Transmission potential of COVID-19 in Iran

We estimated the reproduction number of 2020 Iranian COVID-19 epidemic using two different methods: R0 was estimated at 4.4 (95% CI, 3.9, 4.9) (generalized growth model) and 3.50 (1.28, 8.14) (epidemic doubling time) (February 19 - March 1) while the effective R was estimated at 1.55 (1.06, 2.57) (March 6-19).


Main text 27
Iran has been experiencing a devastating epidemic of COVID-19 in 2020 (1). The virus appears 28 to be spreading rapidly with multiple cases reported among the elite (2). One model estimated that 18,300 29 (95% CI, 3,770 to 53,470) cases would have had occurred by February 24 (3). The Iranian authorities 30 have adopted social distancing measures to slow disease transmission (4). As the epidemic continues, our 31 results call for efforts to minimize COVID-19-related morbidity and mortality. 32 To confront the epidemic, the transmission potential of COVID-19 is a useful metric to guide the 33 outbreak response efforts including the scope and intensity of interventions. During the early transmission 34 phase, the basic reproduction number, R0, represents such a measurement quantifying the average number 35 of secondary cases that primary cases generate in a completely susceptible population in the absence of 36 interventions or behavioral changes. R0>1 indicates the possibility of sustained transmission, whereas 37 R0<1 implies that transmission chains cannot sustain epidemic growth. As the epidemic continues its 38 course, the effective reproduction number, Re, offers a time-dependent record of the average number of 39 secondary cases per case as the number of susceptible individuals gets depleted and control interventions 40 takes effect. Herein we employed two different methods to quantify the reproduction number using the 41 curve of reported COVID-19 cases in Iran and its five regions (Table S1). 42 Using a Wikipedia entry as a starting point, Farsi-speaking coauthor SFR double-checked the data 43 of daily reported new cases, from February 19 through March 19, 2020 (the day before the Iranian New 44 Year), using Iranian official press releases and other credible news sources, and corrected the data as per 45 the official data ( Figure S1, Tables S2 and S3). Incident cases for the 5 regions were missing for two 46 days: March 2-3, 2020, which were not included in our analysis. We noted that the reported national 47 number of new cases did not match the sum of the new cases reported in Iran's five regions on March 5, 48 2020; we decided to treat each time series as independent and used the data as reported. Using Method 1, 49 we estimated R0 using data from February 19 through March 1, 2020. Using Method 2, we estimated R0 50 from the early transmission phase (February 19 through March 1, 2020) and Re based on the growth rate 51 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.08.20030643 doi: medRxiv preprint estimated from March 6 through 19, 2020 during which the epidemic slowed down likely reflecting the 52 impact of social distancing measures. 53 Two methods were utilized to estimate the reproduction number. The first method utilized a 54 generalized growth model (GGM) (5) with the growth rate and its scaling factor to characterize the daily 55 reported incidence, followed by simulation of the calibrated GGM, using a discretized probability 56 distribution of the serial interval and assuming a Poisson error structure. See Technical Appendix for 57 detailed description of Method 1. 58 The second method for estimating the reproduction number is based on the calculation of the 59 epidemic's doubling times, which correspond to the times when the cumulative incidence doubles. We 60 estimated the epidemic doubling time using the curve of cumulative daily reported cases. To quantify 61 parameter uncertainty, we used parametric bootstrapping with a Poisson error structure around the 62 number of new reported cases in order to derive 95% confidence intervals of our estimates (6-8). 63 Assuming exponential growth, the epidemic growth rate is equal to ln(2)/doubling time. Assuming that 64 the pre-infectious and infectious periods follow an exponential distribution, R0 is approximately equal to 1 65 + growth rate × serial interval. See Technical Appendix and Equation 4.14, Table 4.1 in Vynnycky and 66 White (9). 67 In both methods, the serial interval was assumed to follow a gamma distribution (mean: 4.41 68 days; standard deviation: 3.17 days) (10, 11). MATLAB version R2019b and R version 3.6.2 were used 69 for data analyses and creating figures. It was determined a priori that α= 0.05. 70 Using Method 1, we estimated an R0 of 4.4 (3.9, 4.9) for the COVID-19 epidemic in Iran. We 71 estimated a growth rate of 0.65 (95% CI, 0.56, 0.75) and a scaling parameter of 0.96 (95% CI, 0.93, 1) 72 (Table S4). The latter indicated a near-exponential growth of the epidemic (Figure). Using Method 2, we 73 found that from February 19 through March 1, the cumulative incidence of confirmed cases in Iran had 74 doubled 8 times. The estimated epidemic doubling time was 1.20 (95% CI, 1.05, 1.45) days and the 75 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint  Figures S7, S8). 78 Our results are robust, as they are consistent with the Iranian COVID-19 R0 estimates of 4.7 and 4.86 79 generated by Ahmadi et al. and Sahafizadeh and Sartoli (12,13), respectively, but are higher than the R0 80 of 2.72 estimated by Ghaffarzadegan & Rahmandad (14). 81 Our study has limitations. First, our analysis is based on the number of daily reported cases 82 whereas it would be ideal to analyze case counts by date of symptoms onset, which were not available. 83 Second, case counts could be underreported due to underdiagnosis, given subclinical or asymptomatic 84 cases, or limited testing capacity to test mild cases. The rapid increase in case counts might represent a 85 belated realization of the severity of the epidemic and a rapid process of catching up with testing many 86 suspected cases. If the reporting ratio remains constant over the study period, and given the near-87 exponential growth of the epidemic's trajectory, our estimates would remain reliable; however, this is a 88 strong assumption. Third, while data are not stratified according to imported and local cases, we assumed 89 that they were infected locally, as it is likely that transmission has been ongoing in Iran for some time (3). 90 In conclusion, we used two different methods to compute the R0 of COVID-19 epidemic in Iran. 91 Our mean estimate was at 4.4 (95% CI, 3.9, 4.9) for Method 1, and at 3.50 (95% CI, 1.28, 8.14) for 92 Method 2,which

156
. CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.08.20030643 doi: medRxiv preprint Figure. The mean basic reproduction number of COVID-19 epidemic in Iran, with 95% confidence 157 interval. Estimates for growth rate, r, and the scaling of the growth rate parameter, p, are also provided. 158 The plot in the lower panel depicts the fit of the Generalized Growth Model (Method 1) to the Iranian 159 data assuming Poisson error structure as of March 1, 2020. 160 161 162 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10. 1101