Abstract: Background　China ranks third globally in tuberculosis burden and is classified as one of the highest tuberculosis（TB）burden countries.Xinjiang，a multi-ethnic region in northwestern China，has consistently reported one of the highest TB incidence rates nationwide，posing significant challenges to China's TB prevention and control efforts.Therefore，investigating region-specific epidemiological characteristics，quantifying the effects of covariates on pulmonary tuberculosis （PTB）risk，and deriving robust conclusions aligned with Xinjiang's sociodemographic and environmental contexts is critical to formulating targeted，evidence-based strategies for pulmonary tuberculosis control，thereby providing actionable scientific recommendations to mitigate the disease burden in this region. Objective　A hierarchical Bayesian spatiotemporal model integrated with the Integrated Nested Laplace Approximations-Stochastic Partial Differential Equations（INLA-SPDE）acceleration algorithm was employed to analyze the spatiotemporal distribution patterns and associated influencing factors of pulmonary tuberculosis in Xinjiang. Methods　Data on PTB reported incidence counts and population demographics across 14 prefectures in Xinjiang from 2010 to 2019 were retrieved from the China Disease Prevention and Control Information System （CDCIS）. An exploratory statistical analysis was conducted to characterize the spatiotemporal distribution of PTB incidence. Subsequently，a hierarchical Bayesian spatiotemporal model integrated with the INLA-SPDE acceleration algorithm was developed to assess the epidemiological trends and quantify the impact of associated risk factors on PTB transmission dynamics during the study period. Results　The study area encompassed 14 prefectures in Xinjiang，categorized into 10 estimation sites（Urumqi City，Changji Hui Autonomous Prefecture，Karamay City，Hami City，Ili Kazakh Autonomous Prefecture，Bayingolin Mongol Autonomous Prefecture，Kashgar Prefecture，Tacheng Prefecture，Aksu Prefecture and Altay Prefecture）and 4 validation sites（Turpan City，Bortala Mongol Autonomous Prefecture，Kizilsu Kirghiz Autonomous Prefecture and Hotan Prefecture）. From 2010 to 2019，the incidence rate and reported case count of PTB in Xinjiang peaked in 2018（304.945 per 100，000 population，76，846 cases），followed by a declining trend thereafter.The Bayesian posterior estimates indicated that population size，mean temperature，PM2.5 concentration，and longitude were positively associated with tuberculosis incidence，whereas per capita GDP，hospital bed availability，and latitude demonstrated protective effects against disease incidence. Spatial autocorrelation analysis revealed significant geographic clustering of PTB cases（σω 2 =1.806），with a spatial correlation range of 946.053 km，indicating that spatial dependence diminished with increasing distance between sites. Furthermore，short-term temporal persistence of PTB infection rates was strongly supported（α=17.926），suggesting sustained transmission dynamics within localized regions. Conclusion　Significant spatial autocorrelation of PTB incidence was observed in Xinjiang from 2010 to 2019，with a spatial correlation range of 946.053 km，indicating that spatial dependence diminished with increasing distance between geographic locations. Furthermore，population density，mean annual temperature，PM2.5 concentration，and longitude exhibited positive associations with PTB incidence rates. In contrast，per capita GDP，healthcare bed availability，and latitude were identified as protective factors，demonstrating inverse associations with PTB risk.

Key words: Pulmonary tuberculosis, Hierarchical Bayesian spatiotemporal model, INLA algorithm, Xinjiang, Spatiotemporal distribution

摘要： 背景　中国在全球结核病负担排名中位列第三，属于结核病高负担国家之一。新疆是一个多民族聚居地区，作为中国的西北大省，其结核病一直是中国最高地区之一，给中国的结核防控工作带来巨大挑战。故了解该地区区域性特点，评估协变量对肺结核发病风险的量化效应，得出符合新疆地区特性的可靠结论，从而为该地区制定针对性的肺结核防控策略提供科学依据与建议。目的　基于区域分层贝叶斯时空模型并结合集成嵌套拉普拉斯近似算法-随机偏微分方程（INLA-SPDE）加速算法，分析新疆肺结核时空分布特征及相关影响因素。方法　基于中国疾病预防控制信息系统，收集新疆14个地区2010—2019年肺结核报告发病数及人口数据，对肺结核发病情况进行探索性统计分析；建立分层贝叶斯时空模型结合INLA-SPDE加速算法评估2010—2019年新疆地区肺结核流行情况及相关影响因素的影响。结果　研究区域为新疆14个地区，其中估计站点10个：乌鲁木齐市、昌吉回族自治州、克拉玛依市、哈密市、伊犁哈萨克自治州、巴音郭楞蒙古自治州、喀什地区、塔城地区、阿克苏地区、阿勒泰地区；验证站点4个：吐鲁番市、博尔塔拉蒙古自治州、克孜勒苏柯尔克孜自治州、和田地区。2010—2019年新疆肺结核发病率和发病例数在2018年达到峰值（304.945/10万，76 846例），随后开始下降。后验估计值显示，人口数、平均温度、PM2.5以及经度对于肺结核发病率的影响呈正相关；人均GDP、医疗床位数、纬度是影响肺结核发病率的保护因素。肺结核疾病的地理分布数据显示，新疆肺结核疾病数据之间存在空间自相关性（σω2=1.806），两站点间空间相关距离为946.053 km，疾病之间的空间相关性随着距离的增加而减弱；肺结核感染率在短期内具有较强的持续性（α=17.926）。结论　2010—2019年新疆肺结核存在明显空间自相关性，两站点间空间相关距离为946.053 km，疾病之间的空间相关性随着距离的增加而减弱；人口数、平均温度、PM2.5以及经度对于肺结核发病率的影响呈正相关；人均GDP、医疗床位数、纬度是影响肺结核发病率的保护因素。

关键词: 肺结核, 分层贝叶斯时空模型, INLA 算法, 新疆, 时空分布