Spatial spillover effects and regional differences of tuberculosis health education levels among China’s floating population: a spatial lag model study (2014–2017)

Abstract

Background

The cross-regional floating population in China has brought new challenges for tuberculosis prevention and control. Tuberculosis health Education for the floating population can significantly enhance their health awareness and reduce the prevalence of the disease. However, the current status of health education for the floating population is not adequately assessed, highlighting the urgent need to identify the key factors that restrict effective health education and to safeguard the health rights and interests of this group.

Methods

This study utilized the data of China’s migrant population from 2014 to 2017, combined with the China Statistical Yearbook and meteorological data, and conducted literature searches and collected relevant information on tuberculosis health education through databases such as Web of Science and PubMed. It identified the key factors influencing tuberculosis health education from four dimensions: social economy, population, health resources, and meteorological environment. Furthermore, the study also explored the relationships between these indicators and tuberculosis health education at the provincial and urban regional levels respectively. A spatial lag fixed effect model was determined to be the most appropriate analysis model based on Schwarz’s Bayesian Information Criterion and Akaike’s Information Criterion.

Results

The results indicated that at the provincial level a positive correlation between the proportion of the urban population and tuberculosis health education. Conversely, the medical doctors (per 1000 population) negatively correlated with local tuberculosis health education. Additionally, average annual temperature, precipitation, and SO₂ emissions showed positive correlations with tuberculosis health education. At the city level, the results were not statistically significant.

Conclusions

The health education regarding tuberculosis among China’s floating population demonstrates significant spatial clustering from 2014 to 2017. Key determinants identified include economic factors, healthcare resources, and environmental influences, which contribute not only at a local level but also through notable spatial spillover effects. It is essential to adopt targeted public health strategies to enhance interregional collaboration and optimize resource allocation for equitable tuberculosis control.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28718-x.

Introduction

Tuberculosis (TB) remains a significant public health challenge globally, particularly in low-income countries where it poses a considerable disease burden. The Global TB Report 2025 indicated that approximately 10.7 million individuals were living with TB worldwide in 2024. and over 1.2 million individuals worldwide succumbed to the disease. Among the 30 countries with a high burden of tuberculosis, China’s estimated number of tuberculosis cases has dropped from the third to the fourth position, and it has entered the category of moderately to low prevalence countries for the first time [1]. Nevertheless, China’s tuberculosis prevention and control efforts continue to encounter serious challenges. The World Health Organization (WHO) and the global community still need to develop optimal strategies to mitigate the impact of the TB epidemic. Tuberculosis continues to pose a threat to global health, especially among immigrants (non-native births) and the poor and vulnerable, who are at higher risk of contracting the disease [2–4]. With the process of urbanization, floating population within China has become common. Migrant population refers to individuals who have left their registered household locations and reside elsewhere. The statistical definition excludes those experiencing separation between residence and registration within the same city district. Separated population, in this context, includes individuals whose place of residence differs from the township or street where their household registration is located, and who have been away from their registered address for six months or more. This represents internal migration within the country. According to the China Development Report 2023 [5], over one-third of the population lacks household registration in their place of residence, and more than one-fourth constitutes the floating population, making large-scale population mobility a significant challenge for social development [6]. At present, the floating population in China primarily migrates in a family-based manner. The proportions of children, women and the elderly within this group are increasing annually, resulting in a sharp rise in demand for health services [7]. With the growing urban migrant population, tuberculosis cases have significantly increased in recent years in cities such as Beijing, Shanghai, Shenzhen, Guangzhou, Wuhan, and Hangzhou in China. However, standardized screening strategies remain lacking, and the underlying epidemiological mechanisms require further investigation [8–11]. China’s health department issued a guiding document in 2014, stating that the core task is to promote the equalization of basic public health and family planning services for the floating population and steadily achieve full coverage of basic public services for the permanent residents in urban areas [12]. However, due to the high mobility of the floating population, it becomes difficult to obtain the same social resources and medical services as local residents. Furthermore, various factors, including poor living and working conditions, pose serious challenges to their physical and mental health, rendering them susceptible to becoming sources of infectious disease transmission [13, 14]. Consequently, the floating population represents a critical demographic for tuberculosis prevention and control in China. Research has shown that both the proportion of migrants within the total population and the growth rate of new migrants significantly influence the rising incidence of TB [15]. Furthermore, the rates of TB are notably higher among migrants compared to the general population [16]. Moreover, the cure rate of TB in migrant patients is considerably lower than that of the registered population [17–19]. As with any infectious disease, this mass migration facilitates the spread of TB [20, 21].

