京津冀气象条件对学龄前儿童流感的影响及预报方法研究

李怡; 陈仲榆; 柳艳香; 鲁亮

doi:10.11676/qxxb2022.042

京津冀气象条件对学龄前儿童流感的影响及预报方法研究

doi: 10.11676/qxxb2022.042

李怡^1,,
陈仲榆¹,
柳艳香^1, ,,
鲁亮²

1.
中国气象局公共气象服务中心，北京，100081
2.
中国疾病预防控制中心传染病预防控制所，北京，102206

基金项目: 国家科技基础资源调查专项（2017FY101206）、中国气象局公共气象服务中心创新基金项目（Y2020003）

详细信息

作者简介:
李怡，主要从事健康气象服务与气候变化研究。E-mail：liyipmsc@cma.gov.cn

通讯作者:
柳艳香，主要从事应用气象研究。E-mail：liuyx@cma.gov.cn

中图分类号: P49
计量
- 文章访问数: 214
- HTML全文浏览量: 31
- PDF下载量: 79
- 被引次数: 0
出版历程
- 收稿日期: 2021-09-14
- 录用日期: 2022-05-11
- 修回日期: 2022-03-26
- 网络出版日期: 2022-04-12

The influence of meteorological conditions on influenza among preschool children over Beijing-Tianjin-Hebei area and prediction methods

LI Yi^1
,,
CHEN Zhongyu¹,
LIU Yanxiang^{1
, ,},
LU Liang²

1.
Public Meteorological Service Centre，China Meteorological Administration，Beijing 100081，China
2.
National Institute for Communicable Disease Control and Prevention，Chinese Center for Disease Control and Prevention，Beijing 102206，China

摘要

摘要: 基于2014—2016年京津冀地区学龄前儿童流感发病人数和气象观测数据，研究了学龄前儿童流感发病与单个气象要素以及综合气象条件的关系，结果表明：该地区学龄前儿童流感发病人数与一周以内的气温、相对湿度、气压和综合气象条件指标—大气环境人体感知度（BPWI）存在显著线性相关。大气环境人体感知度与学龄前儿童流感具有更稳定的暴露-反应关系：当BPWI≤−11或0≤BPWI<10 时，随着BPWI减小，流感发病风险增大。气压是另一个显著影响流感发病的气象要素：当本站气压>905 hPa，随着气压的升高，流感发病人数增多；当本站气压达到1007 hPa时，流感发病风险达到峰值。在厘清暴露-反应关系的基础上，采用机器学习方法进行预报建模，发现超前3天的BPWI对流感发病人数贡献最大。通过历史回报检验，得到了较好的学龄前儿童流感发病回报效果，为流感的分类人群干预提供了预报依据和科学参考。
- 流感发病特征 /
- 气象条件 /
- 广义相加模型 /
- 机器学习 /
- 预报模型
Abstract: Based on the number of influenza cases amongst preschool children and meteorological observations over Beijing-Tianjin-Hebei area from 2014 to 2016, the relationships between the incidence of preschool children influenza and individual meteorological factor as well as their combined conditions are investigated. The results indicate significant linear correlations between the number of influenza infection amongst preschool children and temperature, relative humidity, atmospheric pressure and a defined comprehensive indicator Body Perception Weather Index (BPWI). The exposure-consequence relationship based on the BPWI is more stable, i.e., while the BPWI values equal or smaller than −11 or within the range of 0—10 correspond to higher risk of influenza. Local atmospheric pressure is another key factor. When the station pressure is higher than 905 hPa, higher pressure causes more infection, and the infection peak corresponds to 1007 hPa. On the basis of these understanding, a machine learning method is used to perform prediction experiment, and it is found that the BPWI with a 3 d lead time contributes the most to influenza incidence. The hindcast evaluation reports a fairly good performance of the prediction model, and this provides valuable evidences and scientific clues for pre-intervention.
- Characteristic of influenza cases /
- Meteorological factors /
- Generalized additive model /
- Machine learning /
- Forecast model

HTML全文

图 1 京津冀流感发病年龄密度分布（a. 2014年，b. 2015年，c. 2016年）

Figure 1. Kernel density estimates of age distribution of influenza occurrence in Beijing-Tianjin-Hebei area （a. 2014，b. 2015，c. 2016）

