呼和浩特市流感样病例与大气环境因素相关性分析与预测研究

Correlation between influenza-like illness cases and atmospheric environmental factors in Hohhot city and the prediction analysis

目的探讨流感样病例与气象因素、空气质量因素的相关关系,用时间序列分析法构建流感样病例预测模型,对流感预警预测技术进行有效探索.方法收集2014年至2017年呼和浩特市流感样病例监测资料及同期气象资料和空气质量因素资料,分析流感样病例与气象因素和空气质量因素的相关性,建立该地区流感样病例季节性自回归移动平均(seasonal autoregressive integrated moving average,SARIMA)模型,分析比较外部因素引入前后模型的拟合优度和预测精度.结果呼和浩特市流感样病例周病例数有明显的季节性及周期性,流感样病例周病例数与气温周均数滞后1周相关(rS=0.550,P<0.01)、与湿度周均数滞后1周相关(rS=-0.304,P<0.01).气温周均数滞后1周纳入流感样病例周病例数SARIMA(1,0,1)(0,1,1)52预测模型,拟合优度和预测精度均有所提高.结论流感样病例与空气质量因素偏相关分析未见有统计学意义;流感样病例的流行与气温周均数和湿度周均数有关,包含气温周均数的SARIMA模型可作为短期预测流感流行的技术方法.

Objective To explore the correlations between influenza-like illness cases and meteorological and air quality factors, and to construct a prediction model for influenza-like illness cases with time series analysis , so as to explore the technology of early warning and prediction for influenza. Methods The surveillance data of influenza-like illness cases from 2014 to 2017, as well as the meteorological data and air quality factors data during the same period in Hohhot were collected. The correlations between the correlations between influenza-like illness cases and meteorological and air quality factors were analyzed. The seasonal autoregressive integrated moving average (SARIMA) model was established for influenza-like illness cases in the region. The goodness of fit and prediction accuracy of the models before and after the introduction of external factors were compared. Results The number of influenza-like illness cases in Hohhot showed obvious seasonality and periodicity features. The weekly number of influenza-like illness cases correlated to one week lag of weekly average temperature (rs=0.550, P<0.01), and to one week lag of weekly average humidity (rs=-0.304, P<0.01). The one week lag of weekly average temperature was included in the SARIMA (1, 0, 1) (0, 1, 1)52 prediction model for the weekly number of influenza-like illness cases. Both goodness of fit and prediction accuracy have been improved. Conclusions Partial correlation analysis of influenza-like illness cases and air quality factors showed no statistical significance.The prevalence of influenza-like illness cases was related to the weekly average temperature and humidity. The SARIMA model contained weekly average temperature could be used as a technical method for short-term prediction of influenza epidemic.