key: cord-0876225-aqozmk1t authors: Cao, H.; Li, B.; Gu, T.; Liu, X.; Meng, K.; ZHANG, L. title: Associations of ambient air pollutants and meteorological factors with COVID-19 transmission in 31 Chinese provinces: A time-series study date: 2020-06-24 journal: nan DOI: 10.1101/2020.06.24.20138867 sha: 2e99f3c410fa9baa9dc913ac15da0461aab83aeb doc_id: 876225 cord_uid: aqozmk1t Background: Evidence regarding the effects of ambient air pollutants and meteorological factors on COVID-19 transmission is limited. Objectives: To explore the associations of air pollutants and meteorological factors with COVID-19 confirmed cases across 31 Chinese provinces during the outbreak period. Methods: The number of COVID-19 confirmed cases, air pollutant concentrations and meteorological factors in 31 Chinese provinces from January 25 to February 29, 2020 were extracted from authoritative electronic databases. The associations were estimated for a single-day lag (lag0-lag6) as well as moving averages lag (lag01-lag05) using generalized additive mixed models (GAMMs), adjusted for time trends, day of the week, holidays and meteorological variables. Region-specific analyses and meta-analysis were conducted in five selected regions with diverse air pollution levels and weather conditions. Nonlinear exposure-response analyses were performed. Results: We examined 77,578 COVID-19 confirmed cases across 31 Chinese provinces during the study period. An increase of each interquartile range in PM2.5, PM10, SO2, NO2, O3 and CO at lag4 corresponded to 1.40 (1.37-1.43), 1.35 (1.32-1.37), 1.01 (1.00-1.02), 1.08 (1.07-1.10), 1.28 (1.27-1.29) and 1.26 (1.24-1.28) odds ratios (ORs) of daily COVID-19 confirmed new cases, respectively. For 1 oc, 1% and 1 m/s increase in temperature, relative humidity and wind velocity, the ORs were 0.97 (0.97-0.98), 0.96 (0.96-0.97), and 0.94 (0.92-0.95), respectively. The estimates of PM2.5, PM10, NO2 and all meteorological factors remained statistically significant after meta-analysis for the five selected regions. The exposure-response relationships showed that higher concentrations of air pollutants and lower meteorological factors were associated with daily COVID-19 confirmed new cases increasing. Conclusions: Higher air pollutant concentrations and lower temperature, relative humidity and wind velocity may favor COVID-19 transmission. As summer months are arriving in the Northern Hemisphere, the environmental factors and implementation of public health control measures may play an optimistic role in controlling COVID-19 epidemic. pollutants. In addition, we included the following potential confounders in the models: daily mean 120 temperature, relative humidity, wind velocity, and categorical variables for day of the week and 121 public holidays. We adjusted for meteorological factors averaged over the same day and the previous 122 day as well as over the seven days preceding this period. The generalized cross-validation (GCV) 123 values were used to select the best averaging period for meteorological factors. We used three 124 degrees of freedom for the spline function of meteorological factors, which reportedly allowed 125 adequate control for their effects on health outcomes (Samet et al. 2000) . In all pollutant models, 126 province was incorporated as a random effect, and the covariates were incorporated as fixed effects. 127 The statistical analyses for the associations of meteorological factors were similar to those for air 128 Regarding the levels of air pollutants and meteorological factors, the associations were estimated 130 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint 6 with different lag structures using a single-day lag from the current day up to the previous six days 131 (lag0-lag6) as well as moving averages of the current and previous days (lag01-lag06). To facilitate 132 comparisons, the associations of air pollutants and meteorological factors with daily confirmed new 133 cases were reported as odds ratio (OR) with 95% confidence interval (95% CI) for an IQR increase 134 in daily average concentrations of air pollutants as well as 1 °C, 1% and 1 m/s increase in daily 135 mean temperature, relative humidity and wind velocity, respectively. 136 Considering the heterogeneity of 31 Chinese provinces, we conducted subgroup analyses to explore 137 region-specific associations. From the north to south of China, five regions (i.e., Heilongjiang, 138 Beijing, Hubei, Guangdong and Hainan) with sufficient sample size of COVID-19 confirmed cases 139 were investigated due to their diversity of air pollution levels and climate conditions. The total 140 effects of air pollutants and meteorological factors in five regions were then calculated through 141 meta-analysis by fixed models or random effect models. 