2004-2014年全国肺结核流行特征分析与多地区发病预测研究

发布时间：2018-06-13 23:48

本文选题：肺结核 + 流行特征　；参考：《兰州大学》2017年硕士论文

【摘要】：目的:分析2004-2014年全国肺结核疾病监测发病数据,反映和揭示全国肺结核病的基本流行特征(高危发病年龄人群、高危发病月份和重点发病地区)和发病趋势。通过对各地区发病率时间轨迹进行相似性聚类分析研究,分离出多个具有不同肺结核发病率时间轨迹特点的地区类别,为全国分类预防和合理规划肺结核疾病医疗卫生资源提供依据。除此之外,为了提前预知我国各地区的肺结核发病情况,一个新型的多地区相互协作的肺结核发病预测模型(MR-GCLSSVM)被提出,对比了新构建的多地区肺结核发病预测模型和两个单一地区发病模型在全国32个地区肺结核发病率数据集上的预测能力,并成功地预测了2015年32个地区的肺结核发病率。研究的结果能为我国各地区肺结核防治提供定量依据,也可为全国公共卫生事业的可持续发展制定切实有效的预防和治理策略提供参考。方法:基于中国疾病预防控制中心(CDC)法定报告传染病数据库的肺结核疫情数据,使用统计、群智能优化参数与神经网络结合的方法对肺结核疫情数据进行处理、分析和建模。本研究主要使用到的方法包括:描述性流行病学法、季节指数法、自组织特征映射聚类方法(SOM)和MR-GCLSSVM模型(多地区的灰狼算法和交叉验证结合优化参数的最小二乘支持向量机模型)。结论:1.总体趋势:全国肺结核病发病率在2005年达到最高峰值后,有明显下降的总体趋势。全国总体发病情况和防控状态均表现良好。2.年龄分布:高危和低危发病人群分别为70-74岁和0-4岁,有明显的年龄特征分布且为先低峰后高峰的双峰分布特点。3.月份分布:肺结核发病率以一年为周期,1-6月是肺结核的流行月份,高危月份为1月、3月和4月,低危月份为9-12月。有明显的月份分布且为自1月起至12月发病率持续下降的分布特点。4.地区分布:高危地区包含广西、海南、贵州、西藏和新疆等经济不发达和医疗水平相对较低的地区。发病低危地区为北京、天津、上海和山东等经济发达和医疗卫生水平较高的地区。肺结核病发病率的高低危地区分布和地区的经济发展和医疗卫生水平可能有一定的关系。5.聚类分析:全国各地区发病率时间轨迹的相似性聚类研究中得出了4个具有不同发病率时间轨迹的地区类。聚类结果表明:贵州和新疆地区被聚类为第1类,这两个地区的发病率轨迹平均值普遍高于其他3类,具有很强的相似性。第4类包含的地区(北京、天津、河北、辽宁、上海、江苏、山东、云南和宁夏)发病率时间轨迹也具有较高的相似性,且有发病率轨迹平均值普遍较低的特点。可以根据不同的地区类包含的特点采取分类策略防控肺结核。6.多地区发病预测:在多地区肺结核发病率预测上,本文提出了一个预测精准度高、预测误差小和建模方便的多地区协同的MR-GCLSSVM模型,为多地区疾病的向前预测提供了一个较先进的模型。
[Abstract]:Objective: to analyze the data of pulmonary tuberculosis surveillance in China from 2004 to 2014, and to reveal the basic epidemic characteristics of pulmonary tuberculosis (age group, month and region) and the trend of the disease. Based on the similarity analysis of incidence time locus in different regions, several regional types with different time trajectories of pulmonary tuberculosis incidence were isolated. To provide the basis for national classification prevention and rational planning of tuberculosis disease medical and health resources. In addition, in order to predict the incidence of pulmonary tuberculosis in various regions of China in advance, a new multi-region cooperative model for predicting the incidence of pulmonary tuberculosis (MR-GCLSSVM) was proposed. In this paper, the predictive ability of the newly constructed multi-region pulmonary tuberculosis incidence prediction model and two single area incidence models on the data sets of 32 regions in the whole country were compared, and the incidence of pulmonary tuberculosis in 32 regions in 2015 was successfully predicted. The results of the study can provide quantitative basis for the prevention and control of pulmonary tuberculosis in various regions of China, and can also provide a reference for the sustainable development of public health in China to formulate effective prevention and treatment strategies. Methods: based on the data of tuberculosis epidemic in the database of infectious diseases reported by the China Center for Disease Control and Prevention, the data of tuberculosis epidemic situation were processed, analyzed and modeled by the methods of statistics, optimization parameters of swarm intelligence and neural network. The main methods used in this study include descriptive epidemiology, seasonal index, Self-organizing feature mapping clustering method (SOM) and MR-GCLSSVM model (multi-region gray wolf algorithm and cross-validation combined with optimized parameters of the least squares support vector machine model). Conclusion 1. General trend: the incidence of pulmonary tuberculosis in the country reached its highest peak in 2005, there is a significant decline in the overall trend. The overall incidence and prevention and control status of the country are good. 2. 2. Age distribution: high risk population and low risk population were 70-74 years old and 0-4 years old respectively. Monthly distribution: the incidence of pulmonary tuberculosis is one year cycle. January, March and April are the high risk months, and the low risk months are September to December in the months of January, March and April, the high risk month is January, March and April, and the low risk month is September December. There is a significant monthly distribution and the distribution of incidence from January to December continued to decline. 4. Distribution: high-risk areas include Guangxi, Hainan, Guizhou, Tibet, Xinjiang and other economically underdeveloped and relatively low level of medical care. The low risk areas are Beijing, Tianjin, Shanghai and Shandong. There may be a certain relationship between the regional distribution of tuberculosis incidence and regional economic development and the level of medical and health care. Cluster analysis: in the study of similarity of incidence time locus in different regions of China, four regional groups with different incidence time trajectories were obtained. The clustering results show that Guizhou and Xinjiang regions are clustered into the first category, and the average incidence rate of the two regions is generally higher than that of the other three groups, which has strong similarity. The time trajectories of incidence in the areas included in the fourth category (Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Shandong, Yunnan and Ningxia) are also similar, and the average values of incidence trajectories are generally lower than those in the regions of Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Shandong, Yunnan and Ningxia. According to the characteristics of different regional classes, classification strategies can be adopted to prevent and control tuberculosis. 6. Multi-region incidence prediction: a MR-GCLSSVM model with high prediction accuracy, small prediction error and convenient modeling was proposed in this paper. It provides a more advanced model for the forward prediction of disease in many areas.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R521;R181.3

【参考文献】