Tuberculosis health education (TB health education) serves as an effective preventive measure to control its prevalence. Studies have shown that the high incidence of TB in certain regions of China can be attributed to inadequate sanitation and lack of knowledge about health prevention of TB [22]. Health education and the management of TB patients are integral components of the nation’s fundamental public health services. These initiatives aim to enhance public understanding of TB prevention strategies, and transmission pathways, while also fostering greater receptiveness to TB-related health services. Additionally, they promote the adoption of healthier lifestyles, thereby improving overall health literacy and reducing the incidence of TB. However, the growing size and increased mobility of the floating population present significant challenges to the implementation of TB health education and management. In contrast to local residents with permanent residence household registration, the floating population exhibits characteristics such as frequent relocations, unstable employment, and high mobility, which hinder their ability to consistently access essential public health services, including TB health education, at a single location. Consequently, this increases the risk of TB transmission. The reasons for the floating population’s unequal access to basic public health services, including TB health education, in their place of residence are multifaceted. These factors not only encompass the mobility characteristics and personal traits of the floating population, such as gender [23–25], ethnicity [26], education level [24–27], migration population proportion [28], occupation [27], and residence [26], but also cover social economy [29–31], the supply of health resources is subject to multiple entities [32] local government [33], population density [34] individual health behaviors [35], family environment [36, 37], health resources [38], natural environmental factors including weather [39–41] and air pollution [40], and other factors such as diabetes [41]. Focusing on a single element cannot effectively address the inadequacy of TB health education services for the floating population. Therefore, the influencing factors of TB health education services for the floating population should be explored from multiple aspects and perspectives to improve the level of health education and reduce the incidence of TB among this population.

Many developing countries utilize Geographic Information Systems (GIS) and spatial statistics to analyze the spatial and temporal distribution of TB [42, 43]. The GIS provides more possibilities for analyzing TB health education indicators. Many spatial approaches, such as spatial global and local autocorrelation and spatial regression methods, have unique advantages in mining spatial properties and have been widely used in the field of public health in recent years [44, 45]. Several studies have identified that public health studies often reveal spatial statistical significance in distribution patterns, highlighting a phenomenon of spatial aggregation. Concurrently, there are significant spatial and regional differences in the prevalence of epidemics [46]. Spatial correlation analysis of TB can assist health or infectious disease managers identify high-risk areas and populations that require immediate intervention [43]. Identifying the factors that influence TB health education and exploring the spatial effects of these indicators can enable health managers to implement targeted interventions that are particularly crucial in resource-constrained areas.

This study established a research framework for the factors influencing TB health education and analyzed the status of TB health education in China from 2014 to 2017 from a spatial perspective. Global and local spatial autocorrelation analyses were performed to explore the spatial relationship between health education and TB prevention. Subsequently, a spatial lag model was developed to investigate the key factors restricting TB health education, examining both direct and indirect effects, thereby providing a reference for alleviating the high incidence of TB and formulating effective TB prevention and control measures. This study is the first spatial study of the influencing factors related to TB health education among the floating population in China. It explores the necessity of tuberculosis health education from a spatial perspective. Tuberculosis requires the attention of the health department and the public. Actively carry out tuberculosis health education services for the floating population to help prevent the spread of tuberculosis.

Materials and methods

Data source

This study covered 31 provincial-level administrative divisions, excluding the Hong Kong Special Administrative Region, Macau Special Administrative Region and Taiwan Province. Spatial geographic data were sourced from the geospatial data cloud, which provides vector maps depicting the fundamental elements of Chinese administration. Floating population data were derived from the China Migrants Dynamic Survey (CMDS) from 2014 to 2017. The CMDS comprises data from the National Health Commission’s annual large-scale sampling survey of the floating population, covering 31 provinces (including autonomous regions and municipalities directly under the central government). The contents cover basic information about the floating population and their family members; the range and trend of the floating population; employment; social security; income and expenditure; housing; basic public health services; marriage; family planning service management; children’s mobility and education; and psychological culture. Economic, population, medical resources, and environmental data were also sourced from the China Statistical Yearbook. In this study, the number of TB health education programs from 2014 to 2017 was analyzed to assess the status of TB health education among the floating population and the spatial disparities in influencing factors.

Measurement of TB health education

Tuberculosis health education for the floating population serves as the primary explained variable in this study. It was selected from the fourth part “Health and Public Services” of the China Floating Population Dynamics Survey (CMDS) data, which inquires “Have you received health education in the following aspects in your current place of residence in the past year?” For this issue, multiple choices are available. The options include a total of 12 health service items, among which TB health education is featured. Among the floating population, those who have received TB health education are assigned a value of 1, and those who have not are assigned a value of 0.

