Exploration On Estimation Method Of Regional Cases Of Seasonal Influenza

Objective:This study aims to explore methods for estimating the regional cases of influenza and the under-reporting region and identify and reduce the degree of under-reporting of reported cases of influenza.Methods:The three sets of influenza surveillance data(influenza reported cases from Nationwide Notifiable Infectious Diseases Reporting Information System(NIDRIS),influenza-like illness consultation rate(ILI%)and influenza virus positive rate from Chinese Influenza Surveillance Information System)of North and South China from 2017 to 2018 were compared using peak comparison,cross correlation and Early Aberration Reporting System C3(EARS-C3)method.Peak comparison,correlation and EARS-C3 method were also used to compare two types of influenza data(influenza reported cases and ILI%)in Guangdong.According to the difference between the influenza report date and the onset date of each province in China and each city in Guangdong,the areas where the influenza reported cases were relatively closer to the true level,that is.the areas with relatively good reporting level,were identified.Based on the influenza reported cases of the relatively well reported cities in Guangdong,combined with the influenza-related Baidu search index of each city in Guangdong,the Biased Sentinel Hospital based Area Disease Estimation(B-SHADE)model and the ratio estimation method were used to estimate weekly influenza cases in Guangdong.and then assessed the degree of influenza under-reporting in Guangdong.We collected the weekly Baidu search index of influenza-related keywords from all cities in China,and used the influenza case data reported by 484 districts and counties containing national influenza surveillance sentinel hospitals to summarize the weekly number of influenza reported cases in each city.Cross correlation analysis was conducted on the weekly Baidu index of each kevword and the weekly influenza reported cases to select keywords.Co-linearity diagnosis and principal component analysis were performed on the Baidu index of the selected keywords.Finally,the main component of Baidu index and the reported cases of influenza were used for geographic weighted(GW)correlation analysis and geographic weighted(GW)regression analysis.According to the GW regression coefficient,the regions that were more likely to be under-reported were roughly judged.The basis for the judgment is that the region with a relatively lower GW regression coefficient was more likely to be under-reported.Results:The three sets of influenza surveillance data in mainland China are highly correlated and could reflect the epidemic trend of influenza,but the timeliness of the three monitoring data is different.Estimated results of the number of regional cases of Guangdong from 2017 to 2018 are as follows:The Pearson correlation coefficient between the B-SHADE estimated case number and the ILI%is 0.831.which is greater than the Pearson correlation coefficient of influenza reported cases and the ILI%of 0.781.The Pearson correlation coefficient between the estimated value of the ratio estimation method and the ILI%is 0.839,which is a!so greater than the Pearson correlation coefficient of influenza reported cases and the ILI%of 0.781.In summary,the correlation between the estimated values of the two methods and the ILI%is higher than reported cases of the surveillance system.The sum of reported cases of influenza in Guangdong from 2017 to 2018 is only 69.47%of the B-SHADE estimated value and 74.32%of the estimated value of the ratio estimation method,and the under-reported cases mainly occur during the peak period.The results of applying the geographical weighted(GW)model to explore the influenza under-reporting area are as follows:using the influenza-related compound Baidu index from September 4.2017 to March 22.2018(the main component one of the six keywords Baidu Index)and the number of cases reported in the same period,the GW correlation analysis provides evidence for the spatial non-stationary relationship between the influenza reported cases and the main component one of the Baidu Index.Using geographical weighted regression(GWR)to explore the relationship is needed in our study.According to R2 and AICc,the performance of the GWR model is significantly better than Eglobal OLS.The GW regression coefficient distribution of the main component one of the Baidu index shows that the GW regression coefficient in Henan,Shanxi,Shandong,Northern Jiangsu,Eastern Sichuan,Gansu,Northern Ningxia,Inner Mongolia,Liaoning,Jilin and Heilongjiang tends to be smaller,suggesting that these regions are more likely to be under-reported.Conclusion:The number of reported cases of influenza can reflect the epidemic trend of influenza,and it has good timeliness of early warning,but it still cannot represent the true level of influenza,as it was under-reported.Combining the Baidu index and the reported cases of influenza in cities of Guangdong,using the B-SHADE method and the ratio estimation method,the influenza reported cases of Guangdong from 2017 to 2018 is only about 70%of the estimated value of influenza cases.Geographically weighted(GW)regression analysis of influenza-related Baidu index and influenza reported cases in cities across the country in the autumn and winter of 2017-2018 revealed that influenza cases of Henan,Shanxi.Shandong,Northern Jiangsu,Eastern Sichuan,Gansu,Northern Ningxia,Inner Mongolia,Liaoning,Jilin and Heilongjiang are more likely to be under-reported.In summary,by combining the influenza-related Baidu index and the number of reported cases of influenza,the B-SHADE model and the ratio estimation method can be used to estimate the regional cases of influenza,and the application of the geographically weighted(GW)model can explore the influenza under-reporting area.These results can provide scientific references for influenza prevention and control.