Trends and Socioeconomic Factors in Smoking and Alcohol Consumption Among Chinese Adults: Evidence From National Health Service Surveys

Background: Smoking and excessive drinking are risk factors for many diseases. With the rapid economic development in China, it is important to identify trends in smoking and alcohol consumption and factors that contribute to these behaviors to ensure the health of the population. Methods: we analyzed pooled cross-sectional data from the fourth, fth, and sixth National Health Service Surveys conducted in Jiangsu Province in 2008, 2013, and 2018, respectively. Trends in smoking and alcohol use were analyzed with descriptive statistics, and bivariate and multivariate logistic regression was used to identify contributing factors. Results: Among total sample, smoking rate was 23.95%, in which the incidence of mild, moderate and severe smoking was 5.75%, 4.63% and 13.56%, respectively; drinking rate was 23.29%, in which non-excessive drinking and excessive drinking were 19.80% and 3.49%, respectively. From 2008–2018, overall and light-to-moderate smoking rates rst increased and then decreased while heavy smoking rate declined; and alcohol consumption increased while excessive drinking increased before decreasing. The varying tendency of smoking and drinking rates in urban area was similar to rural area, however there was a signicant gap in incidence between urban and rural area. Socioeconomic factors, demographic, health-related and year variables were signicant associated with smoking and drinking. Conclusion: Preventive measures such as education and support services along with stricter regulations for tobacco and alcohol use are needed to improve public health in China. with multiple chronic diseases were less likely to smoke (OR = 0.78) and smoke heavily (RRR = 0.75). Compared to respondents who were in very good health, people with poor health were signicantly less likely to smoke (OR = 0.57) and smoke heavily (RRR = 0.49). EQ-5D scores also indicated that the rates of smoking (OR = 0.81) and heavy smoking (RRR = 0.75) were signicantly lower in respondents with incomplete health than in those with complete health. Compared to people who did not have regular physical examinations, those who had undergone a physical examination in the previous 12 months were less likely to smoke (OR = 0.91) and smoke heavily (RRR = 0.9). People who did not engage in physical exercise were more likely to smoke (OR = 1.35)—including smoking moderately (RRR = 1.24) and heavily (RRR = 1.47)—than those who exercised regularly.

are in uenced by demographic factors such as age, sex, marital status, and location of residence; for example, studies conducted in China reported that people living in rural areas were more likely to smoke and drink than urban dwellers [17][18][19].
Most previous studies on the factors in uencing smoking and drinking behavior have analyzed cross-sectional data, which do not re ect long-term trends. Moreover, there is a large gap in development between urban and rural areas in China, but there have been few comparative analyses of smoking and drinking rates according to location. To address these issues, the present study analyzed trends in smoking and alcohol consumption rates and quantities in urban and rural areas as well as factors contributing to these behaviors based on National Health Service Survey (NHSS) data from Jiangsu Province, China. The pooled cross-sectional data encompassed the period of rapid economic and social development and health reforms that occurred between 2008 and 2018. The results of our study can guide the development and implementation of more effective strategies to improve public health in China.

Data source
Since 1993, the NHSS has been organized every 5 years by the National Health Commission of China, with the provincial health commission being responsible for the survey of each region. The NHSS mainly consists of a household survey supplemented with an institutional survey; the former collects data through household interviews, with all permanent residents of surveyed households interviewed by trained and quali ed investigators according to questionnaire items. A multistage strati ed cluster random sampling method was used to select 156 counties (cities and districts) from 31 provinces in China; 5 towns (streets) were randomly selected in each of the sample counties (city or district); 2 villages (neighborhood committees) were randomly selected in each sample town (street), and 60 households were randomly selected in each sample village (neighborhood committees), for a total of 93,600 households (population of nearly 300,000). The fourth NHSS was in June 2008, with 7021 respondents participating in the survey; the fth NHSS was in June 2013, with 10,422 respondents; and the sixth NHSS was in September 2018, with 11,550 respondents.
We used data from 2008, 2013, and 2018 to analyze trends in and socioeconomic factors contributing to smoking and alcohol consumption among Chinese adults. The inclusion criteria were men and women over 15 years of age. After merging the NHSS datasets, the study population comprised 24,939 respondents. To ensure more accurate data analysis, respondents with any missing variables were excluded.
NHSS data provides detailed information on demographic characteristics, socioeconomic status, health status, healthcare use and cost, health behaviors, etc [20]. The structure of the questionnaire has good reliability and validity.

