Transmission dynamics and control of COVID-19 in Chile, March-October, 2020

Since the detection of the first case of COVID-19 in Chile on March 3rd, 2020, a total of 513188 cases, including ~14302 deaths have been reported in Chile as of November 2nd, 2020. Here, we estimate the reproduction number throughout the epidemic in Chile and study the effectiveness of control interventions especially the effectiveness of lockdowns by conducting short-term forecasts based on the early transmission dynamics of COVID-19. Chile’s incidence curve displays early sub-exponential growth dynamics with the deceleration of growth parameter, p, estimated at 0.8 (95% CI: 0.7, 0.8) and the reproduction number, R, estimated at 1.8 (95% CI: 1.6, 1.9). Our findings indicate that the control measures at the start of the epidemic significantly slowed down the spread of the virus. However, the relaxation of restrictions and spread of the virus in low-income neighborhoods in May led to a new surge of infections, followed by the reimposition of lockdowns in Greater Santiago and other municipalities. These measures have decelerated the virus spread with R estimated at ~0.96 (95% CI: 0.95, 0.98) as of November 2nd, 2020. The early sub-exponential growth trend (p ~0.8) of the COVID-19 epidemic transformed into a linear growth trend (p ~0.5) as of July 7th, 2020, after the reimposition of lockdowns. While the broad scale social distancing interventions have slowed the virus spread, the number of new COVID-19 cases continue to accrue, underscoring the need for persistent social distancing and active case detection and isolation efforts to maintain the epidemic under control.


Author summary 38
In context of the ongoing COVID-19 pandemic, Chile has been one of the hardest-hit countries in Latin 39 America, struggling to contain the spread of the virus. In this manuscript, we employ renewal equation to 40 estimate the reproduction number (R) for the early ascending phase of the COVID-19 epidemic and by 41 July 7 th , 2020 to guide the magnitude and intensity of interventions required to combat the COVID-19 42 epidemic. We also estimate the instantaneous reproduction number throughout the epidemic in Chile.

43
Moreover, we generate short-term forecasts based on the epidemic trajectory using phenomenological 44 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 3 models, and assess counterfactual scenarios to understand any additional resources required to contain the CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020.

55
The coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 56 , was declared a global pandemic by the World Health Organization (WHO) on March 57 11 th , 2020 [1,2]. This highly contagious unprecedented virus has impacted government and public 58 institutions, strained the health care systems, restricted people in their homes, and caused country-wide 59 lockdowns resulting in a global economic crisis [3][4][5]. Moreover, as of November 2 nd , 2020, nearly 46 60 million COVID-19 cases in 213 countries and territories have been reported, including more than 1.2 61 million deaths [6]. The social, economic, and psychological impact of this pandemic on much of the 62 world's population is profound [7][8][9][10][11][12][13]. 63 64 Soon after its initial rapid spread in China, the first case of novel coronavirus beyond China was reported 65 in Thailand on January 13 th , 2020 [14]. The first case in the USA was not identified until January 20 th , 66 2020 followed by the detection of the first cases in the European territory on January 24 th , 2020 [15,16]. America has paid a high toll during the COVID-19 pandemic, with some of the worlds' highest death 75 rates [19][20][21]. While home to less than 10% of the world population, Latin America accounts for about 76 one-third of all reported global deaths (~370 thousand) [6]. Several socioeconomic, demographic, and 77 political factors make control of the pandemic in Latin America particularly challenging [22][23][24][25]. Most 78 countries in the region are now facing the stark social and economic costs imposed by large-scale non-79 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 6 and stores, as new infections had reduced between 350-500 per day by the end of April, implying an only 106 apparent flattening of the COVID-19 curve [33][34][35]. Moreover, imposing and lifting lockdowns in small 107 geographical areas (municipalities) proved unsuccessful in areas with high interdependencies such as the 108 Greater Santiago [36] . This strategy resulted in a new wave of infections; with the virus spreading from 109 more affluent areas of Chile to more impoverished, crowded communities, forcing the government to 110 reimpose lockdown measures in Santiago in mid-May ( Figure 1) [23,37,38]. By mid-July, the 111 government implemented the "step by step" strategy, considering five stages of gradual opening, at the 112 municipality level, based on the periodic monitoring of epidemiological and health system indicators. The 113 case counts continued to increase, averaging ~4943 cases per day in June 2020, and started to decline  [40,41]. Moreover, the crude case fatality rate in Chile 120 (~2.8%) resonates with the global average case fatality rate (2.6%) [17,42].

