| Influenza is an acute respiratory infectious disease caused by influenza virus,which continues to threaten global public health security through recurrent seasonal epidemics and irregular pandemics.The seasonal influenza caused by influenza A and B viruses lead to hundreds of millions of cases,3 million to 5 million severe cases,and 290,000 to 650,000 influenza related respiratory deaths,which seriously affects the health of the population.Compared with seasonal epidemics,the influenza pandemic is more serious,of which the morbidity and mortality are higher,the transmission is more rapid,and the affected area is wider.Every pandemic brings disastrous effects to human health and social and economic development.The analysis of epidemiologic characteristics pandemic,seasonal influenza epidemic and human infection with avian influenza virus in Chongqing can provide scientific basis for the planned and systematic prevention and control against seasonal influenza and preparation for pandemic.On the basis of exploring the possible influence factors of influenza activity and previous reports,we conducted the research on forecasting influenza activity using self-adaptive AI model and multi-source data in Chongqing.According to the real-time forecasting results,early warning signals and health prompts could be sent out and early public health actions could be carried out in real time,so that a large amount of surveillance data would be timely transformed into more efficient public health policies,thus providing new theoretical basis and implementation strategies for Chongqing to achieve more accurate influenza prevention and control.Part Ⅰ Epidemiologic characteristics of influenza during pandemic and seasonal epidemicObjectivesTo systematically analyze the epidemiologic characteristics and variation trend of pandemic and seasonal influenza in Chongqing from 2009 to 2018,so as to provide scientific basis for the planned and systematic prevention and control against seasonal influenza and preparation for pandemic,and to explore the possible influence factors of influenza activity and provide theoretical basis for influenza forecasting.MethodsA descriptive epidemiological method was used to analyze the influenza cases and the pathogen surveillance data in Chongqing from 2009 to 2018.Results1.A total of 40,343 influenza cases were reported in Chongqing from 2009 to 2018.The influenza morbidity was 53.44/100,000 during the 2009 pandemic,which was 5.7 times of the annual average morbidity during the seasonal epidemic from 2010 to 2018.2.The ratio of male to female was 1.23:1,the average age was 18.47 years old,73.5% of the cases were 1-19 years old,and 58.4% of the cases were students.The proportions of male(P<0.01),1-19-year-old(P<0.01)and student(P<0.01)influenza cases were higher during pandemic than that of seasonal epidemic.3.The morbidity was mainly from September to November(12,807 cases,84.4%)during the 2009 pandemic,while it was mainly from January to March(8,691 cases,34.5%)and from November to December(6,907 cases,27.4%)during the seasonal epidemic from 2010 to 2018,and the peak month of each year was not completely the same.The monthly distribution was significantly different between the 2009 pandemic and seasonal epidemic from 2010 to 2018(P<0.01).4.The proportion of each subtype of influenza virus was significantly different every year in Chongqing from 2009 to 2018.The dominant influenza viruses were influenza A(H1N1)pdm09 in 2009,2013,2017 and 2018,influenza A(H3N2)in 2012,2014,2015 and 2016,and influenza B in 2010 and 2011.5.The influenza A(H1N1)pdm09 had two successive epidemic peaks every two years,the influenza A(H3N2)virus emerged every year,and the influenza B virus had a single epidemic peak during winter and spring season every two years.ConclusionEpidemiologic characteristics were significantly different between pandemic and seasonal epidemic.The morbidity rose rapidly and significantly,and men,children,young people and students influenza cases had higher proportions during pandemic.Seasonal influenza had more cases in winter and spring,but the epidemic peak occurred at different times every year.Influenza A virus was the dominant pathogen of influenza epidemic in Chongqing.Influenza A(H1N1)pdm09,seasonal influenza A(H3N2),influenza B(Victoria and Yamagata)were the main influenza strains.Different subtypes of influenza viruses had their own epidemiologic characteristics in Chongqing.Part Ⅱ Epidemiologic characteristics of human infection with avian influenza A virus in Chongqing,ChinaObjectivesTo explore epidemiologic characteristics of avian influenza virus and human infections with avian influenza virus in Chongqing,and provide an important basis for the risk assessment of influenza pandemic and the development of influenza activity prediction research strategy.MethodsEnvironmental samples were collected in live poultry markets(LPMs),commercial poultry farms(CPFs)and backyard poultry