Dependent variable
The dependent variables were smoking, smoking quantity, drinking, and drinking quantity. Data on smoking were collected by asking respondents whether they smoked, the number of years they had smoked, and the average number of cigarettes smoked per day. The smoking index (SI) was calculated as number of years of smoking × average number of cigarettes per day; smoking quantity was categorized as low/light smoking (SI ≤ 200), moderate (200 < SI < 400), or high/heavy smoking (SI ≥ 400) [21]. Data on drinking were obtained by asking respondents whether they drink and how often; drinking quantity was converted to standard drinking units by the investigators. Speci cally, 1 can of beer was equivalent to 1 drinking unit; 50 g of a beverage with < 40% alcohol was taken as equivalent to 1.5 drinking units; 50 g of a beverage with ≥ 40% alcohol was equivalent to 2 drinking units; 1 bottle of beer was equivalent to 2 drinking units; 500 g of wine was equivalent to 5 drinking units; and 500 g of yellow rice wine was equivalent to 6.5 drinking units. Drinking quantity was categorized as non-excessive or excessive drinking; the latter was de ned as ≥ 5 and ≥ 4 drinking units at a time for men and women, respectively [22].

Independent variables
The independent variables were demographic information, socioeconomic status, and health-related information. Demographic variables (eg, age, sex, number of siblings, marital status, place of residence, and social health insurance) were included in our Page 4/22 analyses in order to reduce the impact of differences between rural and urban populations. Socioeconomic factors included income level, education level, employment status, type of occupation, and poor or low-security households. We used per capita annual income adjusted by price index as the income variable. Health-related variables included chronic disease, European Quality of Life Scale -5 Dimensions (EQ-5D) score, health status, physical examination, and physical exercise. Self-reported health status was evaluated in the questionnaire with the Visual Analog Scale (VAS) and classi ed into 5 grades as in previous studies [23,24]. Additionally, the wave variable (2008, 2013, and 2018) was included in order to account for uctuations in smoking and drinking rates within each year of the NHSS. Detailed descriptions of dependent and independent variables are shown in Table 1.

= Yes
Multiple chronic diseases 0 = No "Yes" is de ned as having two or more chronic diseases, and "no" is de ned as not suffering from chronic diseases.

= Yes
EQ-5D 0 = Complete health "Complete health" is de ned as EQ-5D score equal to one point, while "incomplete health" is de ned as EQ-5D score less than one point.

Data analysis
All statistical analyses were performed with Stata v14.0 software (Stata Corp, College Station, TX, USA) including descriptive statistics and bivariate and multivariate logistic regression analyses. P < 0.05 was statistically signi cant in all tests. We rst evaluated whether there were statistically signi cant differences between respondents from urban and rural areas with the chisquared test. We then calculated the rate and quantity of smoking and drinking in each year, and analyze the trends from 2008 to 2018. As smoking and drinking were binary variables, we used a bivariate logistic regression model to analyze factors in uencing these behaviors based on the odds ratio (OR); and as smoking quantity and drinking quantity were multi-category variables, we used a multivariate logistic regression model to analyze the in uencing factors based on the relative risk ratio (RRR).

