key: cord-0688945-iy5f4zul authors: Pérez-Reche, Francisco J.; Forbes, Ken J.; Strachan, Norval J. C. title: Importance of untested infectious individuals for interventions to suppress COVID-19 date: 2021-10-20 journal: Sci Rep DOI: 10.1038/s41598-021-00056-5 sha: a02e852cd1e51d3496dc0bfcb7e6f25d9ae17e27 doc_id: 688945 cord_uid: iy5f4zul The impact of the extent of testing infectious individuals on suppression of COVID-19 is illustrated from the early stages of outbreaks in Germany, the Hubei province of China, Italy, Spain and the UK. The predicted percentage of untested infected individuals depends on the specific outbreak but we found that they typically represent 60–80% of all infected individuals during the early stages of the outbreaks. We propose that reducing the underlying transmission from untested cases is crucial to suppress the virus. This can be achieved through enhanced testing in combination with social distancing and other interventions that reduce transmission such as wearing face masks. Once transmission from silent carriers is kept under control by these means, the virus could have been fully suppressed through fast isolation and contact tracing of tested cases. To describe the role of silent carriers in the spread of the virus, we used extensions of the SEIR model 16, 23, 25, 29 to include two types of infected individuals: tested (or reported) and untested (or unreported). The classification of infectious individuals into tested and untested classes intends to be in line with the available data which contains the number of tested cases. We remark that such classification does not make an explicit distinction between symptomatic and asymptomatic cases; in principle, both tested and untested classes may contain symptomatic and asymptomatic cases. We propose two models: model 1 and model 2. Model 1 is a relatively simple extension of the SEIR model that we use to describe the early stages of epidemics in each of the regions. This model is fitted to data to estimate key epidemiological parameters and the proportion of tested and untested infectious individuals in each region (see parameter estimation methods below). Model 2 is a generalisation of model 1 to account for the isolation of infectious individuals and their contacts. The estimated parameters for model 1 are used to parameterise model 2. We now introduce both models, describe the main assumptions of the models and the methods for parameter estimation. Model 1. Model 1 assumes that individuals can be in any of five compartments: Susceptible ( S ), exposed to the virus ( E ), tested infectious ( I t ), untested infectious ( I u ), recovered tested cases ( R t ), dead after testing positive ( D ) and untested infectious that recover or die ( Z u ). The flow between compartments is schematically shown in Fig. 1a . The change of the number of individuals in each compartment is described by deterministic, continuoustime dynamics given by the following differential equations: www.nature.com/scientificreports/ Here, N is the population size (see Table S1 in Additional file 2). According to these equations, susceptible individuals (compartment S ) become exposed to the virus (compartment E ) at a rate β(I t + I u )/N . Here, β is the rate at which an infected individual transmits the infection to a susceptible individual and (I t + I u )/N is the fraction of infectious individuals at time t . Exposed individuals remain in this state during a latent period ζ −1 after which they become infectious. Of those that are infectious, a fraction ρ t are tested for infection and move (1) The number of exposed individuals initially increases, reaches a peak and decreases at later stages of the epidemic. (c) The progression of the number of both tested and untested infected individuals also exhibits a peak. The decay of I t and I u after the peak induces a gradual weakening of the chain of transmission that leads to the end of the epidemic. (d) The number of tested and untested individuals that recover from infection or die increase monotonically during the epidemic. Fig. 2b -d. The reproductive number corresponding to this model can be analytically calculated using the next generation method 24 and conveniently expressed as the sum of two terms, R 0 = R t 0 + R u 0 , giving the contribution of tested and untested individuals, respectively. The dependence of these contributions on the parameters of the model are given by the following expressions: Model 2-Isolation and contact tracing. Model 2 is an extension of model 1 to study the effect of isolation of infected individuals and their contacts. This is achieved by adding two more compartments to model 1, Q t and Q u , which contain isolated tested and untested infectious individuals, respectively (see Fig. 2a ). The fraction of tested and untested infectious individuals who are isolated are denoted as ρ Qt and ρ Qu , respectively. Both types of infectious individuals are assumed to become isolated at the same rate, δ . Isolation of tested individuals is the most natural isolation strategy. Possible isolation of untested individuals, however, was incorporated