Epidemiccal Characterstics Analysis And Mutil-region Morbidity Forecasting Study On Tuberculosis In China, From 2004 To 2014

Objective: Using national monitoring tuberculosis disease incidence data from 2004-2014 to reflect and reveal the basic epidemiological characteristics (high risk of onset age groups, high-risk disease months and high-risk key areas) and incidence trend of the disease. Based on the time trajectory similarity clustering analysis research of regional incidence, different regional categories are isolated according to time trajectory charac-teristics of tuberculosis incidence, which can be used as the scientific basis to make plan for tuberculosis disease prevention. What's more, in order to predict regional tuberculo-sis incidence, a new type of regional cooperating prediction model (MR-GCLSSVM) of tuberculosis disease has been established. The prediction abilities of the new building multi-regional prediction model and two single-regional prediction models on tuberculosis incidences of 32 regions in whole country are compared,which is also successfully applied to forecast the incidences in 2015. The research result can be used as the quantitative basis for effective prevention of TB disease, and can also provide references for the sustainable development of the public health for our country.Methods: 2004-2014 regional TB disease data are downloaded from statutory in-fectious diseases report database in China Centers for Disease Control and Prevention(CDC). Statistical method, combined swarm intelligence optimization parameters and neural network method are used to process, analysis and model the TB epidemic data.The mainly used methods includes: descriptive epidemiological method, seasonal index method, self-organizing feature map (SOM), clustering method and MR-GCLSSVM mod-el (Multi-region least squares support vector machine (SVM) optimizing parameters by combining grey wolf algorithm and cross validation model).Conclusions:1. The overall trend: the incidence of tuberculosis in 2005 reaches the highest peak and the overall trend is decreased in the following year. The prevention and control states in whole nation are performed well.2. The age distribution: high risk and low risk groups respectively are the age grades of 70-74 and 0-4 years old. There is an obvious characteristics of age distribution that is bimodal distribution exists (the later peak is higher than the previous one).3. Month distribution: the period circle of the incidence of tuberculosis is one year.The epidemic period of TB is between January and June. The high-risk months are January, March and April, and the low-risk month is between September and December.In month distribution, it is obvious that incidence of the disease is continuing decline from January to December.4. Regional distribution: high-risk areas includes Guangxi, Hainan, Guizhou, Tibet and Xinjiang, all of which is in underdeveloped regions with low level medical treatment.The low-risk areas includes Beijing, Tianjin, Shanghai, Shandong and other regions, which are economic developed with high level medical development. There are some relationship existing between the TB incidence risk and economic development and the level of medical condition.5. Clustering analysis: Four different incidence time trajectory classes areas are obtained from the similarity clustering research in the whole country area. The result of this study shows: Guizhou and Xinjiang, with very strong similarity, are classified as Class 1,and the incidences of the two region's average trajectory are higher than the other 3 Classes. Class 4 contains 8 areas (Beijing, Tianjin, Hebei, Liaoning, Shanghai,Jiangsu, Shandong, Yunnan and Ningxia), whose incidences have high similarity, and the mean values of their incidence are generally low. According to the different characteristics in each Classes, we can carry out different strategies to prevent and control of the TB disease.6. Multi-regional incidence prediction: For predicting the multi-regional tuberculo-sis incidence, this research proposes a multi-regional cooperation MR-GCLSSVM model.High forecasting precision, low prediction error and convenient to establish prediction model are three advantages of MR-GCLSSVM. It provides an advanced model for before-hand prediction on the multi-regional disease incidence.