投稿系统
Comparisons of different statistical models for analyzing the effects of meteorological factors on COVID-19
doi: 10.2478/fzm-2023-0020
-
Abstract:
Objective This general non-systematic review aimed to gather information on reported statistical models examing the effects of meteorological factors on coronavirus disease 2019 (COVID-19) and compare these models. Methods PubMed, Web of Science, and Google Scholar were searched for studies on "meteorological factors and COVID-19" published between January 1, 2020, and October 1, 2022. Results The most commonly used approaches for analyzing the association between meteorological factors and COVID-19 were the linear regression model (LRM), generalized linear model (GLM), generalized additive model (GAM), and distributed lag non-linear model (DLNM). In addition to these classical models commonly applied in environmental epidemiology, machine learning techniques are increasingly being used to select risk factors for the outcome of interest and establishing robust prediction models. Conclusion Selecting an appropriate model is essential before conducting research. To ensure the reliability of analysis results, it is important to consider including non-meteorological factors (e.g., government policies on physical distancing, vaccination, and hygiene practices) along with meteorological factors in the model. -
Table 1. COVID-19 models incorporating meteorological variable
Author, Date (Reference) Country / Regions Data Collection Range Outcome Meteorological Variables (Exposure) Model Evaluation Metric Statistically Significant Meteorological Variables or the Variables Used for Building the Prediction Model Related Results Fernández, et al. [8] 2021 Spain 31 January – 20 June 2020 COVID-19 density of positive tests Mean Temperature Linear regression model The coefficient β Mean Temperature The coefficient β ranged from - 0.000 848 to - 0.001 19 for every two weeks timestep Pan, et al. [9] 2021 202 locations in Australia, Canada, USA, China, Germany, Italy, Japan, UK NA the basic reproductive number (R0)* Temperature, relative humidity, wind speed, and UV radiation Multiple linear regression models, Meta analysis meta P value All climate variables were not statistically associated with R0 of COVID-19 Temperature (meta P = 0.446), relative humidity (meta P = 0.215), wind speed (meta P = 0.986), and ultraviolet (UV) radiation (meta P = 0.491) Hamdan, et al. [10] 2022 Jordan 2 March – 31 December 2020 COVID-19 cases Average daily temperature, relative humidity, wind speed, pressure, and the concertation of the four pollutants (CO, NO2, PM10, and SO2) Multiple linear regression model Test accuracy All climate variables were used to establish the prediction model for COVID-19 prediction Test accuracy were 52.34%, 32.67%, and 42.37% for the city of Amman, Irbid, and Zarqa, respectively Mehmood, et al. [11] 2021 Pakistan 1 June – 31 July 2020 The daily confirmed COVID-19 cases PM2.5, temperature, dew point, humidity, wind speed, and pressure range. Generalized linear model (GLM) A P value < 0.05 was considered statistically significant PM2.5, temperature, dew point, wind speed, and pressure range Results indicated a significant relationship between COVID-19 cases and PM2.5 and climatic factors at P < 0.05 except for Lahore in case of humidity (r=0.175). Moazeni, et al. [12] 2022 Iran 1 March 2020 – 19 January 2021 The incidence of COVID-19 positive cases Climate region, months of the year, air pollutants, and climatological variables The GLM A P value < 0.05 was considered statistically significant Climate regions, months of the year, and solar energy Climate regions, months of the year, and solar energy were positively related to the yearly data of COVID-19 incidence Nevels, et al. [15] 2021 Wuhan city in China 14 January and 17 March 2020 Effective reproductive number (Reff)** Daily average temperature and average relative humidity GLM and generalized additive model (GAM) with Gamma distribution and logarithm link function Risk ratio (RR) Daily average temperature Daily average temperature is associated with Reff after adjusting for relative humidity (RR 0.92, 95% CI: 0.90, 0.95): Reff would increase by 7.6% (95% CI: 5.4% ~ 9.8%) per 1℃ drop in mean temperature at