空气污染物浓度与呼吸系统疾病的关系研究

Relationship between air pollutant and respiratory diseases

利用杭州市日均空气污染物浓度与呼吸系统疾病门诊人数数据,结合泊松广义相加模型(GAM)和反向传输(BP)神经网络模型,评价该区域主要空气污染物对居民呼吸系统疾病的影响,并进行短期门诊人数预测,结果表明:PM_(2.5)、PM_(10)、NO_2和SO_2每增加1个四分位间距(IQR)时,对呼吸系统疾病门诊人数的相对危险度(RR)最大值分别为1.030(95%置信区间(CI):1.016~1.045)、1.063(95%CI:1.043~1.084)、1.053(95%CI:1.016~1.091)和1.025(95%CI:1.003~1.048),且分别在滞后3、2、4、3d时达到最大值,可见PM_(2.5)、PM_(10)、SO_2和NO_2对呼吸系统疾病存在滞后效应。BP神经网络模型对呼吸系统疾病门诊人数的预测值与实际值接近,且平均相对误差为13.821%,说明BP神经网络模型可用于呼吸系统疾病门诊人数的短期预测。

The impact of major air pollutants on respiratory diseases of residents was evaluated and respiratory diseases outpatient visits in Hangzhou was predicted,based on data of air pollutant concentration and respiratory diseases outpatient visits.A Poisson generalized additive model（GAM）and back propagation（BP）neural networks model were used in this study.The results showed that when PM（2.5）,PM（10）,NO2and SO2increased by an inter-quartile range（IQR）,the maximum relative risk（RR）of respiratory diseases were 1.030（95% confidence interval（CI）：1.016-1.045）,1.063（95%CI：1.043-1.084）,1.053（95%CI：1.016-1.091）and 1.025（95%CI：1.003-1.048）,which appeared after 3,2,4and 3days,respectively,indicating a lag effect on respiratory diseases.The forecast values of respiratory diseases outpatient visits by BP neural networks model was close to the observation values,with its average relative error of 13.821%.Therefore,BP neural networks model could be used to predict the short-term respiratory diseases outpatient visits.