下载: 全尺寸图片幻灯片

图 2 京津冀2014—2016年学龄前儿童流感逐月发病人数与月平均气温时间序列

Figure 2. Monthly time series of preschool children influenza cases and monthly average temperature in Beijing-Tianjin-Hebei area from 2014 to 2016

下载: 全尺寸图片幻灯片

图 3 京津冀学龄前儿童流感各季节日平均发病率特征（a. 春季，b. 夏季，c. 秋季，d. 冬季）

Figure 3. Average daily incidence of influenza in preschool children in Beijing-Tianjin-Hebei area （a. spring，b. summer，c. autumn，d. winter）

下载: 全尺寸图片幻灯片

图 4 2014—2016年京津冀学龄前儿童流感发病人数与（a） BPWI和（b）气压的关系（实线表示流感相对危险度（RR），虚线表示95%置信区间）

Figure 4. Relationships of influenza cases with （a） BPWI and （b） station pressure among preschool children in Beijing-Tianjin-Hebei area during 2014—2016 （the solid line represents the logarithm of the relative risk of influenza，the dotted line represents the 95% confidence interval）

下载: 全尺寸图片幻灯片

图 5 XGBoost模型的特征重要性分析

Figure 5. Analysis of feature importance of the XGBoost model

下载: 全尺寸图片幻灯片

图 6 京津冀学龄前儿童流感预报模型回报拟合度（横坐标为检验集实际发病人数，纵坐标为检验集回报发病人数，蓝实线为回报值与实际值的拟合线，阴影为95%置信区间）

Figure 6. Degree of fitting for prediction of preschool chil-dren influenza cases by prediction model in Beijing-Tianjin-Hebei area （the horizontal axis is the reported cases in test set，the vertical axis is the hindcast cases in test set，the solid blue line is the fitting line between the hindcast value and the reported value，the shade is the confidence interval of 95%）

下载: 全尺寸图片幻灯片

图 7 京津冀学龄前儿童流感历史回报效果检验

Figure 7. Validation of hindcast for preschool children influenza cases in Beijing-Tianjin-Hebei area

下载: 全尺寸图片幻灯片

表 1 2014—2016年逐日各气象要素间的相关系数

Table 1. Correlation coefficients between daily meteorological factors from 2014 to 2016

	气温	相对湿度	风速	日照时数	降水量	气压	BPWI
气温	1.000
相对湿度	0.280^**	1.000
风速	0.027	−0.437^**	1.000
日照时数	0.265^**	−0.584^**	0.300^**	1.000
降水量	0.206^**	0.478^**	0.062^*	−0.430^**	1.000
气压	−0.614^**	−0.194^**	−0.175^**	−0.186^**	−0.218^**	1.000
BPWI	0.993^**	0.253^**	−0.029	0.318^**	0.158^**	−0.596^**	1.000
注： * 表示 P<0.05，** 表示P<0.001。

下载: 导出CSV

表 2 2014—2016年逐日各气象要素区域平均统计量

Table 2. Statistics of daily regional-mean of meteorological factors in Beijing-Tianjin-Hebei area from 2014 to 2016

统计量	BPWI	气温（℃）	相对湿度（%）	风速（m/s）	日照时数（h）	降水量（mm）	气压（hPa）
平均值	−2.95	12.95	58.38	1.97	6.32	1.18	999.05
标准差	10.24	10.77	17.60	0.71	3.55	3.81	23.24
最小值	−31.62	−14.95	16.43	0.60	0.00	0.00	852.40
25%分位数	−12.85	2.25	44.57	1.45	3.49	0.00	992.03
中位数	−1.73	14.67	59.00	1.84	6.89	0.00	1003.98
75%分位数	6.36	22.96	72.33	2.33	9.13	0.22	1013.18
最大值	15.26	31.47	94.50	6.30	13.35	56.53	1034.98

下载: 导出CSV

表 3 2014—2016年京津冀学龄前儿童流感发病与气象要素（发病当日及滞后）相关系数

Table 3. Correlation coefficients between preschool children influenza cases and meteorological factors at the onset day and various lag days from 2014 to 2016

