Urbanization has made cities becoming the ultimate destination for population agglomeration (Vearey et al., 2020). As a result, the labor force characteristics of China’s population are undergoing gradual changes. Urbanization and demographic transitions often lead to a range of health issues. Some studies have suggested that individuals in good health are typically the primary participants in this transition (Chen, 2011). Nevertheless, health risks do not manifest immediately but rather accumulate over time due to low adaptability or ongoing exposure to high-risk environments. Migrant workers often tend to find themselves in low-paying, dirty, harsh, and dangerous jobs compared to local permanent residents. This disparity can be attributed to differences in educational levels and household registration systems (Yue et al., 2013; Shen, 2017). In contrast to physical health issues, mental health concerns are often less apparent and more challenging to detect. Mental health disorders can generate various challenges and risks, including financial difficulties (Anand et al., 2018; Richardson et al., 2021), lethargy (Fibbins et al., 2018), depression, and shorter lifespans (Stubbs and Rosenbaum, 2018). Thus, exploring the mechanisms that influence mental health has important practical implications. Sweetser first introduced the concept of mental health (Sweetser, 1850). This definition has since been further developed through ongoing research (Gamm et al., 2010). As per the definition put forward by the World Health Organization in 2019, mental health in this study is defined as a state of well-being in which an individual can realize their own potential, cope with the normal stresses of life, work productively, and contribute to the community.

As an abstract concept, most researchers generally choose to express mental health using multiple indicators. Consequently, psychological scale tools have emerged as the primary method for diagnosing mental health conditions. Internationally recognized scales that are frequently used include the Symptom Checklist-90 (SCL-90), the Positive and Negative Affect Schedule (PANAS) and the Self-Rating Depression Scale (SDS), etc. On the above mental health scales, the reliability and validity of SCL-90 which is recognized as a classical scale have been tested in China (Zhong et al., 2013), while SDS is difficult to evaluate depression with severe retardation symptoms, and the evaluation effect is poor for people with low educational level or slightly poor intellectual level (Guan et al., 2022). Meanwhile, PANAS has the limitations of an ordinal scale such as low precision and unsuitability for using with parametric statistics (Medvedev et al., 2023). In addition, with the development of data acquisition technology and hardware matching, the information collection mode of mental health has expanded to include wearable electronic devices with sensors (e.g., watch, helmet) and other devices to monitor ecological instantaneous assessments, physiological functions, environmental parameters, physical behavior (Fahrenberg et al., 2007), ultimately forming a longitudinal dataset that allows for the real-time collection, but it needs a long cycle of time consumption. Finally, after comparing the features of the previous scales, Depression Self-Assessment Scale (CES-D20) is employed in this paper to measure mental health, because of its high quality of operability, reliability and recognition degree (Bharadwaj et al., 2017).

The determinants of mental health in previous researches have been mainly summarized into three aspects: built environment (e.g., land development intensity, greening space rate, road network density), work environment (e.g., commuting time, employment status), and subjective perception (e.g., community trust, economic satisfaction, and community security). The influence of these factors is usually related to the unequal distribution of social resources (Marmot et al., 2013). Especially the living conditions of the public, including built environment by which residents are surrounded and the working environment where activities are carried out may make people prone to exert different psychological feelings, thereby affecting people’s mental health (Braveman et al., 2011). Therefore, it is important to study the influencing mechanism of built environment, work environment and subjective perception on mental health.

On the one hand, the built environment, as an activity platform for residents’ daily lives, not only directly affects physical activity and active transportation (Smith et al., 2017), but is also indirectly related to the potential mental health risks caused by unreasonable housing and infrastructure. This has been confirmed in several studies (Melis et al., 2015; Kim and Yoo, 2019; Stone et al., 2022). At the same time, green space can provide important health and environmental services, reduce economic and social inequalities, reduce the mental stress of urban residents, and provide people with better well-being and mental health (Nutsford et al., 2013; Barton and Rogerson, 2017). On the other hand, the impact of the work environment has also gradually come to receive greater attention with regard to its impact on mental health. Poor working conditions (e.g., working overtime, basic salary) can generate stress, fatigue and other negative emotions (Novaco et al., 1979; Evans, 2003; Gatersleben and Uzzell, 2007); while long-term exposure to dangerous workplaces (e.g., to air and noise pollution) can lead to increased health risks (Lundberg, 1976; Feng and Boyle, 2014; Böcker et al., 2016). Long working or commuting hours also reduce the time available for exercise, relaxation, and communication with family and friends, and increase feelings of isolation (Lyons and Chatterjee, 2008; Christian, 2012; Mattisson et al., 2015). Therefore, the improvement of environmental factors is crucial not only to promoting healthy physical activity and reducing the risk of chronic diseases such as obesity and cardiovascular disease, but also to providing people with the opportunity for psychological recovery and spiritual release, thus promoting better mental health and improving overall happiness (Zhao et al., 2022).