Influencing indicators

We conducted a comprehensive search of PubMed, Web of Science, and the China National Knowledge Infrastructure database prior to February 2025. Utilizing the “China Statistical Yearbook” and the migrant population database, we identified four primary factors influencing the acceptance of tuberculosis health education among the migrant population: socio-economic development, medical resources, population characteristics, and environmental and meteorological factors. Notably, meteorological and environmental factors are regarded as potential confounding variables. Subsequently, we summarized the indicators and proxy variables that may impact tuberculosis health education at the provincial level in China. Ultimately, we collected data for 11 indicators, identifying the main influencing factors as follows.

1. Economic indicators included the proportion of the urban population at year-end and urban road lights (thousands), which were obtained from the the Statistical Yearbook covering the years 2014 to 2017.

2. Sociodemographic indicators included population density [47, 48] and illiteracy rate, which were also obtained from the Statistical Yearbook covering the years 2014 to 2017.

3. Healthcare resource indicators included the number of medical doctors per 1,000 population, which is a commonly used proxy in domestic research. Additionally, TB incidence and mortality rates were obtained from the Statistical Yearbook covering the years 2014 to 2017.

4. Environmental and meteorological indicators included smoke (powder) dust emission, sulfur dioxide (SO₂), and average annual temperature (°C) and average annual precipitation, which were sourced from the National Meteorological Information Center.

A spatial database was created in which GIS layers of spatial environmental data were recorded using ArcGIS version 10.8.2 software. An index frame diagram was constructed (Fig. 1).

Fig. 1.

Tuberculosis health education and its influencing factors

Framework

The procedure consisted of three steps. First, spatial global and local autocorrelation methods were used to detect spatial patterns and cluster the data. If a spatial clustering pattern exists in the TB health education model for a floating population, a spatial regression model must be utilized. Second, we conducted a spatial regression analysis to examine the influencing factors and TB health education. A collinearity test was performed. Finally, we utilized spatial regression to analyze the effects of various indicators of TB health education among the floating population and explored their direct and indirect effects.

Step 1: data processing and spatial clustering analysis

Global autocorrelation analysis describes the distribution characteristics of attribute values across an entire region [46]. Moran’s I was employed to determine spatial aggregation within the region. The values of Moran’s I range from − 1 to 1. Moran’s I > 0 represents positive spatial correlation; Moran’s I < 0 represents negative spatial correlation; and Moran’s I = 0 indicates that the entire distribution is random. We visualized the spatial patterns of TB health education, which were classified into five classes using the “natural breaks” method [49].

The global indicator can obscure the non-stationarity of local states [50]. Local indicators are employed to detect spatial autocorrelation. Local autocorrelation analysis can determine whether there is an aggregation between each region and its adjacent regions. The LISA aggregation diagram illustrates four types of spatial aggregation: high-high, low-low, high-low, and low-high. High-high clustering indicates an area with a high level of TB health education is surrounded by an area with a high TB health education level; low-low clustering indicates an area with a low level of TB health education is surrounded by an area with a low TB health education level; high-low clustering indicates an area with a high level of TB health education is surrounded by an area with a low TB health education level; and low-high clustering indicates that an area with a low level of TB health education is surrounded by an area with a high TB health education level [51]. All analyses were conducted using ArcGIS 10.8.2.

Step 2: influencing indicators extraction

No multicollinearity was observed among the influencing factors. Stata17.0 was utilized to assess the inflation factor. A variance inflation factor (VIF) < 10 indicated no serious multicollinearity problems for the independent variables and covariates [52].

Step 3a: spatial regression

We employed panel data models to investigate the association between TB health education and each indicator in China from 2014 to 2017. We used spatial regression to consider the spatial impact of each indicator on TB health education. There are two commonly used spatial regression models: spatial lag regression in Eq. 1 and spatial error regression in Eq. 2 [53].

1

2

where X is the influencing indicator; Y is the number of TB health education programs accepted by each province; β is the coefficient for each indicator; e is the error term; WY is the spatial weight matrix for the dependent variable; We is the spatial weight matrix for the error term; λ is the spatial autoregressive coefficient; and u is the scalar variable(s).

We conducted the Lagrange multiplier (LM) and the robust Lagrange multiplier (Robust LM) test to ensure the rationality of the spatial panel data model construction. The LM and Robust LM tests were employed to ascertain the type of spatial regression by determining which LM values in the regression models exhibited greater significance [54]. We performed the Hausman test to identify fixed and random effects. The Akaike Information Criterion (AIC) and Schwarz’s Bayesian Information Criterion (SBIC) were also applied for model selection [53]. The smaller the AIC and BIC statistics, the better the model. Spatial regression modeling was executed using Stata software. The spatial weight matrix was constructed using Stata’s spatial matrix commands (spmat), based on rook contiguity relationships between geographical units. The matrix was row-normalized to ensure stationarity in the spatial autoregressive process.

