Abstract
Abstract
Objective
Pulmonary tuberculosis (PTB) is a critical challenge worldwide, particularly in China. This study aimed to explore the spatiotemporal transmission patterns and socioeconomic factors of PTB in Dongguan city, China.
Methods/design
An ecological study based on the reported new PTB cases between 2011 and 2020 was conducted in Dongguan city, China. The spatiotemporal analysis methods were used to explore the long-term trend, spatiotemporal transmission pattern and socioeconomic factors of PTB.
Main outcome measures
The number of new PTB cases.
Participants
We collected 35 756 new PTB cases, including 23 572 males and 12 184 females.
Results
The seasonal–trend decomposition indicated a significant downward trend for PTB with a significant peak in 2017 and 2018, and local spatial autocorrelation showed more and more high–high clusters in the central and north-central towns with high incidence. The multivariate spatial time series analysis revealed that the endemic component had a leading role in driving PTB transmission, with a high total effect value being 189.40 (95% CI: 171.65–207.15). A Bayesian spatiotemporal model revealed that PTB incidence is positively associated with the agricultural population ratio (relative risk (RR) =1.074), gender ratio (RR=1.104) and the number of beds in medical institutions (RR=1.028).
Conclusions
These findings revealed potential spatiotemporal variability and spatial aggregation of PTB, so targeted preventive strategies should be made in different towns based on spatiotemporal transmission patterns and risk factors.
Keywords: tuberculosis, epidemiology, bacteriology
STRENGTHS AND LIMITATIONS OF THIS STUDY.
This study contributed to the literature by exploring the long-term trend and spatiotemporal transmission patterns, as well as assessing the socioeconomic risk factors of pulmonary tuberculosis (PTB).
PTB cases were from the China Information System for Disease Control and Prevention, which may underestimate the number of cases in the missing report.
Although we did not include vaccination status in the multivariate model, its effect would be minimal since almost the entire population in China has received the BCG vaccine.
Introduction
Tuberculosis (TB) is a chronic infectious disease caused by Mycobacterium tuberculosis, which typically invades the lungs to lead to pulmonary tuberculosis (PTB). According to the 2021 Global Tuberculosis Report, the global number of people newly diagnosed with TB and reported fell from 7.1 million in 2019 to 5.8 million in 2020, but the number of TB deaths increased in 2020 in most of the high TB burden countries that resulted from reduced access to TB diagnosis and treatment caused by the COVID-19 pandemic.1 As a result, TB has been the leading cause of death from a single infectious agent, ranking above the HIV.1 Globally, an estimated 10.0 million people developed TB in 2020, and eight countries accounted for two-thirds of the global total, which was mainly concentrated in Asia and Africa.1 Notably, China has the third highest TB burden and the second highest multidrug-resistant TB burden worldwide, which poses a major threat to disease control in China.2 Thus, TB is a critical public health challenge worldwide, particularly in China.
Fortunately, PTB is curable and preventable. A better understanding of the spatiotemporal transmission modes and influencing factors of PTB has important guiding significance for formulating effective prevention and control strategies. Previous studies have shown that PTB, known as an airborne infectious disease, has a significant spatiotemporal heterogeneous distribution across regions and over time, but the long-term temporal trends and spatial clustering patterns were various in different provinces or countries.3 4 Because of the high multidrug resistance and low detection, this poses great challenges for PTB prevention and control, so it is very urgent and necessary to pay in-depth attention to a variety of factors contributing to PTB. There have been increasing studies on exploring individual factors affecting PTB susceptibility, such as age, sex, alcohol consumption and diabetes.4 5 However, studies on socioeconomic factors (eg, proportion of migrants, sex ratio and overcrowding) for PTB susceptibility are lacking, which are important for formulating specific public health policies in various regions. In addition, for traditional regression models such as logistic regression used in previous studies, it is difficult to incorporate the effects of socioeconomic factors with the spatiotemporal heterogeneity. Interestingly, the Bayesian spatiotemporal model was superior to traditional methods for exploring the socioeconomic risk factors and spatiotemporal characteristics at the same time, which may provide new guidance for the spatiotemporal analysis of PTB.