Furthermore, subjective perceptions are also important to mental health. It is believed that subjective environmental perceptions affect people’s emotions and behavioral choices (Wang et al., 2019; Liu et al., 2019a). If an environment can promote positive perceptions, people will be more inclined to continue engaging in activities in that environment (Mehrabian and Russell, 1974). Previous research has shown a correlation between subjective environmental perception and healthy activities (e.g., physical activities and social interactions) (Astell-Burt et al., 2014; Cohen-Cline et al., 2015). However, it is worth noting that subjective environmental perception is usually based on the objective environmental conditions, which guide individuals to produce corresponding behaviors (Atari et al., 2009; Liao et al., 2015). Thus, subjective environmental perception may act as a mediator between the environment and behavior, a fact that has been further verified across a number of different theoretical frameworks (Kremers et al., 2006; Sallis et al., 2006). From previous studies, it can be seen that the influencing factors of mental health mostly lie in the relationship between environment and mental health or subjective perception and mental health, and rarely have the relationship between various aspects and mental health at the same time. Therefore, this paper starts from the three aspects of built environment, work environment and subjective perception, and discusses the relationship between these variables and mental health. I hope it can provide some help for future research.

This study uses data from the 2018 CLDS (Liu et al., 2019b; Huang et al., 2023) conducted by the Centre for Social Science Investigation at Sun Yat-sen University. The CLDS project was the first interdisciplinary large-scale tracking social survey on labor status in China, focusing on the status quo and changes in China’s labor force. The project covers education, work, migration, health, social participation, economic activities, grassroots organizations, and other research topics. The project sample population covers 28 provinces and cities in China (with the exceptions of Hong Kong, Macao, Taiwan, Tibet, Hainan, and Xinjiang), and the survey object is the entire labor force (family members aged 15–64 years) within the sample households. This study used survey data from the CLDS in the 2018 specifically as the total sample. The investigation included individual self-assessments of mental health, lifestyle, work environment, and economic status. Based on the needs of the study, the study initially excluded samples with missing or invalid self-rated mental health scale, number of Hukou transfer, and other key indicators. Among the remaining samples, there were 623 missing values for the variable “fun to work” (9.63%) and 481 missing values for the variable “part-time job” (7.44%). Subsequently, multiple imputation was employed to address the missing values, resulting in a complete database with a sample size of 6,464. The sample size distributions in different regions are shown in Figure 1.

Table 1 shows the sociodemographic characteristics of the total sample and subsamples (migrants and local residents) used in this study. This study defines the local residents are those who have not relocated in the previous 6 months, while migrants are those who have left their current place of residence over 6 months. First, in the total sample, the mean value of age was 46.5, and the proportions of male and female respondents were 53.6 and 46.4%, respectively, of which 86.6% were married and 75% had zero Hukou transfers. Based on the socioeconomic characteristics of the research participants, the proportion of people satisfied with their employment status was relatively high (92.6%). The proportion of people who did not have part-time jobs was 92.7%. Among them, the proportion of respondents with higher education qualifications was generally low, at only 8.4%. In the subsample, the average age among the migrants and the local residents were 46.41 and 47.73. The proportion of female among the migrants was relatively high (72.8%), as was the proportion of male among local residents (60.4%). In terms of marital status, a high proportion of both the migrants and the local residents were married, 91.5 and 86.5%, respectively. Furthermore, both migrants and local residents reported a relatively high level of their employment status, at 91.8 and 92.8%, respectively.

Based on previous research, we selected three aspects that impact alterations in individuals’ mental health: built environment, work environment, and subjective perception. These dimensions were taken directly from the CLDS 2018 questionnaire and government-accessible data (National Bureau of Statistics Urban Socioeconomic Survey Department, 2019).

Considering the possibility that mental health is affected by spatial differences in the environment and subjective perceptions (Zhao et al., 2022). In this study, a single-level linear regression model was used to measure the impact mechanism on mental health. This study selected mental health as the dependent variable, used the Ordinary Least Squares (OLS) (Choi and DiNitto, 2013; Mersky et al., 2013) method to establish a linear regression model, and measured the impact of the built environment, work environment, and subjective perception of mental health. The resulting expression is as follows:

In Eq. 1, where Y represents the dependent variable, that is, mental health, x i represents the independent variables (including the built environment, work environment, and subjective perception) and control variables (including age, gender, marital status, hukou transfer and education level), β i represents its correlation coefficient. And α 0 represents intercept, ε is a normal distribution in which the error obeys a mean value of 0.

The average mental health scores are shown in Figure 2. Overall, the average mental health scores gradually increased from the western to the central and eastern regions. The average mental health scores of Jinzhou, Yingkou, Tongling and Taizhou were relatively high, indicating that the residents of these areas have better mental health status, they are happier, more optimistic and more resilient. While respondents from Beihai, Jinchang and Yangquan had relatively low mental health score, they may face more stress and psychological barriers in their daily lives.

