Study On Early Warning Of Influenza In Zhejiang Province

Objective:The monitoring and early warning of influenza has become a focus work of public health worldwide. As a coastal city in the southeast China, Zhejiang Province has the high prevalence rate of influenza which causes large damage to the human health. The lack of relative forecasting model of influenza which adapts to the local situation has also become an urgent problem. This study aims to establish an appropriate forecasting model of influenza in Zhejiang Province, in order to provide scientific evidence for early detection of influenza pandemic and lower the economic loss and social burden to the largest extent.Methods:Number of flu-related illnesses, data of corresponding meteorological factors and positive rates of influenza pathogen were systematically collected among outpatients and emergency patients. The data was from11sentinel hospitals in Zhejiang Province during2012to2013with104weeks in total. The epidemiological characteristics of influenza during the period were then analyzed to describe the distribution. Linear correlation and rank correlation were conducted to explore the association between influenza-like illness and related factors. Finally, Multiple Linear Regression and Support Vector Machine were used to establish the forecasting model of influenza in Zhejiang Province and verified by the historical data. The differences between the two models were studied after that. The related statistical software used in the study including Excel2010, SAS9.2and SPSS Modeler14.1.Results:Correlation analysis indicated that8factors were associated with influenza-like illness occurred in one week, including number of flu-related illnesses and positive rates of influenza pathogen in the same week, average wind velocity3weeks earlier, average relative humidity3weeks earlier, average vapour pressure4weeks earlier, weekly average temperature4weeks earlier, lowest weekly temperature4weeks earlier, highest weekly temperature4weeks earlier. The results showed that several meteorological factors had an obvious lag effect on the occurrence of influenza. The results of influenza forecasting model after verifying revealed that Multiple Linear Regression and Support Vector Machine had the accuracy of46.7%and50.0%respectively with the same level prediction while the accuracy reached100%and96.7%respectively when predicted one higher or lower level. There was not much difference of the forecasting effect between the two models.Conclusions:The recent study demonstrate that number of influenza related diseases patients, meteorological factors and positive rates of influenza pathogen do influence the prevalence of influenza and establish the forecasting model of Zhejiang Province, however, the forecasting effect need to be further studied. The current results could provide scientific evidence for further study on the early-warning of influenza and help establish proper prevention and control strategies.