As an industrial city with a large mobile population, Dongguan has a high incidence of TB, which is a big challenge to public health. Meanwhile, little is known about the spatialtemporal distribution of PTB incidence in Dongguan and the impact of socioeconomic factors on the incidence of PTB. This study aimed to explore the long-term trends and spatiotemporal transmission patterns of PTB and to assess socioeconomic determinants influencing PTB incidence using town-level PTB registered data from 2011 to 2020 in Dongguan City, South China, which is important for understanding its transmission patterns and providing useful information for its prevention and control strategy.
Methods
Study areas and design
This was an ecological study of the reported new PTB cases in Dongguan city between January 2011 and December 2020. Dongguan is located in South China, between longitudes 113°31′ and 114°15′ East and latitudes 22°39′ and 23°09′ North. It has been an important transportation hub and foreign trade port in Guangdong province, China. There were 32 towns in Dongguan city, with 8.46 million inhabitants in 2020 and a total area of 2543 km2.
Data sources
The town-level registration data on new PTB cases were obtained from the China Information System for Disease Control and Prevention. All the cases were geocoded and matched to the town-level map data from the Resource and Environment Science Data Platform (https://www.resdc.cn/data.aspx?DATAID=205). The outcome variable was the number of new PTB cases. The main epidemiological information included sex, age, address and clinical records. The town-level predictor variables were socioeconomic, demographic and healthcare factors, which were obtained from the Dongguan Bureau of Statistics (https://tjj.dg.gov.cn/tjnj/index.html). These predictor variables included the regional gross domestic product (GDP), population density, mobile population, agricultural population ratio (APR), sex ratio, birth rate, number of beds in medical institutions (NBMI) per 10 000 persons and healthcare personnel (HCP) per 10 000 persons (online supplemental file 1).
Temporal trends
Monthly incidence rates were calculated during the study period. The seasonal–trend decomposition using loess (STL) was used to explore the seasonality and trends associated with PTB, which is a filtering procedure for decomposing time series data into different variation components. The formula for the STL model is Yt=St+ Tt + Rt, in which the time series is decomposed into three temporal components including seasonal (St), trend (Tt) and remainder (Rt).6 The function stl and parameter setting periodic were used to decompose the time series data and extract all the above parameters using R software (V.4.0.4).
Spatial autocorrelation analysis
Spatial autocorrelation analysis is a spatial method used to detect spatial autocorrelation based on the locations of study regions. Global Moran’s I index based on the queens’ weights of the neighbourhood rule was used to explore whether there was a significant global spatial autocorrelation in Dongguan.7 Anselin local Moran’s I statistics were used to detect the patterns of spatial autocorrelation (high–high, low–low, high–low and low–high) among local towns and identify the location of spatial clusters. QGIS software (V.3.20.3) was used for spatial description, and GeoDa software (V.1.12) was used for spatial autocorrelation analysis.
