Sociodemographic disparity in health-related behaviors and dietary habits among governmental employees in Hunan, China

Health-related behaviors and dietary habits have played essential roles in shaping population health and are related to morbidity or mortality. We aimed to estimate the distribution of health-related behaviors and dietary habits by socio-demographics among governmental employees in China. A total of 5029 participants were included in this study. Information on their sociodemographic characteristics, health-related behaviors and dietary habits was collected. Socioeconomic status (SES) scores were calculated by multiplying ordinal numerical values assigned to consecutive categories of education level and annual household income. Adjusted multinomial logistic regression analysis and categorical principal component analysis were used to estimate differences in health-related behaviors and dietary habits by socio-demographics.

is a chief contributor to morbidity and the leading cause of mortality [6,7]. Dietary habits are the habitual decisions of individuals or group of people regarding what foods they eat [8]. Some unhealthy dietary habits (e.g., eating midnight snack and consuming pickled foods) have been shown to be associated with a higher risk of cancer [9][10][11]. The importance of healthy behaviors and dietary habits is underscored by the US Affordable Health Care Act, which provides unprecedented universal coverage for preventive behavioral and dietary health services [12].
The distributions of health-related behaviors and dietary habits usually differ by sociodemographic characteristics, such as gender, age and social class, [13][14][15]. Men (including male adolescents), for instance, are more likely than women to smoke, binge drink [16,17]; women (including female adolescents), by contrast, are less likely to engage in physical activity [18]. Younger people were more likely to get involved in risk behaviors and unhealthy dietary habits than older people [19,20]. Those with high socioeconomic status (SES) were less likely to smoke, and have a low level of physically inactivity, but more likely to have high nutritional eating patterns [21][22][23][24][25]. These sociodemographic factors shape social context and day-to-day routines, which in turn in uence health behaviors and dietary habits, as well as people's ability to change. Knowledge of sociodemographic-speci c behavioral characteristics and their interrelationships can help pinpoint intervention targets and guide for development of e cacious interventions.
Although health-related behaviors and dietary habits have been found to be associated with sociodemographics such as gender, age and SES, these factors are usually studied separately. There is a lack of comprehensive study on health-related behaviors and dietary habits by taking account of all sociodemographic factors together, for instance, by combining age and gender. Also, most of the study were conducted in developed countries, it is also important to learn about the situation in developing countries such as China. The current study was conducted to understand the distribution of health-related behaviors and dietary habits by socio-demographics among governmental employees in urban area of China.

Study design and participants
This retrospective study was conducted between January 2018 and December 2018 in Changsha city, Hunan province of China. Brie y, a multistage sampling design was carried out to obtain a representative sample. Firstly, 10 institutions were randomly sampled from the government-run institutions that volunteered to participate in the study in Changsha city. Secondly, in each sampled institution, all the staff were selected to take part in our study using cluster sampling. In China, governmental employees are the workers in the national legislative branch, judicial branch, administrative organs, the Party organs of the Communist Party of China and the democratic parties, people's organizations and public institutions, who perform public duties according to law, such as civil servants, regular employee at university or hospital, etc. After obtaining informed consent, eligible participants were asked to complete a digital self-reported questionnaire. The eligibility criteria for participants included: (1) being able to speak and understand Chinese, (2) being 18-60 years old, (3) working at the selected institution, and (4) having no serious limb diseases. Finally, a total of 5029 governmental employees were include in this study.

Sociodemographic characteristics
The collected sociodemographic data included information on demographic (age, gender, marital status) and socioeconomic characteristics (education and annual household income). SES scores were calculated by multiplying ordinal numerical values assigned to consecutive categories of education level and annual household income [26]. Educational attainment was divided into six groups, and annual household income was divided into seven groups. The SES score ranged from 1 to 42. According to the tertile values of the SES score distribution, study participants were further divided into three groups: low SES (L-SES, 1-12 points), medium SES (M-SES, 13-18 points), and high SES (H-SES, 19-42 points).

Health-related behaviors
Data on health-related behaviors was collected using standardized questionnaires, which included smoking, alcohol drinking, leisure-time physical inactivity, sedentary behavior, late sleeping, and short sleep duration. Smoking was de ned as smoking at least one cigarette a day for more than half a year, which can be further classi ed into three status: never, former, or current smoking. Alcohol drinking was de ned as dinking at least once a week for more than half a year, and can be further classi ed into three status: never, former, or current smoking. Leisure-time physical inactivity was de ned as 30 min/day of physical activity at a frequency of once per week or less. Sedentary behavior was de ned as more than 2 hours/day sitting (such as watching TV, sur ng the Internet, etc.) apart from work and study. Late sleeping was de ned as sleeping after midnight, and short sleep duration was de ned as sleeping time less than 7 hours per night.

