This study reveals temporal, spatial and spatial-temporal TB prevalence distribution in southwestern China at county-level in Yunnan province from 2005 to 2018. In brief, we detected TB high-risk time interval and high-epidemic areas for scanning dynamic spatial-temporal characteristics, simultaneously identified similar patterns of TB prevalence among GMS. To our knowledge, this was the first studies showed the perspective of spatial-temporal PTB, SSP-TB and SSN-TB epidemic in Yunnan province at county-level.
The annual notification of PTB was 59.6 per 100 000 population of Yunnan in 2018. Although significant efforts made an annual decline rate of 1.5% from 2005, the high TB burden with absolute number of 28 618 cases reported in 2018, made a big challenge to achieve the goal of End TB in 2035 without the breakthrough of vaccine or new drug.
Time series decomposed secular trend showed that the peak of notification PTB was in 2005. The presumptive reason is that China launched the direct internet-based reporting system for infectious diseases in 2004, TB was one of the notified diseases, thus greatly increased the TB reporting cases in 2005. From 2003 to 2005, the detection of SSP-TB cases by the public health system more than doubled, from 30% of new cases to 80%. In the following years, plenty efforts of increasing government commitment, intensive DOTS implementation and improved public-health funding. All these measures focused on TB control, and made an acceleration of prevalence decline [4, 23]. The recent PTB prevalence upward was likely attributed to large-scale active case finding strategy implementation in Yunnan from 2016 to 2018. The number of TB suspects with symptoms were 184 618 in 2018, almost doubled from 104 960 in 2015[14, 24, 25].
Seasonality was observed in different counties for TB notification. Interestingly, the peak month and trough month of TB notification were constant in the North Hemisphere regardless of the locations’ longitude. The peak months were roughly the same in American (spring, March), South Korea (spring and summer), Indian (spring, March to May), Singapore (spring and summer, March and July), China (spring, April) , Wuhan city (spring, March) and Xingjiang autonomous prefecture of China (spring, March). Our study was consistent with these researches in the North Hemisphere, seasonal factors were observed and the peak in January and the secondary peak in May. The hypothesis of TB seasonality was related to sunshine and temperature. Vitamin D deficiency in winter, the higher transmission risk for indoor crowding than outdoor in winter, the temperature was inversely and lagged associated with TB incidence , all of which caused seasonality disease for the peak in spring and summer. In China, the Spring Festival effect should also be considered. Which means TB notification significant reduced during Spring Festival holidays, consequently, seasonal factors sharply declined in habitual Spring Festival month of February. Meanwhile, the purely temporal scan revealed the temporal clusters were concentrated in spring and summer in Yunnan each year. In the whole study time frame, the cluster interval for SSN-TB was from 2008 to 2011, though for SSN-TB was more recently from 2013 to 2017, which suggested the ongoing TB control policy should focus on SSN-TB in Yunnan.
Kulldorff’s scan statistics method was developed to evaluate temporal and geospatial distribution, it was applied to detect communicable disease, vector-borne diseases and cancer geospatial aggregation [32–35], meanwhile, the sensitivity of spatial-temporal statistics prompted early detection of disease outbreak and emergency disease from surveillance system [36, 37]. This powerful method showed the strength of statistical robustness and interpretability of analyzed results. Scan statistics were widely applied in study topic related to TB[38–42], whereas, data aggregated into large scales of administrative regions may ignore the disease variation in small size of population, information lose lead to inaccurate and insensitive conclusion, these national-level researches could not preciously detect localized cluster on the resolution of province or prefecture[8, 9, 44].
The purely spatial scan showed that the PTB in Yunnan were not randomly distributed, and the dynamic prevalence of PTB revealed three mainly aggregated regions, the hotspot of Yunnan north-eastern angle in Zhaotong prefecture was high frequently in clusters and hold 8 of 14 years in study interval. Previous studies examined the PTB clusters for Zhaotong prefecture were in towns of Zhengxiong and Weixin county, the strongly associated indicators of the PTB spatial cluster in these regions including the history of chronic bronchitis, living in urban area, smoking and using coal as the main fuel for cooking and heating[13, 45].
Spatial-temporal cluster pattern was in line with pure spatial scanning. Unexpectedly, spatial-temporal scan detected two clusters were implemented active cases finding, one was secondary cluster 2 (Lanping county) and another was cluster 11 (Dongchuan county) in PTB clusters. The time frame of clustering matched with activities of active cases finding [14, 46]. This suggested that by considering the cluster time interval, higher sensitivity and closer to reality outcome for the spatial-temporal scan. Besides, the time frame for SSP-TB clusters concentrated before 2012, though most of the SSN-TB clusters were defined after 2012, which indicated the decline of SSP-TB and the progress and achievement for tuberculosis control in Yunnan.
Our study found that the most recent cluster of PTB and the SSP-TB spatial cluster for the whole interval was in southwestern borders neighbored with Myanmar, Laos, and Vietnam. Furthermore, the correlation of TB prevalence among borders and GMS were relatively high. Strikingly, hierarchical clustering indicated that there were 6 subclasses for TB epidemic pattern among GMS, thus the borders’ TB prevalence was similar to Myanmar TB epidemic pattern. Based on the consistency of traditional and molecular epidemiology evidence which confirmed the relatively lower prevalence of Beijing genotype in the border region of Pu’er, Xishuangbanna, as well as Vietnam and Myanmar[47–49], we speculated that the residents living in the border region moved across the national boundary for livelihood while the air-borne disease of TB was carried beyond foreigner. Recent high TB prevalence and high-risk temporal-spatial clusters in the GMS region suggested that cross-boundary intervention and international control policy should be implemented in these clusters.
Our study has some limitations. Firstly, the surveillance data did not contain covariates of patients’ demographic information for sex, age, etc., yet we did not introduce ecology factors like geographical, meteorological and economic situation, all of which could be possible indicators of TB incidence and prevalence. Secondly, we do not take account for unreported cases when using notifications data, since there was a risk of underestimated prevalence regardless of missing or unreported cases. Thirdly, it is difficult to collect the genetic and lower-level detailed geospatial information for TB among GMS, although it will advance the understanding of TB transmission among GMS. Further study should address these points.