to simulate scenarios in which some untested individuals may self-isolate without being tested if they exhibit symptoms or due to reasons that might not be linked to their infection (e.g. being advised to work from home or voluntarily trying to minimise contact). Contact tracing involves isolation of a fraction q of susceptible individuals that were in contact with infected tested cases (see Fig. 2b ). A fraction b of these contacts would have acquired the virus through the contact (i.e. in the absence of contact tracing and isolation, they would have moved to the exposed compartment). These individuals are quarantined in the Q t compartment before they become infections and will eventually recover or die. The remaining contacts, i.e. a fraction 1 − b , remain susceptible after the interaction with the infected case and their isolation is represented by a compartment S Q where they remain for a period σ −1 Q . After this period, such individuals remain susceptible and return to the S compartment. With these assumptions, the number of individuals in each compartment evolve according to the following differential equations: www.nature.com/scientificreports/ Here, k is the number of contacts per unit time. The transmission rate is given by β = kb . Splitting the transmission rate into the number of contacts per unit time, k , and the fraction of successful transmissions in a contact, b , is necessary to model contact tracing 23 . Note that model 1 can be formally recovered from model 2 when the rates related to isolation and contact tracing are zero, i.e. when ρ Qt = ρ Qu = γ Q = q = σ Q = 0 in Eq. (3). One can again use the next generation method 24 to obtain an expression for the reproduction number of model 2 which can be expressed as the sum of two terms, Control measures. Both model 1 and model 2 can be used to simulate measures such as a lockdown that achieve a reduction of transmission at the population level. This is simply implemented by reducing the intrinsic transmission rate β by a factor r ∈ [0, 1] (i.e. the reduced transmission is β(1 − r)). The effect of the initial lockdown ordered in all studied countries and subsequent relaxation was simulated with a piece-wise dependence of β on time (see details in Sect. 1 of Additional file 3). The gradual increase in testing efforts, isolation of cases and contact tracing was simulated with model 2 assuming sigmoid growth functions for ρ t , ρ Qt and q (see details in Sect. 1 of Additional file 3). Main assumptions of the models. Several simplifying assumptions were made to make the models operational and applicable to settings with limited data which are typical at early stages of epidemics. Here, we describe the main assumptions. Homogeneous populations. The total number of individuals in each compartment of our models represent average values for the whole country. More precise descriptions at smaller scales such as geographical regions within countries or at the level of individuals would require accounting for spatial heterogeneity within the populations 31,46 (e.g. cities and rural areas) as well as differences in both susceptibility and mortality across different age and vulnerability groups 47, 48 or topological heterogeneity of the network of contacts between individuals 49 . Ignoring heterogeneities limits the ability of our models to identify specific ways to make interventions operational. For instance, reductions in transmission are treated at a generic level without specifying if they could be achieved by enhanced social distancing, school closure, etc. Accounting for such details would require using individual-based simulations 21 . Deterministic dynamics. We focused on stages of epidemics in which the number of infectious individuals is large enough as for stochastic effects to be relatively unimportant on average 27 . Our models can be extended to incorporate stochasticity 50, 51 . This would give a more accurate description of epidemics when SARS-CoV-2 has just invaded or at later stages when the number of infected individuals becomes very low. All of the population at the start of the epidemic are susceptible. However, it may be that a proportion of the population are not susceptible for genetic reasons 52 or due to cross immunity 53, 54 . Imported cases. We focused on epidemics that are at a stage in which imported cases are expected to play a secondary role relative to internal transmission. Accounting for imported cases is crucial, however, to prevent reemergence of infection once an area has reached low numbers of infected individuals 55, 56 . Imported cases could www.nature.com/scientificreports/ be included in our models in terms of an influx of infected individuals. In scenarios in which imported cases represent an important fraction of new infections, however, a stochastic version of the model would be more appropriate than the deterministic dynamics used here. The transition times between compartments are exponentially distributed. This memory-less assumption is usual for classical compartmental models 23 . For COVID-19, however, transitions between compartments are better described