prior moving average of 0–8 days lag. Liu, et al. [16] 2022 153 countries and 31 provinces of mainland China 1 March – 5 May 2020 for 153 countries: 17 January – 30 April 2020 for China Daily new cases of COVID-19 Average temperature (AT), maximum temperature (MAXT), minimum temperature (MINT) The GAM RR AT is the primary exposure with a significant association. Repeated analyses bore similar associations when using MAXT and MINT. For the temperate zone or temperatures < 25th percentile or below the smoothing plot peak, the number of confirmed cases increased with increasing temperature, but when in the tropical zone or temperatures were > 25th percentile or above the smoothing plot peak, the correlations were reversed. Nottmeyer, et al. [13] 2021 England 30 January – 31 October 2020 COVID-19 incidence Ambient temperature, absolute humidity, and relative humidity The GLM, and a distributed lag non-linear model (DLNM) RR Ambient temperature, absolute humidity, and relative humidity High risk for COVID-19 at ambient temperature around 12℃ (RR = 1.62, 95%-CI: 1.44: 1.81): relative humidity of 61% (RR = 1.45, 95%-CI: 1.10: 1.90): lower absolute humidity around 6 to 8 g/m3 (RR=1.61, 95% CI: 1.41, 1.83) Liu, et al. [14] 2020 Australia, South Korea, and Italy 31 January – 31 March 2020 for three countries R0* Air pollutants, daily median air temperature, and humidity The DLNM on Quasi-Poisson regression RR Air temperature and humidity have lag and persistence on short-term R0 Seasonal analyses reflect the effect of temperature on R0. 27.6℃ and 8.5℃ are the optimum temperature (lowest RR) for Brisbane (Australia, southern hemisphere) and Gyeonggi-do (South Korea, northern hemisphere), respectively Zhang, et al. [17] 2022 Continental U.S. (CONUS) at county level 23 March – 1 September 2020 The daily COVID-19 cases Five weather-related variables: temperature, specific humidity, shortwave radiation, wind speed, and precipitation Machine learning algorithm (random forest regression [RF] model) Normalized root-mean-square-error (NRMSE) Estimation of COVID-19 cases is more accurate when adding five weather variables in addition to population, population density, and social distance index The performance of the RF model reached the lowest NEMSE (3.46) when using 100-tree model setting, eight predictors in all validation counties Han, et al. [18] 2022 China 10 January – 11 June 2021 Two outcomes: COVID-19 occurrence (COV_O) and COVID-19 intensity (COV_I) In addition to the potential COVID-19-related risk factors, climate variables considered daily minimum temperature (MinT), maximum temerature (MaxT), mean temperature (MeanT), relative humidity (Rh) and minimum relative humidity (MinRh). For COV_O and COV_I, eight machine learning models were applied The accuracy was the main evaluation metric for model performance when COV_O was the outcome, while R2 for COV_I In addition to the risk factors such as air pollutants, the climate variables MinRh, MaxT, and MinT were selected with a high coefficient to predict COV_O: and MaxT was selected to predict COV_I. The optimal model on COV_O prediction presented an accuracy of 91.91% and the best R2 of COV_I prediction reached 0.778. The occurrence was higher under extreme weather and high MinRh. 
下载: 导出CSV
-
[1] Wang Y, Hou H, Wang W. Combination of CT and RT-PCR in the screening or diagnosis of COVID-19. J Glob Health, 2020: 10(1): 010347. doi: 10.7189/jogh.10.010347 [2] World Health Organization. Coronavirus (COVID-19) Dashboard. Available from: https://covid19.who.int/ . Accessed on 07 October 2022.[3] Wang W, Yan Y, Guo Z, et al. All around suboptimal health—a joint position paper of the suboptimal health study consortium and European association for predictive, preventive and personalised medicine. EPMA J, 2021: 12(4): 403–433. doi: 10.1007/s13167-021-00253-2 [4] Wang W. Cardiovascular health in China: low level vs high diversity. The Lancet Regional Health–Western Pacific, 2020: 3: 100038. doi: 10.1016/j.lanwpc.2020.100038 [5] Hou H, Yang H, Liu P, et al. Profile of immunoglobulin G N-glycome in COVID-19 patients: a case-control study. Front Immunol, 2021: 12: 748566. doi: 10.3389/fimmu.2021.748566 [6] Dowell S F, Ho M S. Seasonality of infectious diseases and severe acute