142 The nonlinear exposure-response relationships were examined because linearity assumption of air 143 pollutants and meteorological factors with COVID-19 confirmed cases may not hold. Instead of 144 linear parameters, we included a penalized cubic regression spline of air pollutants or 145 meteorological factors with three degrees of freedom into the GAMMs. To determine the susceptible 146 weather characteristics, we further investigated the three-dimensional exposure-response 147 relationships of temperature, relative humidity and wind velocity by including an interaction term 148 with two of them into the models. 149 To assess the robustness of the associations of air pollutants and meteorological factors with 150 COVID-19 confirmed cases, we conducted sensitivity analyses using four degrees of freedom in 151 meteorological factors and varying the level of smoothness of the time trend function (with six to 152 eight degrees of freedom). 153 All of the statistical analyses were performed using R version 3.5.3 (https://www.r-project.org/). A 154 P-value <0.05 was considered as statistically significant for a two-tailed test. 155 Table 1 . We examined 77,578 159 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. (2.0) μg/m 3 , 18.5 (4.1) μg/m 3 , 80.9 (14.3) μg/m 3 , and 0.8 (0.2) mg/m 3 , respectively. Meteorological 163 factors ranged from 1.4 °C to 11.6 °C in temperature, from 48.7% to 78.8% in relative humidity, 164 and from 2.1 m/s to 4.7 m/s in wind velocity. As shown in Table 2 , the daily concentrations of air 165 pollutants were strongly and positively correlated with each other with the exception of O3 and CO. SO2, NO2 and O3 concentrations were four days earlier than confirmed cases. The time series plots 177 of temperature and relative humidity presented inverse trends compared to confirmed cases, 178 especially after February 16, 2020. 179 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. (which 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 this version posted June 24, 2020. Note: Orange circles: a single-day lag from the current day up to the previous six days (lag0-lag6). 218 Blue triangles: moving averages of the current and previous days (lag01-lag06). 219 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. Note: Orange circles: a single-day lag from the current day up to the previous six days (lag0-lag6). 224 Blue triangles: moving averages of the current and previous days (lag01-lag06). 225 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint (which 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 this version posted June 24, 2020. confirmed new cases at the previous four days lag structure (lag4) in the five selected regions. 246 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. confirmed new cases at the previous four days lag structure (lag4) in the five selected regions. 249 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint (which 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 this version posted June 24, 2020. (which 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 this version posted June 24, 2020. also observed in Hubei, which was the province with the most serious COVID-19 infection. 301 Regarding regional heterogeneity, the positive significant associations of PM2.5, PM10 and NO2 302 remained after the meta-analysis for the five selected regions. The pathogenic mechanisms between 303 air pollutants and infectious diseases are still unclear. Experimental evidence has suggested that the 304 increased transmission ability in the presence of air pollutants may not be caused by improvement 305 of epithelial cell susceptibility to infection but may result from effects on host defenses that prevent 306 the spread of virus (Becker and Soukup 1999) . In addition, positive associations may also be affected 307 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. found that the RRs of the daily confirmed case counts increased to a peak at approximately 8-10 °C 327 (Shi et al. 2020) . Consistent with these results, we examined negative associations between 328 temperature and COVID-19 confirmed new cases with a corresponding OR of 0.97 (0.97-0.98). The 329 nonlinear exposure-response relationship indicated that negative associations were observed for 330 temperatures ranging from -10 °C to 5 °C and above 20 °C, whereas positive associations were 331 observed below -10 °C and 5-20 °C. The positive estimates of temperature reached a peak at 332 approximately 10 °C, which agreed with the previous study. The overall negative estimate observed 333 in our study may be due to the relative low temperature (5.7 ± 3.1 °C) throughout China during the 334 study period. However, Yao et al. found that there were no significant association of COVID-19 335 transmission with temperature (Yao et al. 2020b ). Therefore, more epidemiological and laboratory 336 evidence for the associations of temperature are still needed. 