School data
During the study period, 1,940,735 adolescents were administered a tuberculin skin test, representing 1427 schools in 12 cities (Table 1; Table S1–3; Fig. S1). Figure 1 presents the geographic distribution of school and adolescent participation in tuberculin skin test screening. The number of schools participating in screening ranges from 3 to 47, and there is substantial heterogeneity in screening scale across counties, ranging from 499 adolescents in Guangling District to 98,804 in Binhai County. The number of adolescents administered tests increased with time from 118,831 (Nschools = 311) in 2017 to 734,709 (Nschools = 1186) in 2021 (Table 1). More schools reported indurations cutoffs from students of 15 mm (Nparticipants = 1,922,046) compared to 10 (Nparticipants = 1,196,265) and 5-mm induration cutoffs (Nparticipants = 1,653,995).
This figure presents a choropleth map illustrating the number of participating schools (A) and the total number of schoolchildren (B) across various administrative regions within the study area for the period 2017–2021. Data are categorized into distinct numerical ranges and represented by color intensity, allowing for a visual assessment of participation density and potential spatial disparities. A specifically highlights the distribution of participating schools, with shades of blue, orange, and red indicating increasing numbers of schools (1–9, 10–19, 20–29, 30–39, and 40+, respectively). B mirrors this approach for the schoolchildren, employing a similar color gradient to denote the number of students (1–9999, 10,000–19,999, 20,000–29,999, 30,000–39,999, and 40,000+). Regions marked in gray indicate unavailable data. Data were compiled from annual monitoring records and correspond to the total cumulative participation over the study period. The map is sourced from the National Geographic Information Public Service Platform (https://cloudcenter.tianditu.gov.cn/administrativeDivision). The review number of the map is GS(2024)0650.
Among middle school students, the overall prevalence of tuberculin skin test positivity at 5, 10, and 15 mm cutoffs was 8.2% (95% CI, 8.1–8.2), 3.4% (95% CI, 3.4–3.5), and 1.1% (95% CI, 1.1–1.1), respectively. Among high school students, prevalence at each cutoff was 11.0% (95% CI, 10.9–11.0), 4.6% (95% CI,4.6–4.7), and 1.6% (95% CI, 1.6–1.6), respectively (Tables S4–8).
Overall, we found a reducing prevalence of tuberculin positivity across years of follow-up with decreases occurring during pandemic years, regardless of millimeter induration cutoff (Fig. 2). Among middle school students, the prevalence of tuberculin positivity using a 5 mm induration cutoff was relatively consistent from 2017 to 2019 (9.3% [95% CI, 8.9–9.7], 9.6% [95% CI, 9.3–9.8], 10.2% [95% CI, 9.9–10.4], respectively), but was largely reduced in 2020 and 2021 (8.0% [95% CI, 7.9–8.2] and 7.0% [95% CI, 6.9–7.1]). The percentage prevalence reduction in 2020 and 2021 compared to the expected value was 23.8% (95% PI, −26.6, −20.8) and 36.4% (95% PI, −39.1, −32.7), respectively. From 2017 to 2019, the annual risk of M. tuberculosis infection among middle school students was 7.8, 8.0, and 8.5 infections per 1000 persons (Table 2). In 2020 and 2021, the annual risk of M. tuberculosis infection decreased to 6.7 and 5.8 infections per 1000 persons, respectively. There was a 24.7% (95% PI, −27.2, −21.2) and 37.0% (95% PI, −40.8, −33.3) decreased percentage change in the annual risk of M. tuberculosis infection in 2020 and 2021 compared to the counterfactual model. Among high school students, similar but slightly more modest reductions in tuberculin positivity were seen in post-pandemic years. For example, the prevalence of tuberculin positivity in pre-pandemic years was 11.9% (95% CI, 11.7–12.2), 12.1% (95% CI, 11.9–12.3), and 11.8% (95% CI, 11.7–12.0), respectively. In 2020 and 2021, prevalence was 11.6% (95% CI, 11.5–11.7) and 9.6% (95% CI, 9.5–9.7). In 2020 and 2021, there was a reduction in the annual risk of M. tuberculosis infection of 2.5% (95% PI, −6.5, 1.8) and 19.8% (95% PI, −24.0, −14.5) compared to each respective counterfactual model.
Graph displays the annual percent prevalence of tuberculin skin test positivity at differing millimeter cutoffs for each year from 2017 to 2021 in Jiangsu Province, China. Each point for each year represents the number of students tested during that year in representative schools. Therefore, the denominator for each year may be distinct. Millimeter cutoffs can be seen for 5 mm (left top panel), 10 mm (right top panel), and 15 mm (bottom panel). The gray shaded area on the right is the time period in which Covid-19 restrictions were put in place in Jiangsu Province (starting in February 2020). Gray circles represent the observed percent prevalence of tuberculin skin test positivity during each year of testing in the included cities. The solid red line represents the fitted model of the observed percent prevalence data from pre-COVID (January 2017 to January 2020) and post-COVID (February 2020 to December 2021). The blue dotted line represents the counterfactual expected percent prevalence of tuberculin skin test positivity during the post-COVID time period based on the observed data from the pre-COVID time period. The red transparent shaded area for the observed values represents 95% confidence intervals. The blue transparent shaded area for the predicted values represents 95% confidence intervals from the counterfactual fitted models.