	气温	相对湿度	风速	日照时数	降水量	气压	BPWI
发病当日	−0.762^**	−0.381^**	0.037	−0.136^**	−0.221^**	0.534^**	−0.751^**
发病前1天	−0.763^**	−0.378^**	0.029	−0.150^**	−0.263^**	0.568^**	−0.754^**
发病前2天	−0.767^**	−0.372^**	0.009	−0.144^**	−0.265^**	0.564^**	−0.755^**
发病前3天	−0.768^**	−0.368^**	0.012	−0.141^**	−0.254^**	0.572^**	−0.758^**
发病前4天	−0.769^**	−0.365^**	0.010	−0.143^**	−0.252^**	0.576^**	−0.759^**
发病前5天	−0.770^**	−0.355^**	0.001	−0.143^**	−0.260^**	0.575^**	−0.759^**
发病前6天	−0.781^**	−0.332^**	−0.010	−0.172^**	−0.247^**	0.578^**	−0.771^**
注：** 表示P<0.001。

下载: 导出CSV

参考文献(24)

[1]	段玮,吴双胜,孙瑛等. 2017. CUSUM模型在北京市流感流行起始预警中的应用. 国际病毒学杂志,24(2):111-114 doi: 10.3760/cma.j.issn.1673-4092.2017.02.010 Duan W,Wu S S,Sun Y,et al. 2017. Application of CUSUM model in the early detection of influenza epidemics in Beijing. Int J Virol,24(2):111-114 (in Chinese) doi: 10.3760/cma.j.issn.1673-4092.2017.02.010
[2]	郭倩,陈涛,周罗晶. 2020. 季节性流感与气象因素相关性的系统评价. 疾病监测,35(12):1081-1087 Guo Q,Chen T,Zhou L J. 2020. Correlation between seasonal influenza and meteorological factors:A systematic review. Dis Surveill,35(12):1081-1087 (in Chinese)
[3]	李媛,马辉,崔壮等. 2011. 天津市流行性感冒与气象条件的关系研究. 环境与健康杂志,28(3):227-230 Li Y,Ma H,Cui Z,et al. 2011. Relationship between influenza and meteorological conditions in Tianjin. J Environ Health,28(3):227-230 (in Chinese)
[4]	柳艳香,李宇光,苗蕾等. 2022. 大气环境人体感知度与脑梗塞疾病的关系研究—以黑龙江省五常地区为例. 气象学报,80(3):474-481 Liu Y X, Li Y G, Miao L,et al. 2022. The relationship between body perception of weather indices and cerebral infarction disease:A case study in Wuchang,Heilongjiang province. Acta Meteor Sinica,80(3):474-481 (in Chinese)
[5]	卢汉体,李傅冬,林君芬等. 2015. 基于支持向量机的浙江省流感样病例预警模型研究. 浙江大学学报(医学版),44(6):653-658 doi: 10.3785/j.issn.1008-9292.2015.11.09 Lu H T,Li F D,Lin J F,et al. 2015. Construction of early warning model of influenza-like illness in Zhejiang province based on support vector machine. J Zhejiang Univ (Med Sci),44(6):653-658 (in Chinese) doi: 10.3785/j.issn.1008-9292.2015.11.09
[6]	秦强,申昆玲. 2019. 儿童重症流感发病机制. 中华实用儿科临床杂志,34(2):98-101 doi: 10.3760/cma.j.issn.2095-428X.2019.02.005 Qin Q,Shen K L. 2019. Pathogenesis of severe influenza in children. Chinese J Appl Clin Pediatr,34(2):98-101 (in Chinese) doi: 10.3760/cma.j.issn.2095-428X.2019.02.005
[7]	陶芳芳,阚海东,董晨等. 2010. 上海市流感样病例与气象因素关系的研究. 中华流行病学杂志,31(12):1448-1449 doi: 10.3760/cma.j.issn.0254-6450.2010.12.035 Tao F F,Kan H D,Dong C,et al. 2010. Analysis on the relationship between influenza and meteorology and the establishment of early warning model. Chinese J Epidemiol,31(12):1448-1449 (in Chinese) doi: 10.3760/cma.j.issn.0254-6450.2010.12.035
[8]	于永,王炜翔,周连等. 2015. 南京市流感样病例与气温的关系研究. 环境卫生学杂志,5(5):414-417 Yu Y,Wang W X,Zhou L,et al. 2015. Analysis on relationship between influenza incidence and temperature in Nanjing. J Environ Hyg,5(5):414-417 (in Chinese)
[9]	郑月燕,祝寒松,叶雯婧等. 2019. 福州市流行性感冒与气象因素滞后效应分析. 中国人兽共患病学报,35(12):1154-1159 Zheng Y Y,Zhu H S,Ye W J,et al. 2019. Analysis on the lag effect of influenza and meteorological factors in Fuzhou. Chinese J Zoonoses,35(12):1154-1159 (in Chinese)
