Transmission dynamics of the COVID-19 epidemic in India, and evaluating the impact of asymptomatic carriers and role of expanded testing in the lockdown exit strategy: a modelling approach ============================================================================================================================================================================================ * Mohak Gupta * Saptarshi Soham Mohanta * Aditi Rao * Giridara Gopal Parameswaran * Mudit Agarwal * Mehak Arora * Archisman Mazumder * Ayush Lohiya * Priyamadhaba Behera * Agam Bansal * Rohit Kumar * Ved Prakash Meena * Pawan Tiwari * Anant Mohan * Sushma Bhatnagar ## Abstract **Background** The coronavirus disease 2019 (COVID-19) has caused over 3 200 000 cases and 230 000 deaths as on 2 May 2020, and has quickly become an unprecedented global health threat. India with its unique challenges in fighting this pandemic, imposed one of the world’s strictest and largest population-wide lockdown on 25 March 2020. Here, we estimated key epidemiological parameters and evaluated the effect of control measures on the COVID-19 epidemic in India and its states. Through a modeling approach which accounted for asymptomatics, we assessed the impact of lockdown relaxation and increased testing. **Methods** We estimated the basic reproduction number and effective reproduction number at a national and state level in India after adjusting for imported cases and reporting lag, using time-series data from 4 March to 25 April 2020. Using a dynamic SEIR-QDPA model fitted to data from India, we forecasted the size and temporality of the ongoing first wave while accounting for the interventions in place. We used the model to simulate lockdown relaxation under various scenarios to evaluate its effect on the size and temporality of the second wave. We also evaluated the feasibility of increased testing as a containment strategy after restrictions are relaxed and its impact on epidemic size and resumption of socio-economic activities, while taking into account the changes in transmission dynamics brought about by asymptomatic carriers. **Findings** The median delay from symptom onset to detection (reporting lag) was estimated to be 2·68 days (95% CI 2·00-3·00) with an IQR of 2·03 days (95% CI 1·00-3·00). The R for India was estimated to be 2·083 (95% CI 2·044–2·122; R2 = 0·972), while the Rt gradually down trended from 1·665 (95%CI 1·539–1·789) on 30 March to 1 159 (95% CI 1 *128·1 *189) on 21 April. 60·7% of confirmed COVID-19 cases in our sample were found to be asymptomatic. We observed that delaying the lockdown relaxation increases the time to new rise in active cases after the relaxation in a linear fashion. If lockdown was reintroduced after a fixed relaxation period, the magnitude of the second peak could be reduced by delaying the relaxation and decreasing the duration of relaxation. These benefits were greater in case of a gradual relaxation as compared to a sudden lifting of lockdown. We found that detecting a higher proportion of cases through testing significantly decreases the total infections in future. This positive impact of testing progressively increased at higher transmission rates when restrictions were relaxed. We found that similar containment targets could be achieved by both, a combination of high testing and less social restrictions, and a combination of lower testing with intensive social distancing. **Interpretation** The nationwide social distancing interventions in India since 25 March have reduced the effective transmission levels, though sub-threshold Rt remains to be achieved. If lockdown is to be extended, additional benefits for mitigating the second wave can be achieved if it is extended farther after the peak of active cases has passed. Intensive social distancing is inherently enough to contain the epidemic, however testing will play a pivotal role in the lockdown exit strategy by impeding the epidemic growth enough to allow for greater resumption of socio-economic activities, thus minimizing the social and economic fallout resulting from severe restrictions. Considering that asymptomatics play an undeniable role in transmission of COVID-19, dependence on presence of symptoms for control strategies, behavioral changes and testing should be reduced. Research in context Evidence before this study A nationwide lockdown was imposed in India on 25 March 2020, about three weeks after the first locally transmitted case of novel coronavirus disease (COVID-19) was detected. The lockdown effectively confined 1.3 billion people to their homes, wherein all non-essential activities including offices and schools were shut down, mass gatherings banned, inter-state travel highly restricted, and all international borders sealed. We searched PubMed and preprint archives for articles published up to May 2, 2020, using the terms “coronavirus”, “2019-nCoV”, “COVID-19”, AND “reproduction number”, “transmission”, “lockdown”, “interventions”, AND “India”. We found two studies that reported the time-varying reproduction number (Rt) of COVID-19 based on time-series data from India till early-April, but none accounted for imported cases and reporting lag in the estimation and none analysed the effect of interventions on the reproduction number. We could not identify any study which modelled the role of asymptomatic carriers and the impact of increased testing as a lockdown exit strategy. Added value of this study Our study provides reliable estimates of R and Rt for India and states after correcting for importations and delay to detection. The trend of Rt upto 22 March indicates that control measures implemented across India have been effective in reducing the transmission of COVID-19, though sub-threshold Rt remains to be achieved. Through a modelling approach which accounts for asymptomatic transmission, we evaluated various lockdown exit strategies, including the effect of onset and duration of lockdown relaxation on the second wave, and the impact of increased testing on epidemic size and resumption of socio-economic activities. We found that extending lockdown farther beyond the first peak has additional benefits in mitigating the second wave. Expanded testing significantly reduced transmission, and we found that this impact was greater at higher transmission rates which will be seen once restrictions are relaxed. Implications of all the available evidence To avoid a resurgence in cases, a flexible relaxation approach guided by regional monitoring of effective reproduction numbers is recommended, and this relaxation should be farther beyond the peak of the first wave as feasible. As restrictions are relaxed, increased detection through expanded testing will be essential in limiting the resurgence of cases and will allow greater resumption of socio-economic activities, thus authorities should consider scaling up testing capacity preemptively before lifting restrictions. Considering that asymptomatics play an undeniable role in transmission of COVID-19, dependence on presence of symptoms for control strategies, behavioral changes and testing should be reduced. Further research to determine the asymptomatic proportion and quantify their transmission potential is needed to better inform future control strategies. ## Introduction Originating out of Wuhan, China in December 2019,1 the coronavirus disease 2019 (COVID-19) was declared a pandemic by the WHO on March 11 2020.2 As of 2 May 2020, there have been more than 3 200 000 cases and 230 000 deaths worldwide, and close to 40 000 cases and 1200 deaths in India.3 India reported its first COVID-19 case on 30 January 2020, although actual epidemic growth started from early March.4 For any novel infectious disease, the scale of its public health impact is determined by the basic reproduction number ‘R’ which is the average number of secondary infections generated by an infectious index case in a wholly susceptible population. The R of an infection determines its potential to start an outbreak, the severity of control measures needed to contain the spread, and the fraction of the population that will be infected in the absence of interventions.5 However, once an outbreak is underway, the time-varying effective reproduction number ‘R,’ is more relevant as it tracks the subsequent changes in transmission, and can thus be used to monitor the efficacy of control measures and adjust them accordingly.6–8 However, any given transmission event is reflected in the data only after a delay, which must be accounted for in the estimation of such indicators for accurate interpretation.7 Previous studies have shown that a severe epidemic with R~2·4 can be contained by combining effective quarantine, behavioral change to reduce social mixing, targeted antiviral prophylaxis, and pre-vaccination.5 However, in the absence of targeted therapeutics and vaccination for COVID-19, an unprecedented one-third of the world’s population is currently under lockdowns – with the primary target of reducing the Rt below the threshold of 1.69 India responded to the COVID-19 pandemic rapidly and decisively by imposing a nation-wide lockdown on 25 March 2020, when there were 536 cases and 10 deaths.3,10 This ‘suppression strategy’, though effective, has its limitations-the social and economic cost of such population-wide social distancing is huge, which limits the long-term implementation of these measures.9,11 Additionally, containing COVID-19 in India is a unique challenge due to its high population density, underprepared healthcare system, and wide socio-economic disparity.12 A large proportion of India’s labour force works as daily-wage laborers or migrant workers, and are especially affected during such times, making lockdowns untenable without parallel social support.13 There could be yet unseen adverse effects in the form of non-COVID-19 morbidity and mortality due to aggravation of malnutrition, chronic diseases and lack of access to healthcare during this time.14 At the same time, premature withdrawal of lockdowns without adequately planned interventions for the post-lockdown phase may lead to re-emergence later, or the second wave. 6,9 Thus there arises a need to create a balance to ensure that the disease is contained and the healthcare system remains well prepared while minimizing the collateral damage from intensive blanket interventions. Comprehensive lockdown exit strategies will be central to the future course of the pandemic.10 In such scenarios with limited primary information, dynamic mathematical models can provide actionable insights for researchers and policymakers.7,9,15 Evidence suggests that COVID-19 has a wide clinical spectrum which ranges from asymptomatic to fatal infections, which coupled with high infectivity can lead to a large number of infections and deaths.16 It may be possible that COVID-19 transmission is driven significantly by undetected asymptomatics while fatality is driven by severe cases-a devastating combination.17,18 Some have deemed asymptomatic transmission as the “Achilles’ Heel” of the current control strategies against COVID-19.19 In this study, we estimate the key transmission parameters for COVID-19 in India and its states, and analyse how interventions affected transmission levels across time. Considering that blanket lockdowns are an initial rather than a final step in controlling this pandemic, we model the effect of relaxing public health interventions at various time-points. We evaluate the efficacy of increased detection of asymptomatics through wider testing as a viable lockdown exit strategy, through a modelling approach. ## Methods ### Data sources For estimating the Basic Reproduction Number (R) and Effective Reproduction Number (Rt) for India and various states, we used the publicly available dataset from COVID19India from 04 March to 25 April 2020.4 For estimating the proportion of asymptomatic cases and the delay from symptom onset to confirmation, we obtained data for 140 COVID-19 patients admitted to a tertiary care hospital near Delhi, which included duration from symptom onset to date of admission for symptomatic patients (appendix p5). We used data from Johns Hopkins University (JHU) COVID-19 database on daily cumulative cases, daily cumulative deaths, and daily cumulative recoveries in India from 16 March to 18 April 2020 for model fitting and parameter estimation.3 We used the World Bank Population Database for population data for India.20 A laboratory confirmed case irrespective of symptoms is counted as a confirmed COVID-19 case in India. Testing criteria are provided in appendix p9. ### Estimation of basic reproduction number (R) The best-fit R was calculated for the national and state level incidence data using the R *package* in R 3·6·3 using two independent methods: Maximum Likelihood (ML) method and the Exponential Growth (EG) method after adjusting the incidence data for imported cases.21–23 We assumed the serial interval to be gamma distributed with a mean of 3·96 days (95% CI 3·53-4·39) and a SD of 4·75 days (95% CI 4·46-5·07).24 We analysed the sensitivity of the estimated R to the choice of the time period over which the R was estimated and the serial interval (appendix p8). ### Estimation of reporting lag, lag adjusted incidence, and time-varying effective reproduction number (Rt) A variable delay occurs from symptom onset to COVID-19 confirmation (henceforth referred to as the reporting lag) which is attributed to multiple factors including time taken to seek care (patient dependent) and time taken to detect and test the case (healthcare-system dependent). Since all included patients were tested and confirmed positive within a day of admission, the time from symptom onset to admission obtained from data approximates the reporting lag of these cases. We fitted appropriate distributions to the data and the best fit distribution was chosen. Due to lack of data, we assume that the reporting lag for India and for each state is statistically the same as the estimated reporting lag for the 53 patients from Delhi whose onset date was known. For each reported case, onset dates were sampled to generate 1000 lag-adjusted datasets for incidence by onset (appendix p5·6). Using the reporting lag adjusted incidence, the time-varying Rt, was calculated using *EpiEstim* package in R 