Characteristics of the study population
The characteristics of the study population are shown in Table 2. Of the 24,939 respondents, 10,684 (42.84%) lived in urban areas and 14,255 (57.1%) in rural areas. The overall rate of smoking was 23.95%, with 5.75% light, 4.63% moderate, and 13.56% heavy smokers. Among residents of rural areas, 24.29% were smokers, with 5.55% light, 4.48% moderate, and 14.27% heavy smokers. In urban areas, 23.48% of the population smoked, with 6.03% light, 4.84% moderate, and 12.62% heavy smokers. Light and moderate smoking rates were signi cantly higher whereas the heavy smoking rate was lower in urban areas as compared to rural areas (p < 0.001). The overall rate of alcohol use was 23.29%, with 3.49% of the population engaging in excessive drinking. Drinking rates in urban and rural areas were 23.70% and 77.02%, respectively. The non-excessive drinking rate was higher in urban areas than in rural areas (20.83% vs 19.02%, p < 0.001), while the opposite was true for excessive drinking rate (2.86% vs 3.96%, p < 0.001).
Rural and urban populations differed signi cantly with respect to most of the independent variables. Education level and personal annual income were signi cantly higher for urban respondents than for rural respondents: the proportions of respondents with high or very high income were 30.8% and 30.06%, respectively, for the former group and 15.74% and 8.64%, respectively, for the latter. The proportions of urban respondents with a high school education or university or higher education level were 23.93% and 23.19%, respectively, which were signi cantly higher than the proportions of rural respondents (15.11% and 6.03%, respectively). The rates of employment and unemployment were signi cantly lower in urban areas (50.25% and 17.23%, respectively) than in rural areas (78.23% and 18.6%, respectively), while the proportion of retirees was higher in urban as compared to rural areas (32.52% vs 3.17%). The proportion of poor or low-security households was larger in rural areas (4.39%) than in urban areas (2.59%). In terms of health-related variables, the rate of chronic diseases was higher in the urban population (37.23%) than among rural residents (27.45%); the latter were also less likely to have participated in physical examinations (38.69% vs 57.35%).

