Tuberculosis is one of the top killer diseases in the globe. The aim of this study was to explore the geographic distribution patterns and clustering characteristics of the disease incidence in terms of both space and time with high relative risk locations for tuberculosis incidence in Beijing area. A retrospective space-time clustering analysis was conducted at the districts level in Beijing area based on reported cases of sputum smear-positive pulmonary tuberculosis (TB) from 2005 to 2014. Global and local Moran’s I, autocorrelation analysis along with Ord (Gi*) statistics was applied to detect spatial patterns and the hotspot of TB incidence. Furthermore, the Kuldorff’s scan statistics were used to analyze space-time clusters. A total of 40,878 TB cases were reported in Beijing from 2005 to 2014. The annual average incidence rate was 22.11 per 100,000 populations (ranged from 16.55 to 25.71). The seasonal incidence occurred from March to July until late autumn. A higher relative risk area for TB incidence was mainly detected in urban and some rural districts of Beijing. The significant most likely space-time clusters and secondary clusters of TB incidence were scattered diversely in Beijing districts in each study year. The risk population was mainly scattered in urban and dense populated districts, including in few rural districts.
It has been established that Tuberculosis (TB) is a top killer chronic infectious disease in the world, where 10 million people were ill, and 1.6 million people died from the disease in 2017 [
Geographic Information System (GIS) and its application in epidemiology for the disease interventions have been broadly utilized in spatial epidemiology [
Previously several studies have been conducted about TB in the study region [
A retrospective study conducted in Beijing region based on the reported new (smear positive) pulmonary tuberculosis cases from 2005 to 2014. Beijing, the capital city of China, is located at the northern tip of the North China plainat 39˚54'50"N and 116˚23'30"E surrounded by the mountains. A total area of Beijing has 16,410・54 Km2 and consists 16 administrative districts, and all districts are covered by this study (
All reported cases (smear positive cases) of TB data including age, sex, address, and occupation with clinical records (from 2005 to 2014) were obtained from the National Surveillance System for Notifiable Infectious Diseases, Beijing Information of Tuberculosis Prevention and Control registration database [
All TB cases were inputted into Microsoft Excel 2010 (Microsoft, Redwoods, WA, USA) and geo-coded according to their residential addresses and determining its longitude and latitude coordinate by using ArcGIS 10.1 (ESRI Inc, Redlands, CA USA) and linked to the respected location in Geographic Information system (GIS). To conduct a GIS-based analysis on the spatial distribution of TB, the districts/counties level polygon map at 1:100,000 scales were obtained from National Geometric Center of China [
This study was approved under the Institutional Review Committee for Healthcare Research and Quality protocol at Capital Medical University, Beijing China, with secondary analysis of confidential data from the CDC, Beijing, China.
We accumulated all TB cases by month with their diagnosis dates to identify the temporal patterns of the disease. We also summarized the TB cases including gender, age, occupation and the year of the diagnosis to identify the demographic characteristics of the disease by years. All descriptive information of respondents was summarized annually according to geographic area (districts) and all incidence rates were calculated for each neighborhood, using neighborhood census data as denominators and reported TB cases as the numerator. All descriptive analysis was conducted using SPSS software-v-20.
Spatial autocorrelation (Global Moran’s I) analysis was used to investigate the patterns of TB incidence. It evaluates whether its distribution pattern is clustered, dispersed or at random [
We applied Anselin Local Moran’s I (Local Indicators of Spatial Association-LISA) statistic to determine whether there are positive spatial correlations (high-high or low-low clusters) or negative spatial correlations (high-low or low-high clusters) exist. It also measures the spatial association between each individual spatial district with its neighboring districts [
To explore the additional information about the intensity and existence of core hotspots or cold spot cluster of TB incidence in the study location, we further assessed the local spatial autocorrelation analysis using Getis-Ord Gi* statistic in ArcGIS software [
A retrospective space-time scan analysis based on a discrete Poisson probability model was carried out using SaTScanTM V-9.4 software [
In this study, the maximum size of space-time clusters was defined as twice higher risk at least within the window than outside (RR ≥ 2). First, people within the identified clusters are more likely to be infected with TB bacteria than those who identified within outside the clusters when the RR is greater than one. So, this cluster is a risk factor for the TB infection. On the other hand, the cluster is a protective factor when the RR is less than one. If there is no any effect for the risk on the infection, the RR is one. Therefore, the selection for the maximum size of space-time cluster, needs the RR > 1 at least. In this study, the spatial size of scanning window was setting as 50% of the total population at the risk. The statistical significance of each cluster in the study location was based on comparing the likelihood ratio (LLR) achieved from the Monte Carlo simulation with the maximum number of the replication set to 999 to ensure the sufficient statistical power. The window with the maximum LLR is assumed to be the most likely cluster that is the cluster least likely to be caused by chance, and other windows with a statistically significant LLR were measured as secondary clusters. The relative risk (RR) of the incidence inside and outside the window considered to be statistically significant if P < 0.05 [
In this study, the spatial units of space-time scan analysis were the 16 districts of Beijing, and the time units were 10 years from 2005 to 2014. The time frame of the scan analysis was set to be a month to control the time trends and to observe the cluster changes in the entire study period.