Spatiotemporal analysis of multivariate spatial time series model
The multivariate spatial time series model was used to explore the spatiotemporal components of surveillance data, which takes the Poisson branching process with the immigration method into account and incorporates seasonal effects, long-term trends and overdispersion.8 The model decomposes the incidence of infectious diseases within regions into three components: spatial epidemic component (reflecting inter-regional transmission effects), temporal autocorrelation (reflecting the influence of past incidence on subsequent outbreaks) and local characteristic component (reflecting long-term trends, seasonal effects, and the local risk of sporadic cases from unknown sources), thereby facilitating a more comprehensive analysis of the spatiotemporal transmission characteristics across different regions. The power-law algorithm was used to construct the spatial weight matrix after weighting the adjacency matrix, which obtains deterministic results more easily.9 The multivariate time series model was carried out by the ‘hhh4’ model provided in R package surveillance.8
Bayesian spatiotemporal analysis
Considering the spatial clustering, we used a Bayesian spatiotemporal model to explore potential risk factors for PTB susceptibility. The PTB cases followed a Poisson distribution, and four Bayesian spatiotemporal models were created. Model I contained the main spatial effect, model II contained the main spatial effect, time trend effect and autoregressive time effect, model III contained the main spatial and temporal effects and model IV contained the main spatial, temporal effects and spatiotemporal interactions. In the last model, the model was constructed as follows: log(θit)=α+μi+vi+γt+φt+δit, θit represents relative risk (RR) at the t time point in the i region; α is the intercept, μi is the spatially structured residual and vi is the spatially unstructured residual; γt represents the temporally structured effect, φt represents the temporally unstructured effect and δit represents the spatiotemporal interaction. In this study, the spatial relationships between different regions were based on the Queen’s weight matrix, and the prior distribution for spatial non-structural effects follows a normal distribution, while the prior distribution for spatial structural effects adopts the prior structure of the conditional autoregressive model.10 The hyperprior distribution uses default settings from R-INLA package (R V.4.0.4), and the temporal structural effects employ a RW1-type random walk, with a prior distribution of gamma (1, 0.0005).11 The best-fitting model was selected based on the lowest value of the deviance information criterion (DIC), and socioeconomic factors were not included in the model selection process.
Patient and public involvement
Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this study.
Results
Demographic characteristics
From 2011 to 2020, we collected 35 756 PTB cases in Dongguan, including 23 572 males and 12 184 females, with a sex ratio of 1.93:1. The largest number of cases were reported in the age group of 31–60 years (50.65%), followed by patients aged 0–30 years (44.53%). The annual incidence rate of PTB ranged from 38.10 per 100 000 in 2019 to 54.22 per 100 000 in 2011, with an average rate of 43.22 per 100 000. From 2011 to 2020, we collected 10 179 smear-positive PTB cases, including 6679 males and 3500 females, with a sex ratio of 1.91:1. The annual incidence rate of smear-positive PTB ranged from 8.26 per 100 000 in 2016 to 17.84 per 100 000 in 2020, with an average rate of 12.26 per 100 000 (online supplemental table S1).
Time distribution characteristics
Time series decompositions of raw PTB data showed a clear seasonal pattern, with a peak occurring between March and November each year (figure 1). The interannual pattern showed an overall decreasing trend, but there was a significant peak in 2017 and 2018.
Figure 1. Seasonal–trend decomposition of PTB incidence in Dongguan during 2011–2020 (note: ‘a’ represents time of first COVID-19 case notification).
Regional distribution characteristics
The annual incidence was mapped at the town level in Dongguan from 2011 to 2020 (figure 2). The high-incidence areas were the central and north-central towns such as Shilong, Shipai and Dalang. When comparing the incidence among different years, the high-incidence areas were the central and north-west towns between 2011 and 2014, gradually moved to the central towns between 2015 and 2016, and finally transferred to north-central towns between 2017 and 2020 (online supplemental figure S1). We used cluster analysis to describe the potential heterogeneous distribution of socioeconomic factors and found that Chang'an and Humen towns have high regional GDP (online supplemental figure S2), Shilong and Guancheng towns have high population densities (online supplemental figure S3) and Chang'an and Humen towns are characterised by high numbers of mobile population (online supplemental figure S4). The availability of HCP is higher in Dongcheng, Shilong and Wanjiang, with an average of 169.93 per 10 000 people, compared with Gaobu, Shatian and Shijie with an average of 46.14 per 10 000 people (online supplemental figure S5).
Figure 2. Regional distribution of pulmonary tuberculosis incidence in Dongguan during 2011–2020.