in terms of gamma distributions 5,47 and using models with memory would provide a more precise description of the dynamics 5,23,57 . Tested and untested cases transmit infection at the same rate. The populations of tested and untested cases will typically consist of different proportions of symptomatic and asymptomatic cases. In spite of that, no statistical difference has been observed between the viral load of symptomatic and asymptomatic individuals 58, 59 . It is therefore reasonable to assume that transmission is similar for symptomatic and asymptomatic individuals and therefore similar for tested and untested cases. Differences in overall transmission, however, are incorporated in the model through the assumption that recovery rates for tested and untested cases can be different. Immunity follows after recovery from infection. Whether or not this is the case, and even if it is, the duration of the immunity is still unclear 60 . Our model could easily be extended to account for re-infections, should such data become available, and predictions might significantly change. The latent and incubation periods coincide. We adopted a parsimonious approach which assumes that the latent period (i.e. the time between exposure to communicability) coincides with the incubation period (time between exposure and the appearance of symptoms) 25 . There is, however, a growing body of evidence for presymptomatic transmission [61] [62] [63] [64] [65] . There is the potential to incorporate this in our models by varying the incubation rate parameter, ζ , or including a new compartment for pre-symptomatic infectious individuals 31 . We expect our predictions to be qualitatively independent of these details. Reported deaths. We assume that the number of deaths in the datasets originates from individuals that were tested for infection. Transmission in different regions. It is assumed that transmission of SARS-CoV-2 and the fraction of infectious cases are similar in different regions. This is reflected in the informative prior distributions used for ρ d and R 0 in the parameter estimation procedure (see Table 1 ). Parameter estimation. We fitted the Model 1 to data. The value for the incubation rate was set to 48 ζ = 1/5.2 days -1 . The free parameters in our fits were the rate of transmission, β , proportion of infectious that were tested, ρ t , proportion of tested infectious that die, ρ d , rate to recovery of tested infectious individuals, γ t , rate of recovery of untested infectious individuals, γ u , and initial number of exposed individuals, E(0) . We denote the free parameters by a vector θ = {β, ρ t , ρ d , γ t , γ u , E(0)}. The model was fit to the time series for the number of daily reported infected individuals and cumulative deaths, , in a period of m days in the early stages of epidemics (here, τ is used to denote discrete time in days). In particular, we used m = 15 days since the first data point with a positive number of deaths (see Table S1 in Additional file 2). We used data at early stage of each outbreak to minimise the influence of suppression strategies on our parameter estimates. Using data on deaths is important to obtain reliable descriptions of COVID-19 epidemics since data on deaths is likely to be more accurately recorded than data on infected and recovered individuals 5, 19, 38, 66 . In addition to deaths, we can use data on infected individuals which is represented by the tested infectious compartment, I t , in our models. Our fitting procedure aims at calculating the posterior probability density function for the parameters given the data, π(θ |D obs ) . To this end, we use the procedure proposed in 51 which can be regarded as an approximate Bayesian algorithm 67 . The posterior π(θ |D obs ) is approximated by the empirical distribution of a set of 500 point estimates θ of the model parameters. A point estimate θ is obtained by simulating n e = 3000 epidemics with Table 1 . Assumptions for the prior probability distribution of the estimated parameters. N µ, σ 2 denotes a normal distribution with mean µ and variance σ 2 . U(a, b) denotes a uniform distribution in the interval (a, b). Prior Support www.nature.com/scientificreports/ parameters sampled from a prior probability density π(θ ) . In each realization, a simulation of Model 1 produces deterministic evolution functions I t (t) and D(t) for the number of tested cases and cumulative deaths. The functions I t (t) and D(t) are used to build a random daily time series , where i sim τ and d sim τ are, respectively, the number of tested infected and deaths predicted at day τ . The point estimate θ is defined as the parameter vector corresponding to the realization that gives the closest prediction,D sim , to the observations, D obs . More explicitly, the point estimate for the model parameters is given by where ρ(D obs |D sim (θ)) is a distance function. In particular, we used a weighted quadratic form for the distance: Weighting by the observed values was used to account for the fact that the values taken by i τ and d τ differ by orders of magnitude. We checked, however, that fits with an unweighted