兰州大学机构库 >公共卫生学院
2004-2015年全国肺结核流行趋势时空特征及预测研究
Alternative TitleStudy on Spatial-temporal Characteristics and Prediction for Tuberculosis Trends in China during 2004-2015
毛强
Thesis Advisor刘兴荣
2018-03-10
Degree Grantor兰州大学
Place of Conferral兰州
Degree Name硕士
Keyword肺结核 流行 时空特征 空间自相关分析 空间扫描统计 预测
Abstract

目的:了解我国肺结核疫情现状和一般流行规律,分析肺结核流行时空特征分布,探测肺结核聚集区域、范围和时间,建立肺结核发病趋势预测模型,并将分析结果可视化展示。本研究可为肺结核疫情监测、防控以及编制防治规划提供参考,同时为具有空间属性和季节特征的其它传染病数据分析提供思路和方法学借鉴。

方法:通过国家疾病预防控制中心公共卫生科学数据平台收集2004-2015年我国31个省份(直辖市、自治区,不包括港澳台)的肺结核数据,从国家统计局网站《中国统计年鉴》收集人口数据,从国家基础地理信息系统收集省级行政区域电子地图等数据。研究方法如下:(1)采用一般描述性统计对肺结核流行现状和规律进行分析;采用H-P滤波法、季节指数分析和Cochran-Armitage趋势性检验方法分析肺结核疫情的周期性、季节性特征和发病趋势。(2)采用空间三维趋势面分析、分布方向(标准差椭圆)分析、Maoran`s I指数、LISA值分析进一步了解肺结核空间分布特征、空间格局演化以及全局自相关性和局部区域发病率的分布模式。(3)利用回顾性离散型泊松时空扫描统计量探测肺结核聚集区域、范围和时间,并对聚集区域的危险度进行定量评价。(4)应用单一的SARIMA模型对肺结核月发病率数据进行拟合,并构建SARIMA-GRNN组合预测模型,对模型的适用性和预测效果做出评价,以获取最优的肺结核发病趋势预测模型。

结果:(1)2004-2015年全国31个省份(直辖市、自治区)累计报告肺结核病例12321559例,年均报告发病率为77.31/10万人,年递降率为1.35%,肺结核发病率呈现出不断下降趋势(Z=-231.67,P<0.001)。月发病率呈双高峰分布且存在明显的季节性和周期性,1月的登记率最高,其次是3月,12月最低;1-7月为发病旺季,冬季为发病淡季。人群分布中,男性患者占比69.50%,中老年人群发病率最高,其次是20-24岁年龄组(占比10.21%),各年龄组男性患者数量均高于女性;农民患者人数最多(占比61.84%),农民、工人、家政和待业人员、学生是肺结核高危职业人群而且该人群的报告病例数逐年增加。病例诊断来源中,实验室确诊肺结核数量(占比55.41%)高于临床诊断,菌(-)和仅培阳病例报告比例呈现增长趋势,涂(+)和未痰检病例报告比例呈现下降趋势。省份分布中,年均发病率较低的地区主要集中在沿渤海、黄海的省份(直辖市)以及宁夏回族自治区和云南省,高发病率地区集中在西部和南部地区的部分省份。(2)我国肺结核年发病率均存在全局正向空间自相关性(全局Moran`I在0.2614~0.4607之间,P均小于0.05),为非随机分布,存在区域聚集现象。在东西方向肺结核发病率呈直线分布,东部地区发病率明显低于西部地区;在南北方向肺结核发病率分布呈微“U”形分布,中部地区发病率低于南、北地区。2015年肺结核发病重心轨迹与2004年相比向西南方向移动了157.19km,肺结核疫情在东南-西北方向上扩散且方向性趋势逐渐减弱。High-High分布模式省份增加,主要分布在西部地区(新疆维吾尔自治区、西藏自治区、青海省)以及湖南省和广西壮族自治区,Low-Low分布模式集中在京津冀三省、江苏省和浙江省。(3)时空扫描探测到5个时空聚类区,一级聚集区(LLR=121196.13,RR=1.56)以拉萨市为中心,半径为1601.35km,聚集时间为2005年1月-2010年6月,聚集范围包括西北地区的新疆维吾尔自治区、青海省、甘肃省和西南地区的西藏自治区、四川省、重庆市、云南省和广西壮族自治区,基本囊括了肺结核高发病率地区,部分地区与局部空间自相关分析的热点区域相吻合。(4)SARIMA-GRNN组合模型的MSE为0.1069,RMSE为0.3269,MAE为0.2032,MAPE为0.0313,R2为0.9495,结果提示组合模型预测的精度优于单纯的SARIMA模型。