3·6·3 which uses the Time Dependent Maximum Likelihood approach.25,26 The same serial interval distribution was used as for R estimation. Cases in the dataset not explicitly labeled as ‘imported’ were considered to be locally transmitted. We determined both the imported cases adjusted *Rt* and unadjusted *Rt* for India. The *Rt* trends were overlaid with major epidemic events in the country and mobility information from Google Community Mobility Reports to analyse the correlation between social mobility and transmission rate, if any.27 ### Modelling the pandemic using dynamical compartmental models In order to model the spread of SARS-CoV-2 in the population, we generalize the extensively used SEIR model for epidemics,28 which can fit the early dynamics of the COVID-19 pandemic well it but lacks important social and biological factors that play a critical role in determining the disease’s progression. For introducing the required complexities, we build a model as shown in Figure 1. Model parameters are defined in Table 1. ![Figure 1:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F1.medium.gif) [Figure 1:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F1) Figure 1: Schematic for SEIR-QDPA model Compartments include S(susceptible), E(infected but not yet infectious), Ia(undetected asymptomatic; infectious), Is(undetected symptomatic; infectious), Qa(detected and quarantined asymptomatic), Qs(detected and quarantined symptomatic), Ru(undetected recovered asymptomatic), Ra(recovered detected asymptomatic), Rs(recovered detected symptomatic), D(dead), and P(protected; non-susceptible). Compartments in red are fitted to data; Q=Qa+Qs to active cases, R=Ra+Rs to total recovered cases, and D to total deaths. Transition rates in red are inputs to the model, while others are estimated (Table 1). View this table: [Table 1:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T1) Table 1: Parameters for SEIR-QDPA model. Through a positive protection rate (α), the susceptible population gradually decreases to account for the effect of increasingly intensive social distancing policies and improved public behaviour in reaction to the epidemic.36 We introduce a deprotection rate (a) which increases the susceptible pool once social distancing policies are relaxed. We set the probability of infected case being asymptomatic (pa) to 0·2, 0·4, 0·6, 0·8, as reported estimates for the percent of infections that are asymptomatic range widely from 18%-80%.17,18,32–34 We consider that asymptomatic cases do not exhibit coughing, sneezing or sputum production and are thus expected to show lower infectivity than symptomatics. We conservatively assume an asymptomatic patient is 50% as infective as a symptomatic patient (ai =0·5). We set the fraction of detected asymptomatics (fa) at baseline to 0·1,35 the incubation period (y−1) to 5·1 days,29 and the infectious period for asymptomatics (![Graphic][1]) to 8 days based on virologic and epidemiologic studies.30,31 We assume that all symptomatics are detected, and that no asymptomatic dies from the disease. Further details regarding the construction of the model, including the governing equations are available in appendix p3·4. We fitted for the values of transmission rate (β), protection rate (α), reporting lag (![Graphic][2]), recovery rate (λ), mortality rate (κ), and the initial values of exposed (E) and infected (I) individuals in the population. We estimated the unknown parameters of the model by fitting time series data for active cases (given by cumulative confirmed cases – cumulative recoveries – cumulative deaths), cumulative recoveries and cumulative deaths to the Q(t) = Qs(t) + Qa(t), R(t) = Rs(t) + Ra(t), and D(t) compartments respectively. We then simulate the model with the estimated parameters to predict the key time points and the number of expected cases in the ongoing first wave. We define and predict three key time points which are inherent to epidemic progression-time at peak of daily new reported cases (t1), time at peak of active cases (t2), and time when recovered cases > active cases (t3). Note that these predictions do not take into account the change in transmission dynamics due to relaxation of lockdown. Sensitivity of our results to assumptions of pa, fa and ai was analysed. In the article, we have used the following operational definitions obtained by combining values of various compartments: ‘Symptomatic cases’-cumulative detected symptomatic cases; ‘Detected cases’-cumulative detected cases including symptomatic and asymptomatic cases; ‘Total infections’-cumulative infections including detected cases and undetected asymptomatic infections. ### Simulating the effect of lockdown relaxation To model complete lifting of the nation-wide lockdown, α is set to zero and σ is set to a large value such that the entire protected population is emptied into the susceptible population in a short interval (t½~1·0 day). We trigger this change on 4 May 2020 (tentative date of lockdown relaxation in India) and 7-day intervals thereafter, to compare outcomes if lockdown is lifted at different dates. We also consider the scenario where the lockdown is re-enforced after a fixed relaxation period which we model by setting σ to zero and re-enforcing the same α as before lockdown relaxation. Alternatively, we model a comparatively gradual lockdown relaxation by setting σ=α. ### Simulating the effect of increased testing To model the effect of increased testing capacity and subsequent improved detection, we assume that it will increase the asymptomatic detection rate fa from 0·1 in lockdown, to 0·2, 0·3, 0·4, 0·5, 0·6, and 0·8 after lockdown relaxation starting May 4. To model the effect of varying levels of reduced social mixing and positive behavior change after the lockdown is lifted, we change the effective transmission rate to 90%, 80%, 70%, 60%, 50% and 30% of the original β, starting May 4. Varying levels of social mixing influences the contact rate, while behaviour changes like wearing masks and hand washing decrease transmissibility given a contact-the transmission rate β captures both these changes. We hypothesise that increased testing after lockdown relaxation will decrease the epidemic growth enough to allow for greater resumption of normal social mixing, thus minimising the social and economic fallout resulting from vigorous restrictions, and that this effect will be proportional to the amount of testing done. ### Role of the funding source The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the article. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. ## Results ### Basic Reproduction Number ‘R’ for India The exponential growth (EG) method had a better best-fit R2 over a larger time period and was less sensitive to the choice of the time period (appendix p7-8). The best fit R for India was found to be 2·083 (95% CI 2·044–2·122; R2 = 0·972). Taking into consideration the uncertainty in reported serial intervals, the R ranged from 2–2·5 for serial intervals ranging from 4–4·6 days.24,37–39 Results were found to be sensitive to the serial interval (SI) distribution, thus we provide R estimates based on the most reliable serial interval from 468 infector-infectee pairs in China,24 and also consider a range of possible serial intervals based on other studies.37-39 The R estimates for various states of India have been provided in appendix p7. ### Reporting Lag, Lag adjusted incidence and Time-varying Effective Reproduction Number ‘Rt’ Out of a 140 laboratory confirmed COVID-19 patients, 85 (60.7%) were asymptomatic while 55 (39.3%) were symptomatic. For 53 symptomatic patients, the reporting lag was found to have a mean of 3·40 days (95% CI 2·87–3·96) with standard deviation of 2·09 days (95% CI 1·52–2·56) and a median of 2·68 days (95% CI 2·00-3·00) with an IQR of 2.03 days (95% CI 1.00-3.00). The gamma distribution with shape parameter 3·45 (95% CI 2·42–5·19) and rate parameter 1·02 (95% CI 0·70–1·60) was the best fit to the distribution (appendix p5-6). The first cases of local transmission in India were reported on 4 March, who were family members of an initial imported case. The number of imported cases in India started increasing from early-March, peaked a day after the international travel ban on 23 March,40 and gradually came to a halt on 5 April with a total of 546 imported cases (Figure 2A). ![Figure 2:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F2.medium.gif) [Figure 2:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F2) Figure 2: Transmission dynamics and incidence of COVID-19 in India, along with major events and mobility trends **[A]** Daily new cases by confirmation date in India up to 2 May 2020 stratified as imported (red) and local (dark blue). The dates of testing criteria updates are overlaid since the scope of testing influences the number of confirmed cases (appendix p9).41 **[B]** Daily new cases by onset date (estimated epidemic curve) up to 22 April 2020 in India stratified as imported (red) and local (light blue), along with the instantaneous effective reproduction number R, adjusted for importations (pink) and without adjusting for importations (blue). Dark bands indicate 50% confidence interval and light bands indicate 95% confidence interval for estimated R,. Similar graphs for states of India are provided in appendix. **[C]** Mobility trends in India, compared to a baseline median value for the corresponding day of the week, during the 5-week period Jan 3–Feb 6, 2020. Holiday due to the Holi festival on 10 March 2020 caused a dip in mobility. A sharp dip in mobility is noted at the voluntary public curfew on 22 March and after the nationwide lockdown was enforced on 25 March, except for a rise in residential neighbourhood mobility. The weekly rise in workplace mobility appears to be an artifact due to comparison with normal weekends at the baseline. Source-Google community mobility reports.27 Major interventions are shown, the effects of which are best correlated with Rt trend and mobility changes, since these changes occur in real-time. Rt= time-varying effective reproduction number Incidence by onset and time-varying Rt could be ascertained upto 22 April 2020 since some cases with onset after this date may not have been reported yet in the data, due to the reporting lag. The Rt trends for India showed visible fluctuations over time (Figure 2B). The first uptick in unadjusted Rt (blue band) starting around 13 March 2020 was presumed to be an artifact due to imported cases, since it coincided with increasing imported case onsets, and was not accompanied by a concurrent uptick in import adjusted Rt (pink band). The second uptick in unadjusted Rt correlated with the rise in adjusted Rt, indicating that local transmission was driving this rise. This rise started around the imposition of the nationwide lockdown on 25 March and peaked on 30 March at an adjusted Rt of 1·665 (95%CI 1·539–1·789). After this peak, the Rt continued to decrease to 1·300 (1·247–1·353) on 8 April, 1213 (1·175–1·251) on 15 April and further to its lowest yet value of 1 159 (1128–1 189) on 22 April. A sharp dip in mobility is noted at the voluntary public curfew on 22 March and after the nationwide lockdown was enforced on 25 March, except for a rise in residential neighbourhood mobility (Figure 2C). The daily Rt values for India, and Rt trends for states of India are provided in appendix p11-14. ### Estimated model parameters from data The model was able to fit the data well for the early exponential phase of the growth and also captured the recent slowdown in epidemic growth through the protection rate. Table 2 shows the best-fit parameter estimates for the base assumptions of pa=0·6 (chosen based upon the asymptomatic proportion estimated from primary patient data), ai= 0·5, and fa= 0·1. The reporting lag estimated from model fit was 2·12 days (95% CI 2·07-2·20), which was in agreement with the reporting lag of 2·68 days (95% CI 2·00-3·00) estimated from primary patient data. The sensitivity of the fit parameters to our assumptions was evaluated (appendix p15-18). For the base assumptions, our model predicted 19741 (95% CI 17187–21870) maximum active cases, 34049 (30330–38648) cumulative detected cases, of which 29610 (26377–33609) were symptomatic till the end of first wave. The total infections were 73820 (65746–83811). Key time points were predicted as - time at peak of daily new reported cases (t1) between 16 April and 19 April 2020, time at peak of active cases (t2) between 3 May and 8 May 2020, and time when recovered cases > active cases (t3) between 25 May and 29 May 2020 (Figure 3). 