Analysis On The Epidemic Characteristics Of Influenza In Shanxi Province And The Effect Of Early Warning Based On Wavelet-ARIMA Model

Objective: To study the epidemic and etiology characteristics of influenza in Shanxi Province in 2010-2016,and to understand the epidemic distribution,prevalence,predominant strains,and virus variation of influenza;According the epidemic characteristics results,we established the ARIMA and Wavelet-ARIMA model to make short-time predictions and compared the prediction effects of this two models,in order to explore the effective and reasonable prediction model,achieve dynamic early warning and provide effective theoretical support for prevention and control of influenza in Shanxi province.Methods: To collect influenza-like surveillance data of 17 sentinel hospitals and 12 network laboratories in Shanxi Province from the year 2010 to 2016,and analyze the epidemic distribution characteristics and etiological monitoring conditions;Using the data of influenza-like illnesses accounted for the percentage of outpatient emergency visits(ILI%)from the 14 th week of 2010 to the 13 th week of 2016 to establish the Wavelet-ARIMA model;the data of ILI% from the 14 th to 17 th week of 2016 were used to evaluate the predictive effect of two models.Results: 1.In the year of 2010-2016,17 emergency sentinel hospitals in Shanxi Province collected 16,005,008 emergency monitoring cases,288,627 influenza-like cases,and influenza-like cases accounted for 1.87% of outpatient emergencies.The percentage of ILI% in 2010-2016 was significantly decreasing(trend =6868.846,P<0.001);The composition of influenza-like cases of different monitoring years in different age groups was different(203.53022(28),P<0.001),the group of 0~years old had the largest number of influenza-like cases(54.3%)and lowest in the 60~years old group(3.4%).2.In 2010-2016,12 influenza surveillance network laboratories in Shanxi Province tested a total of 56,643 submitted samples,of which 9637 were positive,and the positive rate was 17.01%.The positive rates of influenza virus isolation in different years were different(=341.367,P<0.001);the influenza virus subtypes in different years were different(=479.757,P<0.001),and the dominant strains in 2010-2016 were H3N2 type,B type,H3N2 type,new H1N1 type,H3N2 type,series Victoria,H3N2 type;the positive rate of virus detection was different in different age groups(=479.757,P<0.001),the positive rate in 5-year-old group was highest(21.18%);there was no significant difference in the positive detection rate of influenza virus among different sexes(=1.706,P=0.192);the positive rate of influenza virus was different in different regions(=669.385,P<0.001).3.Compared with the modeling performance of ARIMA model,the consistency index of wavelet-ARIMA increased by 11.3%,the MAE decreased by 44.4%,the MAPE decreased by 44.9%,the MSE decreased by 64.7%,and the RMSE decreased;Compared with the prediction performance,the MAE decreased by 75%;the MAPE decreased by 77%;the MSE decreased by 90%;and the RMSE decreased by71.0%;the consistency index increased by 51.7%.Conclusions: The data of ILI% in Shanxi province from 2010 to 2016 showed a overall decline trend year-by-year.The peak period was mainly concentrated in winter and spring;the predominant strains prevalent in different years were different.Using the wavelet-ARIMA and ARIMA model to make short-term predictions of influenza surveillance data in Shanxi Province,the combined model's performance is better than the ARIMA model.