337 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint Relative humidity is another meteorological factor that may play a part in infectious disease 338 outbreaks. A previous study found that 12% of influenza virus transmission variability and 36% of 339 influenza virus survival variability could be explained by relative humidity (Shaman et al. 2011) . 340 Our study found that the number of COVID-19 confirmed new cases was negatively associated with 341 relative humidity with a corresponding OR of 0.96 (0.96-0.97). With regard to regional 342 heterogeneity, the negative significant association remained after the meta-analysis for the five 343 selected regions. Furthermore, the exposure-response relationship indicated that the associations 344 became negative and attenuated above 70%, which was the mainly range of relative humidity during Wind velocity is an important meteorological factor affecting the suspension time of pathogens in 358 the air. High wind velocity facilitates dilution and removal of pathogens, thereby reducing the 359 transmission potential. An ecological study found that higher daily average wind velocity 360 corresponded to a lower secondary attack rate of SARS (Cai et al. 2007 ). Consistent with the 361 previous study, we examined the negative associations between wind velocity and COVID-19 362 confirmed new cases in Hubei and throughout China. Regarding regional heterogeneity, the negative 363 significant association remained after the meta-analysis for the five selected regions. The exposure-364 response relationship also indicated that higher wind velocity was associated with lower infection 365 case counts when the wind velocity ranged from 2 m/s to 5 m/s. To our knowledge, we are the first 366 All rights reserved. No reuse allowed without permission. (which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint to investigate the effects of wind velocity on COVID-19 transmission. Therefore, more 367 epidemiological studies and direct laboratory evidence are needed to verify our results. 368 In addition, we considered the associations of meteorological factor combinations on COVID-19 369 infection. The three-dimensional exposure-response relationship indicated that weather with high 370 temperature, low humidity and high wind velocity may decrease COVID-19 transmission but that 371 weather with low temperature, low humidity and low wind velocity may favor its transmission. The 372 survival characteristics of coronavirus support our results. An experimental study on SARS-CoV 373 found that the virus soon lost its activity when the temperature increased to 38 °C and the relative outbreak in China and countries with similar latitude (e.g., U.S., Italy, Japan and Korea), which was 379 in winter and early spring with relative low temperature and low humidity (Rothan and Byrareddy 380 2020). As summer months (hot and humid) are arriving in the Northern Hemisphere, the weather 381 conditions may play an optimistic role in decreasing the transmission. At the same time, it should 382 be noted that COVID-19 is pandemic throughout the world, even in several tropic countries with 383 hot and dry weather conditions (e.g., India, Mexico and Brazil). Therefore, the implementation of 384 extensive public health control measures (e.g., shutting down cities, extending holidays and travel 385 ban) and medication (Pan et al. 2020 ) is more efficient and powerful than weather changing for 386 controlling the COVID-19 infection. 387 Identifying the associations of environmental factors in different lag structures is of great importance 388 in the estimation of health risk. Incubation period, the time of diagnosis and reporting delay should 389 be considered in risk assessment of infectious diseases. For COVID-19, previous four days lag 390 structure (lag4) may be the most appropriate for the OR estimations because at least five days were 391 required for case confirmation with the shortest incubation period (three days of incubation, one day 392 for laboratory diagnosis, and one day delay for case reporting). In addition, the associations of air 393 pollutants and meteorological factors with COVID-19 infection at lag4 were more stable than other 394 lag structures through sensitivity analyses. Therefore, estimates at lag4 were selected for our study. 