The effects of workplace health education on smoking-related behaviour and unequal gains by job position in China: ABWMC programme ndings

Objective: Although the Chinese government has introduced a series of regulations to promote tobacco-related health education in workplaces, the implementation impact has been far from satisfactory. The aim of the present study was to explore the effects of company level tobacco-related health education and the potential unequal gain across job positions. Methods: Data from the 2018 Asia Best Workplace Mainland China programme were employed to address these aims. This was a cross-sectional study that included 14195 employees from 79 companies in mainland China. Spearman correlation tests were used to examine unadjusted correlation between the study variables, and binary logistic regression was used for multivariable analysis. Results: Tobacco-related health education was associated with better smoking harm awareness, lower second-hand smoke exposure, better perceived workplace environment and positive health information-seeking behaviour. Job position interacted with health education, suggesting that positive effects of health education were smaller for general employees than employees who held an administrative position. Conclusions: We conclude that workplace tobacco-related health education was not only associated with tobacco control effects but also had spillover effects, which were signicant for higher-ranking employees. Policy makers should recognize and reduce the potential health disparities. working environment and health effects of health


Introduction
Smoking is one of the main risk factors driving the growing epidemic of noncommunicable diseases worldwide. Tobacco is on track to claim 200 million lives in China this century, predominately among the poorest and most vulnerable people [1][2] . Although the rate of second-hand smoke (SHS) exposure has gradually decreased in China, workplace SHS exposures remain pervasive (i.e., 54.3% in 2015 to 50.9% in 2018) [3][4] . Company level tobacco-related health education for employees has contributed to effective tobacco control or SHS exposure reduction strategies [5][6] .
Since the 1980s, the Chinese government has introduced a series of regulations to reduce SHS in workplaces, including smoking bans in public places and support for companies to conduct health education to employees. However, the implementation impact has been far from satisfactory 5 . Companies may be reluctant to provide health education due to a lack of time and knowledge. There is only a limited number of companies that have a smoke-free policy and that provide tobacco-related health education, resulting in few studies evaluating workplace health education and its impact on employee's smoking behaviour.
Prior to implementing potential robust measures to promote smoking-related health education in workplaces, it may be bene cial for policy makers to have a clear understanding of the impact of such activities in Chinese companies. For example, the most common outcomes assessed following health education is knowledge increase and SHS exposure reduction. However, little is known about how such activities in uence employees, including their perceived workplace environment and regular health information-seeking behaviour.
Another relevant problem is unequal gain from tobacco health education. In western countries, researchers have found that people with different socioeconomic status (SES) but equal health resources have unequal gains. [7][8] For example, one study has reported that African Americans do not gain as much self-rated health as Caucasians with the same resources. 9 Assari and Bazargan found that the protective effects of educational attainment for reducing SHS exposure at work are systemically less for Hispanics than Whites. 10 Some scholars believe that the impact of resources on health outcomes are conditional to factors, such as SES, poverty and residential segregation. [11][12] According to The Minorities' Diminished Returns theory, at least some of the health disparities are due to "less than expected" protective effects, suggesting that population level health disparities are not all due to resource but also due to differential health gain. 9-10 One recent Chinese study has found that higher SES families have better self-reported health (SRH) and fewer activities of daily living (ADL) limitations with the same society resource. 13 However, there are no previously reported studies on tobacco-related health education. Because of the sustained increase in health status disparities in China in the past half century, the difference of population level spillover effects from health education have become increasingly important both for scholars and policy makers. 14 In this paper, we refer to indirect effects of tobacco-related health education beyond reducing SHS exposure and enhancing smoking harm awareness as spillover effects, including the in uence on perceived workplace environment and health information-seeking behaviour.
The aim of the present study was to explore the effects of company level tobacco-related health education and the potential unequal gain across job position. Speci cally, we have the following hypotheses: (1) with company level health education, employees will not only indicate a higher smoking harm awareness and a lower SHS exposure but also show spillover effects about their perceived working environment and health behaviour; and (2) several positive effects of health education are fewer for lower ranking employees than higher ranking employees.

Methods
Design and data Data from the 2018 Asia Best Workplace Mainland China (ABWMC) programme were employed to address these aims. ABWMC is an academic/company partnership programme that aims to support companies in building a healthy workplace. The ABWMC programme was designed by Peking University and organized by the American International Assurance Co. All companies may voluntarily join the programme and are free to withdraw. The inclusion criteria were as follows: (1) registered legal companies in China; (2) agreement to participate in the programme; and (3) at least 100 workers who are full-time employees. We used information from baseline employee questionnaires. The inclusion criteria for participating employees were as follows: (1) aged 18 years old or above; (2) full-time employees; and (3) agreement to participate in the programme.
The human resource departments of each company delivered the questionnaires to all employees. When rst opening the link, content related to informed consent was shown, and employees were able to choose whether to complete the questionnaire or quit. If the employees submitted the questionnaire through the link, we considered that they agreed to participate. The self-check function of the online survey system automatically identi ed missing data, logical errors and illegal characters.
Measures of relevance

Smoking harm awareness
Smoking harm awareness was measured by the following question: 'Do you think smoking can cause any of the following diseases? A: stroke, B: heart disease, C: lung cancer, D: cardiovascular disease, E: chronic obstructive pulmonary disease, F: asthma or G: I don't know.' Only the participants who chose all answers from A to F were classi ed as having smoking harm awareness.

