Application of spatial zero-inflated Poisson model in spatial analysis of Brucellosis in Yunnan Province

doi:10.20199/j.issn.1672-2302.2025.03.006

Abstract

Abstract:

Objective To assess the value of spatial zero-inflated Poisson model applied to the spatial analysis of Brucellosis cases with high zero values in Yunnan Province for methodological references in analysis of the similar data in the field of public health. Methods The data on Brucellosis cases reported in Yunnan Province from January to December of 2022 were collected through the Chinese Disease Prevention and Control Information System. The global Moran’s I, local Moran’s I, and spatial scan statistics were used to analyze global and local spatial autocorrelation and detection of spatial clustering of Brucellosis cases. Traditional Poisson, zero-inflated Poisson, and spatial zero-inflated Poisson models were employed for data fitting, with the optimal model determined based on deviance information criterion (DIC) and Watanabe-Akaike information criterion (WAIC). Results In total, 1 015 cases of Brucellosis were reported in Yunnan Province from January to December of 2022. The number of cases at county-level showed a significant global spatial autocorrelation (Moran’s I=0.40, Z=8.80, P<0.01), and exhibited a serious outbreak in the eastern region of Yunnan, where a high-risk cluster was identified (RR=18.53, LLR=694.21, P<0.01). The spatial zero-inflated Poisson model demonstrated the best fit (DIC=556.055, WAIC=740.752), followed by the non-spatial zero-inflated Poisson model (DIC=815.527, WAIC=1 564.548), both outperforming the traditional Poisson model (DIC=975.799, WAIC=1 613.696). The latter two models showed biased results, leading to inaccurate parameter estimates. The posterior mean of the spatial structure random effects in the optimal model revealed significant spatial autocorrelation (Moran’s I=0.32, Z=5.92, P<0.01), whereas the non-spatial random effects indicated no significant spatial autocorrelation (Moran’s I=0.08, Z=1.52, P>0.05). Conclusion The spatial zero-inflated Poisson model, which incorporates spatial effects, can better address zero-inflated count data with spatial autocorrelation features, and reveal potential spatial structures in disease data, which may provide effective methodological support for the spatial epidemiological analysis of low-prevalence infectious diseases.

Key words: Zero-inflation, Spatial models, Spatial epidemiology, Brucellosis, Yunnan Province

CLC Number:

R181.3，R516.7

YUAN Rui, LI Ke, ZHANG Lele, YU Binbin, YANG Xiangdong, WANG peng, ZHANG Zhijie. Application of spatial zero-inflated Poisson model in spatial analysis of Brucellosis in Yunnan Province[J]. Journal of Tropical Diseases and Parasitology, 2025, 23(3): 160-164,188.