122
In this study, we estimate the transmission potential of COVID-19, including the effective reproduction 123 number, ܴ , during the early transmission phase of the COVID-19 epidemic in Chile and around the mid 124 of the epidemic, by July 7 th , 2020. We also estimate the instantaneous reproduction number throughout 125 the epidemic in Chile. The reproduction number can guide the magnitude and intensity of control 126 interventions required to combat the COVID-19 outbreak [43,44]. We examine the effectiveness of 127 control interventions in Chile (see Table 1) on the transmission rate. To do this, we conduct short-term 128 forecasts using phenomenological growth models calibrated using the early trajectory of the epidemic and 129 by the mid of the epidemic (as of July 7 th , 2020) [45] to anticipate additional resources required to contain 130 the epidemic. These phenomenological growth models are useful in capturing the epidemic's empirical 131 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 148 149 Models 150 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 9 We utilize two phenomenological growth models, the generalized growth model (GGM)

173
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 10 including the sub-exponential growth (0< ‫‬ <1), constant incidence ‫(‬ =0) and exponential growth 174 dynamics ‫(‬ =1). The GLM model is given by the following differential equation:

245
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

264
The epidemic curve showed an increasing trajectory from April-June 2020 and declined thereafter. On CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  To assess the impact of social distancing interventions in Chile given in Table 1, we generated a 20-day 315 ahead forecast for Chile based on the daily incidence curve until March 30 th , 2020. The 28-day calibration 316 period of the GGM model yields an estimated growth rate, r, at 0.8 ( 95% CI: 0.6, 1.0) and a deceleration 317 of growth parameter, p, at 0.8 (95% CI: 0.7, 0.8), indicating early sub-exponential growth dynamics. The 318 20-day ahead forecast suggested that the early social distancing measures significantly slowed down the 319 early spread of the virus in Chile, whose effect is noticeable about two weeks after implementing an 320 intervention, as shown in Figure 6. A case resurgence was observed in Chile in mid-May 2020. As a 321 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ; https://doi.org/10.1101/2020.05.15.20103069 doi: medRxiv preprint 16 consequence of this case resurgence, a total lockdown was imposed in Greater Santiago (representing 322 ~52% of total COVID-19 cases during the epidemic) on May 15 th , 2020. The quarantine in Santiago was 323 gradually eased from August 17, 2020, and was lifted on September 28, 2020, as a part of the move to 324 phase three of a five-step plan of deconfinement that would allow movement on regional transportation 325 and reopening of non-essential businesses and schools [31,69,70]. We generated a 20-day ahead forecast 326 based on the daily incidence curve from May 9 th -July 7 th , 2020. The 60-day calibration of the GLM model CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ; https://doi.org/10. 1101/2020 April 2020 and ~12,959 for May 2020, a 137% increase. The testing rate in Chile further increased in 348 June 2020, testing on average ~17,578 individuals per day, followed by a slight decline in July 2020, 349 testing on average 16587 individuals per day. However, the testing rates continued to increase in August

370
The initial deceleration of the growth parameter in Chile indicates a sub-exponential growth pattern 371 ‫,)8.0~(‬ consistent with sub-exponential growth patterns of COVID-19 that have been observed in 372 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

373
In contrast, studies conducted in Peru, a Latin American country, and Iran have reported a nearly 374 exponential growth pattern of the COVID-19 whereas an exponential growth pattern has been reported in 375 China [72,75,83]. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ; https://doi.org/10.1101/2020.05.15.20103069 doi: medRxiv preprint 20 October, respectively, compared to ~20.09% in May 2020. The high positivity rate at the beginning of the 424 epidemic indicates that the government failed to cast a wide enough net to test the masses early in the 425 pandemic, and there were probably many more active cases than those detected by epidemiological 426 surveillance, underestimating the epidemic growth curve [90][91][92]. The most recent lower testing rates 427 indicate that Chile is testing a comparatively larger segment of the population. This positivity rate for 428 Chile is also consistent with the positivity rate obtained from India, the United States, Canada, and 429 Germany that exhibit moderately high positivity rates (4-8%) for COVID-19, indicating overall limited 430 testing in these countries [89,93]. In comparison, some countries like Mexico and the Czech Republic This study has some limitations. First, our study analyzes cases by the dates of reporting while it is ideal 448 to analyze the cases by the dates of onset or after adjusting for reporting delays. On the other hand, a 449 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ; https://doi.org/10.1101/2020.05.15.20103069 doi: medRxiv preprint 21 substantial fraction of the COVID-19 infections exhibits very mild or no symptoms at all, which may not 450 be reflected by data [101,102]. Moreover, the data are not stratified by local and imported cases; 451 therefore, we assumed that all cases contribute equally to the transmission dynamics of COVID-19.

452
Finally, the extent of selective underreporting, and its impact on these results is difficult to assess. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

517
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review) preprint
The copyright holder for this this version posted December 15, 2020.

569
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

621
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101/2020

673
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint  8  1  .  T  a  r  i  q  A  ,  L  e  e  Y  ,  R  o  o  s  a  K  ,  B  l  u  m  b  e  r  g  S  ,  Y  a  n  P  ,  M  a  S  ,  e  t  a  l  .  R  e  a  l  -t  i  m  e  m  o  n  i  t  o  r  i  n  g  t  h  e  t  r  a  n  s  m  i  s  s  i  o  n  p  o  t  e  n  t  i  a  l  o  f   674   C  O  V  I  D  -1  9  i  n  S  i  n  g  a  p  o  r  e  ,  M  a  r  c  h  2  0  2

726
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 15, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020