结论:(1)肺结核年发病率呈现不断下降趋势,月发病率存在明显的季节性和周期性并呈双峰分布,1-7月为发病旺季;男性、中老年人群,农民、工人、家政和待业人员、学生为肺结核防控重点关注对象;肺结核诊断、登记工作质量逐步提升。西部地区和南部地区部分省份(自治区)为肺结核疫情重点防控区域。(2)肺结核发病率在空间上为非随机分布,存在区域聚集;发病重心向西南方向移动,在东南-西北方向上扩散且方向性趋势减弱。疫情“热点”区域主要分布在西部地区(新疆维吾尔自治区、西藏自治区、青海省)以及湖南省和广西壮族自治区,疫情“冷点”区域分布在京津冀、江苏省和浙江省。时空扫描统计量结果和空间自相关分析、季节性分析基本一致,一级时空聚类区位于西部高发病率地区,聚集时间在发病旺季。(3)SARIMA-GRNN组合预测模型拟合了肺结核时间序列数据的线性和非线性特征,具有较高的预测精度,可用于肺结核疫情预警。(4)季节性特征分析、地理信息系统空间分析和时空扫描统计量相结合分析我国肺结核发病季节性分布、时空特征、空间演化格局,探测聚集区域、范围和时间的效果显著。

Other Abstract

Objectives: The aims of this study were to explore the epidemic situations and the general epidemiological characteristics of tuberculosis in China, and to analyze the temporal-spatial characteristics of it's distribution, and to detect the aggregation area, scope and time. Then, a prediction model for forecasting the trends of tuberculosis incidence was established. Results from the analysis were visualized. This research can be developed for tuberculosis epidemic monitoring, prevention and controlling, so as to provide references for making plans. At the same time, the study could provide analytical methods and references for the infectious disease data which has space attributes and seasonal characteristics.

Methods: In this research, tuberculosis data were collected from the Public Health Scientific data platform of the National Centre for Disease Control and Prevention. The population data were obtained from the Statistical Yearbook of the National Bureau of Statistics. The provincial administrative electronic maps were collected from the National Geomatics Center of China (NGCC).

The research methods are as follows:(1)General descriptive statistics were used to analyze the epidemic situation and characteristics of tuberculosis. H-P filtering method and seasonal index were applied to analyze the periodicity and seasonal characteristics of tuberculosis, and Cochran-Armitage trend test method was used to analyze the incidence trend.(2)The global auto-correlation and local disease distribution pattern were analyzed by the spatial 3D trend surface, distribution direction (standard deviation ellipse), Maoran`s I and LISA value.(3)Using the retrospective discrete Poisson spatio-temporal scanning statistics to detect the aggregation of tuberculosis in space-time. Then, quantitatively evaluate the risk of the aggregation area. (4) A single SARIMA model was used to fit the monthly incidence, and a combined model of the SARIMA-GRNN was constructed to predict the incidence trend of tuberculosis. Finally, the best model was produced by evaluating the applicability and effects of the prediction model.