distance give results that are statistically compatible with those reported in the main text. In addition, in the previous version of this work we obtained similar results by an approximate maximisation of a log-likelihood function in which i sim τ (θ) ∼ Pois(I t (τ )) and d sim τ (θ) ∼ Pois(D(τ )) , i.e. the predicted number of tested infected and deaths were described as random variables obeying a Poisson distribution with mean I t (τ ) and D(τ ) , respectively 68 . The prior probability density is defined as the product of priors for each parameter: The priors used in our fits are summarised in Table 1 . Calibration of Model 1 at early stages of epidemics. Figure 3 shows the estimates for the parameters of model 1 applied to the early stages of each outbreak (see numerical summary statistics in Additional file 2, Table S2 ). A comparison of the predicted trends and the data is shown in Additional file 2, Fig. S1 . The values for the testing coverage, ρ t , reveal that during the early stage of outbreaks, Germany scored the highest in terms of testing for infection (34% [95% CI 10-96%]). Hubei follows Germany in terms of testing coverage, followed by Spain, Italy and the UK. Our prediction for Hubei (32% [95% CI 11-85%]) is statistically consistent with the 65% reporting estimated by Li et al. 16 for China (not only Hubei) in the period 24 Jan-8 Feb 2020 which overlaps with the period 22 Jan-6 Feb 2020 used here. We note, however, that our estimate has a wide 95% CI and the comparison is not highly informative. The high testing percentage predicted for Germany agrees with the higher testing coverage at early stages of epidemics in this country (see Fig. 4 .2 in Additional file 1). Taking the confidence intervals into account, we estimate that for each infected individual tested in the UK, there could have been between 1 and 9 untested infected individuals. At the other end of the testing spectrum, we estimate that for each infected individual tested in Germany, between 0.05 and 9 individuals might have not been tested at the beginning of the epidemic. A higher testing coverage for Germany is in qualitative agreement with estimates given elsewhere 28, 38 . Our estimates for the reporting percentage, however, tend to be higher than www.nature.com/scientificreports/ those obtained by Jagodnik et al. 38 and the differences we found between countries are not as extreme as those given by Chicchi et al. 28 . From our estimates of ρ t , we predict that 66% [95% CI 4-90%] of infected individuals were not tested for infection in Germany. Bearing in mind that 50% of infected individuals might be reported as asymptomatic at testing 10,12 , we conclude that untested cases in Germany would mostly correspond to asymptomatic cases at testing. In contrast, untested cases in other countries might include a significant number of symptomatic individuals in addition to asymptomatic cases. In particular, these may include infections of care home residents that are known to be underreported in many countries at early stages of epidemics 69 . The proportion of tested infected individuals that die, ρ d , is smaller for the outbreak in Germany than for the other outbreaks. This might be a combined effect of the fact that infected individuals in this country were relatively young at the beginning of the outbreak 70 and the high testing rate. Indeed, the COVID-19 fatality rate is lower for the younger than for the elderly 47 and, the higher the testing rate, the more individuals with mild or no symptoms will be included in the tested infected compartment of our model. The lower death rate of individuals with mild symptoms will lead to an effectively lower death rate for the whole set of infected individuals in this compartment. Accordingly, a lower value of ρ d does not necessarily mean a lower overall infection fatality rate. In fact, our estimates for IFR = ρ t ρ d does not vary much among countries (between 0.5% [95% CI 0. 16, 35 . In general, the values we obtained are smaller than the infectious period (time from infection to death or recovery) reported elsewhere for COVID-19 5, 47, 72 . Our estimates thus probably reflect a reporting delay in all the studied outbreaks, in agreement with data on the onset of symptoms and reporting 70, 73, 74 . Under this hypothesis, our model predicts the smallest reporting delay for Germany, in agreement with a higher testing effort in this country at early stages of the epidemic. The removal period for untested infected individuals, γ −1 u , ranges between 3.1 [95% CI 2.5-6.1] days for Hubei and 3.5 [95% CI 2.5-8.0] days for the UK. Comparing with the reporting-to-recovery period γ −1 t and bearing in mind the reporting delays in all outbreaks, our estimates of γ −1 u suggest that untested individuals remain infectious for a shorter time than tested individuals. We predict that the number of exposed individuals at the beginning of our simulations, E(0) , is of the order of several thousand for all the countries, in qualitative agreement with estimates of a previous study for China 16 . The estimated reproduction number R 0 is statistically similar