SHS exposure
In the survey, the participants were asked the following question: 'How many days a week do you usually suffer from SHS exposure at workplace for more than 15 minutes? A: almost every day, B: 4-6 days, C:1-3 days or D: never'. Only the participants who chose D were classi ed as having no SHS exposure.

Tobacco-related health education
We de ned tobacco-related health education as follows: (1) organized by company level; (2) all employees have opportunities to join; and (3) the contents should be related to tobacco control or smoking cessation. This de nition was explained to the respondents when conducted the survey. We further classify all respondents into three categories: have tobacco-related health education and also presence such activities (both of the questions answered 'Yes'=2); have tobacco-related health education but absence such activities ( rst question answered 'Yes', second answered 'No'=1 ); without tobacco-related health education (Otherwise=0)

Perceived workplace environment
There were two variables in this study. The rst variable was for the employees to believe that they work in a healthy environment, and the second variable was for the employees to believe that the company policy protects health. For the rst variable, participants were asked the following question: 'Do you think your working environment is healthy? A: I totally agree, B: I Agree, C: Just ok, D: I do not agree, or E: I totally disagree. Only the participants who chose A and B were classi ed as believing that they work in a healthy environment.
For the second variable, participants were asked the following question: 'Do you think your company's policy can protect your health? A: I totally agree, B: I Agree, C: Just ok, D: I do not agree, or E: I totally disagree. Only the participants who chose A and B were classi ed as believing that their workplace policy protects the health of employees.

5.Health information-seeking behaviour
In the survey, the participants were asked the following question: 'How often do you search for health knowledge? A: Always, B: Very often, C: Sometimes, D: Occasionally, or E: Never. Only the participants who chose A and B were classi ed as having such behaviour regularly.

Other covariates
We controlled for several variables of individual characteristics, such as gender, age, marital status, education, ethnicity and job position. For job position, we further classi ed all employees into two categories as follows: administrative employees (participants with administrative rank) and nonadministrative employees (participants without administrative rank).

Data analytical plan
Our data have a hierarchical structure, therefore we rstly try to use multilevel analyses by setting individual-level and company-level factors. This type of analysis will take into account the fact that workers' response are correlated within companies. There are four standardized models should be run (Null model, Random coe cients regression model, Intercepts as model; Slopes as outcomes model)" However, when we nished the Null model, we nd intraclass correlation coe cient (ICC) is too low (lower than 0.059), is 0.051, indicating that only about 5.1% of the total variation on SHS exposure was attributable to differences between companies/clusters. 15 In other word, we can use usual method to perform analyses. Therefore, we use logistic regression for our statistic Our data analysis was conducted in three steps. First, we examined the distribution of the categorical and continuous variables. Second, Spearman correlation tests were used to examine unadjusted correlation between study variables. Third, we performed binary logistic regression for multivariable analysis.
The dependent variables included the smoking-related variables (smoking harm awareness and SHS exposure), working environment variable (perceived workplace environment) or health information-seeking behaviour. The explanatory variable indicated if the company provided tobaccorelated health education (yes=1 or otherwise=0).
To examine whether there was an interaction effect between job position and health education, we further conducted regression analysis using two models. Model 1 only entered the main effects of health education, job position and covariates. Model 2 also added an interaction term between job position and health education.
We used SPSS 24.0 SPSS Inc, Beijing, China statistical software to conduct all analyses.

Ethics
All participants were informed that the research team would analyze the data anonymously. This study was approved by the Peking University (ethical approval number: IRB00001052-18055).

Descriptive statistics
The total number of participants was 14195 employees from 79 companies in mainland China. The companies included 51.9% private companies, 32.9% foreign companies, 7.6% state-owned companies, 6.3% joint ventures, and 1.3% other companies. The respondents included 5802 (40.9%) employees who reported working in companies that have tobacco-related health education and 8393 (59.1%) employees who reported working in companies without such activities. Among all the respondents who reported that have health education, 2317(39.9%) reported they did not participant in such activities. Table 1 shows the descriptive statistics of the overall sample.  Figure 1 shows the relationship between tobacco-related health education and other key characteristics. Employees in companies with health education reported lower proportion of SHS exposure, higher proportion of smoking harm awareness, higher proportion of perceived safe workplace environment and more positive health information-seeking behaviour.
Bivariate analysis Table 2 shows bivariate correlations between the study variables. Health education was positively correlated with smoking harm awareness, perceived safe workplace environment and health information-seeking behaviour. Moreover, health education was inversely correlated with SHS exposure. Education attainment was also associated with the positive effects, except SHS exposure. Multivariable models (outcome: smoking harm awareness and SHS exposure) Table 3 presents the summary of the results for both logistic regression models with health education as the independent variable as well as SHS exposure and smoking harm awareness as the dependent variables. Based on Model 1, health education was associated with better smoking harm awareness and lower odds of workplace SHS exposure. No signi cant interactions were identi ed between job position and health education for these tobacco control effects. Note: *p<0.05 Model 1 only entered the main effects of health education, job position and covariates. Model 2 was also added interaction terms between health education and job position. Table 4 presents the summary of the results of both logistic regression models with health education as the independent variable as well as perceived workplace environment and health information-seeking behaviour as the dependent variables. Based on Model 1, heath education was associated with higher odds of perceived healthy workplace environment and health information-seeking behaviours. Model 2 showed signi cant interactions between job position and health education with spillover effects, suggesting that company level tobacco-related health education has larger positive effects on perceived workplace environment and health information-seeking behaviours for administrative employees than general employees, which was evidenced by the odds ratio being greater than 1 for the interaction terms.
In addition, compare with the respondents who absence of company level tobacco related health education, both model show that presence respondents have more signi cate positive in uence on all the outcome variables. Model 1 only entered the main effects of health education, job position and covariates. Model 2 was also added interaction terms between health education and job position