Step 3b: effect decomposition

We selected the spatial lag model as the spatial regression model. To narrow down the differences in the level of TB health education among provinces and promote equity in health services, we studied the direct, indirect, and total effects of each index. The direct effects represented local impacts, and the indirect effects represented the impacts on neighbors. We utilized spatial spillover effects to elucidate the spatial influence of a certain factor on a neighboring region. Specifically, spatial spillover effects include a direct impact on TB health education from the unit itself and an indirect impact on TB health education due to geographic proximity, eventually affecting the unit in reverse.

Results

TB health education in 31 provinces of China from 2014 to 2017

This study concentrated on the annual number and trend of TB health education recipients, without emphasizing the continuity of TB health education among the follow-up individuals. The average number of people surveyed during the 4 years was 186,482, and the number of people receiving TB health education was 48,623. Xinjiang, Guangdong, Sichuan, Chongqing, and Jiangsu were the top five provinces in terms of average annual TB health education (Table 1).

Table 1.

TB health education prevention and control among the floating population from 2014 to 2017

Year	Total number of respondents	Area of coverage	TB health education number
2014	200,937	31	48,279
2015	206,000	31	77,460
2016	169,000	31	48,674
2017	169,989	31	20,078
Annual average	186,482	31	48,623

TB health education exhibited significant variation across the 31 provinces of China. The use of darker colors in the accompanying map indicates higher average levels of TB health education. Notably, the central and western regions of China, a larger proportion of the migrant population has received tuberculosis health education, indicating a higher level of tuberculosis health education in these areas, while the level is relatively lower in the eastern regions.Xinjiang (6.30%), Guangdong (5.30%), and Sichuan (4.96%) reported the highest levels of TB health education. TB health education levels were low in Hebei (1.20%), Anhui (1.70%), Liaoning (1.74%), Jilin (1.82%), and Shanxi (1.84%) (Fig. 2(a)).

Fig. 2.

a Average distribution of TB health education in China from 2014 to 2017. b TB health education among the floating population from 2014 to 2017

From 2014 to 2017, the TB health education of all provinces in 2015 improved (except for Heilongjiang and Xinjiang), and the TB health education of Guangdong in 2015 was the highest (5,219) (Fig. 2(b)). The year with the lowest TB health education was 2017, and the lowest TB health education was in Shanghai (75). Xinjiang consistently ranked highest among all provinces from 2014 to 2017(except for 2015). The blue color dominating the 2017 column indicates low TB health education values across most provinces in that year (Fig. 3).

Fig. 3.

Heat map of TB health education for different provinces in China, 2014–2017

Spatial autocorrelation

The results of the global spatial autocorrelation analysis showed that TB health education had no spatial autocorrelation in 2014 (Moran’s I > 0, P > 0.05), indicating no obvious spatial aggregation. TB health education had a positive spatial autocorrelation (Moran’s I > 0, P < 0.05), indicating that TB health education had spatial aggregation from 2015 to 2017 (Table 2).

Table 2.

Global autocorrelation of TB health education prevention and control among the floating population from 2014 to 2017

Year	Moran’s I	Z value	P value
2014	0.180	1.743	0.081
2015	0.217	2.062	0.039
2016	0.291	2.691	0.007
2017	0.440	4.170	0.000

Local spatial clustering for TB health education

A local autocorrelation analysis was conducted on the acceptance of TB health education across 31 provinces and autonomous regions in China, resulting in the generation of a LISA cluster map. Two regions with agglomerates were identified. There were 8 high-high clusters, including Qinghai, Guizhou, Hunan, Guangxi, Gansu, Yunnan, Xizang, and Sichuan. Conversely, nine low-low clusters were identified in Inner Mongolia, Hebei, Liaoning, Beijing, Tianjin, Shanxi, Shandong, Jiangsu, and Anhui (Fig. 4). Since 2014, the high-high clustering has notably shifted southward and westward. The low-high cluster areas from 2014 to 2017 included the provinces of Xizang, Guizhou, Jiangxi, Shanghai, Ningxia and Gansu. During this period, the high- and low-clustering provinces included Henan and Beijing. Notably, Henan Province consistently appeared in the high-low cluster from 2014 to 2017. The provinces represented by the low-high and high-low clusters indicate areas of TB health education transition, warranting attention due to their proximity to provinces with high or low TB health education levels. Agglomerates classified with high-low clusters present a high TB health education level and can affect neighboring provinces with a low TB health education level. Provinces characterized by significant high-low clustering have policy significance, which can improve TB health education in other provinces. In addition, provinces with low-high clustering refer to provinces with low TB health education levels surrounded by provinces with high TB health education levels. Attention should be paid to provinces with low TB health education levels to explore the underlying reasons why they are close to provinces with high TB health education but low TB health education acceptance levels.

Fig. 4.