Dietary habits
Dietary habits included four eating behavior habits and 15 kinds of food consumption. Eating behavior habits included irregularly meals habit (yes or no), eating midnight snack habit (yes or no), eating gluttony habit (yes or no), and salty taste habit (yes or no). Food consumption was assessed by asking whether consuming each of the following food groups for more than 3 times/week: (i) rice; (ii) wheat; (iii) grains; (iv) sh; (v) vegetables; (vi) poultry meat; (vii) livestock meat (pork, beef/lamb); (viii) Soybean and products; (ix) fruits; (x) eggs; (xi) dairy products; (xii) nuts, (xiii) desserts (xiv) pickle food, and (xv) smoked sh or meat.

Data Analysis
Crude frequencies and weighted percentages were calculated for distribution of socio-demographic characteristics. Among these socio-demographic variables, gender, age and SES were considered as important factors for health-related behaviors and dietary habits. Thus, analyses of health-related behaviors and dietary habits were strati ed by these three variables. Participants were grouped into 8 categories based on gender (male/female) and four age subgroups (≤30, 31-40, 41-50 and ≥51), and 3 categories based on SES. Weighted percentage of the six health-related behaviors and four eating habits were calculated by these categories. Adjusted multinomial logistic regression analysis was used to determine frequencies (health-related behaviors and eating habits) in L-SES and H-SES.
Categorical principal components analysis (CATPCA) was used to examine the different consumption patterns for the 15 food groups mentioned previously (Table 1, see line 541). The analysis allowed for simultaneous evaluation of the relationships between various variables. The analysis was considered reliable for Cronbach's alpha coe cient >0.7. All the statistical analyses were performed using SPSS version 19.0 software. A two-sided p-value <.05 was considered as being of statistical signi cance.

Sample characteristics
A total of 5029 participants were included in the study (3006 for female; 2023 for males), and their characteristics were shown in Table 2 (see line 543). The age distribution of the sample was as follows: 19.6% of participants were ≤30 years, 40.1%were 31-40 years, 24.8% were 41-50 years, and 15.6% were ≥51years old. Most participants were married/cohabited (84.8%), with an education level of undergraduate/college or above (94.6%), and with a family income level of 110,000-200,000 (36.2%). According to the de nition of SES for the current study, 35.3% were in L-SES group, 36.6% were in M-SES group, and 28.1% were in H-SES group. The prevalence of current smoking, current alcohol drinking, leisure-time physical inactivity, sedentary behavior, late sleeping and short sleep duration among the participants was 12.0%, 6.7%, 54.8%, 74.8%, 16.6% and 34.4%, respectively. Additionally, the prevalence of four eating behavior habits was 38.3% for irregular meals habit, 26.1% for salty taste habit, 4.8% for eating gluttony and 2.0% for eating midnight snack, respectively ( Table 2, see line 543).