When using alternative millimeter cutoffs, we consistently found decreases in the annual risk of M. tuberculosis infection in 2021 compared to counterfactual estimates (Fig. 3; Table S5; Table S7). Using a 10 mm cutoff, there was a 27.7% (95% PI, −39.4, −10.5) reduction in the annual risk of M. tuberculosis infection among high school students and a non-significant 32.2% (95% PI, −55.2, 37.0) decrease among middle school students (Table S5). When using a 15 mm induration cutoff, the estimated percentage reduction in the annual risk of M. tuberculosis infection in 2020 and 2021 among middle school students was 23.6% (95% PI, −40.8, 7.5) and 39.9% (95% PI, −58.5, 9.1). Among high school students, the percentage reduction in the annual risk of M. tuberculosis infection in 2020 and 2021 was 7.4% (95% PI, −8.7, −6.1) and 30.4% (95% PI, −31.7, −29.0) compared to the expected model (Table S7).
The figure presents the observed percentage reduction in the annual risk of Mycobacterium tuberculosis infection for 2020 and 2021 among middle school and high school students compared to predicted values based on historic data from 2017 to 2019. The annual risk of M. tuberculosis infection was calculated by transforming the percent prevalence of tuberculin positivity in each measured year as described in the methods. The mean age of middle school students was 12.5 years and high and school students was 15.5 years. All participants from each year were included in this specific presented analysis. Other analyses included only schools with full data on all years and are presented in Supplementary Tables 4, 6 and 8. They were broadly consistent with these results.
When restricting our population to only schools with complete annual survey data in all years, our results were broadly consistent (Table S4; Table S6; Table S8). Using a 5 mm induration, there was a 48.6 (95% PI, −55.7, −39.0) and 19.9 (95% PI, −29.5, −7.5) percentage reduction in the annual risk of M. tuberculosis infection among middle and high school students in 2021. Using a 10 mm induration, this percentage reduction in 2021 was 9.9 (95% PI, −10.1, −9.7) among high school students and no statistically significant reduction among middle school students. Using a 15 mm induration, there was a non-statistically significant percentage reduction in 2021 of 42.7 (95% PI, −62.1, 16.4) among high school students compared to the counterfactual model.
We performed several sensitivity analyses. First, we estimated the number of new student entries in participating schools from the 12 cities; we found that this number remained largely consistent over the study time period (Fig. S2). The mean number of student entries per school was also similar over time. In 2017, the mean number of student entries per school was 341 (95% CI, 314–368), similar to the mean school entries in 2019 (371; 95% CI, 353–389) and 2021 (370; 95% CI, 354–386). This suggests that the number of students entering school was not substantially altered pre- and post-pandemic. Second, we conducted a quantitative bias analysis considering the sensitivity and specificity of the tuberculin skin test and potential influence on annual trends in tuberculin skin test positivity. When adjusting for diagnostic characteristics of the tuberculin skin test, we found relatively similar adjusted versus observed proportions in 2017 (11.1% versus 11.0%), 2018 (11.5% versus 11.4%), and 2019 (11.5% versus 11.4%). Adjusted proportions were reduced compared to observed proportions in 2020 (10.1% versus 10.4%) and 2021 (8.7% versus 7.7%).
In most cities, we found that a decreasing trend in tuberculin positivity was seen in 2020 and 2021 (compared to 2019). However, this was not the case for schools in every city. For example, decreasing trends were seen in Lianyungang, Nantong, Suzhou, Wuxi, amongst others. However, we did not see decreasing trends in Changzhou, Huaian, and others (Table S9; Table S10), suggesting there was some city-level heterogeneity in our findings.
Tuberculosis disease notification data
In total, there were 149,970 tuberculosis diagnoses in the province from 2017 to 2021. The absolute number of tuberculosis diagnoses ranged from 28,379 (2365 per month) in 2017 to 22,827 (1902 per month) in 2021 (Fig. S3).
We compare key characteristics of students and schools tested in each time period (2017–2019 and 2020–2021). These 12 cities are located in different parts of Jiangsu Province, north, central, and south, and have different incidence rates of tuberculosis and characteristics, which is also the case in this study (Table S11; Table S12).
Overall, provincial tuberculosis case notifications dropped 23% (95% PI, 18–27) in the first 4 months of 2020 compared to the counterfactual model (5666 observed versus 7319 expected notifications) (Fig. S4, left panel). However, observed notifications quickly rebounded and by the end of 2020 there was no statistical reduction in observed cases per month (2029 observed versus 2085 expected notifications [95% PI, 1944–2227]; –2.7% change, 95% PI, –4.4, 8.9). There continued to be no statistical reduction in observed versus expected notifications in 2021 (1902 observed versus 2004 expected notifications [95% PI, 1801–2207]; –5.1% change, 95% PI, –13.0, 5.6). The notification rate followed a similar trend as the absolute number of notifications (Fig. S4, right panel). There was a significant drop in the observed tuberculosis notification rate in the first months of 2020 followed by a rebound to the expected values found through the counterfactual model.