Local spatial autocorrelation analysis of TB health education among the floating population from 2014 to 2017

The variance inflation factor (VIF) analysis indicated that multicollinearity was not a critical concern in the model, as all VIF values were well below the threshold of 10 (ranging from 1.27 to 5.05), with a mean VIF of 3.19. Consequently, all variables were retained for the analysis. We performed model tests (including the LM, LM error test, Robust LM, and Robust LM error test). We concluded that the spatial lag model was the most suitable (Table 3). The Hausman test (Hausman test statistics = 55.11, P < 0.001) proved the rationality of using a spatial fixed effects model. Model (2) was the best, and we chose the spatial lag fixed (individual) effects model (Table 4). Subsequently, the model of the difference specification of TB health education was evaluated using the AIC and BIC. The results of TB health education (2), TB health education (3), and TB health education (4) were similar, but the spatial regression of the TB health education (2) model was significant; therefore, we finally chose TB health education (2) as the research model.

Table 3.

Lagrange multiplier test results

Test	Statistic	p-value
Spatial error:
Moran’s I	7.547	0.000
Lagrange multiplier	52.764	0.000
Robust Lagrange multiplier	2.215	0.137
Spatial lag:
Lagrange multiplier	56.633	0.000
Robust Lagrange multiplier	6.084	0.014

Table 4.

TB health education 2014 − 2017: estimated from spatial panel data models

	Model 1	Model 2	Model 3	Model 4
Main
Population density	0.023	0.009	0.033	0.011
	(0.028)	(0.024)	(0.026)	(0.022)
Illiteracy	9.659	70.278	-20.799*	5.924
	(17.817)	(50.795)	(11.202)	(47.650)
Smoke dust emission	-13.323***	-17.031***	-14.943***	-14.722***
	(4.094)	(5.155)	(2.844)	(4.792)
Urban Road Lights (thousands)	0.111	-1.331**	0.365***	-1.709***
	(0.141)	(0.589)	(0.092)	(0.541)
Proportion of the urban population at year-end	27.763**	130.183***	-3.703	97.700***
	(12.648)	(31.349)	(8.259)	(30.744)
SO2	18.195***	21.519***	8.935***	13.742***
	(4.235)	(5.195)	(3.118)	(5.078)
Medical doctors (per 1000 population)	-52.925	-1437.935***	348.910*	-1533.999**
	(272.969)	(426.530)	(179.351)	(697.797)
Average Annual Temperature	2.233***	5.400***	1.241***	3.705***
	(0.687)	(1.385)	(0.477)	(1.407)
Average Annual Precipitation	-0.005	0.898***	-0.032	0.589*
	(0.175)	(0.307)	(0.117)	(0.321)
TB incidence	13.955***	9.973	10.568***	9.721
	(4.070)	(9.442)	(2.847)	(8.549)
TB mortality	-323.850	-840.147	185.188	-804.201
	(514.882)	(594.162)	(379.198)	(554.109)
_cons	-2331.557***
	(897.106)
Spatial rho	0.616***	0.550***	-0.051	0.028
	(0.065)	(0.064)	(0.114)	(0.117)
R2	0.342	0.057	0.174	0.001
AIC	-5930	-5934	-5934	-5934
BIC	-5887.696	-5897.336	-5897.336	-5897.336

Standard errors are indicated in parentheses; AIC Akaike Information Criterion, SBIC Schwar’s Bayesian Information Criterion; (1) Random effects model, (2) individual fixed effects model, (3) time fixed effects, and (4) double fixed effects

^* P < 0.1, ^** P < 0.05, ^*** P < 0.01

The first TB health education determinant component is the economic development dimension; the proportion of the urban population at year-end (130.18, confidence interval [CI]: 68.74, 191.63) had a significant positive effect on TB health education, and Urban road lights (thousands) (-1.33, CI: -2.49, -0.18) had a significant negative effect on TB health education. In terms of health resources, medical doctors (per 10 000 population) (-1437.94, CI: --2273.92, -601.95) had significant negative effect on TB health education. In the environmental and meteorological dimensions, smoke dust emission (-17.03, CI: -27.14, -6.93) had significant negative effect on TB health education, while both the average annual temperature(5.40, CI: 2.69, 8.12) and average annual precipitation (0.90, CI: 0.30, 1.50) of each province had significant positive effect on TB health education.