Spatial autocorrelation analysis
Except for 2011 (Moran’s I=0.147, p=0.059), the global spatial autocorrelation of PTB showed significant global correlation (range from 0.186 to 0.504, all p<0.05; table 1), indicating that PTB has a positive spatial correlation on the town scale. The Local Indicators of Spatial Association (LISA) maps of local spatial autocorrelation analysis were used to detect the clustering patterns of PTB (online supplemental figure S6). This shows that high-incidence clusters (high–high pattern with dark red colour) were mostly from central and north-central towns including Shilong, Chang'an and Liaobu.
Table 1. Global spatial autocorrelation analysis of pulmonary tuberculosis incidence rate from 2011 to 2020.
| Years | Moran’s I | Z | P value | Mean | SD | Distribution pattern |
| 2011 | 0.147 | 1.625 | 0.059 | −0.032 | 0.110 | Random |
| 2012 | 0.186 | 1.964 | 0.036* | −0.032 | 0.111 | Cluster |
| 2013 | 0.255 | 2.620 | 0.010* | −0.032 | 0.109 | Cluster |
| 2014 | 0.448 | 4.543 | <0.001* | −0.032 | 0.106 | Cluster |
| 2015 | 0.217 | 2.251 | 0.019* | −0.032 | 0.110 | Cluster |
| 2016 | 0.245 | 2.514 | 0.012* | −0.032 | 0.110 | Cluster |
| 2017 | 0.260 | 2.647 | 0.006* | −0.032 | 0.110 | Cluster |
| 2018 | 0.504 | 4.929 | <0.001* | −0.031 | 0.109 | Cluster |
| 2019 | 0.332 | 3.375 | 0.002* | −0.031 | 0.108 | Cluster |
| 2020 | 0.284 | 2.954 | 0.003* | −0.029 | 0.106 | Cluster |
The P value is statistically significant.
Spatiotemporal analysis of multivariate spatial time series model
A multivariate spatial time series Power-law model was constructed based on monthly data from 2011 to 2020, in which the cases were divided into autoregressive, spatialtemporal and endemic components. The total effect values were 3.67 (95% CI: 1.97–5.34) for the spatialtemporal component, 0.22 (95% CI: 0.18–0.25) for the autocorrelation component and 189.40 (95% CI: 171.65–207.15) for the endemic component. These results indicated that there was much variation in the endemic component, but little variation observed in the spatialtemporal and autocorrelation components. When comparing the fitted components among different towns, some special transmission characteristics were observed: (1) PTB cases in all towns have big endemic components, indicating the high infection risk from the local transmission, (2) PTB cases in eastern towns (such as Fenggang, Qingxi and Zhangmutou) have significantly big autocorrelation components, suggesting that these towns were predominantly affected by the previous infection in their regions and (3) PTB cases in central towns (such as Dalianshan, Liaobu, Dongkeng, Changping, Huangjiang, Gaobu and Qishi) have large spatialtemporal components, indicating the high transmission risk from the adjacent towns (online supplemental figure S7).
Influencing factors for PTB
According to the lowest value of the DICs in table 2, we selected the model IV (spatiotemporal interactions model) as the best-fitted model to explore potential influencing factors for PTB. The socioeconomic variables were added to the model IV dependently to perform the univariate analysis (table 3). There were significant associations in terms of regional GDP, APR, mobile population, sex ratio, birth rate, NBMI and HCP. Based on the magnitude of the correlation coefficients between variables (online supplemental table S2) and the univariate results, we screened the best-fitted multivariate Bayesian model (table 3). There were significantly positive associations in terms of APR (RR=1.074; 95% CI: 1.058–1.090), sex ratio (RR=1.104; 95% CI: 1.090–1.118) and NBMI (RR=1.028; 95% CI: 1.017–1.040).