395 Effect of nitrogen dioxide on respiratory viral infection in airway epithelial 438 cells Influence of meteorological factors and air 440 pollution on the outbreak of severe acute respiratory syndrome The effects of temperature and relative 443 humidity on the viability of the sars coronavirus Air pollution and case fatality of sars in the 446 people's republic of china: An ecologic study. Environmental health : a global access science source 2:15 Air pollution and the 449 development of asthma from birth until young adulthood. The European respiratory journal Short-term elevation 452 of fine particulate matter air pollution and acute lower respiratory infection. American journal of 453 respiratory and critical care medicine Low temperature 456 and low uv indexes correlated with peaks of influenza virus activity in northern europe during Relationship between ambient air pollution and daily 459 mortality of sars in beijing Early transmission dynamics in wuhan, china, of 461 novel coronavirus-infected pneumonia Impact of meteorological factors on the covid-19 464 transmission: A multi-city study in china Independent and interactive effects of ambient 467 temperature and absolute humidity on the risks of avian influenza a(h7n9) infection in china Climate change-related migration and infectious disease Acute effects of ambient air 485 pollution on lower respiratory infections in hanoi children: An eight-year time series study Association of public health interventions with 488 the epidemiology of the covid-19 outbreak in wuhan, china Effects of absolute humidity, relative 491 humidity, temperature, and wind speed on influenza activity in toronto, ontario, canada. Applied and 492 environmental microbiology 85 Cardiovascular mortality and 494 exposure to airborne fine particulate matter and cigarette smoke: Shape of the exposure-response 495 relationship The epidemiology and pathogenesis of coronavirus disease (covid-19) 498 outbreak The national morbidity, mortality, and 501 air pollution study. Part i: Methods and methodologic issues Absolute humidity and pandemic versus epidemic influenza Impact of temperature on the dynamics of the covid-506 19 outbreak in china An initial investigation of the association between the 509 sars outbreak and weather: With the view of the environmental temperature and its variation Stability of middle east respiratory syndrome 513 coronavirus (mers-cov) under different environmental conditions. Euro surveillance : bulletin Europeen 514 sur les maladies transmissibles = Temperature significant change covid-19 517 transmission in 429 cities Impact of climate change on human infectious diseases: Empirical 520 evidence and human adaptation Ambient nitrogen dioxide pollution and spread All rights reserved. No reuse allowed without permission.(which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint 23 There were several strengths in the present study. First, the study period (from January 25 to 396February 29, 2020) covered the COVID-19 outbreak in China, which allowed representative 397 association estimates between environmental factors and disease infection. Second, 31 Chinese 398 cities with diverse air pollution levels (e.g., daily mean PM2.5 ranging from 3.2 μg/m³ to 311.5 μg/m³) 399 and meteorological conditions (e.g., daily mean temperature ranging from -21.3 °C to 26.3 °C) were 400 included in our study, which provided enough span for association estimates and exposure-response 401 analyses. Third, to our knowledge, we are the first to explore the associations of meteorological 402 factor combinations on COVID-19 infection. The three-dimensional exposure-response relationship 403 suggested that low temperature, low humidity and low wind velocity likely favor COVID-19 404 Our study also had some limitations. First, this is a province-level ecological study without 406 considering the implementation ability of COVID-19 control policy, urbanization rate and 407 availability of medical resources, which may affect COVID-19 transmission and confound our 408 results. Second, demographic variables (e.g., age, gender and history of cardiopulmonary diseases) 409were not considered in our study due to a deficiency of detailed information for each infectious case. (which 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 this version posted June 24, 2020. (which 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 this version posted June 24, 2020. All rights reserved. No reuse allowed without permission.(which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.(which 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 this version posted June 24, 2020. All rights reserved. No reuse allowed without permission.(which 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 this version posted June 24, 2020. . https://doi.org/10.1101/2020.06.24.20138867 doi: medRxiv preprint