farms(BPFs)and tested for avian influenza virus by real-time RT-PCR.Human infections with avian influenza virus and the close contacts were investigated.Genetic sequencing and phylogenetic analysis were also conducted.Results:1.The avian influenza virus A(H5 & H9)was detected every year from 2013 to 2018 poultry-related places in Chongqing,with the sample positive proportions as 14.0%(920/6,576)and 17.8%(1,168/6,576)respectively.While avian influenza virus A(H7N9)was detected for the first time until in February 2017,and the sample positive proportion peaked in April,2017(9.67%,49/507),and 21 districts or counties(53.8%)had the virus detected from February to June in 2017.2.No human infection with influenza virus A(H5 & H9)was confirmed from 2013 to 2018 in Chongqing.Since the confirmation of the first patient infected with influenza A(H7N9)virus on March 5,2017,nine patients had been identified within four months in Chongqing.Their mean age was 45 years,77.8% were male,66.7% were urban residents and 55.6% were of poultry related occupation.All patients became infected after exposure to live poultry or the related places,and no close contacts got infected.The median time interval from initial detection of influenza A(H7N9)virus in Chongqing to the patients’ onset was 75 days.3.The proportion of positive samples was 2.94%(337/11,451)from February 2017 to May 2018,and was higher(P<0.01)in LPMs(3.66%,329/8,979)than that in CPFs(0.41%,5/1,229)and BPFs(0.24%,3/1,243).The proportion of positive samples(34.4%,22/64)in the premises to which the patients were exposed was significantly higher than that(5.7%,257/4,474)in premises with no patients.4.Phylogenetic analysis indicated that influenza virus A(H7N9)isolated in Chongqing belonged to the Yangtze River Delta lineage,originated from these virus named A/Shanghai/1/2013(H7N9)and(A/shanghai/05/2013(H7N9)and resembled those circulated in Jiangsu and Anhui provinces between late 2016 and early 2017.ConclusionAvian influenza virus A(H5 & H9)had existed for many years in Chongqing while no human infection with these viruses was confirmed.Avian influenza virus A(H7N9)was introduced into Chongqing most likely between late 2016 and early 2017 from Jiangsu and Anhui province with long-distance live poultry transportation,which swept across half of Chongqing territory,and was more likely to break through the species barrier and resulted in human infections than influenza virus A(H5 & H9).The risk factor for human infection with influenza virus A(H7N9)was the exposure to live poultry or the related places with higher level of virus pollution.H7N9 cases were highly scattered and no human-to-human transmission was found.Part Ⅲ Forecasting influenza activity using multi-source data and self-adaptive AI model in Chongqing,ChinaObjectivesTo explore the forecast of influenza activity using self-Adaptive AI model and multi-source data in Chongqing,China and to provide a new theoretical basis and implementation strategy for Chongqing to achieve more accurate influenza prevention and control.MethodsMulti-source electronic data from 2012 to 2018,including percentage of influenza-like illness(ILI%),weather data and influenza-related public sentiment on internet of Chongqing were collected.With the ILI% of the next week in Chongqing as the forecasting target,data from 2012 to 2016 were used as the training data set to construct an innovative Self-adaptive AI Model(SAAIM),which integrated Seasonal Autoregressive Integrated Moving Average model and XGBoost model using a self-adaptive weight adjustment mechanism.SAAIM was applied to ILI% forecast in Chongqing from 2017 to 2018,of which the performance was compared with three previously available forecasting models.Results1.ILI% showed an irregular seasonal trend from 2012 to 2018 in Chongqing.2.Compared with three reference models,SAAIM achieved the best performance on forecasting ILI% of Chongqing with the mean absolute percentage error(MAPE)of 11.9%,7.5% and 11.9% during the periods of the year 2014-2016,2017 and 2018 respectively.3.Among the three categories of source data,historical ILI% contributed the most to the forecast accuracy by decreasing the MAPE by 19.6%,43.1% and 11.1%,followed by weather information(MAPE reduced by 3.3%,17.1% and 2.2%),and influenza-related public sentiment on internet(MAPE reduced by 1.1%,0.9% and 1.3%).ConclusionFrom 2012 to 2018,the influenza activity in Chongqing presented an irregular seasonal high incidence.SAAIM was successfully constructed based on multi-source big data and artificial intelligence.Historical ILI% contributed the most to the prediction.Weather data were very important to accurately estimate the subtle changes of influenza epidemic.However,influenza-related public sentiment on internet including Baidu Index and Sina Weibo had the least impact on forecasting.Accurate ILI% forecast in Chongqing can be made by SAAIM,with better performance than other prediction models. |
PDF Full Text Request