Distribution of health-related behaviors and eating behavior habits by gender and age
We further examined distribution of health-related behaviors and eating behavior habits by age for men and women separately ( Table 3, see line 574). Smoking and alcohol drinking behaviors were observed for males, but hardly for females. For males, the age of ≥51 years group had the highest rate of current smoking (36.2%), current alcohol drinking (25.7%) and short sleep duration (39.0%). Younger males (age ≤40 y) had a high rate of leisure-time physical inactivity (age ≤30 y:52.8%; age 31-40 y: 59.6%). The youngest age group (age ≤30 y) had the highest rate of sedentary behavior (80.4%) and late sleeping (30.1%). As for eating behavior habits, the youngest age group (age ≤30 y) had the highest rate of irregular meals habit (50.4%), eating midnight snack habit (3.5%), eating gluttony habit (10.3%) and salty taste habit (35.8%).
For females, the youngest age group (age ≤ 30 y) had the highest rate of sedentary behavior (81.7%), leisure-time physical inactivity (69.4%) and late sleeping (20.7%). The ≥ 51 age group had the highest rate of short sleep duration (45.4%). As for eating behavior habits, the youngest age group (age ≤30 y) had the highest rate of irregular meals habit (61.1%), eating midnight snack habit (3.9 %), eating gluttony habit (7.0%) and salty taste habit (25.6%). A negative dose-response trend increasing age was found in irregular meals habit, eating midnight snack habit and salty taste habit Distribution of health-related behaviors and eating behavior habits by SES Table 4 (see line 592) presented the distribution of health-related behaviors and eating behavior habits among different SES groups after adjusting for age, gender and marital status. For health-related behaviors, compared to the L-SES group, the H-SES group had lower rate of current smoking (11.4% vs 13.7%, p = 0.006) and leisure-time physical inactivity (51.9% vs 54.4%, p = 0.028), but higher rate of late sleeping (18.5% vs 15.2%, p < 0.001). No signi cant differences were found in the rate of current alcohol drinking, sedentary behavior and short sleep duration. Additionally, the rates of the four dietary behaviors were similar in the three different SES groups.
Food consumption patterns by gender, age and SES CATCPA was performed to identify the patterns of food consumption for 15 food groups by gender, age and SES (Figure 1). The Cronbach's alpha was 0.887 and the total eigenvalue was 5.812, indicating good reliability for assessment of simultaneous relationships among the selected variables. Among the 15 food groups, rice consumption habit was located in the centroid. Four food consumption patterns were identi ed. Pattern 1 named 'traditional foods' pattern was characterized by more consumption of vegetables, livestock and poultry meat, fruits and eggs. Pattern 2 named 'cereals and dairy product' pattern was characterized by more consumption of grain, wheat, soybean and products, and dairy product. Pattern 3 named ' sh and nut' pattern was characterized by more consumption of sh and nut.
Pattern 4 named 'smoked and pickled foods, and dessert' pattern was characterized by more consumption of smoked sh or meat, pickle food and dessert. CATCPA indicated the 'traditional foods' and 'cereals and dairy product' patterns were mainly observed in the 41-50 age group with H-SES. "Fish and nut" and "smoked and pickled foods, and dessert" patterns were mainly observed among the ≥51 y males in L-SES group. There is an absence of difference in four food consumption patterns among young females (aged ≤ 40 years old) in the M-SES group.