A total of 44,408 new sputum smear positive pulmonary tuberculosis cases were reported in Beijing from 2005 to 2014. We excluded 3530 cases due to unavailable residence information or not residing in the study area. Finally, we employed 40,878 cases for the final analysis. The annual average incidence rate of TB observed as 22.11 per 100,000 populations, which is appeared almost stable from 2006 to 2008, (25.65, to 25.71 per 100,000 populations), and displayed lowest in 2014 with 16.55/100,000 populations (
Likewise, the high incidence of TB was observed in highly densely populated districts (
Global Moran’s I autocorrelation analysis was used to detect TB incidence patterns such as clustered, dispersed and randomly distributed in the study region. The global spatial correlation analysis exhibited the presence of positive spatial clustering of TB incidence in Beijing region (Z-score from 0.842 to 2.324) (
Anselin’s Local Moran’s I analysis disclosed the diverse patterns of the core clustering of high TB incidence districts with HH (high-high) and LH (low-high) cluster during the study period (
The clusters of TB incidences, including hotspots and cold spots were estimates using the local Getis-Ord-Gi* statistics. Location and size of hotspots clusters varied in each study year. The most hotspots incidence locations of TB incidence were detected in the urban districts e.g., Chaoyang, Xicheng, Fengtai and Dongcheng districts. The hotspots cluster further divided into primary (GiZ score > 2.58), secondary (GiZ score = 1.96 - 2.58 SDs) and tertiary (GiZ score = 1.64 - 1.96) according to the values of GiZ score for the clearly defining the intensity of clusters (
A Space-time cluster analysis of TB incidence was conducted in Beijing districts by using space-time permutation model to identify the statistically significant monthly spatiotemporal clusters from 2005-2014 (
The most likely clusters of TB incidence were detected varied in each study year. A large circular most likely cluster was detected in the northern part of Beijing covering three districts (Yanqing, Huairou, and Miyun) with cluster time ranged from 1st to 31st January, clusters center were 40.628˚N, 116.578˚E, and 40.524˚N, 116.986˚E, radius 37.05 Km and 37.73 Km with the relative risk (RR)
Years | Moran’s-I | Z Score | P-Value |
---|---|---|---|
2005 | 0.199 | 1.920 | 0.033 |
2006 | 0.235 | 2.324 | 0.017 |
2007 | 0.118 | 1.627 | 0.059 |
2008 | 0.064 | 1.072 | 0.149 |
2009 | 0.054 | 0.842 | 0.191 |
2010 | 0.115 | 1.263 | 0.103 |
2011 | 0.149 | 1.623 | 0.066 |
2012 | 0.163 | 1.751 | 0.051 |
2013 | 0.100 | 1.354 | 0.107 |
2014 | 0.104 | 1.142 | 0.119 |
Scan Year | Cluster Time Frame | Center (Latitude, Longitude)/Radius (km) | Cluster Districts | Relative Risk | P-Value |
---|---|---|---|---|---|
2005 | 01 - 31 Jan | 116.5787 E, 40.628 N/36.25 km | 2 | 0.23 | <0.001 |
2006 | 01 - 31 Jan | 40.6281 N, 116.578 E)/36.25 km | 2 | 2.29 | <0.001 |
2007 | 01 - 31 Jan | 40.524 N, 116.986 E/0 km | 1 | 0.16 | <0.001 |
2008 | 01 Nov - 31 Dec | 39.911 N, 116.410 E/0 km | 1 | 2.03 | <0.001 |
2009 | 01 - 31 Jan | 40.6281 N, 116.578 E/37.05 km | 3 | 0.30 | <0.001 |
2010 | 01 - 31 Jan | 40.524 N, 116.986 E/37.73 km | 3 | 0.20 | <0.001 |
2011 | 01 - 31 Jan | 40.214 N, 116.204 E/21.03 km | 2 | 2.89 | <0.001 |
2012 | 01 - 31 Jan | 40.214 N, 116.204 E/21.03 km | 2 | 2.84 | <0.001 |
2013 | 01 - 31 Jan | 40.214 N, 116.204 E/21.03 km | 2 | 0.29 | <0.001 |
2014 | 01 - 31 Jan | 40.214 N, 116.204 E/21.03 km | 2 | 3.47 | <0.001 |
value 0.30 and 0.20 (P < 0.001) respectively in 2009 and 2010. Followed by second largest circular clusters were detected in 2005 and 2006 covering two districts (Huairou and Miyun district), and similarly, Haidian and Changping district in 2013 and 2014, whereas only one small cluster was found in Miyun and Xicheng district in 2007 and 2008 respectively in the study region. There were several secondary clusters were also detected apart from the most likely clusters (
This study declared that the most likely and secondary clusters of TB incidence were scattered both in rural and urban districts in Beijing (
likely to prone to have a disease infection than those who reside outside the low relative risk area. The study found the most likely cluster of TB incidence with high Relative Risk values (RR > 1) in the year 2006, 2008, 2011-2012 and 2014, whereas low relative risk cluster of TB incidence with low RR values (<1) were detected in other remaining years (