Based on the Model 2 estimation, we conducted direct/indirect effects analyses to interpret how the factors influence TB health education (Table 5). For direct effects, smoke dust emission (-18.19, CI: -28.77, -7.61), urban road lights (thousands) (-1.47, CI: -2.71, -0.23), proportion of the urban population at year-end (143.79, CI: 78.96, 208.62), SO₂ emissions (23.39, CI: 13.22, 33.57), medical doctors (per 10 000 population) (-1572.28, CI: -2495.29, -649.28), average annual temperature (5.90, CI: 3.14, 8.66), and average annual precipitation (1.02, CI: 0.38, 1.66) had significant spatial differences in TB health education among provinces. For indirect effects, smoke dust emission (-19.32, CI: -33.53, -5.11), proportion of the urban population at year-end (154.22, CI: 45.45, 262.99), SO₂ emissions (24.78, CI: 9.81, 39.74), medical doctors (per 1000 population) (-1656.64, CI: -2864.17, -449.11), average annual temperature (6.33, CI: 1.85, 10.80), and average annual precipitation (1.08, CI: 0.22, 1.95) had significant spatial differences in TB health education among provinces; the above indicators had an impact not only on the local TB health education of each province but also on the TB health education of other neighboring provinces.

Table 5.

Direct and indirect effects decomposition of TB health education associated with determining factors in China, TB health education 2014–2017

Variable	Coef.	Standard error	z	P > z	Confidence interval
Direct effect
Population density	0.011	0.027	0.400	0.687	-0.042	0.063
illiteracy rate	74.781	53.448	1.400	0.162	-29.974	179.536
Smoke dust emission	-18.193	5.398	-3.370	0.001	-28.773	-7.613
Urban road lights (thousands)	-1.471	0.631	-2.330	0.020	-2.708	-0.234
Proportion of the urban population at year-end	143.792	33.076	4.350	0.000	78.964	208.619
SO2	23.392	5.191	4.510	0.000	13.217	33.566
Medical doctors (per 1000 population)	-1572.284	470.928	-3.340	0.001	-2495.287	-649.282
Average annual temperature	5.903	1.408	4.190	0.000	3.143	8.663
Average annual precipitation	1.018	0.325	3.130	0.002	0.381	1.655
TB incidence	10.719	10.189	1.050	0.293	-9.251	30.689
TB mortality	-937.564	637.503	-1.470	0.141	-2187.047	311.918
Indirect effect
Population density	0.012	0.030	0.400	0.689	-0.046	0.070
illiteracy rate	75.932	57.290	1.330	0.185	-36.353	188.218
Smoke dust emission	-19.320	7.250	-2.660	0.008	-33.529	-5.110
Urban road lights (thousands)	-1.576	0.812	-1.940	0.052	-3.168	0.015
Proportion of the urban population at year-end	154.222	55.497	2.780	0.005	45.449	262.995
SO2	24.777	7.635	3.250	0.001	9.812	39.741
Medical doctors (per 1000 population)	-1656.642	616.098	-2.690	0.007	-2864.173	-449.111
Average annual temperature	6.326	2.285	2.770	0.006	1.848	10.805
Average annual precipitation	1.082	0.441	2.450	0.014	0.218	1.946
TB incidence	11.532	11.346	1.020	0.309	-10.706	33.771
TB mortality	-988.381	741.949	-1.330	0.183	-2442.573	465.812
Total effect
Population density	0.023	0.056	0.400	0.686	-0.087	0.132
illiteracy rate	150.713	108.739	1.390	0.166	-62.412	363.838
Smoke dust emission	-37.513	11.881	-3.160	0.002	-60.799	-14.226
Urban road lights (thousands)	-3.047	1.392	-2.190	0.029	-5.775	-0.319
Proportion of the urban population at year-end	298.013	82.756	3.600	0.000	135.816	460.211
SO2	48.168	11.605	4.150	0.000	25.424	70.913
Medical doctors (per 1000 population)	-3228.927	1017.538	-3.170	0.002	-5223.264	-1234.589
Average annual temperature	12.229	3.451	3.540	0.000	5.466	18.992
Average annual precipitation	2.100	0.724	2.900	0.004	0.680	3.519
TB incidence	22.252	21.226	1.050	0.294	-19.351	63.854
TB mortality	-1925.945	1353.239	-1.420	0.155	-4578.244	726.354

Supplementary analysis at the prefecture-level city scale

To address potential concerns regarding within-province heterogeneity, we conducted a supplementary city-level fixed-effects regression utilizing prefecture-level panel data from 16 provinces and 208 cities, encompassing 832 observations from 2014 to 2017. The dependent variable was the number of individuals from the floating population receiving tuberculosis (TB) health education, while the independent variable was GDP per capita. The model incorporated city fixed effects, year fixed effects, and province-year fixed effects to control for all time-varying unobservables at the provincial level (e.g., policy shifts, fiscal allocations). Robust standard errors were clustered at the city level.