95% CI of 1000 bootstrapped predictions are reported here. The predicted case counts increased and key points were delayed with higher assumptions of asymptomatic proportion (pa). ![Figure 3:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F3.medium.gif) [Figure 3:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F3) Figure 3: Model simulation of the first wave of COVID-19 in India, assuming the lockdown continues indefinitely with the initial stringency Bands represent 95% CI for the mean prediction over 1000 bootstraps. Three key time points in epidemic progression are shown-time at peak of daily new reported cases (t1), time at peak of active cases (t2), and time when recovered cases > active cases (t3). Results shown for the baseline scenario (asymptomatics are 50% infectious compared to symptomatic, 60% of total infections are asymptomatic, 10% asymptomatics are detected and quarantined). Results for other assumptions in appendix. **[A]** Simulated values of model compartments over time. Quarantined cases are equivalent to the active cases at a particular time. ‘+’ represents data with which the model was trained. **[B]** Predicted total infections, detected cases, and symptomatic cases over time. **[C]** Predicted daily new cases over time. View this table: [Table 2:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T2) Table 2: Estimated parameters for the model After comparing with recent data, we note that the model underpredicted the first wave because it did not consider a parallel leakage of protected compartment back into the susceptible compartment, which tends to happen in reality as lockdowns are not absolute. This also leads to a higher estimated transmission rate (to account for transmission otherwise caused by leakage of protected people), and thus our modeling results for lockdown relaxation and impact of testing should be considered as worst-case scenarios. ### Impact of Lockdown relaxation and its temporality On a complete removal of the lockdown irrespective of the date of relaxation, we observed that the number of active cases will start to rise exponentially after a variable delay (Figure 4A). We varied the time when the lockdown relaxation was initiated from 4th May to 13th July at one week intervals and quantified the delay (from the relaxation date) after which the number of active cases started rising again. We observed that delaying the lockdown relaxation increases the time to new rise in a linear fashion, with Pearson’s R = 0·997 (95% CI 0·989–0·999; p<0·0001) (Figure 4B). ![Figure 4:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F4.medium.gif) [Figure 4:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F4) Figure 4: Effect of complete lockdown relaxation under various scenarios Results are shown for the baseline scenario (asymptomatics are 50% infectious compared to symptomatics, 60% of total infections are asymptomatic, 10% of asymptomatics are detected and quarantined). Bands represent 95% CI for the mean prediction over 1000 bootstraps for all model plots. **[A]** Simulated values of the model compartments Q (active cases), Qs (active symptomatic cases) and R (recovered) under complete and sustained lockdown relaxation starting 4 May, 25 May, and 15 June 2020; showing increasing delay to start of the second wave with later relaxation. Inlay shows the underlying depletion of undetected infectious pool as the first wave crosses the peak. **[B]** Days to new rise in active cases (time delay after respective relaxation date) at different dates for lockdown relaxation. This effect is expected to be generalizable when interpreted with respect to the actual date of peak of active cases (compare with red dot). Black line represents the line joining the mean lag for 1000 bootstrapped simulations, and bands represent 95% confidence intervals. **[C]** Simulated values of the model compartments Q (active cases), Qs (active symptomatic cases) and R (recovered) under complete relaxation lasting 7 days, starting 4 May, 25 May, and 15 June 2020; showing increasing delay to start of the second wave and lower magnitude of the second wave with later relaxation. **[D]** Heatmap for the peak active cases under different lockdown relaxation durations and dates of start of relaxation. These values are counterfactuals in the worst-case scenario, where lockdown has been completely lifted across the country. When we simulated limited time duration relaxation periods, we found that there is a rise in active cases observed in all simulations but the extent of the rise is highly dependent on when the relaxation was started and the duration of the relaxation period (Figure 4C and 4D). Both delays in the lockdown relaxation and smaller relaxation periods reduced the number of active cases at the peak, suggesting that small relaxation periods enacted once the peak has been crossed may help reduce the number of active cases at any future point in time. In the case of a gradual lockdown release, we found that there was both a delay to new rise in active cases and the reduction in the second peak was more pronounced (appendix p26-27). The results shown in Figure 4 consider that the peak of active cases is expected to be on 4 May 2020. Since our model underpredicted the first wave, the actual peak of active cases may occur later than 4 May. The findings remain unaffected in essence, though these results will be better interpreted in the context of the actual peak. ### Effect of improved testing on epidemic control and restoration of normal social mixing We observed that both increased testing and increasingly vigorous social distancing have a positive impact on reducing the number of total infections and symptomatic cases. With increased testing, the fraction of total infections that were detected increased, ie, the ascertainment rate improved (Figure 5A and 5B). Although the number of detected cases may remain almost constant in some scenarios of increased testing, a lower proportion of the same are symptomatic, which highlights the significance of detecting more asymptomatic infections. ![Figure 5:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F5.medium.gif) [Figure 5:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F5) Figure 5: Effect of expanded testing and varying social mixing after complete lockdown relaxation. Results shown for the baseline scenario (asymptomatics are 50% infectious compared to symptomatic, 60% of total infections are asymptomatic, 10% asymptomatics are detected before lockdown relaxation). Any increase in testing or any decrease in social mixing starts from the day of lockdown relaxation. Results for other assumptions in appendix. Error bars represent 95% confidence intervals for 1000 bootstrapped predictions. All values are given in thousands of individuals. **[A]** and **[B]** Total number of infections, detected cases and symptomatic cases at 15 days and 45 days after lockdown relaxation with varying levels of testing. **[C]** Effect of increasing testing and decreasing social mixing on the number of symptomatic cases at 15 days after the lockdown relaxation. **[D]** Heatmap for total symptomatic cases after 15 days under different reductions in transmission rate (proxy for social distancing policies) and asymptomatic detection rate (proxy for testing policy). An example of a feasible combination of testing and social distancing policy is indicated by the area between two watershed lines (grey) for a containment target of 50,000-100,000 cases. Similar heatmap for total infections is given in appendix p24. We further found that the positive impact of increased testing becomes more prominent at progressively higher values of transmission rate β (Figure 5C). In the baseline scenario with detection increased from 10% to 20% for 15 days-for every extra asymptomatic detected by increased testing, the number of infections prevented are 8·3 at β, 6·4 at 0·8β and 3·5 at 0·5β, while the number of symptomatic cases prevented are 1 at p, 0·9 at 0·8β, 0·6 at 0·5β. This benefit increases further with higher increases in testing. For an increase in detection from 10% to 20%, the symptomatic cases decreased by 6·6% at β, by 5·5% at 0·8β, and by 3% at 0·5β. For an increase in testing from 10% to 50%, the symptomatic cases decreased by 32·3% at β, by 26·8% at 0·8β, and by 14·5% at 0·5β. After lockdown relaxation, lower levels of social restrictions (high β) when coupled with increased testing, can achieve similar results as a more restrictive social distancing regime where testing was not increased (Figure 5D). That is, increased testing allowed greater resumption of normal social mixing after lockdown relaxation. An example of a feasible combination of testing and social restrictions is indicated by the area between two watershed lines (grey) in Figure 5D. Due to uncertainty in the percent of infections that are asymptomatic, we evaluated the effect of testing across the range of pa (appendix p22-25). ## Discussion The trend of effective reproduction number (Rt) of COVID-19 in India indicates that control measures have been effective in slowing down the spread of COVID-19 across the country. To achieve sustained suppression, monitoring of the time varying Rt at district, state and national level should be done to reach and maintain an Rt close to the threshold value of 1. If lockdown is to be extended, additional benefits can be achieved if it is extended farther after the peak of active cases has passed. As these restrictions are relaxed, increased detection through testing will be essential in limiting the resurgence of cases and thus testing capacity should be ramped up preemptively before lifting restrictions. Considering that asymptomatics play an undeniable role in transmission of COVID-19, dependence on presence of symptoms for control strategies, behavioral changes and testing should be reduced. The range of R of SARS-CoV-2 in India was found to be 2–2·5, with 2·083 being the best fit. Our results align with recent studies which estimate the R to be 2–2·7.42–44 In comparison, the R was 1·4–1·6 for the 2009 influenza (H1N1) pandemic, 2·0–3·3 for the 2003 SARS epidemic, and 2·0–3·0 for the 1918 Spanish flu pandemic, which reflects the seriousness of the current pandemic.45-47 The proportion of population that must become immune in order to halt the epidemic is given by 1 –1/R, the herd immunity threshold.48 For COVID-19, our estimates imply that approximately 50–60% of the population must be infected or vaccinated in order to attain long-term epidemic control. In reality, this threshold is usually higher due to non-homogenous mixing in populations.48 In early stages of the epidemic in India, we found that restrictions on international travel were effective in limiting the number of imported cases in India, although this is of limited importance once local chains of transmission had been established.49 Since testing of travelers was based on appearance of symptoms,41 asymptomatic imported infections that remained undetected may have played some role in the early spread of COVID-19. A ‘suppression’ strategy (eg: lockdown) aims to arrest epidemic growth by reducing Rt below 1.6,9 After the nationwide lockdown was imposed on 25 March 2020, the mobility levels quickly dropped to low levels, but the Rt continued to increase till 30 March (Figure 2) probably due to inflation of estimated transmission by the Nizammudin cluster (a super-spreading event originating in Delhi)– which represented about 30% of total COVID-19 cases in India in early April, with latest data linking the cluster to 4291 cases across more than 15 Indian states.50 This event adds to the list of multiple COVID-19 super-spreader events around the world, which have caused unexpected spikes in cases.51 It should be noted that clusters may disproportionately inflate transmission estimates, because targeted testing of people linked to the cluster leads to higher test positivity rates. The Rt down trended 30 March onwards, with the most recent estimated Rt of 1·159 (95% CI 1·128–1·189) on 22 April which was the lowest value of Rt yet. Since there was no significant susceptible depletion, this decrease in transmission can be attributed to the intensive social restrictions in place. The trend of Rt from 23 April onwards is of particular interest and it remains to be seen, as to whether the Rt can reach sub-threshold levels (below 1) before the lockdown is relaxed. India is under one of the strictest lockdowns in the world for more than five weeks now,52 and a comprehensive lockdown exit strategy will be needed to consolidate and build upon the gains achieved this far. A sudden and complete lifting of the ongoing nationwide lockdown is not a feasible option since it will lead to a rapid exponential increase in cases due to absence of herd immunity. A lockdown of adequate length and efficacy eventually causes the active cases to peak and then gradually decrease. Once the peak of active cases is reached, extending lockdown farther beyond the peak may have additional benefits due to progressive exhaustion of the infectious pool in the population, which is practically comparable to a lower pre-relaxation prevalence of COVID-19. This has 2 effects–first, the rebound epidemic growth is initially slower which delays the resurgent rise in cases after relaxation of lockdown. This seems to imply that though extending lockdown inherently buys time for preparation, it also adds progressively longer preparation time after the lockdown is relaxed. Second, we find that if lockdown is to be reimposed after a fixed relaxation period, the magnitude of the second peak can be reduced by relaxing the lockdown farther from the first peak. This is of particular interest if an intermittent lockdown strategy is implemented in the future, where measures need to be imposed and relaxed repeatedly. The time gained should be used to strengthen surveillance systems, ramp-up testing capacity and increase health-system preparedness. It is optimal to prevent a second wave from occurring at all, by fine-tuning lockdown relaxation based on serial monitoring of Rt to keep its value under 1.6,8 In this scenario, a later relaxation will allow stabilisation of disease prevalence at a lower value, which can provide a buffer for response if and when a resurgent rise in cases is seen (maintaining Rt=1 implies that the prevalence will remain constant at the pre-relaxation level). These observations may increase the benefit of lockdowns above what is widely known, and can better inform the delicate balance of cost and benefits of such intensive policies. Massive scaling up of testing has been proposed as a lockdown exit strategy.17,53 In this study, we present quantitative evidence based on modelling for the same. Extremely low transmission rates during intensive restrictions are inherently enough to contain the epidemic. However, as transmission rates increase with progressive restoration of normal socio-economic activities post lockdown relaxation, testing assumes an increasingly substantial role in containment. The extent of relaxation that will be possible without causing an untenable rebound in infections, will highly depend on the amount of testing that is done, especially after lockdown relaxation. While having both intensive social distancing policies and very expansive testing may be nonviable, combining the effects of both to a feasible extent can effectively keep the epidemic under control. Our findings align with results seen in countries with an aggressive testing approach, like South Korea and Taiwan where severe restrictions have been avoided. 54 Even if the amount of testing being done during lockdown is deemed to be sufficient, a rapid and massive scaling up of testing capacity is needed preferably before relaxing restrictions. The monetary cost of expanding testing even at a large scale, is expected to be smaller than the cost of implementing intensive social distancing for long periods.11 In addition to supporting the economy, this approach can ameliorate the humongous social and humanitarian implications of imposing population-wide lockdowns, especially in a country such as India. 