in all the studied outbreaks. This reflects our prior assumption that transmission of SARS-CoV-2 is intrinsically similar in different regions. The transmission rate, β , was derived from the estimates of ρ t , γ t , γ u and R 0 and it takes values that are around 1 for all countries. Suppression strategies. Suppression of the virus is achieved when the reproductive number is smaller than 1. In terms of model 2, the condition R Q < 1 can be achieved in several different ways, i.e. there are many combinations of the parameters of the model that can lead to suppression of the virus (see Eq. 4). A simple suppression strategy consists in reducing the transmission rate β . We studied this strategy in April 2020 to make predictions related to the lockdowns that were active in Germany, Italy, Spain and the UK; the lockdown had just being lifted in Hubei 68 . We predicted that early reduction of transmission would delay the outbreaks but resurgence of infection was likely after relaxing the lockdowns unless transmission was kept reduced by a factor larger than 70% of its intrinsic values (see an update to these results in Sect. 2 of the Additional file 3). These results were prescient for Germany, Italy, Spain and the UK where further waves of infection occurred later in 2020. In addition to nationwide lockdowns, all studied regions have implemented isolation and contact tracing programmes to suppress the virus. Following this, the reductions of β mentioned above for scenarios without isolation and contact tracing should be interpreted as an effective measure of the reduction of transmission. Indeed, when isolation and contact tracing are taken into account, one needs a smaller reduction in β to suppress the virus. Below we use model 2 to account for the combined effect of a reduced β and enhanced testing, isolation and contact tracing in the design of suppression strategies bearing in mind the presence of silent carriers of the virus. From Eq. (4) one can see that interventions that only isolate tested cases and their contacts can reduce the reproductive number R t Q associated with tested infected individuals. However, such strategy cannot suppress the virus since the reproductive number of untested cases is R u Q > 1 for all the studied epidemics (Fig. 3h) . Ensuring that R u Q < 1 is then a crucial step for suppression of the virus. The reproduction number R u Q decreases with the testing rate, physical distancing and isolation of untested cases. Even in optimistic scenarios in which all tested cases would isolate in 0.5 days, we estimate that more than 25% of untested cases should isolate to suppress the virus. This is unlikely to occur by spontaneous isolation of untested cases (see Sect. 3 of the Additional file 3). We now focus on more feasible suppression approaches that combine enhanced testing and physical distancing with isolation and contact tracing of tested cases (i.e. no isolation is assumed for untested infectious individuals). The condition R u Q < 1 can be achieved through physical distancing and/or enhanced testing that are described by a reduction in β and an increase in ρ t , respectively. Figure 4a shows the threshold line corresponding to R u Q = 1 in the space (ρ t , r) , where r is the reduction factor of β . In particular, R u Q < 1 can be achieved without the need of physical distancing ( r = 0 ) if more than ~ 80% of infected individuals are tested. The condition R u Q < 1 could also be achieved without any testing (i.e. with ρ t = 0 ) if a severe reduction of transmission with www.nature.com/scientificreports/ r ∼ 80% is imposed. The later corresponds to a lockdown scenario studied above. Between these two extremes, the condition R u Q < 1 requires a combination of physical distancing and enhanced testing. Once the condition R u Q < 1 is satisfied for untested cases, suppression can be achieved by reducing R t Q through isolation and contact tracing of tested cases. To illustrate this, we take the UK as an example and consider three scenarios with different levels of testing and social distancing. Testing 30% of cases and reducing transmission by 70% through physical distancing leads to R u Q < 1 (circle in Fig. 4a ) and suppression is possible if ∼ 20% of infected cases were isolated in δ −1 = 0.5 days (see black line in Fig. 4b) . In this case, contact tracing would not be needed. For slower isolation of tested cases, however, contact tracing would be needed even if 100% of tested cases were isolated. This is illustrated in Fig. 4b by the lines corresponding to isolation in δ −1 = 2 or 3 days. Our results echo other studies in terms of the need to quickly isolate infected individuals 75 . In order to allow for a reasonable time between reporting of infection and isolation, carriers of the virus should be identified as early as possible, i.e. reporting delays should be minimised. Suppression is also possible if 50% of cases are tested and strong physical distancing is imposed which keeps transmission reduced by a 60% (square in Fig. 4a ). In this case, suppression without contact tracing is only possible for prompt isolation of tested cases (see black line in Fig. 4c) . For slower isolation rates, a high coverage of www.nature.com/scientificreports/ contacts is needed even if all tested cases were isolated. For instance, if 100% of tested cases were isolated after 2 days, more than 60% of contacts should be traced to suppress the virus (see the red line in Fig. 4c ). In principle, mass testing and high compliance with isolation would make suppression possible without the need for physical distancing. For instance, suppression would be possible if > 80% of the cases were tested (triangle in Fig. 4a) . As in the scenarios described above, high isolation and contact tracing would be needed to suppress the virus even if isolation is very fast (see Fig. 4d ). Detecting 80% of cases as assumed in the last scenario could be possible if, for instance, only truly asymptomatic cases escaped testing (around 20% of infected individuals are expected to be truly asymptomatic 15 ). We tested the feasibility of this scenario for the second wave of infection observed in the UK after the relaxation of the initial lockdown. Stay-at-home and physical distancing policies remained active once the lockdown was relaxed in the UK. It is interesting, however, that the second wave of infection can be explained by model 2 in terms of a hypothetical scenario consisting of complete relaxation of the lockdown combined with enhanced testing and launching of test and trace schemes throughout the country. More explicitly, the scenario is as follows: Testing only misses asymptomatic cases (i.e. 15 ρ t = 80%), 84% of those that tested positive self-isolate 76 after 2 days of being tested 77 and 60% of contacts are traced and isolated 78 (this corresponds to the diamond in Fig. 4d that is below the red line, i.e. it is in the region with R Q > 1 ). Full details of this analysis and a comparison of the predictions of model 2 with data are given in Sect. 4 of the Additional file 3. We checked that a similar scenario in which 95% of contacts are traced instead of 60% leads to suppression of the virus (see the star symbol in Fig. 4d which is above the red line where R Q < 1). Even though a scenario that assumes a complete relaxation of the lockdown can reproduce the data in the UK, a scenario with a partial relaxation of the lockdown is more realistic since stay-at-home and physical distancing policies remained active after the relaxation of the lockdown. This would indicate that the levels of testing, isolation and contact tracing were probably less effective than those used above for complete relaxation of the lockdown. For example, we checked that a scenario with post-lockdown transmission reduced by 20% and an isolation rate of 20% also captures the observed trend for the number of dealy deaths after the lockdown (see Fig. S4 (c) in Sect. 4 of the Additional file 3). In general, there can be many combinations of the parameters of model 2 leading to a reasonable description of the data. In spite of that, our results suggest that high levels of testing, isolation and contact tracing could suppress the virus even if 20% of cases were asymptomatic and not tested for infection. The main aim of our modelling work is to contribute to the understanding of the epidemiological patterns of SARS-Cov-2. The models should be viewed as a general guide of how the outbreak and interventions may play out rather than as an exact representation of COVID-19 epidemics. In spite of our simplifying assumptions, there are two main implications from the models which are relevant for health policy in dealing with the outbreak. The first, involves the existence of a significant proportion of cases that are not tested and act as silent carriers of the infection. We found that the predicted percentage of untested infected individuals at the early stages of epidemics may have represented 60-80% of the cases. The specific percentage depends on the country and we found the lowest proportion of unreported cases in Germany. The levels of underreporting are expected to have gradually decreased during the course of the studied epidemics since testing capacity increased in all the studied regions. It is likely, however, that underreporting remains significant due to asymptomatic individuals or cases with mild symptoms. Our model predicts that a resurgence of the virus was likely after relaxing the initial lockdown in 2020 in all the studied regions. This would be similar to second waves of infection observed in the 1918 influenza epidemics 79 . In fact, we made this prediction in April 2020 when lockdowns were still active in the studied European countries and it had just been relaxed in Hubei 68 . This prediction was confirmed by the resurgence of cases observed in European countries in August 2020 80 . The second implication involves the finding that unreported cases play an important role in the control of COVID-19 epidemics. In particular, unreported cases act as silent carriers and control strategies need to account for them or be prone to the risk of re-emergence or ineffective suppression of spread. For instance, we predict that isolation and contact tracing of tested