Discussion
This retrospective study showed that the distributions of health-related behaviors and dietary habits varied by gender, age and SES among governmental employees in China. Males in the meddle-aged group (41-60) were more likely to smoke, use alcohol and have short sleep duration, but males in the 31-40 age group had the highest rate of leisure-time physical inactivity (59.6%). Young females (age ≤ 30 y) were more likely to have leisure-time physical inactivity, but females aged ≥ 51 y were more likely to have short sleep duration. Sedentary behavior, late sleeping, as well as all four risk eating behavior habits were more prevalent in youngest age group (age ≤ 30 y), regardless of gender. In addition, we found that people in the H-SES group were less likely to be current smoker and have physical inactivity, but were more likely to have late sleeping. We also identi ed four food consumption patterns, which varied by gender, age and SES.
We found gender and age differences in the occurrence of health-related behaviors. For instance, the rate of smoking and alcohol drinking were much lower in females than males, which may be explained by socio-cultural factors. The social gender role in culture consider it acceptable, appropriate, or even desirable to engage in certain behaviors such as smoking and drinking for males, but almost not for females [27][28][29]. In this study, a higher rate of smoking and alcohol drinking was found in middle-aged (41-60y) males than that in younger males (≤ 40 y). In recent years, the Chinese government has been gradually strengthening the policies against tobacco use, and the group intervention on smoking cessation based on social cognitive theory could be effective to reduce smoking [30,31]. Male smokers in the high age group lacking health knowledgeable and ignoring the proper health behaviors, cannot su ciently understand the need for preventive measures [32,33], even reduce the stress of social isolation, economic hardship, prior trauma, and the loss of power and status by smoking [34]. In addition, we found young females and males had a high rate of leisure-time physical inactivity, late sleeping and sedentary behavior, which was consistent with previous research [35]. With the rapid socioeconomic development in China and changes in urbanization, work stress emerged in the urban population, especially among the working young population due to a rise in living costs, which may result in long working time and lacking leisure-time to participant in physical activity. Modern media use has also become intricately connected with our bedtime routine and sleep [36]. Additionally, individuals in the ≥ 51 y age group were more likely to have a high level of short sleep duration. These ndings strongly suggest that planning of effective health promotion programs should target health-related behaviors that are modi ed by gender and age.
The association between SES and various health-related behaviors are multiple and complex [37]. Past studies have consistently shown SES as the strongest predictor of tobacco use [38][39][40]. In the current study, we found higher rate of smoking and leisure-time physical inactivity in the H-SES group than L-SES group, which were consistent with previous research [41,42]. Our study, for the rst time, found higher rate of late sleeping in H-SES group than L-SES group among governmental employees in China. Tobacco control policies, leisure-time physical inactivity and late sleeping interventions in China should be increasingly focused on populations with low-SES in order to break the link between socioeconomic disadvantage and these unhealthy behaviors. Education, one of the components of SES, remained the strongest predictor of these unhealthy behaviors [38][39][40]42], suggesting future intervention programs to take targeted measures to reduce unhealthy behaviors, according to education attainment status. Although no signi cant differences in short sleep duration and sedentary behaviors were observed by SES in this study, we found much higher rate of short sleep duration and sedentary behaviors in our study population than the general population in southeast of China in other studies [41,43], drawing our attention to the impact of short sleep duration and sedentary behaviors on health among this special population [9]. We found differences in dietary habits including eating behavior habits and food consumption patterns by gender, age and SES. For eating behavior habits, both females and males in the younger age group (≤ 30 years) have the highest rate of these four unhealthy dietary habits. Younger individuals may get involved easily in unhealthy eating behaviors due to poor control awareness and health literacy [19,20,44]. Although there were absent differences in four eating behaviors among three SES groups, combination of some other studies, SES remained an important factor to impact unhealthy eating behaviors [45][46][47].
We also explored food consumption patterns by socio-demographics using CATCPA analysis, which is the rst study of its kind among governmental employees in China. We found the 'traditional foods' and 'cereals and dairy product' patterns were mainly observed in the 41-50 age group with H-SES. In China, traditional foods such as vegetables, fruits, meats and eggs were considered as healthy food pattern in line with Chinese traditions. Research has consistently shown that high intake of vegetables and fruits, whole grains, and eggs may decrease the risk of chronic diseases, even cancer [48,49]. In addition, we found ' sh and nut' and 'smoked and pickled foods, and dessert' patterns were mainly observed among the ≥ 51 y males in L-SES group. High intake of sugars, and pickled or smoked foods may increase such risk [9,22,50,51], but sh consumption was associated with a reduced risk of all-cause mortality [52]. Available study has pointed out that males have a higher liking for sweets than females [53]. Younger age, low education, and low family income were associated with a poor health literacy status, which may result in a low nutrient food consumption [44]. The ndings have implications for future intervention programs to take targeted measures to reduce unhealthy food consumption habits, according to speci c sociodemographic characteristics.
To the best of our knowledge, this is the rst study using a large sample from a representative urban governmental employee population. This is also the rst study to use the CATPCA method to study health-related behaviors and dietary habits by socio-demographic characteristics such as gender, age and SES groups. However, some limitations should be considered when interpreting our results. First, many variables were self-reported and the ndings may be subject to recall bias. Second, the study participants came from Hunan province and China, and the generalizability of the study ndings may not be able to generalize to other parts of China despite its large sample size. Future multiprovince or multinational surveillance studies are needed to further test ndings in our study. Third, the retrospective design of the current study may preclude any observation of behaviors and dietary habits changes over time. Future longitudinal studies may overcome such a limitation.

Conclusions
This study con rmed the various distribution of health-related behaviors and dietary habits by sociodemographic characteristics among governmental employees. Notably, this study found that middle-aged group (41-60 years) had the highest rates of smoking and alcohol drinking, and young-aged group (≤ 40 years) with L-SES were more likely to have multiple unhealthy behaviors and dietary habits. We also identi ed four food consumption patterns, which varied by gender, age and SES. Our ndings of the disparity distribution of health-related behaviors and dietary habits by speci c-gender, age and SES among Chinses governmental employees have important policy implications for developing targeted health interventions to facilitate health-related behaviors and dietary habits in this population.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.     Variables in association: Pattern 1 named 'traditional foods' pattern was characterized by more consumption of vegetables, livestock and poultry meat, fruits and eggs. Pattern 2 named 'cereals and dairy product' pattern was characterized by more consumption of grain, wheat, soybean and products, and dairy product. Pattern 3 named ' sh and nut' pattern was characterized by more consumption of sh and nut. Pattern 4 named 'smoked and pickled foods, and dessert' pattern was characterized by more consumption of smoked sh or meat, pickle food and dessert; SES, socioeconomic status; L-SES: lowsocioeconomic status; M-SES, medium-socioeconomic status; H-SES, high-socioeconomic status.