12,13 Blanket testing of HCWs can be a judicious use of the expanded capacity, considering they are highly exposed personnel, and risk spreading the infection to patients, co-workers and family members if infected. This will limit depletion of an already scarce workforce due to unnecessary quarantine, while also reducing spread from unrecognised asymptomatic infections in HCWs.55 Other essential workforce like law enforcement personnel, grocery vendors, sanitation workers, etc with high contact rates should also be considered. SARS-CoV-1 did not reach the scale of SARS-CoV-2 despite a comparable R due to low community transmissibility and onset of infectivity well after symptom onset which allowed optimal efficacy of traditional control measures like symptom-triggered isolation and contact tracing.19,47 Pre-symptomatic transmission occurs before the onset of symptoms in an eventually symptomatic patient, while asymptomatic transmission occurs through patients who never become symptomatic. Presence of both these features in COVID-19 is a significant deterrent for control strategies.16,19,37,56 In such a scenario and R~2.5, modeling studies indicate that controlling COVID-19 outbreaks through contact tracing and isolation alone is not possible.15 However, contact tracing systems should be strengthened since they are a prerequisite for expanded testing of contacts, and they may achieve significant containment at lower Rt. Based on our findings, it is possible that detecting more asymptomatics through testing impedes transmission to an extent where the total number of infections and thus the number of symptomatic cases decreases (Figure 5), relieving burden upon the healthcare system and reducing mortality. This finding will increasingly approximate reality if asymptomatics play a larger role in transmission. A case in point is a blanket testing study done in a small town of Italy which achieved almost complete outbreak control.17 Although blanket testing is not practical for larger implementation, it further highlights the importance of detecting and isolating asymptomatics in controlling COVID-19 outbreaks. A symptom-based monitoring approach during quarantine will miss asymptomatic infections, who will escape the quarantine net and go on to spread the disease. With emerging evidence of infectious asymptomatics, it is prudent to modify the public health response to address these concerns. Thus, all contacts should ideally be tested at the end of quarantine irrespective of symptoms. In settings where testing all contacts is not yet possible, extended quarantine periods upto 28 days may be considered, which have two-fold benefits. First, almost all asymptomatics finish their infectious period before 28 days, and second, more symptomatics can be detected by day 28 (only 2 out of 10000 symptomatic cases are missed by day 28, compared to 101 cases by day 14).29 Such extended quarantines are already in place in certain parts of India and China.57,58 Currently, a 14 day quarantine is recommended based on studies of incubation period of COVID-1 9,29,42 but studying the incubation period inherently assumes an onset of symptoms. It is encouraging to note that the need for expanded testing can be supported by high-throughput machines and by pooling of samples.41 Pooling can also be used for community surveillance, and has the potential to drastically increase detection capabilities while saving costs and resources and should be used wherever possible, while also enhancing research to boost pool size and testing accuracy.59 While contact tracing, isolation and testing are important, the role of behaviour change in reducing transmission must not be underestimated. Asymptomatic people are themselves less likely to take appropriate precautions, and people use less caution around other people who don’t have symptoms. Universal mask wearing in public spaces should be encouraged, and mandated by policy if required.60 Since ensuring long-term compliance of citizens to health advisories and public restrictions will be another challenge, transparent and proactive communication by authorities along with continued social support for vulnerable groups will be essential. There is looming uncertainty regarding the burden of asymptomatics and the role they play in transmission. Estimates for asymptomatics range from 18% to 80% of total infections.17,18,32–34 With India and China reporting asymptomatics in the higher ranges, it may be possible that young and developing countries have a high proportion of asymptomatic carriers. These estimates are cross-sectional, and thus do not differentiate asymptomatics from pre-symptomatics. It is only retrospectively that the true burden of asymptomatics can be ascertained through serological studies, which will also help us to understand the true fatality rate of COVID-19.61 Epidemiologic and virologic studies have established that asymptomatics are infectious, and have similar upper respiratory viral loads as symptomatic patients.16,17,30,31 This is in comparison to influenza, where asymptomatics have lower viral loads and thus are less infectious.19 Studies to evaluate pre-symptomatic transmission of COVID-19 have shown that 25–50% of total transmission occurred before the index case showed symptoms,31,56 although quantitative evidence of asymptomatic transmission is lacking and deserves further research. There is an urgent need to identify these gaps in the understanding of SARS-CoV-2 in order to grasp the true size and severity of this pandemic and plan future strategies accordingly. Blanket interventions have been effective to suppress the pandemic till now, but targeted interventions will be key as we move forward. Various interventions need to be stratified based on how effectively they suppress viral transmission and the amount of disruption they cause. Cost effectiveness analysis must be done, and bundles of interventions that together achieve high efficacy with least accompanying disruption should be deployed. Highly effective and disruptive interventions should be targeted to areas with active hotspots and high community transmission. It will be essential to build robust disease surveillance systems to assess the relative impact of each intervention in real-time and reduce the time delay to response. Expanded testing and strengthened contact tracing will enable this by reducing the reporting lag and rapidly detecting any surge in cases. Instead of adopting an intermittent lockdown policy, where lockdowns are treated as either ‘on’ or ‘off’,9 some countries have adopted a staged alert system for responding to the COVID-19 pandemic,62 where a geographical area may move up and down alert levels, to reflect the level of suppression that the local outbreak situation demands. Similarly, India has recently stratified its districts into red, orange and green zones based on surveillance trends, in preparation for a staggered relaxation of lockdown.63 Such social distancing policies which are dynamic with respect to geography and time are direly needed as we move into a time of relative uncertainty post lockdown relaxation. Several limitations of our study should be noted. First, we assumed the delay from symptom onset to confirmation to be similar to the delay from symptom onset to isolation due to lack of data, and that this delay is uniform across India. Though the latter approximates the former, further studies to ascertain the true reporting lag in India are needed since it is critical for identifying Rt changes at correct points in time.6 Second, our model does not factor in pre-symptomatic transmission, which along with a short estimated reporting lag may underestimate the transmission by symptomatic cases and thus inflate the effect of detecting asymptomatics. Third, since our primary goal was to evaluate the effect of identifying asymptomatics, we assumed that increased testing increases the detection of asymptomatics only, while in reality it would detect more cases across the clinical spectrum. However, the interpretations regarding impact of increased testing are not sensitive to this assumption, which has been discussed through an alternate interpretation of the model (appendix p4). Fourth, we assumed a constant death rate (κ), in contrast to reality where the death rate gradually decreases during an epidemic to ultimately converge at the near-actual death rate 61. Thus, we refrained from forecasting deaths due to obvious bias in the prevailing death rate. Notwithstanding the limitations, this study reports a reporting delay estimate from India for the first time, which can be used in future modeling studies. Here, we built a mathematical model which can account for the dynamics of lockdown imposition and relaxation, varying levels of case detection, lag to symptom onset and case reporting, while simultaneously allowing to test the range of asymptomatic burden and transmissibility. Since we have presented findings across the range of uncertainty regarding asymptomatics (appendix), our results are robust to emerging evidence. Though our model is fitted to data from India, we expect the insights into lockdown relaxation and testing impact to be generalisable. In conclusion, though disruptive, the world’s largest lockdown in India has been effective in reducing the transmission levels of COVID-19. To avoid a resurgence in cases, a dynamic relaxation approach guided by regional monitoring of effective reproduction numbers is recommended, and this relaxation should be farther from the peak of active cases as feasible. Asymptomatic infectives could be a considerable challenge to long-term containment efforts, and increased detection will play an increasingly pivotal role once restrictions start to be lifted. The amount of testing will dictate the extent of resumption of socio-economic activities, and authorities should scale-up testing capacity as a priority. Further, control measures should be appropriate in the social context of the population, as this pandemic brings a humanitarian crisis in addition to a public health one, especially in vulnerable populations across the world. ## Data Availability The code for the simulations and detailed results are available at the following GitHub repository: [https://github.com/technosap/SEIR\_QDPA-COVID\_19](https://github.com/technosap/SEIR_QDPA-COVID_19). The primary data for calculation of reporting lag is available in the appendix. All other data is from publicly available datasets. ## Funding No funding was received for this study. ## Contributors MG and GGP conceptualised the study. MG and SSM designed the model. SSM programmed the model and performed formal analyses. MG wrote the original draft. SSM, AR, MAg, MAr and AMa assisted in manuscript preparation and reviewed existing evidence. GGP, AL, PB and AB consulted on the analyses and reviewed the draft. SSM and MG made the figures. RK, VPM, PT, SB and AMo collected and curated primary patient data and enabled inter-department coordination at the hospital. GGP and MG supervised the project. All authors interpreted the results, contributed to writing the manuscript, and approved the final version for submission. ## Data sharing and Code availability The code for the simulations and detailed results are available at the following GitHub repository: [https://github.com/technosap/SEIRQDPA-COVID19](https://github.com/technosap/SEIRQDPA-COVID19). The primary data for calculation of reporting lag is available in the appendix. All other data is from publicly available datasets. ## Declaration of interests We declare no competing interests. ## Ethics Approval All relevant ethical guidelines have been followed; necessary IRB and/or ethics committee approvals have been obtained (through *Institute Ethics Committee AIIMS Delhi; Ref. No. IECPG-166/23.04.2020*). All necessary patient/participant consent has been obtained. ## SUPPLEMENTARY APPENDIX CONTENTS 1. ***Model Description*** * 1.1. From SEIR to SEIR-QDPA Model * 1.2. An alternate interpretation of testing 2. ***Reporting Lag and Incidence by Onset*** * 2.1. Table S1. Reporting Lag Raw Data (n=53) * 2.2. Methodology for choosing the best fit distribution * 2.3. Table S2. Reporting Lag Distribution Fit Parameters * 2.4. Table S3. Reporting Lag Distribution Goodness-of-Fit Analysis * 2.5. Sampling of Lag-adjusted Onset Dates 3. ***R0 Estimates with EG and ML methods and R0 for states of India*** * 3.1. Methodology for Estimating R0 * 3.2. Table S4. Fit Parameters for R estimation * 3.3. R0 Estimation for States with more than 50 cases on 20/04/20 * 3.4. Table S5. R0 estimates for 12 Indian States * 3.5. Figure S1. Sensitivity of India R0 estimates to Serial Interval and Time Period. 4. ***Testing criteria for COVID-19 in India*** * 4.1. Table S6. Federal testing recommendations provided by the Indian Council of Medical Research (ICMR) 5. ***Estimation of Effective Reproduction Number* (***Rt***)** * 5.1. Additional Information about estimation of Rt * 5.2. Table S7. Raw Data for Imported Cases * 5.3. Rt estimates for States * 5.4. Table S8. Rt point estimates for national COVID-19 incidence data (without import adjustment) * 5.5. Table S9. Rt point estimates for national COVID-19 incidence data (import adjusted * 5.6. Figure S2: Incidence by onset and effective reproduction rates (Rt) up to 11 April 2020 for Indian states of Maharashtra, Gujarat, Kerala, Delhi, Rajasthan, and Madhya Pradesh. * 5.7. Figure S3. Rt for other Indian States up to 11 April 2020. 6. ***Model fitting and Fit Sensitivity Analysis*** * 6.1. Fitting Methods, Initial Conditions and Uncertainty estimation * 6.2. Figure S4. Best Fit to Data for Base Assumptions (ai = 05, pa = 06, fa = 01). * 6.3. Assumption of constant Recovery Rate and Mortality Rate * 6.4. Sensitivity of estimated parameters to assumptions of ai, pa, and fa * 6.5. Figure S5. Sensitivity of fit parameters α, β, δs to assumptions of ai, pa, and fa. * 6.6. Figure S6. Sensitivity of fit parameters λ and κ to assumptions of ai, pa, and fa. * 6.7. Figure S7. Sensitivity of fit parameters δs to assumptions of ai, pa, and fa 7. ***Prediction Sensitivity*** * 7.1. Figure S8. Sensitivity of predictions to assumptions of ai, pa, and fa. 8. ***Calculation of R0 from the Model Parameters*** * 8.1. Calculating R0 from SEIR-QDPA Model * 8.2. Figure S9. Estimation and Sensitivity of Model R0 to assumptions. 