[10]	Baumgartner A E,Dao C N,Nasreen S,et al. 2012. Seasonality,timing,and climate drivers of influenza activity worldwide. J Infect Dis,206(6):838-846 doi: 10.1093/infdis/jis467
[11]	Chen T Q,Guestrin C. 2016. XGBoost:A scalable tree boosting system∥Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco:ACM,785-794
[12]	Dai Q G,Ma W,Huang H D,et al. 2018. The effect of ambient temperature on the activity of influenza and influenza like illness in Jiangsu province,China. Sci Total Environ,645:684-691 doi: 10.1016/j.scitotenv.2018.07.065
[13]	Huang X D,Mengersen K,Milinovich G,et al. 2017. Effect of weather variability on seasonal influenza among different age groups in Queensland,Australia:A Bayesian spatiotemporal analysis. J Infect Dis,215(11):1695-1701 doi: 10.1093/infdis/jix181
[14]	Iuliano A D,Roguski K M,Chang H H,et al. 2018. Estimates of global seasonal influenza-associated respiratory mortality:A modelling study. Lancet,391(10127):1285-1300 doi: 10.1016/S0140-6736(17)33293-2
[15]	Li Y P,Wang X L,Zheng X Y. 2018. Impact of weather factors on influenza hospitalization across different age groups in subtropical Hong Kong. Int J Biometeorol,62(9):1615-1624 doi: 10.1007/s00484-018-1561-z
[16]	Liu W D,Dai Q G,Bao J,et al. 2019a. Influenza activity prediction using meteorological factors in a warm temperate to subtropical transitional zone,Eastern China. Epidemiol Infect,147:e325 doi: 10.1017/S0950268819002140
[17]	Liu Z D,Zhang J,Zhang Y,et al. 2019b. Effects and interaction of meteorological factors on influenza:Based on the surveillance data in Shaoyang,China. Environ Res,172:326-332 doi: 10.1016/j.envres.2019.01.053
[18]	Lu Y,Wang S,Wang J Y,et al. 2019. An epidemic avian influenza prediction model based on google trends. Lett Org Chem,16(4):303-310 doi: 10.2174/1570178615666180724103325
[19]	Polozov I V,Bezrukov L,Gawrisch K,et al. 2008. Progressive ordering with decreasing temperature of the phospholipids of influenza virus. Nat Chem Biol,4(4):248-255 doi: 10.1038/nchembio.77
[20]	Qi L,Liu T,Gao Y,et al. 2021. Effect of meteorological factors on the activity of influenza in Chongqing,China,2012–2019. PLoS One,16(2):e0246023 doi: 10.1371/journal.pone.0246023
[21]	Silvennoinen H,Peltola V,Vainionpää R,et al. 2011. Incidence of influenza-related hospitalizations in different age groups of children in Finland:A 16-year study. Pediatr Infect Dis J,30(2):e24-e28 doi: 10.1097/INF.0b013e3181fe37c8
[22]	Wang Y F,Xu K,Kang Y,et al. 2020. Regional influenza prediction with sampling Twitter data and PDE model. Int J Environ Res Public Health,17(3):678 doi: 10.3390/ijerph17030678
[23]	Worby C J,Chaves S S,Wallinga J,et al. 2015. On the relative role of different age groups in influenza epidemics. Epidemics,13:10-16 doi: 10.1016/j.epidem.2015.04.003
[24]	Yu H J,Alonso W J,Feng L Z,et al. 2013. Characterization of regional influenza seasonality patterns in China and implications for vaccination strategies:Spatio-temporal modeling of surveillance data. PLoS Med,10(11):e1001552 doi: 10.1371/journal.pmed.1001552