We investigated the spatial patterns and space-time clusters of TB incidence and highlighted those geographic areas in Beijing districts from 2005 to 2014. This study established that Xicheng district followed by Mentouhou district had the highest TB incidence during in the study period (
The declining of TB incidence was noted since 2006 in the study region, that could be due to apply the standard diagnosis and treatment system with improved notification of TB cases after setup the MDGs Goal in the year 2000. This declining of TB incidence reveals the effectiveness of TB control program and implementation of Direct Observed Treatment Short-Course (DOTS) strategy in district level over the last two decades [
Through this study detected seasonal trends with apparent peaks in all season. Almost, the disease incidence peaks were detected in summer between March to July until late autumn and in winter. Such seasonal patterns revealed by the previous studies in different regions as like: in China [
The estimates of the global Moran’s I indicated that occurrence of TB was distributed randomly at the district level, which is consistent with the finding of Lan Li et al. [
The analysis of space-time scan statistic highlighted the significant most likely and several secondary clusters, which were distributed diversely in each study year. Three most likely clusters were detected in Ecological Preservation Development districts in 2009 (Yanqing, Huairou, Miyun) and 2010 (Huairou, Miyun, and Pinggu districts) and followed by two most likely clusters in 2005 and 2006 such as g, Huairou, and Miyun, districts (
Several studies have stated that overcrowding conditions and social disturbance could all be risk factors for the burden and variations of disease occurrence. Furthermore, patient care factors [
Spatial analysis has been extensively utilized to detect the distribution patterns of various communicable diseases along with non-communicable diseases by GIS technology, ArcGIS, SaTScan and other relevant software, and assimilated meaningful results [
Though our study demonstrated the usefulness of spatial and temporal clustering analysis in Beijing, where there were some limitations. First, the estimated risk of TB infection might be underestimated in some areas because cases may not be reported or recorded in the health system. Second, we used space-time scan statistics to detect clusters in different space and time. The method that depends on circular spatial scanning windows and space-time cylinders, that does not allow for irregular space [
We investigated the space-time scan characteristic of pulmonary TB incidence in Beijing region and identified the hotspot clusters in urban and some rural districts over the study period by using GIS analysis. The most likely significant clusters and relative risk of TB were detected by space-time analysis in the urban districts and few clusters in rural districts. Therefore, our findings suggest that TB control measures should be focused on those “hot spot” districts including most likely clusters districts to allocate public health resources more precisely to reduce the burden of TB incidence.
The authors thank the Ministry of Science and Technology of Beijing, China, National Center for Public Health Surveillance and Information Services, China Center for the Disease Control and Prevention, Beijing, China for providing the tuberculosis case data.
Conceptualization, Gehendra Mahara and Xiuhua Guo; Data curation, Gehendra Mahara, Mina Karki, Kun Yang and Sipeng Chen; Formal analysis, Gehendra Mahara, Kun Yang and Sipeng Chen; Methodology, Gehendra Mahara and Mina Karki; Project administration, Xiuhua Guo; Resources, Kun Yang and Xiuhua Guo; Software, Gehendra Mahara and Kun Yang; Writing―original draft, Gehendra Mahara; Wei Wang, Writing―review & editing, Gehendra Mahara, Xiuhua Guo and Wei Wang.
We the authors declare that we have no conflict of interests.
Mahara, G., Karki, M., Yang, K., Chen, S.P., Wang, W. and Guo, X.H. (2018) Space-Time Cluster Analysis of Tuberculosis Incidence in Beijing, China. Journal of Tuberculosis Research, 6, 302-319. https://doi.org/10.4236/jtr.2018.64027