the number of individuals discharged from the hospital, here indicated with ![Graphic][10]. Under the assumption that the recorded deaths for COVID19 are all from the hospitals, ![Graphic][11] is obtained by modeling as random variables the days *τ* spent in a hospital before death, whose probability density function (PDF) is indicated with *pD*(*τ*), and the time in a hospital before discharge, whose PDF is indicated with *pR*(*τ*). Our procedure consists of the following steps. Sampling a random value from *pD*(*τ*) for each individual in *D*out, we obtain the days of entrance of individuals that will die, thus the sequence ![Graphic][12]. We estimate the number of individuals entering on day *k* that will be eventually discharged as: ![Formula][13] Then ![Graphic][14] is obtained by sampling an exit time from *pR* (*τ*) for each individual in ![Graphic][15]. Finally: ![Formula][16] Reports by ISS indicate that for COVID-19 casualties the median residence time at a hospital is about 5 days.12 We use this information to parameterize the distribution *pD*(*τ*) as an exponential function of mean 5 (hence, a median of 3·5 days; 0·05–0·95 quantiles: 1·4–15·0 days). We also assume that *pr* (*τ*) follows a gamma distribution with shape parameter 14 and scale 1, which has a median of 13·7 days (0·05–0·95 quantiles: 8·5–20·7 days), in agreement with the recovery rate previously estimated.4 Final data adopted for parameter estimation of the regional model is the median over 100 random generations of ![Graphic][17], downscaled to the province level and smoothed by using a moving average of 7 days. A sensitivity analysis of ![Graphic][18] on the parameters of the *pD* (*τ*) and *pR*(*τ*) showed that the time series considered have only marginal variations. ### Parameter estimation The effect of the containment measures was parameterized in our previous application4 by assuming that the transmission parameters (*βP*, *βI* and *βa*) had a sharp decrease after the containment measures announced on February 24 and March 8. We update here such description to fully account for the set of progressively more restrictive measures that were introduced form March 8 to March 22, when also industrial and production activities were stopped. We describe the temporal changes in the *βP*’s (the remaining transmission parameters, *βI* and *βa*, are assumed to be proportional to *βP*, see table 1) using 4 values: The value before February 24 ![Graphic][19], the values achieved right after (within two days) the measures introduced on February 24 ![Graphic][20] and the first set of lockdown measures implemented on March 8 ![Graphic][21]. Finally, we assume that due to the progressive implementation of the lockdown and the introduction of more restrictive measures, the transmission rates further linearly decreased from March 10 to March 22, eventually achieving the value ![Graphic][22], which is then held constant. We let ![Graphic][23] possibly vary among different Italian regions to reflect possible heterogeneity in disease transmission. Specifically, we estimate the hyperparameters controlling the prior of the parameters ![Graphic][24] (a Gaussian distribution truncated between 0 and 1) in a hierarchical Bayesian framework. Moreover, we fix some parameters and update others based on our previous estimation and on new pieces of evidence that became available in the meanwhile. Parameters are estimated comparing data and simulation of the flux of hospitalization ((1 − *ζ*)*ηI*) at provincial scale. We assume that each data point follows a negative binomial distribution with mean *μ*, equal to the value predicted by the model, and variance equal to *ωμ* (NB1 parametrization13,14). Parameter values are summarized in Table 1. View this table: [Table 1:](http://medrxiv.org/content/early/2020/05/05/2020.04.30.20083568/T1) Table 1: Model parameters. The posterior distribution of the parameters marked as estimated was sampled through the DREAMZS implementation of the Markov chain Monte Carlo algorithm.16 For all estimated parameters we used uninformative priors within biologically meaningful boundaries. Moreover, following our previous application,4 we assumed *σ =* 0 25, *rS =* 0 5 and *rE = rP = rA = rR = rS*, while *rI = rQ = rH* = 0. Moreover: *γq = γI = γH*, *γa = 2γI*, and *αH = αI*. ## Supplementary Results ## Acknowledgments EB, DP and AR acknowledge funding from Fondazione Cassa di Risparmio di Padova e Rovigo (IT) through its grant 55722 (April 2020). * Received April 30, 2020. * Revision received April 30, 2020. * Accepted May 5, 2020. * © 2020, Posted by Cold Spring Harbor Laboratory This pre-print is available under a Creative Commons License (Attribution-NonCommercial-NoDerivs 4.0 International), CC BY-NC-ND 4.0, as described at [http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/) ## References 1. 1.Leung K, Wu JT, Liu D, and Leung GM. First-wave COVID-19 transmissibility and severity in China outside Hubei after control measures, and second-wave scenario planning: a modelling impact assessment. The Lancet 2020;395:1382–93. 2. 2.Kissler SM, Tedijanto C, Goldstein E, Grad YH, and Lipsitch M. Projecting the transmission dynamics of SARS-CoV-2 through the postpandemic period. Science 2020:eabb5793. 3. 