Trends in smoking and alcohol consumption
The smoking rate increased from 23.95% in 2008 to 25.33% in 2013 but decreased to 25.33% in 2018 (Fig. 1). In 2008 and 2013, the smoking rate was higher in rural areas (26.11% and 24.4%, respectively) than in urban areas (24.41% and 22.45%, respectively), but this was reversed in 2018 (22.24% vs 23%). Similar trends were observed in light and moderate smoking rates (Figs. 2 and 3). However, the rate of heavy smoking showed a continuously declining trend over time (Fig. 4) higher in rural as compared to urban areas throughout the survey period; however, the rates in urban areas increased more dramatically from 2008 to 2013, such that the differences between urban and rural areas shrank from 2013 to 2018.
Factors associated with smoking rate and quantity After controlling for confounding variables, we found that smoking rate and quantity differed signi cantly between rural and urban areas (Table 3). Rural respondents were 30% less likely to smoke (OR = 0.07), 22% less likely to be light smokers (RRR = 0.78), 36% less likely to be moderate smokers (RRR = 0.64), and 30% less likely to be heavy smokers (RRR = 0.3). Other demographic factors also in uenced smoking rate and quantity. Older respondents (≥ 46 years) were more likely to smoke, less likely to be light or moderate smokers, and more likely to be heavy smokers than respondents who were ≤ 45 years old. Smoking was less common in women than in men (OR = 0.02), and the rates of light, moderate, and heavy smoking were lower in women than in men (RRR = 0.18, 0.01, and 0.01, respectively). Married people were more likely to smoke than those who were unmarried (OR = 2.56); this was true for light (RRR = 1.24), moderate (RRR = 10.07), and heavy (RRR = 4.53) smoking. People with social health insurance were less likely to smoke (OR = 0.7), whether lightly (RRR = 0.58), moderately (RRR = 0.72), or heavily (RRR = 0.7). Socioeconomic status in uenced smoking rate and quantity: compared to people with very low income, those with a very high income were more likely to smoke (OR = 1.18) and to be moderate (RRR = 1.26) or heavy (RRR = 1.28) smokers. There was a signi cant inverse correlation between smoking rate and education level, with an especially close correlation observed for the rate of heavy smoking. Compared to respondents who were employed, those who were unemployed or retired people were ~ 50% less likely to smoke, whether lightly, moderately, or heavily.
People with multiple chronic diseases were less likely to smoke (OR = 0.78) and smoke heavily (RRR = 0.75). Compared to respondents who were in very good health, people with poor health were signi cantly less likely to smoke (OR = 0.57) and smoke heavily (RRR = 0.49). EQ-5D scores also indicated that the rates of smoking (OR = 0.81) and heavy smoking (RRR = 0.75) were signi cantly lower in respondents with incomplete health than in those with complete health. Compared to people who did not have regular physical examinations, those who had undergone a physical examination in the previous 12 months were less likely to smoke (OR = 0.91) and smoke heavily (RRR = 0.9). People who did not engage in physical exercise were more likely to smoke (OR = 1.35)-including smoking moderately (RRR = 1.24) and heavily (RRR = 1.47)-than those who exercised regularly. Factors associated with drinking rate and quantity Demographic variables including age, sex, and marital status were signi cantly associated with drinking rate and quantity (Table 4). People ≥ 46 years old were more likely to consume alcohol than those ≤ 45 years old, whereas the 46-59 year age group was more likely to drink excessively than people ≤ 45 years (RRR = 1.55). Women were less likely to drink than men (OR = 0.05), including drinking excessively (RRR = 0.01). Married people were more likely to drink than those who were unmarried people (OR = 3.26) and to engage in non-excessive (RRR = 3.2) and excessive (RRR = 3.49) drinking. Alcohol consumption was positively correlated with income level and negatively correlated with education level. People with high school-level education or higher were less likely to drink than those with primary school or lower education, whether this consisted of non-excessive or excessive drinking.
Compared to employed respondents, those who were retired (OR = 0.64) or unemployed (OR = 0.73) were less likely to drink and to drink excessively (RRR = 0.54, RRR = 0.57).
Chronic disease, health status, and physical exercise were signi cantly associated with drinking rate and quantity. People with a chronic disease were signi cantly more likely to drink excessively than those without a chronic disease (RRR = 1.18), while those with ≥ 2 chronic diseases were signi cantly less likely to drink than those without chronic diseases (OR = 0.73), whether the drinking was excessive (RRR = 0.49) or not (RRR = 0.76). A lower level of health was associated with a lower alcohol consumption rate. People who did little physical exercise were more likely to drink and drink excessively than those who exercised regularly (OR = 1.10, RRR = 1.25).