9. ***Sensitivity Analysis of the effect of testing rate and social distancing policies*** * 9.1. Figure S10. Sensitivity of effect of testing rate and social distancing policies on total symptomatic cases to probability of asymptomaticity at 15 days after lockdown relaxation. * 9.2. Figure S11. Sensitivity of effect of testing rate and social distancing policies on total symptomatic cases to probability of asymptomaticity at 45 days after lockdown relaxation. * 9.3. Figure S12. Sensitivity of the number of symptomatic cases to probability of asymptomaticity over a 75-day period after lockdown relaxation if the testing rate and social distancing policies are not changed after relaxation. * 9.4. Figure S13. Sensitivity of effect of testing rate and social distancing policies on total infections to probability of asymptomaticity at 15 days after lockdown relaxation. * 9.5. Figure S14. Sensitivity of effect of testing rate and social distancing policies on total infections to probability of asymptomaticity at 45 days after lockdown relaxation. 10. ***Effect of Gradual Lockdown Relaxation*** * 10.1. Simulating gradual lockdown relaxation * 10.2. Figure S15. Gradual complete lockdown relaxation at three different points of time. * 10.3. Figure S16. Lag before new rise in active case after Slow Lockdown Relaxation. * 10.4. Figure S17. Gradual lockdown relaxation for 14 days at two different points of time. 11. *Lockdown Relaxation and the Healthcare System Capacity* * 11.1. Estimating the Indian Healthcare System Capacity * 11.2. Estimating the time taken to reach the Healthcare system capacity under lockdown relaxation scenarios. * 11.3. Figure S18. Time taken to reach the healthcare system capacity under lockdown relaxation scenarios. 12. ***Google COVID-19 Mobility Reports*** * 12.1. Background Information on how Google’s mobility reports are generated ## 1. Model Description ### 1.1 From SEIR to SEIR-QDPA Model In order to accurately describe the progression of the epidemic in the population, we generalize and redefine the 4 compartments of a SEIR model ie. [S(t), E(t), I(t), R(t)]. We redefine the I(t) compartment as the number of undetected infective people and split it into two compartments signifying the undetected symptomatic infectives Is(t) and undetected asymptomatic infectives Ia(t). Exposed patients go to either of the compartments at a fixed ratio of asymptomatics to symptomatics given by r = pa/(1-pa) where pa is the probability of being asymptomatic. We set the rates of transition from E(t) to Ia(t) and Is(t) as rγ and γ respectively. This results in the distribution of exposed cases into the two infective compartments such that for each symptomatic case, there are r asymptomatic cases, causing a fixed fraction (pa) of the cases being asymptomatic. In our model, we further assume that a fixed fraction of asymptomatic infectious cases gets detected (fa) and go to the asymptomatic quarantine compartment Qa(t) with detection rate atδa where at = fa/(1-fa), and the rest go into an undetected recovery compartment Ru(t) with recovery rate for asymptomatics δa. The choice of rates leads to the distribution of asymptomatics into detected quarantined and undetected recovered in the ratio of at, i.e. a fixed fraction (fa) of asymptomatics are detected. He et. al. determined that infectiousness declines significantly after a period of 7 days by studying the distributions of incubation period and serial interval.1 Wölfel et al found that live virus could not be isolated after day 8 in spite of continuing high viral loads in young cases with mild symptoms.2 Backed by both epidemiological and biological evidence, we set the time spent by an asymptomatic being infectious as 8 days. The δa then becomes ⅛ days-1. We assume that all symptomatic cases get detected with an average reporting lag ![Graphic][3] after which they go to the symptomatic quarantine compartment Qs(t). We refer to the symptomatic and asymptomatic quarantine compartment together as the quarantine compartment Q(t). We assume that transition of a case from Ia(t) and Is(t) to Q(t) entails isolating, testing and reporting the case and that this isolation is 100% effective. From the symptomatic quarantine compartment, cases go to either the Rs(t) or D(t) denoting symptomatic recovered cases and fatal cases with recovery rate λ and mortality rate κ respectively. We assume that the recovery of asymptomatic cases in quarantine occurs at the same rate λ as symptomatics. Thus, from the asymptomatic quarantine compartment, cases go to Ra(t) denoting detected asymptomatic recovered cases with the recovery rate λ. We assume here that no asymptomatic case dies due to the infection. We refer to the symptomatic and detected asymptomatic recovered compartment together as the recovered compartment R(t). ![Formula][4] ![Formula][5] Finally, we introduce an extension previously described by Peng et al where a new compartment P(t) is added to account for the combined effect of social-distancing by individuals and lockdown of infected regions within the country thus reducing the pool of susceptible individuals in the population at a protection rate α.3 Further, to model the effect of the relaxation of such measures, we introduce a deprotection rate σ with which individuals can leave the P(t) compartment thus increasing the susceptible population S(t). The value of σ is set to zero till any preventive policies are withdrawn. The dynamics of the model are described by the following ordinary differential equations (ODEs): ![Formula][6] where N is the total population of the country and is assumed to be constant. The compartments [S(t), E(t), Is(t), Ia(t), Qs(t), Qa(t), Rs(t), Ra(t), D(t), Ru(t), P(t)] denote the susceptible, exposed, undetected symptomatic, undetected asymptomatic, quarantined symptomatic, quarantined asymptomatic, recovered symptomatic, recovered asymptomatic, fatalities, undetected recovered, and protected individuals respectively at time t. ### 1.2. An alternate interpretation of testing in our model An increase in testing criteria and capacity would naturally have the largest impact on increasing the detection of infections with no or mild symptoms, considering that most infections with moderate-severe symptoms are more likely to be detected even at baseline testing. Since our primary goal in this study was to evaluate the effect of detecting more asymptomatics, we assumed that increased testing increases the detection of asymptomatics only, while in reality it would detect more cases across the clinical spectrum. However, the interpretations regarding impact of increased testing are not sensitive to this assumption, which has been discussed here through an alternate interpretation of the model. Our 11-compartment model allows some flexibility in terms of interpretation. If instead of interpreting the Is and Ia compartment as the Symptomatic and Asymptomatic infected individuals, we consider them as the Detectable and Undetectable infections (irrespective of symptoms), our model parameter pa now represents 1-ascertainment proportion (ascertainment proportion is the proportion of total infections that are considered detectable at baseline) and increasing the fa now becomes equivalent to improving the testing by reducing the bias towards detectable cases and improves detection of all cases irrespective of symptoms. Our model takes into account that undetected infected spend more time in an infective state in the population![Graphic][7] compared to the detected cases![Graphic][8], which reflects reality. Thus, this alternate view allows us to generalize our results with respect to how the testing policies affect the epidemic after lockdown relaxations. One limitation of this interpretation is that we assume asymptomatic cases (here, undetectable cases) have lower infectivity than symptomatics (here, detectable cases) in a homogenous social mixing environment. Although undetectable cases are more likely to be asymptomatic, this assumption does not necessarily hold true in this alternate interpretation. Still, this limitation would result in an underestimation of the impact of testing rather than an overestimation. ## 2. Reporting Lag and Incidence by Onset ### 2.1 Table S1. Reporting Lag Raw Data (n=53) View this table: [Table3](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T3) ### 2.2 Methodology for choosing the best fit distribution Gamma and Exponential distributions were fit to the Reporting Lag data with Maximum Likelihood Estimation using the *fitdistrplus* package in R 3·6·3.4 Due to the small sample size and lack of reliable prior information about the distribution of reporting lag, the maximum likelihood method was chosen. Since the data was found to have a heavy tail, we chose to fit gamma and exponential distributions. The goodness of fit was estimated using the *Akaike information criterion* (*AIC*) and *Bayesian Information Criterion* (*BIC*). The distribution with a better BIC was selected as the appropriate descriptor of the Reporting Lag distribution. ### 2.3 Table S2. Reporting Lag Distribution Fit Parameters View this table: [Table4](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T4) ### 2.4 Table S3.Reporting Lag Distribution Goodness-of-Fit Analysis View this table: [Table5](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T5) ### 2.5 Sampling of Lag-adjusted Onset Dates From the estimated distribution of Reporting Lag, 1000 samples of the fitted distribution parameters (φi) were drawn taking into account the uncertainty in the distribution parameters ie. shape and scale for the gamma distribution and rate for the exponential distribution to serve as the posterior distribution of reporting lag. For each of the 1000 samples of fitted parameters, the reporting dates (ri) were transformed to give the symptom onset date (oi) by the formula: ![Formula][9] where li ~ Gamma(φi) or Exp(φi) resulting in 1000 lag adjusted datasets. This process was applied to the reporting incidence data for the nation and the different states. ## 3. R Estimates with both EG and ML methods and R for states of India ### 3.1 Methodology for Estimating R For the estimation of R, we use Maximum Likelihood (ML) method as described by White & Pagano and the Exponential Growth (EG) method as described by Wallinga & Lipsitch in R 3·6·3 using the *R0* package.5,6 The best fit time period was chosen for calculation of confidence intervals for R. As we had adjusted for imported cases, we reported the results of the method which gave a more reliable fit with a better R2 score as the R. We also analysed the sensitivity of R to the serial interval distribution chosen by keeping standard deviation fixed at 3 and varying the mean from 2 days to 7 days. ### 3.2 Table S4. Fit Parameters for R estimation View this table: [Table6](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T6) ### 3.3 R Estimation for States with more than 50 cases on 20/04/20 The R could be estimated for the states of Kerala, Maharashtra, Delhi, Rajasthan, Tamil Nadu, Uttar Pradesh, Telangana, Andhra Pradesh, Karnataka, Jammu & Kashmir, Haryana and Punjab. The ML algorithm failed to converge for the states of Gujarat, West Bengal, Madhya Pradesh and Bihar. The fits for the states had a very low R2 due to multiple factors such as a major impact of import of new cases from other states (significant for some states due to import of cases from the Delhi cluster) in the estimation of R and policies implemented by the government to curb the spread. When calculating the R, we were unable to adjust for imported cases as state-wise imported case data was not available. Thus, we used Maximum Likelihood approach for estimating R for the states as it is more robust for unaccounted import of cases. ### 3.4 Table S5. R estimates for 12 Indian States View this table: [Table7](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T7) ![3.5 Figure S1](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F6.medium.gif) [3.5 Figure S1](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F6) 3.5 Figure S1 Sensitivity of India R estimates to Serial Interval and Time Period (A,B) Sensitivity Analysis of R estimates with respect to the choice of the time period over which the R was calculated for EG and ML methods. (C,D) Goodness of Fit across a choice of time periods over which the R was calculated for EG and ML methods. Red represents the best fit. (E,F) Sensitivity Analysis of R estimates with respect to the distribution of Serial Intervals across which the R was calculated for EG and ML methods. All graphs were generated in R package with R 3·6·3. ## 4. Testing criteria for COVID-19 in India7 ### 4.1 Table S6. Federal testing recommendations provided by the Indian Council of Medical Research (ICMR) View this table: [Table8](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T8) ## 5. Estimation of Effective Reproduction Number (Rt) ### 5.1 Additional Information about estimation of Rt When evaluating the Rt in *EpiEstim* package in R 3·6·3 with the method described by Wallinga and Teunis, we consider a gamma prior with mean 2·6 and SD 2, and 5-day sliding windows.8 The serial interval distribution was assumed to follow a gamma distribution with mean 3·96 days (95% CI 3·53-4·39, SD = 0·215) and a standard deviation of 4·75 (95% CI 4·46–5·07, SD = 0·145) days based on a study done in Wuhan on 468 infector-infectee pairs.9 Correction for the imported cases was done for the national incidence data after adjusting for reporting lag. The estimates of the effective reproduction number for each day were combined for the 1000 lag adjusted datasets by calculating pooled mean and pooled standard deviation and a net estimate for 50% and 95% confidence intervals were calculated. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ### 5.2 Table S7.Raw Data for Imported Cases. (Source-COVID19India10) View this table: [Table9](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T9) ### 5.3 Table S8. Rt point estimates for national COVID-19 incidence data (without import adjustment) View this table: [Table10](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T10) ### 5.4 Table S9. Rt point estimates for national COVID-19 incidence data (import adjusted) View this table: [Table11](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/T11) ### 5.5 Rt trends for States The Rt was evaluated for states with more than 50 cases on 20th April 2020 other than Odisha due to a lack of active cases until very recently. The method to estimate was the same as the method used for the national data. Imported cases could not be adjusted due to lack of state-wise data of internationally imported cases and for interstate transport of COVID-19 cases. The estimates of the effective reproduction number for each day were combined for the 100 lag adjusted datasets by calculating pooled mean and pooled standard deviation and a net estimate for 50% and 95% confidence intervals were calculated. All bootstraps were done in Python 3·6 and plots were made in matplotlib 2·2·2. We examined the Rt curves of the states of Maharashtra, Gujarat, Delhi, Rajasthan and Madhya Pradesh which together accounted for 66% of total cases in India as on 28 April 2020. Delhi, Maharashtra, Rajasthan and Madhya Pradesh showed the Nizammudin cluster-related rise in Rt during late-March, with Delhi being the origin of the cluster showing the highest rise. Gujarat, in contrast, showed a rise later in early-April which was probably linked to increased local transmission in the state capital of Ahmedabad. We also included Kerala since it was the first state to reach Rt =1 in early April and has sustained low transmission levels since then. All states except Madhya Pradesh showed a continuing downtrend on the last examined date. The composite of these trends roughly depicts the trend of Rt of India. Kerala’s ability to maintain a low Rt may be attributed to its high literacy rate, active community participation and decentralisation of power-village councils taking the initiative to contain cases locally, use of drones for cluster containment, and social welfare initiatives which provided isolation rooms for people with cramped homes which didn’t allow social distancing from family. Kerala’s health-system response has been proven to be robust in earlier outbreaks like the Nipah virus outbreak in 2018. ![5.5 Figure S2:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F7.medium.gif) [5.5 Figure S2:](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F7) 5.5 Figure S2: Incidence by onset and effective reproduction rates (Rt) up to 11 April 2020 for Indian states of Maharashtra, Gujarat, Kerala, Delhi, Rajasthan, and Madhya Pradesh The state name is green if the most recent 50% CI of Rt is below 1 as on 11 April 2020, otherwise red. State-level Incidence by Onset and Effective Reproductive Number (Rt) number calculated with time-dependent Maximum Likelihood Method with 5 day sliding windows and 100 lag adjusted datasets. All bootstraps were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![5.6 Figure S3](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F8.medium.gif) [5.6 Figure S3](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F8) 5.6 Figure S3 Rt for other Indian States up to 11 April 2020 The state name is green if the most recent 50% CI of Rt is below 1, otherwise red. State-level Incidence by Onset and Effective Reproductive Number (Rt) number calculated with time-dependentMaximum Likelihood Method with 5 day sliding windows and 100 lag adjusted datasets. All bootstraps were done in Python 3 6 and plots were made in matplotlib 2·2·2. ## 6. Model fitting and Fit Sensitivity Analysis ### 6.1 Fitting Methods, Initial Conditions and Uncertainty estimation Fitting was done using Least-Square Fitting with Trust Region Reflective algorithm in *Scipy* package (Python 3·7). This method was chosen as it allows for a bound on the fitted parameters and was found to converge reliably. The data was fit from the date the number of active cases crossed 100 individuals. Due to lack of data for E(t) and Is(t) compartment, we also fit for the initial conditions for these compartments. Initial conditions for Ia(t) and Ru(t) were set in the ratio of asymptomatics to symptomatics with respect to the initial values of Is(t) and R(t) compartments respectively. Since the number of detected recovered cases was very small we assume all of them were asymptomatic and thus Ra(t) was initially set to zero. All other compartments were set to zero with the exception of S(t) which was set to ensure the sum of all compartments was N. Newer values were given higher weightage for the fitting by supplying the fitting algorithm with errors given by e = w (1-w)t where t is the index of the sample and w = 0·5 is an arbitrary error weightage parameter as described by Chowell. 11 In order to estimate the uncertainty in the parameter fit we adapt the Bootstrap approach described by Chowell using a Poissonian error structure without overdispersion and determine the 95% confidence intervals of the parameters with 1000 bootstrap samples. 11 All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![6.2 Figure S4](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F9.medium.gif) [6.2 Figure S4](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F9) 6.2 Figure S4 Best Fit to Data for Base Assumptions (ai = 0·5, pa = 0·6, fa = 0·1) ‘+’ represent data with which the model was trained, ‘o’ refers to data points for dates after the training limit. Bands represent 95% confidence intervals for the mean prediction. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ### 6.3 Assumption of constant Recovery Rate (λ) and Mortality Rate (κ) Our model fails to predict the increased rise in recovery as it assumes a constant recovery rate (λ) but in reality, the recovery rate (λ) is often found to increase during the progress of an outbreak due to multiple factors. Similarly, mortality (κ) is also found to reduce. Due to lack of sufficient data about the nature of the increase in these rates, the model was assumed to have a constant value that fits well to early dynamics of epidemics. In reality, a more flexible form such as a sigmoid may be a better assumption. ### 6.4 Sensitivity of estimated parameters to assumptions of ai, pa, and fa The sensitivity of the fit parameters to our assumptions were evaluated and it was found that recovery rate (λ) and mortality rate (κ) is independent of the assumptions, which is expected as these parameters are directly dependent on the relation between quarantine, recovery and fatality data. On the other hand, while there was some variability in the values of protection rate (between 0·032–0·048) and reporting rate (between 0·40–0·48) many of which were within the confidence interval, a large variability was observed in the value of transmission rate (0·8–2·0) with higher transmission rates favoured when the ratio of asymptomatic was low and their relative infectivity is also low. ### 6.5 Figure S5. Sensitivity of fit parameters α and β to assumptions of ai, pa, and fa. All values are reported with 95% CI (empirical estimate) with 1000 bootstrapped fits to the dataset as described in Chowell (2012)11. All plots were made in matplotlib 2·2·2. ![6.5 Figure S5](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F10.medium.gif) [6.5 Figure S5](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F10) 6.5 Figure S5 Sensitivity of fit parameters k and k to assumptions of ai, pa, and fa All values are reported with 95% CI with 1000 bootstrapped fits to the dataset as described in Chowell (2012)11. All plots were made in matplotlib 2·2·2. ![6.6 Figure S6](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F11.medium.gif) [6.6 Figure S6](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F11) 6.6 Figure S6 Sensitivity of fit parameters δ to assumptions of ai, pa, and fa All values are reported with 95% CI with 1000 bootstrapped fits to the dataset as described in Chowell (2012)11. All plots were made in matplotlib 2·2·2. ![6.7 Figure S7](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F12.medium.gif) [6.7 Figure S7](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F12) 6.7 Figure S7 Sensitivity of fit parameters δS to assumptions of ai, pa, and fa All values are reported with 95% CI with 1000 bootstrapped fits to the dataset as described in Chowell (2012)11. All plots were made in matplotlib 2·2·2. ## 7. Prediction Sensitivity ![7.1 Figure S8.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F13.medium.gif) [7.1 Figure S8.](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F13) 7.1 Figure S8. Sensitivity of predictions to assumptions of ai, pa, and fa All values are reported with 95% CI with 1000 bootstrapped fits to the dataset as described in Chowell (2012).11 Cases counts are reported in thousands and dates are reported with respect to 18th April 2020. All plots were made in matplotlib 2·2·2. ## 8. Calculation of R from the Model Parameters ### 8.1 Calculating R from SEIR-QDPA Model The Basic Reproduction number (R) can be calculated for the model using the next-generation matrix method as described by van den Driessche (2017).12 When there is a non-zero probability of asymptomaticity (pa), the R is given by: ![Formula][10] When the probability of asymptomaticity is zero, the model reduces to an SEIR at the initial time point and the R is simply given by: ![Formula][11] A very important factor that should be taken into consideration when interpreting the value of R given by the model is that unlike other simpler models such as SIR or SEIR models, R is not a descriptor of the dynamics at all points of time of the epidemic. This is because of the introduction of the “Protected” compartment. The better descriptor of the dynamics is given by the effective R given by: ![Formula][12] which is a function of time that takes into account the reduction in the number of susceptibles in the population as they enter the “Protected” compartment. The true effective reproductive number Rt is given by: ![Formula][13] Since our reduction in susceptible population necessarily decays exponentially from N and asymptotic to zero, our estimates of R from the model are artificially higher than R directly estimated from data in order to explain the dynamics of the epidemic where the spread is very fast in early stages but saturates before the spread to less than the growth limit of Rt = 1. While our assumption of the constant protection rate with no leakage, may stand true for small ideal communities, for a country with non-homogenous responses to the crisis this assumption may not stand true. ![8.2 Figure S9.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F14.medium.gif) [8.2 Figure S9.](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F14) 8.2 Figure S9. Estimation and Sensitivity of Model R to assumptions All values are reported with 95% CI with 1000 bootstrapped fits to the dataset as described in Chowell (2012).11 The R is found to be lowest in the scenarios of low asymptomatic infectivity, high probability of asymptomaticity and high detection rate and conversely the highest at high asymptomatic infectivity, low probability of asymptomaticity and low detection rate. All plots were made in matplotlib 2·2·2. ## 9. Sensitivity Analysis of effect of testing rate and social distancing policies ![9.1 Figure S10.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F15.medium.gif) [9.1 Figure S10.](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F15) 9.1 Figure S10. Sensitivity of effect of testing rate and social distancing policies on total symptomatic cases to the probability of asymptomaticity at 15 days after lockdown relaxation All values are reported in thousands of cases. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. When the proportion of asymptomatics is 80%, and detection is increased from 10% to 20% for 15 days– for every extra asymptomatic detected by increased testing, the number of infections prevented is 7.2 at β, 5.5 at 0·8β and 3.8 at 0·5β. When the proportion of asymptomatics is 40%, and detection is increased from 10% to 20% for 15 days– for every extra asymptomatic detected by increased testing, the number of infections prevented is 9.2 at β, 7.1 at 0·8β and 3.8 at 0·5β. ![9.2 Figure S11](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F16.medium.gif) [9.2 Figure S11](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F16) 9.2 Figure S11 Sensitivity of effect of testing rate and social distancing policies on total symptomatic cases to probability of asymptomaticity at 45 days after lockdown relaxation All values are reported in thousands of cases. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![9.3 Figure S12](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F17.medium.gif) [9.3 Figure S12](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F17) 9.3 Figure S12 Sensitivity of the number of symptomatic cases to the probability of asymptomaticity over a 75-day period after lockdown relaxation if the testing rate and social distancing policies are not changed after relaxation It is observed that the order of the trends reverses over time. This can be explained by the fact that when the epidemic was under control due to a large fraction of the population being protected, lower values of pa reduced the total number of infections due to decreased asymptomatic spread. But when the lockdown was relaxed, the incidence of symptomatic cases reflected the probability of being symptomatic which is higher when pa is lower and vice versa. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![9.4 Figure S13.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F18.medium.gif) [9.4 Figure S13.](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F18) 9.4 Figure S13. Sensitivity of effect of testing rate and social distancing policies on total infections to the probability of asymptomaticity at 15 days after lockdown relaxation All values are reported in thousands of cases. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![9.5 Figure S14.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F19.medium.gif) [9.5 Figure S14.](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F19) 9.5 Figure S14. Sensitivity of effect of testing rate and social distancing policies on total infections to probability of asymptomaticity at 45 days after lockdown relaxation All values are reported in thousands of cases. