Impact of mitigating interventions and temperature on the instantaneous reproduction number in the COVID-19 epidemic among 30 US metropolitan areas

Background: After more than three months into the coronavirus disease (COVID-19) epidemic, over 170,000 people had died worldwide. The current study aims to evaluate how mitigating interventions affected the epidemic process in the 30 largest metropolitan areas in the US and whether temperature played a role in the epidemic process. Methods: Publicly available COVID-19 cases and deaths data and weather data were analyzed at the metropolitan level. The time-varying reproductive numbers were used to explore the trends. Results: We found that virus transmissibility, measured by instantaneous reproduction number (Rt), had declined significantly since the end of March for all areas and almost all of them reached a Rt of 1 or below by April 15, 2020. Cities with warm temperature tended to have a lower peak Rt than that of cities with cold temperature. However, large geographic variations exist. Conclusions: Though the end of epidemic of COVID-19 is near, temperature may have some weak effects on the virus transmission, and the return of the coronavirus outbreak is still possible.


Introduction
This was to compare the declining patterns of R t among metropolitan areas. We also realigned 1 2 3 the time scale from the peak of the outbreak. The first two weeks of R t estimates were excluded, 1 2 4 as the first week R t were zeros, and second week estimates were too variable due to small 1 2 5 number of cases. Descriptive statistics and bivariate associations were reported. Pearson correlation coefficients and student t-tests were used for comparisons. The sizes of total population and people aged 65 1 2 8 or older, and the percent of positive tests at each date were used for adjustment. R package 1 2 9 EpiEstim was used [11] to estimate the instantaneous reproduction numbers. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which this version posted May 1, 2020. . https: //doi.org/10.1101//doi.org/10. /2020 There were many statistical comparisons involved. Although we did not adjust for multiple 1 3 1 comparisons, we were cautious about over interpretations and conducted statistical tests only 1 3 2 between prior selected pairs (e.g., southern versus northern metropolitan areas).

3 3
In this study, the author has no financial and conflict of interest to disclose. The ethics approval 1 3 4 was exempted for this study, as no human subjects were involved, and all data were publicly 1 3 5 available. The statistical codes and data will be available online (github address after blind 1 3 6 review). The basic characteristics of metropolitan areas were presented in Table 1. All metropolitan areas 1 3 9 had at least 1.5 million people in 2019 and over 1,000 confirmed cases. The case-fatality rates  The trends of R t for 30 metropolitan areas were shown in Figure 1a locations and temperature conditions. Overall, the instantaneous R t s in all areas reached peaks or 1 4 9 some stable points after two to three weeks, decreased significantly since the end March, and 1 5 0 most areas reached a R t of 1 or less by April 15. It is of note that around the week of March 25, 1 5 1 many schools were closed and many companies started offering employees working from home. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which this version posted May 1, 2020. . https://doi.org/10. 1101/2020 The US government has issued COVID-19 coping guideline to all US citizens, and many states 1 5 3 also issued stay-home rule. Chicago, New York and Philadelphia started the epidemic earlier, had higher peak R t s than that  On the other hand, the R t curves were indistinguishable between upper midwestern cities and interventions in the mid-March (appendix Table 1 for dates stay-home rule issued). In addition, 1 6 6 the west coastal cities had an early start of the epidemic, and R t curves were less volatile than 1 6 7 that of other cities during the study period ( Figure 1c). The unusually high R t in Salt Lake City 1 6 8 in the early epidemic may be due to small number of cases during that period (Figure 1d).  To evaluate the association between R t and temperature across regions, we compared the highest 1 7 3 R t (occurred after the first two weeks) among them (Figure 2), most cities had a peak R t between 1 7 4 All rights reserved. No reuse allowed without permission.
was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which this version posted May 1, 2020. peak R t s than the rest of cities. The average peak of R t s in Boston, Chicago, New York, and 1 7 8 Philadelphia were marginally higher than that of Houston, Los Angeles, Orlando and Miami (p = 1 7 9 0.07). In addition, we also arbitrarily examined the R t patterns on the 15 th day after the outbreak  Miami, and Orlando were relatively stable (Figure 1a). After the end of March, due to national 1 9 1 effort in mitigating the epidemic, all R t curves started declining. The state of Florida, however, 1 0 declining early, while cities like Pittsburg and Detroit had much higher R t at the beginning, and 1 9 8 R t declined later. Even for cities like Chicago and New York, the intervention effects were 1 9 9 evident based on the sharp decline of R t since the mid-March, despite they had much higher R t s 2 0 0 in the beginning of epidemic. It has been suggested that like many other respiratory virus infections, a seasonal pattern may 2 0 2 exist for SARS like coronavirus [24,25]. However, as demonstrated in this study, the association other external factors. In this study, we found the peak R t s in warm cities were lower on average 2 0 5 than those in cold cities, suggesting that the virus transmissibility might be lower in warm 2 0 6 temperature than cold temperature. interventions such as travelling restriction, social distancing and stay-home rules will change the 2 1 0 epidemic process [26][27][28]. The availability of testing, diverse case ascertainment criteria, the 2 1 1 delay of diagnosis and case isolation, incomplete contact tracing, the percent of asymptomatic 2 1 2 cases, and the infectivity of asymptomatic cases will profoundly affect our ability to understand 2 1 3 the epidemic. In this study, we observed a possible negative association between temperature and 2 1 4 virus transmissibility (Figure 1a and 2). However, there were large variations in the peak R t s 2 1 5 among regions with lower temperature, partly due to different intervention effects and also may 2 1 6 be due to cultural and social differences. It is also likely that other environmental factors such as  was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which this version posted May 1, 2020.  was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which this version posted May 1, 2020.