3.Lavezzo E, Franchin E, Ciavarella C, et al. Suppression of COVID-19 outbreak in the municipality of Vo, Italy. 2020. doi: doi:10.1101/2020.04.17.20053157. medRxiv: 2020.04.17.20053157. url: [https://doi.org/10.1101/2020.04.17.20053157](https://doi.org/10.1101/2020.04.17.20053157). [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoibWVkcnhpdiI7czo1OiJyZXNpZCI7czoyMToiMjAyMC4wNC4xNy4yMDA1MzE1N3YxIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjAvMDUvMDUvMjAyMC4wNC4zMC4yMDA4MzU2OC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 4. 4.Gatto M, Bertuzzo E, Mari L, et al. Spread and dynamics of the COVID-19 epidemic in Italy: effects of emergency containment measures. Proceedings of the National Academy of Sciences of the USA 2020. Published online April 23, 2020. 5. 5.World Health Organization. Rational use of personal protective equipment (PPE) for coronavirus disease (COVID-19): interim guidance. 2020. url: [https://apps.who.int/iris/handle/10665/331498](https://apps.who.int/iris/handle/10665/331498) (visited on 03/25/2020). 6. 6.Li R, Pei S, Chen B, et al. Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV2). Science 2020:eabb3221. 7. 7.Grassly N, Pons-Salort M, Parker E, et al. Report 16: Role of testing in COVID-19 control. Imperial College London. 2020. url: [https://www.imperial.ac.uk/media/imperial-college/medicine/mrc-gida/2020-04-23-COVID19-Report-16.pdf](https://www.imperial.ac.uk/media/imperial-college/medicine/mrc-gida/2020-04-23-COVID19-Report-16.pdf) (visited on 04/28/2020). 8. 8.Zhang J, Litvinova M, Wang W, et al. Evolving epidemiology of novel coronavirus diseases 2019 and possible interruption of local transmission outside Hubei Province in China: a descriptive and modeling study. 2020. doi: 10.1101/2020.02.21.20026328. medRxiv: 2020.02.21.20026328. url: [https://doi.org/10.1101/2020.02.21.20026328](https://doi.org/10.1101/2020.02.21.20026328). [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoibWVkcnhpdiI7czo1OiJyZXNpZCI7czoyMToiMjAyMC4wMi4yMS4yMDAyNjMyOHYxIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjAvMDUvMDUvMjAyMC4wNC4zMC4yMDA4MzU2OC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 9. 9.Cereda D, Tirani M, Rovida F, et al. The early phase of the COVID-19 outbreak in Lombardy, Italy. 2020. arXiv: 2003.09320v1[q-bio.PE]. url: [https://arxiv.org/abs/2003.09320v1](https://arxiv.org/abs/2003.09320v1). 10. 10.He X, Lau EH, Wu P, et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nature medicine 2020: 1–4. 11. 11.Ferguson N, Laydon D, Nedjati Gilani G, et al. Report 9: Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand. Imperial College London, 2020. doi: doi:10.25561/77482. url: [http://hdl.handle.net/10044/1/77482](http://hdl.handle.net/10044/1/77482). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.25561/77482&link_type=DOI) ## Supplementary References 1. 12.Palmieri L, Andrianou X, Barbariol P, et al. Caratteristiche dei pazienti deceduti positivi a COVID-19 in Italia. Istituto Superiore di Sanità. 2020. url: [https://www.epicentro.iss.it/coronavirus/bollettino/Report-COVID-2S19\_26\_marzo.pdf](https://www.epicentro.iss.it/coronavirus/bollettino/Report-COVID-2S19_26_marzo.pdf) (visited on 04/28/2020). 2. 13.Cameron AC and Trivedi PK. Econometric models based on count data. Comparisons and applications of some estimators and tests. Journal of Applied Econometrics 1986;1:29–53. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/jae.3950010104&link_type=DOI) 3. 14.Lindén A and Mäntyniemi S. Using the negative binomial distribution to model overdispersion in ecological count data. Ecology 2011;92:1414–21. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1890/10-1831.1&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21870615&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F05%2F2020.04.30.20083568.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000292814300005&link_type=ISI) 4. 15.Task force COVID-19. Epidemia COVID-19, Aggiornamento nazionale: 23 aprile 2020. Dipartimento Malattie Infettive e Servizio di Informatica, Istituto Superiore di Sanità. url: [https://www.epicentro.iss.it/coronavirus/bollettino/Bollettino-sorveglianza-integrata-COVID-19\_23-aprile-2020.pdf](https://www.epicentro.iss.it/coronavirus/bollettino/Bollettino-sorveglianza-integrata-COVID-19_23-aprile-2020.pdf) (visited on 04/24/2020). 5. 16.ter Braak C and Vrugt J. Differential evolution Markov chain with snooker updater and fewer chains. Stat. Comput. 2008;18:435–46. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s11222-008-9104-9&link_type=DOI) [1]: /embed/graphic-8.gif [2]: /embed/graphic-9.gif [3]: /embed/graphic-10.gif [4]: /embed/inline-graphic-1.gif [5]: /embed/inline-graphic-2.gif [6]: /embed/inline-graphic-3.gif [7]: /embed/graphic-11.gif [8]: /embed/inline-graphic-4.gif [9]: /embed/inline-graphic-5.gif [10]: /embed/inline-graphic-6.gif [11]: /embed/inline-graphic-7.gif [12]: /embed/inline-graphic-8.gif [13]: /embed/graphic-12.gif [14]: /embed/inline-graphic-9.gif [15]: /embed/inline-graphic-10.gif [16]: /embed/graphic-13.gif [17]: /embed/inline-graphic-11.gif [18]: /embed/inline-graphic-12.gif [19]: /embed/inline-graphic-13.gif [20]: /embed/inline-graphic-14.gif [21]: /embed/inline-graphic-15.gif [22]: /embed/inline-graphic-16.gif [23]: /embed/inline-graphic-17.gif [24]: /embed/inline-graphic-18.gif