Discussion
The present study evaluated the possible in uence of company level tobacco-related health education, and it showed that tobacco-related health education is not only associated with better tobacco control effects but is also related to some spillover effects. Consistent with predictions, health education was associated with better smoking harm awareness, lower SHS exposure, better perceived workplace environment and positive health information-seeking behaviour. In addition, the interaction between job position and health education moderated the spillover effects.
Change in knowledge is often targeted because it is recognized as fundamental to change health behaviour in various behavioural theories. 16 The demonstrated associations are particularly notable given that the awareness of tobacco harm is a commonly cited barrier to conduct more active tobacco control measures. By addressing this barrier through health education, the implementation of workplace health promotion measures is more practicable. However, the increase in the awareness of tobacco harm suggests that the delivered education session could be an effective method to increase the understanding of the company level smoke-free workplace policy and encourage compliance with the policy.
Although several studies have found that participation in workplace tobacco-related health education has led to improving smoking harm awareness and quitting intention 17 , the possible association in the present study were much more comprehensive, resulting in positive change of health behaviour and health attitude. Therefore, as this study indicated, implementation of a workplace health education could be an opportunity to improve the health of employees via multiple levels.
Although health education has overall positive association with employee perception of work environment and health information-seeking behaviour, such associations should not be considered equal between common employees and administrative employees. Some recent studies have found similar patterns for the associations between a wide range of SES indicators and health outcomes. [18][19][20] In the United States, economic resources and psychological assets systemically result in a smaller health gain for some populations, suggesting that the mechanism generating health disparities is more than differential exposures to resources. 8 Our ndings have practical implications. The present study highlights the importance of offering health education from a company level with the optimal method being to incorporate short time health education into routine activities. The lack of health education may translate into a 'missed opportunity' to promote population health in a cost-effective way. The rst step towards the universal adoption of health education provision in workplaces is to educate policy makers and leaders of workplaces. Therefore, the survey report will be disseminated to the participating companies and government agencies, which will aim to encourage those struggling to catch up and provide accessible options and implement other key health education measures.
Moreover, given the existing unequal gain of equal health education, policies that merely focus on the equal distribution of resources and ignore the differential distribution of barriers across groups could be a potential problem. The spillover effects associated with tobacco-related health education were not equal. Further studies should explore other health indicators related to this topic because the ultimate objective of such activities is to promote health and reduce health disparities at the same time. Thus, a related health program should avoid the unintended effect of exacerbating the existing health inequities rather than reducing them.
The present study had several limitations. First, we used only cross-sectional data for estimation. After the ABWMC programme conducts additional follow-up surveys, it will be possible to perform a longitudinal study to obtain more convincing ndings. Second, the current study was limited to interested companies, thereby potentially introducing selection bias, and we did not recruit any participants from other areas, such as government employees. Third, our results were based on self-report information. It was not possible to objectively verify the survey answers, and some respondents may not have provided real information. Fourth, As the study was not an experimental one but a cross-sectional survey, it is possible that some other factors caused the observed effects.

Conclusion
Taken together, the results of this empirical analysis not only contribute to identifying the possible in uence of workplace tobacco-related health education but also provide further evidence regarding the spillover effects of such activates in a developing country. Further analyses suggested that job position is among the channels moderating such bene cial spillover effects. Policy makers should recognize and reduce the potential health disparities.

Declarations
Ethics approval and consent to participate This study was approved by the Peking University (Ethical approval number: IRB00001052-18055).

Competing interests
We declare no competing interests.

Funding
This study was supported by AIA Company limited.
Author's Contributions Haoxiang Lin and Meijun Chen nished the rst draft, should contributed equally to this study. Zhao Liu performed the statistical analysis. Chun Chang managed all the research. All authors contributed to the subsequent revisions of the paper and all authors have approved the nal paper for submission.

Availability of Data and Materials
The data of the studies is accessible via School of Public Health, Peking University Health Science Center.

Consent for publications
Not applicable