Discussion
In this study, we analyzed pooled cross-sectional data from the NHSS of Jiangsu Province to identify trends in smoking and alcohol consumption and associated socioeconomic factors. The overall smoking rate and light and moderate smoking rates were higher in rural as compared to urban areas from 2008 to 2018, from 2008 to 2013, the rates increased more precipitously in urban areas such that by 2018, the difference between the 2 locations had shrunk. In Chinese culture, drinking is an important way to relieve stress and socialize, and many people regard drinking as a normal habit in rural areas or as serving a social facilitator role in city life. Thus, in order to reduce the rates of alcohol-related diseases and injuries, regulatory policies restricting alcohol use as well as health education are needed. Income level was signi cantly associated with smoking and drinking rates and quantities.
Smoking rate was reported to be higher among people who were poor than among the wealthy [27,28]. However, in accordance with previous ndings [29], we found that income level was positively associated with smoking rate, with the highest income group having more smokers, especially moderate and heavy smokers. The rate of alcohol use-especially excessive drinking-also increased with income level, which is supported by earlier studies [16,30,31]. A possible explanation for these observations is the rapid economic development in China and associated increase in income levels, which has led to increased discretionary spending on tobacco and alcohol, especially among people who lack awareness of the adverse health consequences.
We found that education level was signi cantly associated with smoking and drinking rates: respondents with a higher education level were less likely to smoke and drink (and engage in heavy smoking or excessive drinking) than those with a lower education level, which is consistent with earlier ndings [12][13][14]32]. A higher educational level may be associated with a greater capacity for translating information into behaviors-in this case, controlling smoking and alcohol consumption. Therefore, improving education level is one strategy to increase health awareness and discourage tobacco and alcohol use in the Chinese population. Employment status was also associated with smoking and drinking behavior: both smoking and drinking rates were higher in employed respondents compared to those who were unemployed or retired, which could be related to pressure to socialize and engage with others (colleagues etc) through drinking. In terms of demographic and health-related variables, older people, men, and married people were more likely to smoke and drink alcohol than young people, women, and unmarried people, which is in agreement with other reports [33][34][35]. Heavy smoking and excessive drinking were also positively correlated with age. This may be because older people in China typically have a low level of education and are thus less likely to be aware of the negative health effects of smoking and drinking, and instead consider these as normal life activities. Additionally, smoking and drinking in women is poorly regarded in Chinese culture; hence, these behaviors are far more common in men. Finally, people who were in poor health (ie, had chronic disease[s]) were less likely to smoke and drink; this is likely because these individuals were incapable of tolerating the effects of tobacco and alcohol.
Compared to 2008, the rate of light and moderate smoking increased signi cantly in 2013 and 2018, while the rate of heavy smoking decreased. The latter likely re ects the implementation of tobacco control measures in China as well as increased health awareness in the population. However, there is still no national legislation on smoking in China and the smoking control standards set by the WHO have not been reached. In addition, drinking rate and quantity increased in 2013; by 2018, the rates of drinking and non-excessive drinking had increased signi cantly but the rate of excessive drinking was unchanged. Thus, excessive alcohol consumption remains a serious problem in China but may be di cult to curb, given the importance of drinking as a means of social bonding in Chinese culture.
The results of our analyses indicate that measures are needed to control smoking and alcohol use and promote healthy behaviors in China. Firstly, the government should use public awareness campaigns targeting people with middle or high income levels or low education level, or those living in urban areas. Secondly, the education level of people in rural areas and health awareness among the elderly should be promoted. Thirdly, health facilities (eg, hospitals) should provide more support services such as smoking cessation programs. Fourthly, smoking bans in public places must be strictly enforced by the authorities. Finally, China's tobacco tax is relatively low compared to that of other countries; increasing taxes on tobacco and alcohol is one way to discourage their consumption.
Our study had certain limitations. Firstly, we analyzed only NHSS data from Jiangsu Province, which may limit the generalizability of the observed trends. Additionally, we examined correlations but were unable to make causal inferences regarding the data.
Finally, our regression analysis focused on individual variables but did not examine the effects of interactions among variables on tobacco and alcohol consumption [36,37].

Conclusions
Our results showed that the rate of heavy smoking declined from 2008 to 2018, while overall smoking and light and moderate smoking rates have recently decreased. Similar trends were observed in the rate of excessive alcohol consumption, but overall drinking rate increased continuously from 2008 to 2018. The trends in smoking and drinking rates were similar between urban and rural areas, although smoking rates were initially higher and then lower in rural areas as compared to urban areas. While drinking rates were higher in rural areas throughout the survey period, the rates rose steeply in urban areas such that the difference between the 2 locations had decreased by 2018. Socioeconomic factors such as income and education levels and work status as well as demographic and health-related variables in uenced smoking and alcohol consumption rates. Preventive measures such as increasing education and awareness in key groups and providing technical and support services to control smoking and alcohol use are needed to more effectively promote public health in China.  Rates of moderate smoking from 2008-2018.