Impact of complex environmental exposures on acute symptoms in Jinan: Based on LASSO variable selection and generalized additive mixed models
- VernacularTitle:济南市复合环境暴露对急性症状的影响：基于LASSO变量筛选与广义加法混合模型评估
- Author:
Yongxue CUI
1
;
Fangyi WANG
1
;
Qi ZHANG
2
;
Caixia MA
1
;
Xingyi GENG
1
Author Information
- Publication Type:Selectedarticle
- Keywords: air pollution; meteorological factor; acute symptom; least absolute shrinkage and selection operator regression; generalized additive mixed model
- From: Journal of Environmental and Occupational Medicine 2025;42(10):1177-1184
- CountryChina
- Language:Chinese
- Abstract: Background Air pollution and meteorological factors exert complex nonlinear effects on acute symptoms in the population, with intricate interactions among these factors. Traditional statistical methods struggle to simultaneously address complex nonlinear relationships and multicollinearity issues. Objective To delineate the dynamic effects of air pollutants and meteorological parameters on acute symptoms in three distinct populations with the multicollinearity being addressed and to generate reliable scientific evidence for prevention and control of health risk factors. Methods A time-series study design was employed to collect data on air pollution (daily mean temperature, daily precipitation, daily mean relative humidity, and daily mean wind speed), meteorological factors [Air Quality Index (AQI), fine particulate matter (PM2.5), inhalable particulate matter (PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and 8-hour maximum ozone (O3)], and acute symptoms such as fever, cough, and sore throat in Jinan from June to December 2023. Key variables were selected using least absolute shrinkage and selection operator (LASSO) regression, followed by generalized additive mixed modeling (GAMM) to analyze the health effects of combined environmental exposures to air pollution and meteorological factors. Linear variables were modeled using linear mixed-effects function, nonlinear variables were smoothed using thin-plate regression splines, and variables with interaction effects were smoothed using low-rank scale-invariant tensor product splines. Fluctuations in independent variables following a normal distribution were treated as sampling errors and incorporated as random effects in the GAMM. Results For fever, the daily mean temperature, daily mean relative humidity, daily mean wind speed, and ambient SO2 were statistically significant (P<0.05), with daily mean wind speed being a linear influencing factor. When the daily mean temperature was below 3 °C, each 10 °C increase corresponded to a relative risk (RR) of 2.64 (95%CI: 2.50, 2.79). When the daily mean temperature was ≥3 °C, each 10 °C increase corresponded to an RR of 0.86 (95%CI: 0.83, 0.89). Each 10% increase in daily mean relative humidity was associated with an RR of 0.93 (95%CI: 0.89, 0.97). Each 1 m·s−1 increase in daily mean wind speed corresponded to an RR of 1.06 (95%CI: 1.02, 1.10). Within the concentration ranges of <10 μg·m−3, 10–<12.5 μg·m−3, and ≥12.5 μg·m−3, each 1 μg·m−3 increase in ambient SO2 corresponded to RR values of 1.01 (95%CI: 0.98, 1.05), 1.21 (95%CI: 1.17, 1.24), and 0.97 (95%CI: 0.94, 0.99), respectively. For cough, the daily mean temperature, daily mean relative humidity, PM10, and SO2 were statistically significant (P<0.001), with PM10 being a linear influencing factor. When the daily mean temperature was below 1 °C, each 10 °C increase corresponded to an RR of 1.47 (95%CI: 1.42, 1.52). When the daily mean temperature was ≥1 °C, each 10 °C increase corresponded to an RR of 0.85 (95%CI: 0.82, 0.87). Each 10% increase in daily mean relative humidity was associated with an RR of 0.95 (95%CI: 0.92, 0.98). Each 50 μg·m−3 increase in PM10 concentration corresponded to an RR of 1.05 (95%CI: 1.02, 1.08). Within the concentration ranges of <10 μg·m−3, 10–<12.5 μg·m−3, and ≥ 12.5 μg·m−3, each 1 μg·m−3 increase in ambient SO2 corresponded to RR values of 1.00 (95%CI: 0.97, 1.03), 1.12 (95%CI: 1.09, 1.16), and 0.98 (95%CI: 0.95, 1.00), respectively. For sore throat, the daily mean temperature, daily mean relative humidity, daily mean wind speed, PM10, and SO2 were statistically significant (P<0.05), with daily mean wind speed and PM10 being linear influencing factors. When the daily mean temperature was below 2 °C, each 10 °C increase corresponded to an RR of 1.82 (95%CI: 1.69, 1.96). When the daily mean temperature was ≥2 °C, each 10 °C increase corresponded to an RR of 0.81 (95%CI: 0.77, 0.87). Each 10% increase in daily mean relative humidity was associated with an RR of 0.94 (95%CI: 0.88, 1.00). Within the concentration ranges of <10 μg·m−3, 10–<12.5 μg·m−3, and ≥12.5 μg·m−3, each 1 μg·m−3 increase in ambient SO2 corresponded to RR values of 1.02 (95%CI: 0.97, 1.08), 1.13 (95%CI: 1.08, 1.19), and 0.98 (95%CI: 0.94, 1.02), respectively. Each 1 m·s−1 increase in daily mean wind speed and each 50 μg·m−3 increase in PM10 concentration were associated with RR values of 1.06 (95%CI: 1.00, 1.12) and 1.04 (95%CI: 0.98, 1.10), respectively. An interaction effect was observed between daily mean wind speed and PM10: increasing daily mean wind speed non-linearly reduced the impact of PM10, on sore throat whereas PM10 had no significant effect on wind speed. Conclusion This study, by combining LASSO and GAMM, largely eliminates the multicollinearity among selected variables. It reveals complex non-linear effects and interactions between air pollutants, meteorological factors, and acute symptoms in different population groups in Jinan. The symptoms like fever, cough, and sore throat are non-linearly associated with daily mean temperature and SO2 concentration, while PM10 and wind speed show a linear relationship or interactive effects. These findings provide a new basis for the precise prevention and control of health risk factors.