Table 2. Comparison of Bayesian spatiotemporal models.
| Models | RR without covariates | DIC |
| Model I: spatial effect | log (θit)=α+μi+vi | 3840.29 |
| Model II: spatial+time trend+autoregressive time effects | log(θit)=α+μi+vi+(β+δi)×t | 3842.22 |
| Model III: spatial+temporal effects | log(θit)=α+μi+vi+ γt+ φt | 3844.58 |
| Model IV: spatial+temporal+spatiotemporal interaction | log(θit)=α+μi+vi+ γt+ φt+ δit | 3818.18 |
: ,: ,Where, θit relative riskRR at the t time point in the i region; α , intercept; μi , spatially structured residual; vi , spatially unstructured residual; β, linear time trend; δi, autoregressive time effect; γt, temporally structured effect; φt , temporally unstructured effect; δit , spatiotemporal interaction.
DICdeviance information criterionRRrelative risk
Table 3. Effect estimations of influencing factors in univariate and multivariate Bayesian models of PTB.
| Variables | Coefficient (95% CI) | RR (95% CI) | DIC |
| Univariate models for PTB | |||
| Regional GDP* | −0.026 (−0.037, 0.016) | 0.974 (0.964, 0.985) | 3803.445 |
| APR* | 0.051 (0.037, 0.065) | 1.052 (1.037, 1.067) | 3768.316 |
| Population density | 0.006 (−0.006, 0.018) | 1.006 (0.994, 1.018) | 3819.277 |
| Mobile population* | −0.029 (−0.039, 0.018) | 0.972 (0.962, 0.982) | 3799.089 |
| Sex ratio* | 0.095 (0.082, 0.108) | 1.100 (1.085, 1.114) | 3591.437 |
| Birth rate* | −0.023 (−0.039, -0.009) | 0.978 (0.962, 0.991) | 3806.746 |
| NBMI* | 0.023 (0.012, 0.034) | 1.023 (1.012, 1.035) | 3803.291 |
| HCP* | 0.016 (0.006, 0.027) | 1.017 (1.006, 1.027) | 3810.578 |
| Multivariate model for PTB | |||
| Regional GDP | −0.006 (−0.018, 0.005) | 0.994 (0.982, 1.005) | 3487.426 |
| APR* | 0.071 (0.056, 0.086) | 1.074 (1.058, 1.090) | |
| Sex ratio* | 0.099 (0.086, 0.112) | 1.104 (1.090, 1.118) | |
| Birth rate | −0.014 (−0.033, 0.005) | 0.986 (0.968, 1.005) | |
| NBMI* | 0.028 (0.016, 0.039) | 1.028 (1.017, 1.040) |
Represents statistically significant (Pp<0.05).
APR, agricultural population ratio; DIC, deviance information criterionGDP, gross domestic product; HCP, healthcare personnel (per 10 000 persons); NBMI, number of beds in medical institutions (per 10 000 persons); PTBpulmonary tuberculosisRR, relative risk
Discussion
In this study, the average incidence of PTB in Dongguan from 2011 to 2020 was 43.22 per 100 000, with a downward trend, which may be related to the implementation of Directed Observed Treatment Short-Course (DOTS) and the national TB control programme. A well-established government-led public health system had a positive impact on the control of TB. Interestingly, the STL models in this study predicted that there was a significant decreasing trend in PTB incidence after February 2020, which may be due to the fact that improved control measures (such as the increased use of face masks and physical distancing) during the COVID-19 outbreak may cut-off the transmission route of TB.12 However, the number of TB deaths increased in 2020 in most high-burden countries resulting from reduced access to TB diagnosis and treatment caused by the COVID-19 pandemic.1 Since the long-term effect of the COVID-19 outbreak on the development of TB has not yet been explored, the above phenomenon may provide an important direction for future predictive analyses. Seasonal analyses based on the STL method showed clear seasonality in PTB, with a low incidence in winter and peaks in other seasons. Winter’s rise in respiratory infections can obscure or postpone PTB diagnosis,13 while the dislike of seeking medical care in cold conditions impedes its prompt detection. Furthermore, the Chinese cultural practice of avoiding hospital visits during the Spring Festival further compounds these issues, aligning with the observed spike in TB cases post festival.14
