Zhejiang Province is in the eastern region of China with a permanent population of 64.57 million in 2020, which covers an area of 100,000 square km, including plains, mountains, seashores, islands, and lakes (15). There are two sub-provincial cities of Hangzhou and Ningbo, and nine prefecture-level cities, including Wenzhou, Huzhou, Shaoxing, Jiaxing, Taizhou, Quzhou, Jinhua, Lishui, and Zhoushan (16). There are a total of 90 counties/cities within the Zhejiang Province.

Data for all notified PTB cases among students in Zhejiang Province were collected in the Tuberculosis Information Management System (TBIMS) from January 2007 to December 2020. We excluded extrapulmonary TB cases and cases identified as nontuberculous mycobacteria (17). All students included kindergartens, primary schools, high schools, and universities (18). Details for each PTB case included basic demographic information, clinical diagnosis information, laboratory outcomes, and treatment outcomes. Annual student population data for each administrative district were obtained from the Zhejiang Statistical Yearbook, education administrative department and Local Statistical Yearbook.

Combined with the features of age group and grade in China, students aged 3–6, 7–12, 13–15, and 16–18 years old were defined as kindergarten, primary school, and junior high, and senior high school students. Students greater than 19 years old were denoted as college students and above in this study. Notified PTB cases included laboratory-confirmed and clinically diagnosed cases. Clinically diagnosed PTB cases were defined as those with PTB-specialized chest imaging, clinical manifestations (coughing, expectoration for ≥2 weeks, or hemoptysis), and no response to diagnostic anti-inflammatory therapy with negative laboratory test results or missing results (anti-TB drug was excluded) (19). Laboratory confirmation of PTB was based on sputum smear, culture, and GeneXpert results indicating infection with M. tuberculosis. Completion of treatment course was defined as: (1) clinically diagnosed PTB patients completed the standardized course and had negative sputum smear/culture results and (2) laboratory-confirmed patients who completed the prescribed course of treatment, with a negative sputum test result at the latest, and no sputum test result at the end of the course of treatment. Cure was defined as persistent negative sputum smear or culture results for two times, one of which was at the completion of the standardized course of treatment in laboratory-confirmed patients (20). Multidrug-resistant tuberculosis (MDR-TB) was defined as tuberculosis that was resistant to two or more drugs, such as isoniazid and rifampin (21). Health-seeking delay was defined as the time interval from the occurrence of the first symptoms to the first visit to a designated hospital, when the first visit was >2 weeks after the occurrence of the first symptoms. Diagnostic delay referred to the interval between the visit to a designated hospital and diagnosis of PTB, if the diagnosis took >2 weeks (22).

Based on bacteriological or clinical diagnoses, the results of notified PTB cases among students were described in terms of sex, age group, city, anti-TB treatment, source of patients, classification of treatment, delays in seeking medical care, and treatment outcome.

We used the Joinpoint regression method to assess the yearly time trend of the notification rates of PTB and evaluate the annual percentage changes (APCs) and their confidence intervals (23). The Joinpoint regression model builds a segmented regression based on the temporal characteristics of the disease distribution. By partitioning the study time into different intervals, the Joinpoint regression method can fit and optimize the trend for each interval, ultimately evaluating the details among different interval-specific disease characteristics (24). The model analyzes statistically significant turning points and calculates the corresponding t-value and P-value through Monte Carlo permutation testing and selects the best-fit models according to Bayesian information criterion. In this study, we took the nature logarithm of annual notification incidence of PTB among students as the dependent variable and the year of notification as the independent variable. The trend change in this time segment was statistically significant when the APC had a P < 0.05 and the 95% confidence interval (CI) did not include 0 (25).

Spatial autocorrelation analysis, which includes global spatial autocorrelation and local spatial autocorrelation, is used to reveal the spatial distribution patterns of diseases, and identify disease risk areas. The global spatial autocorrelation analysis is a description of the spatial characteristics of the attribute values of the entire region and determines the aggregation characteristics of the studied variables in the overall space (26). Moran's I was used to assess the spatial autocorrelation, taking values in the range of−1 to 1. Moran's I >0 indicates a positive spatial autocorrelation, <0 means a negative autocorrelation, and 0 indicates random spatial distribution (no spatial autocorrelation). Moran's I of larger absolute values indicates a stronger spatial autocorrelation (27). A Moran's I was statistically significant at P < 0.05 and Z-score ≥ 1.96. Meanwhile, local spatial autocorrelation analysis was used to identify the spatial correlation between the study variables and the area variables of the neighboring regions (28). Local indicator of spatial autocorrelation (LISA) was used to test the local autocorrelation results to evaluate the existence of local clusters and detect regions with significant spatial correlations (29).

Kulldorff's space–time scan statistics are based on the Poisson probability distribution model to detect the spatial–temporal aggregation of diseases. The space–time scan statistic uses the moving windows to create a cylinder with the bottom surface as the scan area, height as the possible clustering time, and radius as the scan risk population (30). The size and location of scan window was constantly changed in space and time for searching possible clusters. In this method, the expected incidence is calculated based on the actual incidence number and population size of each window, and it is compared with the actual incidence number to build the log likelihood ratio (LLR) for identifying clusters using the Monte Carlo randomization method to examine the statistically significant cluster at 95% CI (31).