As demonstrated in Appendix Table S1, the coefficient for GDP per capita is positive but not statistically significant (β = 0.736, p = 0.147), with an extremely low within-R² value of 0.0032. To further test whether this null finding is driven by provincial capitals, we repeated the analysis excluding all capital and sub-provincial cities; the coefficient remained virtually unchanged (β = 0.745, p = 0.155). This finding indicates that, after controlling for provincial-level macro influences, economic disparities within the same province do not significantly account for variations in TB health education coverage across cities. Consequently, the primary determinants of disparities in TB health education appear to function at the provincial level and through cross-provincial spatial spillover effects. This further substantiates our choice of the provincial-level spatial lag model as the primary analytical framework.

Discussion

This study employed a spatial lag model to explore, for the first time at the provincial level, the influencing factors and spatial spillover effects of TB health education among China’s floating population. Our analysis revealed a significant spatial dependency of TB health education across Chinese provinces and identified that socioeconomic, health resource, and environmental factors not only have a direct impact on local TB health education levels but also exert indirect spatial spillover effects on neighboring regions.

Spatial patterns and policy implications

The results of global and local spatial autocorrelation analyses indicated that from 2015 to 2017, TB health education was not randomly distributed but exhibited distinct spatial clustering characteristics. High-high clusters are predominantly located in the southwestern and northwestern provinces, while low-low clusters are concentrated in parts of northern and eastern China. This spatial pattern suggested that the effectiveness of tuberculosis health education demonstrated regional interconnectedness. The TB health education level of a province could be influenced by the policies and practices of neighboring provinces. For instance, Henan, identified as a persistent high-low outlier province, might be significantly impacted by its status as a major province for tuberculosis. As a populous province and a significant labor exporter, its response to national prevention and control policies targeting key populations, such as migrant workers and students, had directly resulted in large-scale health education campaigns. However, it did not effectively disseminate its high TB health education level to surrounding provinces. This underscored the importance of enhancing inter-regional cooperation and sharing experiences in formulating prevention and control strategies [55, 56]. Such collaboration is essential for dismantling administrative barriers and achieving synergistic effects in public health initiatives.

Drivers and spillovers of socioeconomic factors

The study found that the proportion of the urban population had a significant positive direct effect and spatial spillover effect on TB health education. This finding confirmed that urbanization, a critical indicator of socio-economic development, was typically accompanied by more comprehensive public service systems and higher health investments [57]. In regions with elevated levels of urbanization, the floating population was more likely to integrate into the local public health service network, thereby enhancing the accessibility of TB health education. More importantly, this positive impact tended to diffuse to neighboring provinces, likely stemming from the policy demonstration effect or resource spillover from regions with higher economic development levels [58], prompting adjacent areas to increase their investment in public health.

The “treatment over prevention” tendency in health resource allocation

A counterintuitive yet crucial finding was that the medical doctors (per 1000 population) showed a significant negative correlation with local TB health education levels and also exhibited a negative spatial spillover effect on neighboring regions. This discovery contradicted the conventional wisdom that “more resources lead to better services”; however, it profoundly revealed the long-standing structural contradiction in China’s healthcare system, which emphasized treatment over prevention [59, 60]. Under fiscal decentralization and a GDP-centric evaluation system, local governments might tend to allocate resources to medical infrastructure construction that can quickly demonstrate political achievements [61], rather than to slow-yielding public health projects such as health education. Provinces with abundant medical resources might have a more mobile population, increasing the difficulty of covering health education. Meanwhile, this distortion in resource allocation generated negative externalities, as neighboring provinces might emulate this model under competitive pressure, further exacerbating the imbalance in regional basic public health services.

The “forcing mechanism” of environmental risks

The positive correlation between average annual temperature, precipitation, SO₂ emissions, and tuberculosis health education among the floating population might reveal that environmental risks exacerbated this group’s health vulnerability, thereby generating stronger demand for health education. This interpretation is supported by established biological evidence: Exposure to SO₂ and smoke dust has been demonstrated to impair respiratory host defense. SO₂ can damage the respiratory mucosal epithelium and weaken the mucociliary clearance of the respiratory tract, which serves as the first line of defense against Mycobacterium tuberculosis (MTB). Furthermore, smoke dust particles can penetrate deep into the alveoli, reducing the phagocytic activity of alveolar macrophages and compromising T-cell mediated immunity, thereby increasing susceptibility to MTB infection and progression to active tuberculosis disease.Behaviorally, the Health Belief Model provides a framework for understanding the observed association [62–64]. The floating population were often concentrated in outdoor or poorly ventilated industries such as construction, manufacturing, and sanitation, where their working and living environments directly exposed them to extreme temperatures, precipitation and air pollution, posing significantly higher health risks compared to general residents [65, 66]. Consequently, their direct experiences with these environmental factors intensified concerns about respiratory health issues, particularly anxiety regarding susceptibility to infectious diseases like tuberculosis. This anxiety motivated them to actively seek and accept health education. Additionally, adverse environmental conditions created a perceived threat that enhanced the relevance and effectiveness of health education, making the floating population more inclined to adopt protective measures. This demand for health education, driven by environmental risks in a pressure-induced manner, compelled public health service providers to intensify health education interventions for this key demographic in high-risk areas. Such effects also exhibited spatial spillover [67]; enhanced health education measures for migrants in provinces with elevated environmental risks can serve as a model for neighboring provinces facing similar risks and demographic structures, thereby fostering coordinated intervention efforts.