Accurate spatial monitoring data on PTB is essential for effective prevention and intervention strategies. Importantly, we found that the spatial differences in PTB incidence were significant. For example, the high-incidence areas were the central and north-west towns between 2011 and 2014, gradually shifted to the central towns between 2015 and 2016, and finally moved to north-central towns between 2017 and 2020. Moreover, we observed that PTB has a long history of high–high aggregation in central and north-central towns such as Shillong and Chang'an areas. This result is consistent with previous studies,15 16 indicating that TB high–high clusters mainly occur in areas with low economic levels and poor health service capacity. According to data from the Dongguan Bureau of Statistics, the large agricultural population may be an important factor for the high–high clusters of PTB in these areas.16
Multivariate spatial time series analysis is an important method to analyse the spatiotemporal patterns of infectious disease epidemics, providing useful information for future prevention and intervention. In this study, the multivariate spatial time series model found that the endemic component played a leading role in the incidence of PTB, suggesting that the occurrence of PTB in these areas is associated with local risk factors such as socioeconomic and demographic characteristics.15 16 However, previous research revealed that the autoregressive component predominated in all the provinces, and the spatiotemporal component was mainly located in the well-developed provinces.17 The above findings suggest that there is a big gap in the spatiotemporal patterns between provincial-level and town-level studies, and our study may provide more useful information for town-based precise preventive strategies. When comparing the components among different towns, a large proportion of autoregressive and spatialtemporal components also existed in specific towns. The autoregressive component was mainly located in eastern towns, indicating that these towns were mainly affected by the previous infections in their own regions. So the early protection measures may aid in reducing the risk of PTB transmission in these areas. The spatialtemporal components were mainly located in central towns, indicating the high transmission risk from neighbouring areas with high incidence. So it is necessary to monitor PTB infection of the floating population in the adjacent areas.18 Therefore, the government needs to adjust measures to local conditions and take corresponding prevention measures according to the spatiotemporal patterns of the local epidemic, so as to achieve accurate prevention and control of TB.
The multivariate nature of socioeconomic factors for PTB incidence, along with the inherent spatiotemporal clustering, poses new challenges for data analysis. The Bayesian spatiotemporal model was superior to the traditional methods in revealing the potential risk factors and also considering the spatialtemporal clustering. This model has been used to explore the risk factors for COVID-19 and malaria19 20 ; however, few studies have applied it in the PTB field. In the present study, we used the Bayesian spatiotemporal model to explore the effects of socioeconomic factors on the incidence of PTB and found that the risk of PTB incidence was significantly associated with APR, sex ratio (male to female) and NBMI. For example, the higher risk of PTB incidence was associated with higher APR, which may be affected by low health awareness as well as high delays of diagnosis and treatment.21 22 The higher rate of males was the risk factor for PTB, which may be related to the high smoking rate and low immunity in males.23 In addition, in many cases, there was strong evidence that males had significantly high delays in seeking and/or obtaining TB treatment in many settings.24 NBMI was the risk factor for PTB, indicating that the greater the number of beds in health facilities in this area, the stronger the ability to detect patients with PTB. Therefore, we should take effective preventive measures to control PTB and reduce its disease burden, such as health education, improving the medical service network to ensure that patients with TB can be diagnosed and cured in their own counties and quitting smoking strategies.