respiratory syndrome–what we don't know can hurt us. Lancet Infect Dis, 2004: 4(11): 704–708. doi: 10.1016/S1473-3099(04)01177-6 [7] Yuan J, Yun H, Lan W, et al. A climatologic investigation of the SARS-CoV outbreak in Beijing, China. Am J Infect Control, 2006: 34(4): 234–236. doi: 10.1016/j.ajic.2005.12.006 [8] Loché Fernández-Ahúja J M, Fernández Martínez J L. Effects of climate variables on the COVID-19 outbreak in Spain. Int J Hyg Environ Health, 2021: 234: 113723. doi: 10.1016/j.ijheh.2021.113723 [9] Pan J, Yao Y, Liu Z, et al. Warmer weather unlikely to reduce the COVID-19 transmission: An ecological study in 202 locations in 8 countries. Sci Total Environ, 2021: 753: 142272. doi: 10.1016/j.scitotenv.2020.142272 [10] Hamdan M, Dabbour L, Abdelhafez E. Study of climatology parameters on COVID-19 outbreak in Jordan. Environ Earth Sci, 2022: 81(8): 228. doi: 10.1007/s12665-022-10348-2 [11] Mehmood K, Bao Y, Abrar M M, et al. Spatiotemporal variability of COVID-19 pandemic in relation to air pollution, climate and socioeconomic factors in Pakistan. Chemosphere, 2021: 271: 129584. doi: 10.1016/j.chemosphere.2021.129584 [12] Moazeni M, Maracy M R, Dehdashti B, et al. Spatiotemporal analysis of COVID-19, air pollution, climate, and meteorological conditions in a metropolitan region of Iran. Environ Sci Pollut Res Int, 2022: 29(17): 24911–24924. doi: 10.1007/s11356-021-17535-x [13] Nottmeyer L N, Sera F. Influence of temperature, and of relative and absolute humidity on COVID-19 incidence in England - A multi-city time-series study. Environ Res, 2021: 196: 110977. doi: 10.1016/j.envres.2021.110977 [14] Liu Y, Lin X, Qin S. The short-term seasonal analyses between atmospheric environment and covid-19 in epidemic areas of cities in Australia, South Korea, and Italy. preprint arXiv: 2022: 194(3): 195. [15] Nevels M, Si X, Bambrick H, et al. Weather variability and transmissibility of COVID-19: a time series analysis based on effective reproductive number. Exp Results, 2021: 2: e15. doi: 10.1017/exp.2021.4 [16] Liu M, Li Z, Liu M, et al. Association between temperature and COVID-19 transmission in 153 countries. Environ Sci Pollut Res Int, 2022: 29(11): 16017–16027. doi: 10.1007/s11356-021-16666-5 [17] Zhang X, Maggioni V, Houser P, et al. The impact of weather condition and social activity on COVID-19 transmission in the United States. J Environ Manage, 2022: 302(Pt B): 114085. [18] Han Y, Huang J, Li R, et al. Impact analysis of environmental and social factors on early-stage COVID-19 transmission in China by machine learning. Environ Res, 2022: 208: 112761. doi: 10.1016/j.envres.2022.112761 [19] Auler A, Cássaro F, Silva V, et al. Evidence that high temperatures and intermediate relative humidity might favor the spread of COVID-19 in tropical climate: a case study for the most affected Brazilian cities. Sci Total Environ, 2020: 729: 139090. doi: 10.1016/j.scitotenv.2020.139090 [20] Su X, Yan X, Tsai CL. Linear regression. WIREs, 2012: 4(3): 275–294. [21] Grissom R J, Kim J J. Effect sizes for research: univariate and multivariate applications. Routledge, 2012. [22] Gotway C A, Stroup W W. A generalized linear model approach to spatial data analysis and prediction. J Agric Biol Environ Stat, 1997: 157–178. [23] Nelder J A, Wedderburn R W. Generalized linear models. J R Stat Soc Ser A, 1972: 135(3): 370–384. doi: 10.2307/2344614 [24] Karadağ M, Kulb S, Yoloğluc S, et al. Comparison of GAM and DLNM Methods for Disease Modeling in Environmental Epidemiology. Turkiye Klinikleri Journal of Biostatistics, 2021: 13(1). [25] Liu H. Generalized additive model., Duluth, USA: Department of Mathematics and Statistics, University of Minnesota, 2008: 55812. [26] Gasparrini A. Distributed lag linear and non-linear models in R: the package dlnm. J Stat Softw, 2011: 43(8): 1. [27] Zheng Y L, Guo Z, Zhang Y B, et al. Rapid triage for ischemic stroke: a machine learning-driven approach in the context of predictive, preventive and personalize medicine. EPMA J, 2022: 1–14. [28] Wu Z, Li L, Jin R, et al. Texture feature-based machine learning classifier could assist in the diagnosis of COVID-19. Eur J Radiol, 2021: 137: 109602. doi: 10.1016/j.ejrad.2021.109602 -
点击查看大图
表(1)
计量
- 文章访问数: 294
- HTML全文浏览量: 176
- PDF下载量: 14
- 被引次数: 0