Results: (1) From 2004 to 2015, a total of 12,321,559 tuberculosis cases were reported in 31 provinces (municipalities and autonomous regions), with an annual reported incidence of 77.31/100,000. The annual decreasing rate of tuberculosis incidence was 1.35%, showing a declining trend (Z=-231.67, P < 0.001). There were obvious seasonal and periodic characteristic in monthly incidence with a double peak distribution. The highest registration rate appeared in January, followed by March, the lowest appeared in December. Peak season of morbidity was January to July, the off-season of morbidity arised in winter. Obviously, males account for 69.50%. The number of middle and old people were the highest, followed by the 20-24 age groups (10.21%). The numbers of males in all age groups were higher than females. The most patients were farmers (61.84%). Farmers, factory workers, students, housekeeping workers and unemployed people were higher risk in all kinds of occupations, and the reported case of these occupations increased by years. The number of laboratory confirmed tuberculosis cases (55.41%) were higher than clinical diagnosis. The reported incidence of +TB and unsputum cases showed a downward trend. The areas with relatively low annual incidence were mainly located along the Bohai Sea and Huanghai (municipalities directly under the Central Government), Ningxia hui autonomous region and Yunnan province.(2) The incidence rate of tuberculosis in China was non-random distribution in space (the overall Moran`I were 0.2614~0.607, P < 0.05). The distribution of tuberculosis in the east and west direction was linear, and the incidence of tuberculosis in the east is obviously lower than in the west. The distribution of tuberculosis in the north and south shown shaped nearly like the letter "U", with the incidence of the central region was higher than the north and the south. In 2015, the center of tuberculosis incidence moved towards the southwest direction 157.19 km compared with 2004. The incidence of tuberculosis spread from southeast to northwest and the direction of incidence showed a weaker trend. The provinces of High-High distribution pattern had progressively increased, mainly distributed in the western regions (Xinjiang Uygur autonomous region, Tibet autonomous region and Qinghai province), Hunan province and Guangxi zhuang autonomous region. The Low-Low distribution patterns concentrated in Jingjin-hebei district (Beijing city, Tianjing city and Heibei province), Jiangsu province and Zhejiang province. (3) Five spatial-temporal clustering areas were detected by space-time scanning. The first concentration area was located in Lhasa city with a radius of 1601.35 km (LLR, RR were 121196.13 and 1.56, respectively). The aggregation range included the Tibet autonomous region, Xinjiang Uygur autonomous region, Guangxi zhuang autonomous region, Qinghai province, Gansu province, Sichuan province, Chongqing city and Yunnan province. Some areas properly coincided with hot spots in local spatial auto-correlation analysis.(4) The MSE, RMSE, MAE, MAPE, R2 of the SARIMA-GRNN model were 0.1069, 0.3269, 0.2032, 0.0313 and 0.9495, respectively. The results of indicates that the prediction accuracy of the combined model were better than the SARIMA model.

Conclusions:(1)The annual incidence rate showed a downward trend, indicating that the effectiveness of prevention and control of tuberculosis in China was significant. The monthly incidence had an evidently seasonal and cyclical with a bimodal distribution, Simultaneously the onset season from January to July. Male, middle aged people, farmers, workers, domestic and unemployed workers and students were the main targets of prevention and control of tuberculosis. The quality of diagnosis and registration of tuberculosis had gradually improved. Some provinces (autonomous regions) in the western region and the southern region were key prevention and control regions for tuberculosis epidemics.(2)The incidence was non random distribution in space, and there was an aggregation region. The center of the gravity moves to south-west. Incidence diffused in the direction of southeast to northwest and the direction trend was weakens. The “hot spot” epidemic regions were mainly distributed in the western regions (Xinjiang Uygur autonomous region, Tibet autonomous region, Qinghai province) as well as Hunan province and Guangxi zhuang autonomous region. The “cold spots” epidemic area were located in Beijing, Tianjin, Hebei, and Zhejiang provinces. The statistical results of spatio-temporal scans were consistent with the spatial auto-correlation analysis and seasonal analysis. The first-order spatio-temporal cluster area was located in the high-incidence area in the west, and the aggregation time was the peak season of onset.(3)The SARIMA-GRNN model fitted the linear and non-linear characteristics of tuberculosis time series data and possessed a higher prediction accuracy, which could be used for the early warning of the epidemic situation.(4)The conclusions of this study are notable by the combination of seasonal characteristics analysis, geographic information system spatial analysis, and space-time scan statistics analysis to detect the seasonal variation, spatial-temporal characteristics distribution, space time evolution and spatiotemporal aggregation. 

URL查看原文
Language中文
Document Type学位论文
Identifierhttp://ir.lzu.edu.cn/handle/262010/211769
Collection公共卫生学院
Recommended Citation
GB/T 7714
毛强. 2004-2015年全国肺结核流行趋势时空特征及预测研究[D]. 兰州. 兰州大学,2018.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Altmetrics Score
Google Scholar
Similar articles in Google Scholar
[毛强]'s Articles
Baidu academic
Similar articles in Baidu academic
[毛强]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[毛强]'s Articles
Terms of Use
No data!
Social Bookmark/Share
No comment.
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.