In conclusion, this study provided insight into the long-term trend and spatiotemporal transmission patterns of PTB, indicating that there was a significant downward trend for PTB incidence, with high–high clusters in the central and north-central towns. Notably, the multivariate time series analysis revealed that the endemic components played an important role in driving PTB transmission, suggesting that a great number of cases are influenced by local risk factors. Moreover, the Bayesian spatiotemporal model found that PTB was significantly associated with local socioeconomic factors including APR, sex ratio and NBMI. These findings suggest that targeted preventive efforts should be made in different towns based on their spatiotemporal transmission patterns and risk factors.
supplementary material
Footnotes
Funding: This work was supported by the Guangdong Basic and Applied Basic Research Foundation (No. 2023A1515011583), the Key Scientific Research Foundation of Guangdong Educational Committee (No. 2022ZDZX2033) and the National Science and Technology Major Project (No. 2018ZX10103001). The funders had no role in the study design, data collection and analysis, and interpretation of the data.
Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2024-085733).
Provenance and peer review: Not commissioned; externally peer reviewed.
Patient consent for publication: Not applicable.
Map disclaimer: The inclusion of any map (including the depiction of any boundaries therein), or of any geographic or locational reference, does not imply the expression of any opinion whatsoever on the part of BMJ concerning the legal status of any country, territory, jurisdiction or area or of its authorities. Any such expression remains solely that of the relevant source and is not endorsed by BMJ. Maps are provided without any warranty of any kind, either express or implied.
Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Ethics approval: This study was approved by the ethics committee of the Sixth People's Hospital of Dongguan (No. Z2020-007). Written informed consent from each study participant was not required because we did not include any data of patients’ personal information, including name, telephone number, identity information and so on. This study only used secondary aggregated data from the China Information System for Disease Control and Prevention.
Contributor Information
Jingfeng Zhang, Email: 1393482611@qq.com.
Minghao Zhong, Email: 380082268@qq.com.
Jiayin Huang, Email: 786342344@qq.com.
Wenjun Deng, Email: gabyddd@163.com.
Pingyuan Li, Email: 15944466506@163.com.
ZhenJiang Yao, Email: 497345240@qq.com.
Xiaohua Ye, Email: smalltomato@163.com.
Xinguang Zhong, Email: 13798192031@163.com.
Data availability statement
Data are available on reasonable request.
References
- 1.World Health Organization . Global tuberculosis report 2021. Geneva, Switzerland: World Health Organization; 2021. [Google Scholar]
- 2.WHO global lists of high burden countries for tuberculosis (TB), TB/HIV and multidrug/rifampicin-resistant TB (MDR/RR-TB), 2021–2025: background document. 2021.
- 3.Chen J, Qiu Y, Yang R, et al. The characteristics of spatial-temporal distribution and cluster of tuberculosis in Yunnan Province, China, 2005-2018. BMC Public Health. 2019;19:1715. doi: 10.1186/s12889-019-7993-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Zhang Y, Wang XL, Feng T, et al. Analysis of spatial-temporal distribution and influencing factors of pulmonary tuberculosis in China, during 2008-2015. Epidemiol Infect. 2018;147:e25. doi: 10.1017/S0950268818002765. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Li X-X, Wang L-X, Zhang H, et al. Spatial variations of pulmonary tuberculosis prevalence co-impacted by socio-economic and geographic factors in People’s Republic of China, 2010. BMC Public Health. 2014;14:1–12. doi: 10.1186/1471-2458-14-257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Cleveland RB, Cleveland WS, Mcrae JE. STL: A seasonal-trend decomposition. J Off Stat. 1990;6:3–73. [Google Scholar]
- 7.Moran PAP. Notes on continuous stochastic phenomena. Biometrika. 1950;37:17–23. doi: 10.2307/2332142. [DOI] [PubMed] [Google Scholar]