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ## 10. Effect of Slow Lockdown Relaxation ### 10.1 Simulating gradual lockdown relaxation In order to simulate a slow lockdown release, instead of setting the value of deprotection rate (σ) as a large value, we set it equal to the value of protection rate (α) such that the lockdown reverts itself at the same rate as it occurred. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![10.2 Figure S15](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F20.medium.gif) [10.2 Figure S15](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F20) 10.2 Figure S15 Gradual complete lockdown relaxation at three different points of time The delayed gradual relaxation of the lockdown buys more time before the number of active cases starts rising again as observed in the fast lockdown relaxation. Bands represent 95% confidence intervals for the mean prediction. All simulations were done in Python 3 6 and plots were made in matplotlib 2·2·2. ![10.3 Figure S16](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F21.medium.gif) [10.3 Figure S16](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F21) 10.3 Figure S16 Lag before new rise in active case after Slow Lockdown Relaxation. The time gained by delaying the relaxation is found to be linearly increasing (Pearson’s R: 0·987; 95% CI 0·949–0·997; p<0·0001). The gain in time in the case of a slow lockdown relaxation is found to be more than the case of a fast lockdown relaxation. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ![10.4 Figure S17.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F22.medium.gif) [10.4 Figure S17.](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F22) 10.4 Figure S17. Gradual lockdown relaxation for 14 days at two different points of time. The delayed gradual relaxation of the lockdown reduced the number of active cases in the second as observed in the fast lockdown relaxation. A slow relaxation reduced the effect of relaxation on the growth of the second peak compared to the fast lockdown relaxation. Bands represent 95% confidence intervals for the mean prediction. All simulations were done in Python 3·6 and plots were made in matplotlib 2·2·2. ## 11. Lockdown Relaxation and the Healthcare System Capacity ### 11.1 Estimating the Indian Healthcare System Capacity Previous studies have estimated the number of ICU beds in India to be around 35,699 to 57,119 ICU beds.13 We assume the number of ICU beds in India to be around 70,000 to account for growth in the healthcare system. If we assume that 50% of these beds are equipped with ventilators and are allocated exclusively to COVID-19 patients, that gives us 35,000 COVID-19 ready ICU beds. Previous studies suggest that approximately 5% of symptomatic patients need ICU support,14 that gives us an upper bound for the healthcare system capacity of 7,00,00 active symptomatic cases. ### 11.2 Estimating the time taken to reach the Healthcare system capacity under lockdown relaxation scenarios We simulate the model under the different (fast) lockdown relaxation scenarios we described earlier and determined the time taken to reach the healthcare system capacity of 7,00,000 active symptomatic cases from the date of relaxation. We find that for all finite relaxation durations, there exists a delay in lockdown relaxation that prevents the scenario of the health capacity being overloaded. For smaller relaxation periods, this delay is found to be smaller. It should be noted that these assumptions assume fast and complete lockdown relaxations periods for the entire country at once. In reality, the lockdown relaxations will likely be introduced on a state-by-state basis and may be more of a slow staggered response. ![11.3 Figure S18](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/18/2020.05.13.20096826/F23.medium.gif) [11.3 Figure S18](http://medrxiv.org/content/early/2020/05/18/2020.05.13.20096826/F23) 11.3 Figure S18 Time taken to reach the healthcare system capacity under lockdown relaxation scenarios. Blank blocks represent the scenario where the healthcare system capacity is never reached and the epidemic is under control. Figure shows the heatmap for the mean time taken to hit the healthcare capacity under different lockdown durations and dates of start of lockdown relaxations. ## 12. Google COVID-19 Mobility Reports ### 12.1 Background Information on how Google’s mobility reports are generated15 The mobility scores are calculated by comparing visits and length of stay at different places change compared to a baseline value for the day of the week. The baseline values were calculated from the median values during the 5-week period Jan 3–Feb 6, 2020. These values were calculated on the basis of anonymized data from users who have opted-in to Location History for their Google Account. We acknowledge that this mobility data may be biased towards users of this service, and may or may not represent the exact behavior of a wider population. For the analysis of mobility, Google has divided the mobility into 6 categories relevant to social distancing policies and practices. They are: Grocery & Pharmacy (places like grocery markets, food warehouses, farmers markets, speciality food shops, drug stores, and pharmacies), Parks (places like local parks, national parks, public beaches, marinas, dog parks, plazas, and public gardens), Transit stations (places like public transport hubs such as subway, bus, and train stations), Retail & Recreation (places like restaurants, cafes, shopping centres, theme parks, museums, libraries, and movie theatres), Residential (places of residence), Workplaces (places of work such as offices). ## Acknowledgements We thank all members of the India COVID-19 Apex Research Team (iCART) for sharing their expertise. We thank our professors at the All India Institute of Medical Sciences (AIIMS), New Delhi for their mentorship. SSM received a KVPY fellowship and support from IISER Pune. We thank Dr Hemant Deepak Shewade from the International Union Against Tuberculosis and Lung Disease (The Union) for his valuable inputs. The views expressed in this publication are those of the authors and not necessarily those of their affiliated institutes. We express our gratitude to all personnel who are at the frontlines of this pandemic across the globe. * Received May 13, 2020. * Revision received May 13, 2020. * Accepted May 18, 2020. * © 2020, Posted by Cold Spring Harbor Laboratory This pre-print is available under a Creative Commons License (Attribution-NonCommercial 4.0 International), CC BY-NC 4.0, as described at [http://creativecommons.org/licenses/by-nc/4.0/](http://creativecommons.org/licenses/by-nc/4.0/) ## References 1. Zhu N, Zhang D, Wang W, et al. A Novel Coronavirus from Patients with Pneumonia in China, 2019. N Engl J Med 2020; 382: 727–33. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1056/NEJMoa2001017&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 2. WHO Director-General’s opening remarks at the media briefing on COVID-19 - 11 March 2020. [https://www.who.int/dg/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19\---|11-march-2020](https://www.who.int/dg/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19\---|11-march-2020) (accessed April 2, 2020). 3. 2019 Novel Coronavirus COVID-19 Data Repository by Johns Hopkins CSSE. [https://github.com/CSSEGISandData/COVID-19](https://github.com/CSSEGISandData/COVID-19). 4. COVID19India Database API. [https://api.covid19india.org/](https://api.covid19india.org/). 5. Longini IM, Nizam A, Xu S, et al. Containing Pandemic Influenza at the Source. Science 2005; 309: 1083–7. [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6Mzoic2NpIjtzOjU6InJlc2lkIjtzOjEzOiIzMDkvNTczNy8xMDgzIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjAvMDUvMTgvMjAyMC4wNS4xMy4yMDA5NjgyNi5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 6. Leung K, Wu JT, Liu D, 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; . DOI:10.1016/S0140-6736(20)30746-7. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(20)30746-7&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32277878&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 7. Kucharski AJ, Russell TW, Diamond C, et al. Early dynamics of transmission and control of COVID-19: a mathematical modelling study. Lancet Infect Dis 2020; . DOI:10.1016/S1473-3099(20)30144-4. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S1473-3099(20)30144-4&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32171059&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 8. Pan A, Liu L, Wang C, et al. Association of Public Health Interventions With the Epidemiology of the COVID-19 Outbreak in Wuhan, China. JAMA 2020; published online April 10. DOI:10.1001/jama.2020.6130. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1001/jama.2020.6130&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32275295&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 9. Ferguson N, Laydon D, Nedjati Gilani G, et al. Report 9: Impact of non-pharmaceutical interventions (NPIs) to reduce COVID19 mortality and healthcare demand. Imperial College London, 2020 DOI:10.25561/77482. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.25561/77482&link_type=DOI) 10. McFall-Johnsen JK Lauren Frias, Morgan. A third of the global population is on coronavirus lockdown. Bus. Insid. [https://www.businessinsider.com/countries-on-lockdown-coronavirus-italy-2020-3](https://www.businessinsider.com/countries-on-lockdown-coronavirus-italy-2020-3) (accessed May 1, 2020). 11. Extended lockdown to cause $234.4 billion economic loss. Econ. Times. [https://economictimes.indiatimes.com/news/economy/indicators/extended-lockdown-to-cause-234-4-billion-economic-loss-says-barclays/articleshow/75136935.cms?from=mdr](https://economictimes.indiatimes.com/news/economy/indicators/extended-lockdown-to-cause-234-4-billion-economic-loss-says-barclays/articleshow/75136935.cms?from=mdr) (accessed April 29, 2020). 12. Pradhan B. Social distancing is a luxury Indians earning just Rs 150 a day cannot afford. ThePrint. 2020; published online March 23. [https://theprint.in/india/social-distancing-is-a-luxury-indians-earning-just-rs-150-a-day-cannot-afford/386063/](https://theprint.in/india/social-distancing-is-a-luxury-indians-earning-just-rs-150-a-day-cannot-afford/386063/) (accessed May 1, 2020). 13. More than 40 crore workers in India’s unorganized sector may be affected due to the lockdown: ILO. Gaonconnection Your Connect. Rural India. 2020; published online April 10. [https://en.gaonconnection.com/more-than-40-crore-workers-in-indias-unorganized-sector-may-be-affected-due-to-the-lockdown-which-will-affect-their-employment-and-earnings-ilo/](https://en.gaonconnection.com/more-than-40-crore-workers-in-indias-unorganized-sector-may-be-affected-due-to-the-lockdown-which-will-affect-their-employment-and-earnings-ilo/) (accessed May 1, 2020). 14. Covid-19: collateral damage of lockdown in India I The BMJ. [https://www.bmj.com/content/369/bmj.m1711.long](https://www.bmj.com/content/369/bmj.m1711.long) (accessed May 10, 2020). 15. Hellewell J, Abbott S, Gimma A, et al. Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. Lancet Glob Health 2020; 8: e488-96. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S2214-109X(20)30074-7&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32119825&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 16. Hu Z, Song C, Xu C, et al. Clinical characteristics of 24 asymptomatic infections with COVID-19 screened among close contacts in Nanjing, China. Sci China Life Sci 2020; published online March 4. DOI:10.1007/s11427-020-1661-4. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s11427-020-1661-4&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32146694&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 17. Lavezzo E, Franchin E, Ciavarella C, et al. Suppression of COVID-19 outbreak in the municipality of Vo, Italy. *medRxiv* 2020;: 2020.04.17.20053157. 18. Gudbjartsson DF, Helgason A, Jonsson H, et al. Spread of SARS-CoV-2 in the Icelandic Population. N Engl J Med 2020; : null. 19. Gandhi M, Yokoe DS, Havlir DV. Asymptomatic Transmission, the Achilles’ Heel of Current Strategies to Control Covid-19. N Engl J Med 2020; : null. 20. World Bank. World Bank Population Database. [http://api.worldbank.org/v2/en/indicator/SP.POP.TOTL?downloadformat=csv](http://api.worldbank.org/v2/en/indicator/SP.POP.TOTL?downloadformat=csv). 21. Obadia T, Haneef R, Boëlle P-Y. The R0 package: a toolbox to estimate reproduction numbers for epidemic outbreaks. BMC Med Inform Decis Mak 2012; 12: 147. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/1472-6947-12-147&link_type=DOI) 22. White LF, Pagano M. A likelihood-based method for real-time estimation of the serial interval and reproductive number of an epidemic. Stat Med 2008; 27: 2999–3016. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/sim.3136&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18058829&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000257567900002&link_type=ISI) 23. Wallinga J, Lipsitch M. How generation intervals shape the relationship between growth rates and reproductive numbers. Proc R Soc B Biol Sci 2007; 274: 599–604. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1098/rspb.2006.3754&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=17476782&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000243354200019&link_type=ISI) 24. Du Z, Xu X, Wu Y, Wang L, Cowling BJ, Meyers LA. Early Release - Serial Interval of COVID-19 among Publicly Reported Confirmed Cases - Volume 26, Number 6—June 2020 - Emerging Infectious Diseases journal - CDC. DOI:10.3201/eid2606.200357. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.3201/eid2606.200357&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32191173&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 25. Cori A, Ferguson NM, Fraser C, Cauchemez S. A New Framework and Software to Estimate Time-Varying Reproduction Numbers During Epidemics. Am J Epidemiol 2013; 178: 1505–12. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/aje/kwt133&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=24043437&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 26. Wallinga J, Teunis P. Different Epidemic Curves for Severe Acute Respiratory Syndrome Reveal Similar Impacts of Control Measures. Am J Epidemiol 2004; 160: 509–16. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/aje/kwh255&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=15353409&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000223938000001&link_type=ISI) 27. Google LLC. Google COVID-19 Community Mobility Reports. [https://www.google.com/covid19/mobility/](https://www.google.com/covid19/mobility/) (accessed April 24, 2020). 28. Sun C, Hsieh Y-H. Global analysis of an SEIR model with varying population size and vaccination. Appl Math Model 2010; 34: 2685–97. 29. Lauer SA, Grantz KH, Bi Q, et al. The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application. Ann Intern Med 2020; published online March 10. DOI:10.7326/M20-0504. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.7326/M20-0504&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32150748&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 30. Wölfel R, Corman VM, Guggemos W, et al. Virological assessment of hospitalized patients with COVID-2019. Nature 2020; published online April 1. DOI:10.1038/s41586-020-2196-x. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/s41586-020-2196-x&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32235945&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 31. He X, Lau EHY, Wu P, et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nat Med 2020; 1-4. 32. Mizumoto K, Kagaya K, Zarebski A, Chowell G. Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship, Yokohama, Japan, 2020. Eurosurveillance 2020; 25: 2000180. 33. Coronavirus: 80% cases asymptomatic, but no need to revise testing criteria, says ICMR. India Today. [https://www.indiatoday.in/india/story/80-of-coronavirus-cases-in-india-are-asymptomatic-icmr-1669073-2020-04-20](https://www.indiatoday.in/india/story/80-of-coronavirus-cases-in-india-are-asymptomatic-icmr-1669073-2020-04-20) (accessed April 30, 2020). 34. Day M. Covid-19: four fifths of cases are asymptomatic, China figures indicate. BMJ 2020; 369. DOI:10.1136/bmj.m1375. [FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiYm1qIjtzOjU6InJlc2lkIjtzOjE4OiIzNjkvYXByMDJfMTEvbTEzNzUiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyMC8wNS8xOC8yMDIwLjA1LjEzLjIwMDk2ODI2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 35. Russell TW, Hellewell J, Abbott S, Golding N, Gibbs H, Jarvis CI. Using a delay-adjusted case fatality ratio to estimate under-reporting. CMMID Repos. 2020; published online March 22. [https://cmmid.github.io/topics/covid19/global\_cfr\_estimates.html](https://cmmid.github.io/topics/covid19/global_cfr_estimates.html) (accessed April 28, 2020). 36. Peng L, Yang W, Zhang D, Zhuge C, Hong L. Epidemic analysis of COVID-19 in China by dynamical modeling. *medRxiv* 2020;: 2020.02.16.20023465. 37. Tindale L, Coombe M, Stockdale JE, et al. Transmission interval estimates suggest pre-symptomatic spread of COVID-19. *medRxiv* 2020;: 2020.03.03.20029983. 38. Nishiura H, Linton NM, Akhmetzhanov AR. Serial interval of novel coronavirus (COVID-19) infections. Int J Infect Dis 2020; 93: 284–6. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.ijid.2020.02.060&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32145466&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 39. You C, Deng Y, Hu W, et al. Estimation of the Time-Varying Reproduction Number of COVID-19 Outbreak in China. *medRxiv* 2020;: 2020.02.08.20021253. 40. Advisory: Travel and Visa restrictions related to COVID-19. Bur. Immigr. Gov. India. [https://boi.gov.in/content/advisory-travel-and-visa-restrictions-related-covid-19-0](https://boi.gov.in/content/advisory-travel-and-visa-restrictions-related-covid-19-0) (accessed April 21, 2020). 41. ICMR. Testing Strategy Updates for COVID-19 in India. Indian Counc. Med. Res. [https://www.icmr.gov.in/cteststrat.html](https://www.icmr.gov.in/cteststrat.html) (accessed April 30, 2020). 42. Li Q, Guan X, Wu P, et al. Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia. N Engl J Med 2020; 382: 1199–207. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1056/NEJMoa2001316&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 43. Report of the WHO-China Joint Mission on Coronavirus Disease 2019 (COVID-19). [https://www.who.int/publications-detail/report-of-the-who-china-joint-mission-on-coronavirus-disease-2019-(covid-19)](https://www.who.int/publications-detail/report-of-the-who-china-joint-mission-on-coronavirus-disease-2019-(covid-19)) (accessed April 27, 2020). 44. Wu JT, Leung K, Leung GM. Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling study. The Lancet 2020; 395: 689–97. 45. Coburn BJ, Wagner BG, Blower S. Modeling influenza epidemics and pandemics: insights into the future of swine flu (H1N1). BMC Med 2009; 7: 30. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/1741-7015-7-30&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=19545404&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 46. Mills CE, Robins JM, Lipsitch M. Transmissibility of 1918 pandemic influenza. Nature 2004; 432: 904–6. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/nature03063&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=15602562&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000225733500053&link_type=ISI) 47. WHO. Consensus document on the epidemiology of severe acute respiratory syndrome (SARS). WHO. [https://www.who.int/csr/sars/WHOconsensus.pdf?ua=1](https://www.who.int/csr/sars/WHOconsensus.pdf?ua=1) (accessed May 2, 2020). 48. Fine P, Eames K, Heymann DL. “Herd Immunity”: A Rough Guide. Clin Infect Dis 2011; 52: 911–6. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/cid/cir007&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21427399&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000288802600016&link_type=ISI) 49. Mandal S, Bhatnagar T, Arinaminpathy N, et al. Prudent public health intervention strategies to control the coronavirus disease 2019 transmission in India: A mathematical model-based approach. Indian J Med Res 2020; published online March 23. DOI:10.4103/ijmr.IJMR\_504\_20. [CrossRef](http://medrxiv.org/lookup/external-ref?access\_num=10.4103/ijmr.IJMR_504_20&link_type=DOI) 50. Ministry of Health and Family Welfare, Government of India. Press Release on COVID-19 situation in India. pib.gov.in/Pressreleaseshare.aspx?PRID=1615800 (accessed April 27, 2020). 51. South Korea’s Patient 31, Punjab Priest, Bhilwara Doc, Tablighi Jamaat: Infamous Covid-19 Super-Spreaders - News18. [https://www.news18.com/news/india/south-koreas-patient-31-punjab-priest-bhilwara-doc-tablighi-jamaat-infamous-covid-19-super-spreaders-2573717.html](https://www.news18.com/news/india/south-koreas-patient-31-punjab-priest-bhilwara-doc-tablighi-jamaat-infamous-covid-19-super-spreaders-2573717.html) (accessed May 3, 2020). 52. OxCGRT: Oxford COVID-19 Government Response Tracker. [https://covidtracker.bsg.ox.ac.uk/stringency-scatter](https://covidtracker.bsg.ox.ac.uk/stringency-scatter) (accessed April 29, 2020). 53. Peto J, Alwan NA, Godfrey KM, et al. Universal weekly testing as the UK COVID-19 lockdown exit strategy. The Lancet 2020; . DOI:10.1016/S0140-6736(20)30936-3. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(20)30936-3&link_type=DOI) 54. Pueyo T. Coronavirus: How to Do Testing and Contact Tracing. Medium. 2020; published online April 29. [https://medium.com/@tomaspueyo/coronavirus-how-to-do-testing-and-contact-tracing-bde85b64072e](https://medium.com/@tomaspueyo/coronavirus-how-to-do-testing-and-contact-tracing-bde85b64072e) (accessed April 29, 2020). 55. Black JRM, Bailey C, Przewrocka J, Dijkstra KK, Swanton C. COVID-19: the case for health-care worker screening to prevent hospital transmission. The Lancet 2020; . DOI:10.1016/S0140-6736(20)30917-X. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(20)30917-X&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32305073&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 56. Liu Y, Centre for Mathematical Modelling of Infectious Diseases nCoV Working Group, Funk S, Flasche S. The contribution of pre-symptomatic infection to the transmission dynamics of COVID-2019. Wellcome Open Res 2020; 5: 58. 57. Are 14 Days Enough? Kerala’s Case For 28-Day Quarantine. NDTV.com. [https://www.ndtv.com/kerala-news/coronavirus-kerala-thats-the-logic-behids-28-day-isolation-period-senior-kerala-doctor-2213520](https://www.ndtv.com/kerala-news/coronavirus-kerala-thats-the-logic-behids-28-day-isolation-period-senior-kerala-doctor-2213520) (accessed April 30, 2020). 58. China’s Harbin orders 28-day quarantine after rise in imported cases of coronavirus. Deccan Her. 2020; published online April 12. [https://www.deccanherald.com/international/chinas-harbin-orders-28-day-quarantine-after-rise-in-imported-cases-of-coronavirus-824407.html](https://www.deccanherald.com/international/chinas-harbin-orders-28-day-quarantine-after-rise-in-imported-cases-of-coronavirus-824407.html) (accessed April 30, 2020). 59. Evaluation of COVID-19 RT-qPCR test in multi-sample pools I medRxiv. [https://www.medrxiv.org/content/10.1101/2020.03.26.20039438v1](https://www.medrxiv.org/content/10.1101/2020.03.26.20039438v1) (accessed April 21, 2020). 60. Abaluck J, Chevalier JA, Christakis NA, et al. The Case for Universal Cloth Mask Adoption and Policies to Increase Supply of Medical Masks for Health Workers. Rochester, NY: Social Science Research Network, 2020 DOI:10.2139/ssrn.3567438. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.2139/ssrn.3567438&link_type=DOI) 61. Spychalski P, Błażyńska-Spychalska A, Kobiela J. Estimating case fatality rates of COVID-19. Lancet Infect Dis 2020; . DOI:10.1016/S1473-3099(20)30246-2. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S1473-3099(20)30246-2&link_type=DOI) 62. New Zealand Government. COVID-19 Alert System. Unite COVID-19. [https://covid19.govt.nz/alert-system/covid-19-alert-system/](https://covid19.govt.nz/alert-system/covid-19-alert-system/) (accessed April 29, 2020). 63. Ministry of Home Affairs, Government of India. MHA Order to extend Lockdown period for 2 weeks w.e.f. 4.5.2020 with new guidelines. [https://www.mha.gov.in/sites/default/files/MHA%20Order%20Dt.%201.5.2020%20to%20extend%20Lockdown%20period%20for%202%20weeks%20w.e.f.%204.5.2020%20with%20new%20guidelines.pdf](https://www.mha.gov.in/sites/default/files/MHA%20Order%20Dt.%201.5.2020%20to%20extend%20Lockdown%20period%20for%202%20weeks%20w.e.f.%204.5.2020%20with%20new%20guidelines.pdf). ## References for supplementary appendix 1. He X, Lau EHY, Wu P, et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nat Med 2020;: 1-4. 2. Wölfel R, Corman VM, Guggemos W, et al. Virological assessment of hospitalized patients with COVID-2019. Nature 2020; published online April 1. DOI:10.1038/s41586-020-2196-x. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/s41586-020-2196-x&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32235945&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 3. Peng L, Yang W, Zhang D, Zhuge C, Hong L. Epidemic analysis of COVID-19 in China by dynamical modeling. *medRxiv* 2020;: 2020.02.16.20023465. 4. Delignette-Muller ML, Dutang C. fitdistrplus: An R Package for Fitting Distributions. J Stat Softw 2015; 064. [https://ideas.repec.org/a/jss/jstsof/v064i04.html](https://ideas.repec.org/a/jss/jstsof/v064i04.html) (accessed May 2, 2020). 5. White LF, Pagano M. A likelihood-based method for real-time estimation of the serial interval and reproductive number of an epidemic. Stat Med 2008; 27: 2999–3016. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/sim.3136&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18058829&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000257567900002&link_type=ISI) 6. Wallinga J, Lipsitch M. How generation intervals shape the relationship between growth rates and reproductive numbers. Proc R Soc B Biol Sci 2007; 274: 599–604. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1098/rspb.2006.3754&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=17476782&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000243354200019&link_type=ISI) 7. ICMR. Testing Strategy Updates for COVID-19 in India. Indian Counc. Med. Res. [https://www.icmr.gov.in/cteststrat.html](https://www.icmr.gov.in/cteststrat.html) (accessed April 30, 2020). 8. Wu JT, Leung K, Leung GM. Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling study. The Lancet 2020; 395: 689–97. 9. Du Z, Xu X, Wu Y, Wang L, Cowling BJ, Meyers LA. Early Release - Serial Interval of COVID-19 among Publicly Reported Confirmed Cases - Volume 26, Number 6—June 2020 - Emerging Infectious Diseases journal - CDC. DOI:10.3201/eid2606.200357. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.3201/eid2606.200357&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32191173&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 10. COVID19India Database API. [https://api.covid19india.org/](https://api.covid19india.org/). 11. Chowell G. Fitting dynamic models to epidemic outbreaks with quantified uncertainty: A primer for parameter uncertainty, identifiability, and forecasts. Infect Dis Model 2017; 2: 379–98. 12. van den Driessche P. Reproduction numbers of infectious disease models. Infect Dis Model 2017; 2: 288–303. 13. Jayaram R, Ramakrishnan N. Cost of intensive care in India. Indian J Crit Care Med Peer-Rev Off Publ Indian Soc Crit Care Med 2008; 12: 55–61. 14. Wu Z, McGoogan JM. Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention. JAMA 2020; 323: 1239–42. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1001/jama.2020.2648&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F18%2F2020.05.13.20096826.atom) 15. Google LLC. Google COVID-19 Community Mobility Reports. [https://www.google.com/covid19/mobility/](https://www.google.com/covid19/mobility/) (accessed April 24, 2020). [1]: /embed/inline-graphic-1.gif [2]: /embed/inline-graphic-2.gif [3]: /embed/inline-graphic-3.gif [4]: /embed/graphic-8.gif [5]: /embed/graphic-9.gif [6]: /embed/graphic-10.gif [7]: /embed/inline-graphic-4.gif [8]: /embed/inline-graphic-5.gif [9]: /embed/graphic-14.gif [10]: /embed/graphic-29.gif [11]: /embed/graphic-30.gif [12]: /embed/graphic-31.gif [13]: /embed/graphic-32.gif