Descriptive analysis was conducted using R software (version 4.2.1, R Foundation for Statistical Computing) and Microsoft Excel. Joinpoint software (Joinpoint Regression Program, Version 4.9.1.0, April 2022; Statistical Methodology and Applications Branch, Surveillance Research Program, National Cancer Institute) was used to calculate the APC and its 95% CI. Moran's I was calculated using GeoDa (version 1.20) and SaTScan software (version 10.1, Boston, MA, USA) was used to determine the spatial and temporal aggregation of PTB. All visual results were presented using ArcGIS software (version 10.8, SERI Inc., Redlands, CA, USA).

Our study was approved by the ethics committee of the Zhejiang Center for Disease Control and Prevention and exempted from the requirement of informed consent. All personal information in this study was kept confidential as required.

A total of 17,500 PTB cases among students were notified in the Zhejiang Province from 2007 to 2020, accounting for 3.75% of all identified PTB cases (466,958). Among all diagnosed PTB cases, 68.77% (12,035) were derived from the resident population and the rest (5,465, 31.23%) were from the migrant population. The notified PTB incidence among students ranged from 9 cases per 100,000 to 17 cases per 100,000, with an average annual notified incidence of 12.89 cases per 100,000 population (Figure 1). From 2007 to 2013, the annual incidence of student PTB remained higher than the average incidence during the study period. A peak notified PTB incidence of 17.08 per 100,000 was observed in 2009. Among all the diagnosed PTB cases, 59.69% (10,445) were male and 40.31% (7,055) were female, with a male to female ratio of 1.48:1. In terms of age distribution, 12.86% were 13–15 years old, 39.90% were 16–18 years old and 43.65% were more than 19 years old. In terms of geographic distribution, more than 10% of all notified student cases were found in Hangzhou, Wenzhou, Ningbo, and Jinhua; the highest proportion of PTB bacteriological diagnosis (272, 40.42%) was observed in Jiaxing city while the highest proportion of clinical diagnosis (605, 76.01%) was observed in Quzhou city. Referral and passive findings were the main sources of student PTB in the Zhejiang province (Figure 2). 97.69% of PTB cases in students were initially treated and 70.16% were clinically diagnosed. Almost all student PTB patients received anti-TB treatment and nearly 95% of PTB cases completed PTB treatment or were cured. While 7,931 student cases (45.32%) experienced health-seeking delays, there was a low rate of diagnostic delay (2,453, 14.02%). In addition, 24.49% (4,286) of the students' cases had a treatment duration of ≥9 months (Table 1).

Joinpoint regression analysis revealed a downward trend in notified incidence of PTB cases among students, which decreased from 13.87 cases per 100,000 to 10.32 cases per 100,000 from 2007 to 2020. Furthermore, the annual percentage change was −3.68% (95% CI, −5.4 to −1.9) (Table 2, Figure 3A). However, the notified incidence of bacteriologically confirmed PTB cases showed a different trend and decreased from 3.99 cases per 100,000 to 2.67 cases per 100,000 from 2007 to 2017 but increased from 2.67 cases per 100,000 to 5.30 cases per 100,000 from 2017 to 2020, with APCs of −5.2% (95% CI, −9.1 to −1.1, P = 0.018) and 28.6% (95% CI, 2.1 to 62.0, P = 0.036), respectively (Table 2, Figure 3B).

The results of the global autocorrelation analysis indicated that the annual Moran's I was statistically significant in 2007–2010, 2012–2015, 2017, and 2019–2020, while there was no statistically significant difference in 2011, 2016, and 2018 (Table 3). The Moran's I was between 0.180 and 0.407, indicating a spatially positive correlation among students PTB in the Zhejiang Province. The LISA cluster map in county level revealed that hot spot regions of PTB epidemics among students were mainly concentrated at the border of Hangzhou, Jinhua, and Quzhou cities, including Jiande, Tonglu, Chunan, Linan, Lanxi, Pujiang, Changshan, and Kecheng (Figure 4). Although the hot spot regions changed dynamically per year, Jiande and Tonglu remained as hot spots throughout. The cold spots also changed during the study period. For example, the cold spots were mainly concentrated in Jiaxing, Zhoushan, and the western regions of Lishui in 2010, 2012, 2013, and 2015, whereas they were concentrated in Jiaxing, southern Wenzhou, and southeastern Jinhua in 2014.

One most likely cluster and three secondary clusters were identified by a spatial–temporal scan among students in the Zhejiang Province during the study period. For the most likely cluster, the cluster period was from January 2007 to December 2013 and its regions were mainly clustered in the western part of the Zhejiang Province including Jiande, Tonglu, Chunan, Lanxi, Kaihua, and Kecheng, with 691 notified PTB cases. The risk ratio of the most likely cluster was 2.07 (LLR = 14.16, P < 0.001) (Table 4 and Figure 5).