- 8.Meyer S, Held L, Höhle M. Spatio-temporal analysis of epidemic phenomena using the R package surveillance. J Stat Softw. 2017;77:1–55. doi: 10.18637/jss.v077.i11. [DOI] [Google Scholar]
- 9.Geilhufe M, Held L, Skrøvseth SO, et al. Power law approximations of movement network data for modeling infectious disease spread. Biom J. 2014;56:363–82. doi: 10.1002/bimj.201200262. [DOI] [PubMed] [Google Scholar]
- 10.Lee D. A comparison of conditional autoregressive models used in Bayesian disease mapping. Spat Spatiotemporal Epidemiol. 2011;2:79–89. doi: 10.1016/j.sste.2011.03.001. [DOI] [PubMed] [Google Scholar]
- 11.Blangiardo M, Cameletti M, Baio G, et al. Spatial and spatio-temporal models with R-INLA. Spat Spatiotemp Epidemiol. 2013;4:33–49. doi: 10.1016/j.sste.2012.12.001. [DOI] [PubMed] [Google Scholar]
- 12.Peng F, Tu L, Yang Y, et al. Management and treatment of COVID-19: the Chinese experience. Can J Cardiol. 2020;36:915–30. doi: 10.1016/j.cjca.2020.04.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Li X-X, Wang L-X, Zhang H, et al. Seasonal variations in notification of active tuberculosis cases in China, 2005–2012. PLoS ONE. 2013;8:e68102. doi: 10.1371/journal.pone.0068102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hong C-Y, Wang F-L, Zhang Y-T, et al. Time-trend analysis of tuberculosis diagnosis in Shenzhen, China between 2011 and 2020. Front Public Health. 2023;11:1059433. doi: 10.3389/fpubh.2023.1059433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Mahara G, Yang K, Chen S, et al. Socio-economic predictors and distribution of tuberculosis incidence in Beijing, China: a study using a combination of spatial statistics and GIS technology. Med Sci (Basel) 2018;6:26. doi: 10.3390/medsci6020026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Li Q, Liu M, Zhang Y, et al. The spatio-temporal analysis of the incidence of tuberculosis and the associated factors in mainland China, 2009-2015. Infect Genet Evol. 2019;75:S1567-1348(18)30835-9. doi: 10.1016/j.meegid.2019.103949. [DOI] [PubMed] [Google Scholar]
- 17.Zuo Z, Wang M, Cui H, et al. Spatiotemporal characteristics and the epidemiology of tuberculosis in China from 2004 to 2017 by the nationwide surveillance system. BMC Public Health. 2020;20:1284. doi: 10.1186/s12889-020-09331-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Liao W-B, Ju K, Gao Y-M, et al. The association between internal migration and pulmonary tuberculosis in China, 2005-2015: a spatial analysis. Infect Dis Poverty. 2020;9:5. doi: 10.1186/s40249-020-0621-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Sahu SK, Böhning D. Bayesian spatio-temporal joint disease mapping of Covid-19 cases and deaths in local authorities of England. Spat Stat. 2022;49:100519. doi: 10.1016/j.spasta.2021.100519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Rotejanaprasert C, Ekapirat N, Sudathip P, et al. Bayesian spatio-temporal distributed lag modeling for delayed climatic effects on sparse malaria incidence data. BMC Med Res Methodol. 2021;21:287.:287. doi: 10.1186/s12874-021-01480-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lin Y, Enarson DA, Chiang C-Y, et al. Patient delay in the diagnosis and treatment of tuberculosis in China: findings of case detection projects. Public Health Action. 2015;5:65–9. doi: 10.5588/pha.14.0066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Lei X, Huang K, Liu Q, et al. Are tuberculosis patients adherent to prescribed treatments in China? Results of a prospective cohort study. Infect Dis Poverty. 2016;5:38.:38. doi: 10.1186/s40249-016-0134-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Amere GA, Nayak P, Salindri AD, et al. Contribution of smoking to tuberculosis incidence and mortality in high-tuberculosis-burden countries. Am J Epidemiol. 2018;187:1846–55. doi: 10.1093/aje/kwy081. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Horton KC, MacPherson P, Houben RMGJ, et al. Sex differences in tuberculosis burden and notifications in low- and middle-income countries: a systematic review and meta-analysis. PLoS Med. 2016;13:e1002119. doi: 10.1371/journal.pmed.1002119. [DOI] [PMC free article] [PubMed] [Google Scholar]