Figures/Tables 3

Figure 1

Figure 2

Table 1

References 30

[1]	Di Pierdomenico A, Borgia SM, Richardson D, et al. Brucellosis in a returned traveller[J]. CMAJ, 2011, 183(10):E690-E692.
[2]	World Health Organization. Brucellosis[EB/OL]. (2020-07-29)[2024-09-16]. https://www.who.int/news-room/fact-sheets/detail/brucellosis.
[3]	马云龙, 刘英, 李小龙, 等. 中国布鲁菌病的时空分布及影响因素[J]. 中华疾病控制杂志, 2023, 27(11):1241-1246,1295.
[4]	Lai SJ, Zhou H, Xiong WY, et al. Changing epidemiology of human brucellosis, China, 1955-2014[J]. Emerg Infect Dis, 2017, 23(2):184-194. doi: 10.3201/eid2302.151710 pmid: 28098531
[5]	国家卫生健康委员会统计信息中心. 2022中国卫生健康统计年鉴[EB/OL]. (2023-05-17)[2024-09-16]. https://www.nhc.gov.cn/mohwsbwstjxxzx/tjtjnj/202305/49beded3bd984669bfe9089c6f231cf5/files/1740022550235_90098.pdf.
[6]	张夏男, 张洪伟, 叶萌, 等. 我国南北方2004—2017年布鲁氏菌病发病趋势的Joinpoint回归分析[J]. 中国人兽共患病学报, 2020, 36(9):758-762. doi: 10.3969/j.issn.1002-2694.2020.00.116
[7]	张莉. 云南统计年鉴2023[M]. 北京:中国统计出版社: 135.
[8]	乾莲, 孟媛媛, 孙海龙, 等. 云南省布鲁氏菌病流行状况[J]. 中国人兽共患病学报, 2021, 37(2):138-145. doi: 10.3969/j.issn.1002-2694.2021.00.009
[9]	杨秋菊, 于彬彬, 张青, 等. 云南省2010—2019年人间布鲁氏菌病监测及流行趋势分析[J]. 中国人兽共患病学报, 2023, 39(3):258-262. doi: 10.3969/j.issn.1002-2694.2023.00.007
[10]	Ridout M, Demétrio CGB, Hinde J. Models for count data with many zeros[C]// Proceedings of the XIXth international biometric conference. Cape Town: International Biometric Society Invited Papers, 1998, 19(1):179-192.
[11]	Lambert D. Zero-inflated Poisson regression,with an application to defects in manufacturing[J]. Technometrics, 1992, 34(1):1-14.
[12]	陈秋兰. 我国布鲁氏菌病流行时空分析与患者健康相关生命质量研究[D]. 北京: 中国疾病预防控制中心, 2017.
[13]	国家卫生健康委员会. 布鲁氏菌病诊断:WS 269—2019[EB/OL]. (2019-01-02)[2024-09-16]. http://www.nhc.gov.cn/wjw/s9491/201905/b109b71e7a624256985b573944b5d292.shtml.
[14]	云南省统计局. 2022云南统计年鉴[EB/OL]. (2023-05-16)[2024-09-16]. https://stats.yn.gov.cn/Pages_22_2309.aspx.
[15]	Liang DY, Liu D, Yang M, et al. Spatiotemporal distribution of human brucellosis in Inner Mongolia, China, in 2010-2015, and influencing factors[J]. Sci Rep, 2021, 11(1):24213. doi: 10.1038/s41598-021-03723-9 pmid: 34930982
[16]	张辉国, 梁韵婷, 胡锡健. 基于贝叶斯LASSO矩阵指数空间模型辨识布鲁氏菌病发病的关键风险因素[J]. 中国人兽共患病学报, 2023, 39(12):1233-1238. doi: 10.3969/j.issn.1002-2694.2023.00.141
[17]	洪功飞, 刘华, 车跃光, 等. 安徽省牛羊布鲁氏菌病防控现状与思考[J]. 中国动物检疫, 2021, 38(11):64-68.
[18]	Nguyen LT, Stevenson MA, Firestone SM, et al. Spatiotemporal and risk analysis of H5 highly pathogenic avian influenza in Vietnam, 2014-2017[J]. Prev Vet Med, 2020,178:104678.
[19]	Anselin L. Local indicators of spatial association—LISA[J]. Geogr Anal, 1995, 27(2):93-115.
[20]	Kulldorff M. A spatial scan statistic[J]. Commun Stat Theory Meth, 1997, 26(6):1481-1496.
[21]	Besag J, York J, Mollié A. Bayesian image restoration,with two applications in spatial statistics[J]. Ann Inst Stat Math, 1991,43:1-20.
[22]	Aktekin T, Musal M. Analysis of income inequality measures on human immunodeficiency virus mortality: a spatiotemporal Bayesian perspective[J]. J R Stat Soc Ser A Stat Soc, 2015, 178(2):383-403.
[23]	Arab A. Spatial and spatio-temporal models for modeling epidemiological data with excess Zeros[J]. Int J Environ Res Public Health, 2015, 12(9):10536-10548.
[24]	Rue H, Held L. Gaussian Markov random fields: theory and applications[M]. New York: Chapman and Hall/CRC, 2005.
[25]	U.S. Centers for Disease Control and Prevention. Introduction to Public Health[EB/OL]. (2025-01-10)[2025-03-10]. https://www.cdc.gov/training-publichealth101/php/training/introduction-to-public-health.html.
[26]	Gómez-Rubio V, López-Quílez A. Statistical methods for the geographical analysis of rare diseases[J]. Adv Exp Med Biol, 2010,686:151-171.
[27]	Costa ACC, Codeço CT, Honório NA, et al. Surveillance of dengue vectors using spatio-temporal Bayesian modeling[J]. BMC Med Inform Decis Mak, 2015,15:93.
[28]	Kanankege KST, Errecaborde KM, Wiratsudakul A, et al. Identifying high-risk areas for dog-mediated rabies using Bayesian spatial regression[J]. One Health, 2022,15:100411.
[29]	Dhewantara PW, Zhang WY, Al Mamun A, et al. Spatial distribution of leptospirosis incidence in the Upper Yangtze and Pearl River Basin, China: tools to support intervention and elimination[J]. Sci Total Environ, 2020,725:138251.
[30]	Agarwal DK, Gelfand AE, Citron-Pousty S. Zero-inflated models with application to spatial count data[J]. Environ Ecol Stat, 2002,9:341-355.

变量	SZIP模型		ZIP模型		P模型
变量	系数（95%CI）	RR（95%CI）	系数（95%CI）	RR（95%CI）	系数（95%CI）	RR（95%CI）
α（截距）	-1.353 （-2.495~-0.748）	0.274 （0.083~0.471）	0.187 （0.053~0.321）	1.208 （1.055~1.377）	-0.391 （-0.513~-0.268）	0.678 （0.599~0.764）
β₁ （国内生产总值，百亿元）	-0.001 （-0.002~-0.000）	0.999 （0.998~1.000）	-0.249 （-0.270~-0.229）	0.751 （0.731~0.771）	-0.229 （-0.248~-0.209）	0.772 （0.752~0.791）
β₂ （羊存栏量，万头）	0.039 （0.011~0.072）	1.040 （1.011~1.047）	0.026 （0.022~0.031）	1.027 （1.022~1.031）	0.037 （0.033~0.041）	1.038 （1.033~1.042）
β₃ （2021年发病数）	0.016 （0.006~0.026）	1.016 （1.006~1.026）	0.023 （0.022~0.024）	1.023 （1.022~1.024）	0.025 （0.024~0.026）	1.026 （1.024~1.027）
零膨胀参数p_i（95%CI）	0.114（0.088~0.141）		0.357（0.275~0.443）
DIC	556.055		815.527		975.799
WAIC	740.752		1 564.548		1 613.696