To explore the spatial differentiation and autocorrelation characteristics of mental health in city scale, Global Moran’s I (Moran, 1950) and Local Moran’s I (Anselin, 1995) were implemented. The Moran ‘s I index was 0.103 with a Z-score of 4.162 and a value of p <0.01, indicating that the mental health in study area has a positive correlation. As shown in Figure 3, we further explored the spatial differentiation characteristics of mental health using Local Indicators of Spatial Association (LISA) clustering at the city scale in the study area. The results of the LISA clustering significance test showed that, first, eastern regions, for example, cities in Shandong province, Hangzhou, Suzhou, Nantong in Yangtze River Delta and Dalian, Yingkou in Liaodong Peninsula generally showed a positive local spatial correlation with mental health and belonged to high-high clustering areas. This indicates that the mental health scores of the residents in and surrounding cities were higher overall, and they may generally have a better mental health status. Second, cities such as Shanghai, Chengde, Zhengjiang and Huai’an are low-high clustering areas, indicating that people living in those areas may have worse mental health status than surrounding cities. The high-low clustering areas including Shuozhou and Qinzhou, indicating that the mental health scores in these areas were higher than those in the surrounding areas. Low-low aggregation areas mainly included in cities in Gansu and Qinghai provinces such as Wuwei, Zhangye and Xining, indicating that the residents in these areas have relatively poor mental health.

As shown in Table 4, regression analysis of the baseline model shows that there is a significant positive correlation in subjective perception factors, among which the coefficients of work feelings, community trust, economic satisfaction and community security are 0.235, 0.254, 1.525, and 0.921, respectively. This means that better subjective feelings contribute to better mental health status, because better work experiences, friendly interpersonal relationships, higher income and higher sense of security mean lower levels of work stress and greater enthusiasm for life, a finding which is consistent with those recorded in previous studies (Jakobsen et al., 2022; Shields-Zeeman and Smit, 2022). Second, in terms of the built environment, an increase in the greening space rate causes an increase of 0.001 point of mental health. Generally, a higher green space rate means higher environmental quality and more parks and other leisure places to promote social activities among residents, especially the elderly (Home et al., 2012; Bertram and Rehdanz, 2015). An increase of Road network density also causes an increase of 1.153 point of mental health. A higher road network density may reduce vehicle speed in the community, decreasing peoples’ concerns regarding traffic accidents. Higher road network density can also improve accessibility to the community, providing more walking options to make it easier to reach service facilities (Tao et al., 2020). In addition, in terms of the work environment, commuting time had a significant negative correlation (p < 0.001) with mental health. For control variables, age, marital status and low education have significant negative correlations with mental health with coefficients of 1.547, −0.055, and −0.830, respectively. This result indicates that mental health may decline when people get aged, get married or have a low education level.

The mean of the mental health scores for local residents and migrants are 52.52 and 51.58, respectively. The t-test results for the differences between local residents and migrants’ groups are statistically significant, as the significance (Sig.) of Hukou transfer is higher than 0.05 for Levene’s test and less than 0.001 in the t-test (Table 5). The mental health of local residents’ group (t = 3.417) seems higher than those of migrants group (t = 3.356). Thus, the following section further delves into the impact mechanism of mental health and explores the impact of the built environment, work environment and subjective perception on mental health between migrants and local residents.

As shown in Table 6, by constructing a subsample model for migrants and local residents groups, we found that work feeling (p < 0.01), community trust (p < 0.01) and economic satisfaction were all significantly and positively correlated with mental health. However, there are some differences in the factors that affect the mental health of migrants and local residents. For the migrants group, the coefficient of economic satisfaction ( β = 2.034, p < 0.001) is higher than the local residents group ( β = 1.393, p < 0.001). The reason for this phenomenon may be that migrants usually desire a higher level of income to settle down, so they face more pressure, creating a significant psychological burden (Zhao, 1999), especially for female migrants (Bardasi and Francesconi, 2004; Jackson et al., 2019). Better economic satisfaction may help migrants improve their mental health. Besides, among local residents, community security was found to be significantly related to mental health ( β = 1.095, p < 0.001), while the migrants group shows no significance among community security. Furthermore, for the local residents group, the coefficient of community trust ( β = 0.270, p < 0.001) is higher than migrants group ( β = 0.185, p < 0.01). The scope of activities among local resident communities is relatively stable, meaning where they live is a very important consideration. The safety and trust of the community are important requirements for local residents who usually have children, and children are usually more susceptible to potential artificial hazards (e.g., violence) (Jackson et al., 2019) or natural hazards (e.g., pollution) (Tortorella et al., 2022). For the built environment and work environment, both the migrants group and the local residents group show positive significance in the greening space rate and road network density to mental health. Previous studies found that community green spaces can alleviate the homesickness of the migrants by increasing social cohesion and have a positive effect on the mental health, and local residents may pay attention to the greening of their living environment due to their long-term stable living characteristics (South et al., 2018; Yang et al., 2020; Zhang et al., 2022). Commuting time has a negative impact on mental health. Commuting drains energy and attention. Long-time commutes to work are likely to be much more tiring and stressful than shorter time commutes (Evans et al., 2002). Excessively long commutes may reduce people’s mental health and may increase the risk of depression (Wang and Liu, 2022). For other variables, the number of hospitals, land development intensity, PM2.5 index, part-time or not and employment status show no significance.