Study On The Epidemiological Characteristics And Prediction Models For AIDS In China During 2005-2018

Objectives:Based on the national historical surveillance data of AIDS in China from 2005 to 2018,this study aimed（1）to describe the epidemiological characteristics of AIDS,（2）to analyze the spatial epidemiological characteristics of AIDS and explore high-risk areas,（3）to construct prediction models that estimate AIDS incidence in China.Our study will hopefully contribute to the improvement of AIDS prevention and control strategies in China.Methods:We extracted the incidence and mortality data of AIDS in China from 2005 to 2018,published by the National Public Health Science Data Center.Firstly,the distribution of AIDS by people,time,and place（namely known as three-dimension）was used to describe the epidemiological characteristics.Secondly,we used global spatial auto-correlation,local spatial auto-correlation,and the centroid transferring curve model to analyze the spatial epidemiological characteristics of AIDS in China,which aimed to explore the high incidence area of AIDS.Finally,the seasonal auto regressive integrated moving average（SARIMA）model and the nonlinear auto regressive（NAR）neural network model were built to predict the trend of the AIDS epidemic.We also used the mean absolute error（MAE）,mean square error（MSE）,and mean absolute percentage error（MAPE）to evaluate the optimal prediction model.Results:（1）From 2005 to 2018,a total of 437,205 AIDS cases were reported in China,and the incidence of AIDS increased from0.4324/100,000 in 2005 to 46,195/100,000 in 2018,with an average annual incidence of 2.3109/100,000.A total of 131,422 deaths were reported,and the mortality rate increased from 0.1012/100,000 in 2005 to1.3420/100,000 in 2018,with an annual mortality rate of 0.6946/100,000.The overall AIDS incidence and mortality rate were increasing year by year（P<0.001）.From 2005 to 2018,the incidence and mortality of AIDS showed an increasing trend（P<0.001）,and the annual incidence of AIDS was always higher than the annual mortality of AIDS in China.From2005 to 2018,the top five provinces with the total number of cases were Guangxi（66094 cases,15.12%）,Sichuan（65374 cases,14.95%）,Yunnan（51889 cases,11.87%）,Henan（34518 cases,7.81%）,and Guangdong（32968 cases,7.54%）.AIDS occurred in all age groups.The high-risk population were people aged 30-44 years（166,512 cases,38.09%）and45-59 years（113,378 cases,25.93%）.The proportion of elder cases（≥60years）increased significantly from 4.87%in 2005 to 22.92%in 2018.The male-to-female ratio of AIDS cases was 2.96:1.Farmers and migrant workers accounted for the largest population of AIDS cases（189,027cases,50.67%）.（2）The global spatial auto-correlation results showed the AIDS epidemic followed a spatial pattern of clustered distribution across the country from 2008 to 2018,and the degree of clustering has continued to increase.The local spatial auto-correlation results showed that the high-high clustering pattern and the low-low clustering pattern were more apparent.The results of the analysis of the centroid transferring curve model showed that the center of AIDS incidence in China was mainly located in the west of Hubei and the east of Sichuan.（3）Compared with SARIMA（1,0,4）（0,1,2）₁₂,nonlinear auto regressive neural network model had lower MAE（0.0425 VS 0.0473）,lower RMSE（0.0504 VS0.0579）,and lower MAPE（11.3370%VS 12.6310%）.The fitting and prediction effect of the NAR neural network model was better than that of SARIMA.Then two models were used to predict the monthly incidence of AIDS from 2019 to 2022,and the results showed that the incidence increases continuously.Conclusions:（1）From 2005 to 2018,the incidence of AIDS in China showed an increasing pattern.The top five provinces with the total number of AIDS cases were Guangxi,Sichuan,Yunnan,Henan,and Guangdong.The number of males was significantly higher than that of females.The high incidence age was mainly concentrated in the 30～44and 45～59 years old.Many AIDS cases were reported among farmers and migrant workers.（2）The AIDS epidemic presented a spatial pattern of clustered distribution nationwide.The incidence of AIDS in China could be potentially divided into two regions,the northern region（“cold spot”region）represented by Inner Mongolia and the southwest region（“hot spot”region）represented by Guangxi.The results of the centroid transferring curve model showed that the AIDS epidemic increased unevenly in various regions of China.（3）SARIMA（1,0,4）（0,1,2）₁₂model and NAR neural network model can be used in the prediction of monthly AIDS incidences in China.After the quantitative comparison of the two models,the NAR neural network model had better fitting and prediction effects than the SARIMA model.