cases can have a limited impact on the suppression of spread unless the underlying transmission of silent carriers is suppressed. The latter can be achieved by combining physical distancing and thorough testing of case contacts. Related to this, we found that physical distance might not be essential in an ideal scenario in which testing only misses 20% of cases (e.g. asymptomatic), there is a high adherence to self-isolation policies and contact tracing is highly effective. In the absence of pharmaceutical interventions, however, a certain level of physical distance is likely to be necessary for suppression. To summarise, in line with previous work 19, 81 and our predictions in April 2020, we suggest that widespread testing combined with contact tracing 26, 27 , isolation of infected individuals and social distancing are necessary to suppress SARS-CoV-2 using non-pharmaceutical interventions without severe lockdowns. Vaccinations are now available and our models could be extended to study the combined effect of vaccination and non-pharmaceutical interventions on epidemics while accounting for underreporting. Information on the effects of vaccines is still limited but it is expected that they will reduce the risk of individuals becoming infected and will protect against COVID-19 symptoms and severe illness [82] [83] [84] . Gradual reductions of the risk of infection and risk of death can be readily simulated by assuming gradually decreasing the transmission rate ( β ) and proportion of infected individuals that die ( ρ d ). A reduction in β leads to a reduction of the reproduction number of both tested and untested infectious individuals and this aids suppression of the virus. Enhanced protection against symptoms plays a less clear role in the suppression and would require a more detailed analysis. Indeed, protection against symptoms might lead to an increased proportion of silent carriers of the virus. This might www.nature.com/scientificreports/ lead to an increase in the reproduction number that might ultimately reduce some of the benefits of the vaccine. Based on early evidence that vaccines reduce the risk of serious illness 82 , prioritising the vaccination of the most vulnerable individuals is crucial to make sure that a potential resurge of transmission associated with silent carriers occurs when those individuals have been already vaccinated. In addition to that, keeping high levels of testing and isolation of cases will be crucial to prevent potential negative effects associated with unreported cases. All data used in this work are available from the cited sources. The models were analysed implemented in Mathematica. A notebook that retrieves the data and runs the calculations can be downloaded from https:// doi. org/ 10. 6084/ m9. figsh are. 14636 199. Point estimates of the parameters of model 1 and an R script to generate Fig. 3 can also be downloaded from this link. www.nature.com/scientificreports/ A novel coronavirus from patients with pneumonia in China World Health Organization World Health Organization Estimating the number of infections and the impact of non-pharmaceutical interventions on COVID-19 in 11 European countries Evidence of initial success for China exiting COVID-19 social distancing policy after achieving containment Contact tracing for COVID-19: current evidence, options for scale-up and an assessment of resources needed How will country-based mitigation measures influence the course of the COVID-19 epidemic Contagion: The Economic and Social Impacts of Covid-19 on our Region Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship Covid-19: identifying and isolating asymptomatic people helped eliminate virus in Italian village Iceland lab's testing suggests 50% of coronavirus cases have no symptoms Sentinel surveillance of SARS-CoV-2 in wastewater anticipates the occurrence of COVID-19 cases Report of the WHO-China Joint Mission on coronavirus disease 2019 (COVID-19 Occurrence and transmission potential of asymptomatic and presymptomatic SARS-CoV-2 infections: A living systematic review and meta-analysis Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV2) Super-spreaders in infectious diseases Coronavirus may have infected half of UK population-Oxford study Fundamental principles of epidemic spread highlight the immediate need for large-scale serological surveys to assess the stage of the SARS-CoV-2 epidemic European Centre for Disease Prevention and Control. Immune responses and immunity to SARS-CoV-2 Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand Infectious diseases of humans: dynamics and control Modeling infectious diseases in humans and animals Mathematical Tools for Understanding Infectious Disease Dynamics Phase-adjusted estimation of the number of coronavirus disease 2019 cases in Wuhan China Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts The Efficacy of Contact Tracing for the Containment of the 2019 Novel Coronavirus (COVID-19) Analysis and forecast of COVID-19 spreading Novel coronavirus 2019-nCoV: early estimation of epidemiological parameters and epidemic predictions Early dynamics of transmission