Pathways of spatial spillover effects

Beyond identifying the existence of spatial spillovers, it is essential to elucidate the specific pathways through which these effects operate. Drawing upon the mechanisms of interregional economic interaction, public policy diffusion, and population mobility, we delineate three primary transmission channels underlying the spatial spillovers observed in our model. First, concerning economic factor flows, the positive spatial spillover of the urbanization rate is likely realized through two interconnected sub-pathways. At the macro level, industrial radiation from highly urbanized provinces stimulates economic growth and fiscal revenue in adjacent provinces, thereby expanding their fiscal space for public health investment [68, 69]. At the micro level, the enhancement of cross-regional transportation and information infrastructure reduces the physical and transactional costs of disseminating health education resources—such as training materials, personnel expertise, and program protocols—across provincial boundaries, thereby facilitating the spatial diffusion of effective health education models [70]. Second, intergovernmental policy diffusion: In the context of China’s fiscal decentralization, provincial health policies interact through mechanisms of intergovernmental learning, competition, and emulation [71]. The negative spatial spillover associated with the physician density variable can be interpreted through the “resource siphon” effect, where specialized medical personnel tend to concentrate in economically advanced provinces, thereby weakening the health education service capacity of neighboring regions. Concurrently, competitive pressures among provincial governments regarding health resource allocation may lead to imbalanced efficiency gains, further shaping the spatial configuration of tuberculosis (TB) health education [61]. Third, population migration and disease transmission: Environmental factors exert spatial spillovers through the cross-provincial mobility of the floating population and the transboundary spread of TB [21]. Climate conditions and air quality influence both the direction and magnitude of population flows, thus altering the demographic composition and health education needs of the floating population in adjacent provinces [72]. Given that TB is transmitted via airborne droplets, its transmission dynamics are not confined by administrative boundaries; environmental conditions in one province that facilitate TB transmission may pose imported risks to neighboring provinces through population movements. This perceived threat incentivizes neighboring provincial governments to intensify health education interventions targeting the floating population [73].

Research limitations

This study acknowledged several limitations. Firstly, the CMDS data were only updated until 2017, which meant they did not account for recent policy changes. Furthermore, the model excluded critical variables, such as provincial-level special funding for tuberculosis prevention and control, as well as the capacity of primary healthcare services. These omissions may have impacted the accuracy of the estimation results.

Conclusions

In conclusion, tuberculosis health education among China’s floating population exhibits significant spatial clustering and regional disparities from 2014 to 2017. Socioeconomic factors, healthcare resource, and environmental conditions not only exert direct local effects but also demonstrate notable spatial spillover effects on neighboring regions. These findings underscore the importance of adopting a regionally coordinated and spatially sensitive approach in designing and implementing TB health education programs. Future interventions should aim to enhance cross-regional collaboration, optimize resource distribution, and tailor health education strategies to local spatial and socioeconomic contexts. Such efforts will contribute to reducing TB incidence and advancing health equity for China’s floating population.

Authors’ contributions

Conceptualized the study: YaJie Feng, Lihan Zhang, Junping Liu, Weijian Song, Juan Guan, Yunkai Li, Wei Liu, Yanfu Wang, Libo Liang. Methodology: YaJie Feng, Lihan Zhang, Junping Liu, Liang Li-Bo. Supervised data collection: YaJie Feng, Lihan Zhang. Directed data analysis: YaJie Feng. Interpreted the findings: YaJie Feng, Liang Li-Bo, Yanfu Wang. Drafted the manuscript: Yajie Feng, Lihan Zhang. All authors reviewed the manuscript and approved the final version.

Funding

The authors would like to thank the National Natural Science Foundation of China (72274047,72574057) for providing funding.

Data availability

The data include tuberculosis health education data across each province and spatial data of China. Spatial geographic data were obtained from the geospatial data cloud (URL: https://www.gscloud.cn/) floating population data were derived from the China Migrants Dynamic Survey (CMDS) from 2014 to 2017 (URL: https://www.ncmi.cn/). The meteorology data are available in the China Meteorological Data Network (URL: https://data.cma.cn/) and other data can be obtained from the China Statistical Yearbook (URL: https://www.stats.gov.cn/). The data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.