and control of COVID-19: a mathematical modelling study A mathematical model for the spatiotemporal epidemic spreading of The end of social confinement and COVID-19 re-emergence risk Modelling the COVID-19 epidemic and implementation of population-wide interventions in Italy Spread and dynamics of the COVID-19 epidemic in Italy: Effects of emergency containment measures Estimation of the transmission risk of the 2019-nCoV and its implication for public health interventions Effective containment explains subexponential growth in recent confirmed COVID-19 cases in China A modelling framework to assess the likely effectiveness of facemasks in combination with 'lock-down' in managing the COVID-19 pandemic Correcting under-reported COVID-19 case numbers: estimating the true scale of the pandemic Epidemic Data for Novel Coronavirus COVID-19" from the Wolfram Data Repository Protect (contact tracing for coronavirus (COVID-19)) Test and Protect rolled out nationally -gov COVID-19 Government Response Tracker COVID-19: Testing and Contact Tracing How many confirmed cases are there in your area? Estimates of the severity of coronavirus disease 2019: a model-based analysis Early transmission dynamics in Wuhan, China, of Novel coronavirus-infected pneumonia Networks: an introduction Synergy in spreading processes: From exploitative to explorative foraging strategies Prediction of invasion from the early stage of an epidemic Initial Review and analysis of COVID-19 host genetics and associated phenotypes Possible Biological Explanations for Kids' Escape from COVID-19 Targets of T cell responses to SARS-CoV-2 coronavirus in humans with COVID-19 disease and unexposed individuals China coronavirus infections rise as some Spanish companies prepare to restart work Impact of international travel and border control measures on the global spread of the novel 2019 coronavirus outbreak Appropriate models for the management of infectious diseases Presymptomatic SARS-CoV-2 Infections and transmission in a skilled nursing facility SARS-CoV-2 viral load in upper respiratory specimens of infected patients Can you catch the coronavirus twice? We don't know yet Transmission of 2019-nCoV infection from an asymptomatic contact in Germany Asymptomatic cases in a family cluster with SARS-CoV-2 infection Asymptomatic and presymptomatic SARS-CoV-2 infections in residents of a long-term care skilled nursing facility-king County Presumed asymptomatic carrier transmission of COVID-19 Presymptomatic Transmission of SARS-CoV-2-Singapore Inferring the number of COVID-19 cases from recently reported deaths. medRxiv Markov chain Monte Carlo without likelihoods Importance of untested infectious individuals for the suppression of COVID-19 epidemics Mortality associated with COVID-19 outbreaks in care homes: early international evidence COVID-19) Daily Situation Report of the Robert Koch Institute -11/04/2020 A systematic review and meta-analysis of published research data on COVID-19 infection fatality rates Incubation period and other epidemiological characteristics of 2019 novel coronavirus infections with right truncation: A statistical analysis of publicly available case data QueHa cemos/ Servi cios/ Vigil ancia Salud Publi caREN AVE/ Enfer medad esTra nsmis ibles/ Docum ents/ INFOR MES/ Infor mes% 20COV ID-19/ Infor me% 20n% C2% BA% 2021 Integrated surveillance of COVID-19 in Italy Impact of delays on effectiveness of contact tracing strategies for COVID-19: A modelling study Weekly statistics for NHS Test and Trace (England): 3 December to 9 December Efficacy of contact tracing for the containment of the 2019 novel coronavirus (COVID-19) Public health interventions and epidemic intensity during the 1918 influenza pandemic Rapid Risk Assessment: Coronavirus disease 2019 (COVID-19) in the EU/EEA and the UK -eleventh update: resurgence of cases World Health Organization. WHO Director-General's opening remarks at the media briefing on COVID-19 Effectiveness of the Pfizer-BioNTech and Oxford-AstraZeneca vaccines on covid-19 related symptoms, hospital admissions, and mortality in older adults in England: Test negative case-control study COVID-19) Vaccination and non-pharmaceutical interventions for COVID-19: A mathematical modelling study World Health Organization. WHO Director-General's opening remarks at the media briefing on COVID-19 -3 We acknowledge fruitful discussions with Oliver Carrillo, Bruno Lopes, Christopher McGuigan, Veronica Morales, Stefano Polizzi, Ovidiu Rotariu, John Strachan and Eduard Vives. F.P.R., K.F. and N.S. planned the project and designed the research. F.P.R. retrieved the data and run the numerical simulations. F.P.R. N.S. and K.F wrote the manuscript. All authors reviewed the manuscript and approved the final version of it. The authors declare no competing interests. The online version contains supplementary material available at https:// doi. org/ 10. 1038/ s41598-021-00056-5.Correspondence and requests for materials should be addressed to F.J.P.-R.Reprints and permissions information is available at www.nature.com/reprints.Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.