Explaining Income-Related Inequalities in Dietary Knowledge: Evidence from the China Health and Nutrition Survey

Lack of adequate dietary knowledge may result in poor health conditions. This study aims to measure income-related inequality in dietary knowledge, and to explain the sources of the inequality. Data were from the China Health and Nutrition Survey (CHNS) conducted in 2015. A summary of the dietary knowledge score and dietary guideline awareness was used to measure the dietary knowledge of respondents. The concentration index was employed as a measure of socioeconomic inequality and was decomposed into its determining factors. The study found that the proportion of respondents who correctly answered questions on dietary knowledge was significantly low for some questions. Compared to rural residents, urban residents had a higher proportion of correctly answered dietary knowledge questions. In addition, there are pro-rich inequalities in dietary knowledge. This observed inequality is determined not only by individual factors but also high-level area factors. Our study recommends that future dietary education programs could take different strategies for individuals with different educational levels and focus more on disadvantaged people. It would be beneficial to consider local dietary habits in developing education materials.


Introduction
Overweight and obesity are a worldwide public concern. The World Health Organization (WHO) estimated that 39% of women and 39% of men aged 18 and over were overweight in 2016 [1]. Previous evidence has shown that an unhealthy diet was admittedly one of the critical contributors to obesity. Previous systematic reviews which focused on the relationship between the nutrition knowledge and dietary intake showed that, as an integral component of health literacy, low nutrition knowledge has a significant association with the poor eating habits and bad health conditions [2]. Health literacy includes the level of knowledge, personal skills and confidence to take action to improve personal and community health by changing personal lifestyles and living conditions [3].
As result of adopting open up policy and reformation in recent decades, China has experienced expansive economic growth. The dietary problems remarkably transited from malnutrition due to food scarcity in whole country to the excessive and unbalanced nutrition due to dietary structure and better off having more dietary related knowledge in China, and many determinants, such as income, education level, and geographic location, are contributors to this income-related inequality. This study will contribute to current studies by providing readers with a preliminarily understanding of the inequality in dietary knowledge. This study will support policy makers and public health workers with evidence-based information to address inequality as well as in developing health policy and public intervention measures.

Study Design
This is a cross-sectional study designed to measure the degree of inequality in dietary knowledge and decompose this inequality into its determinants.

Data
This study used data from the China Health and Nutrition Survey (CHNS). The CHNS is an ongoing international collaborative project between the Carolina Population Center at the University of North Carolina at Chapel Hill and the National Institute of Nutrition and Food Safety at the Chinese Center for Disease Control and Prevention. The CHNS collects the nutritional and health status, as well as the demographic and socioeconomic data of Chinese population [16]. The CHNS was designed to examine the effects of health policies and programs on the health and nutritional status of the Chinese population. The first round of the CHNS was conducted in 1989, and its successive rounds of the survey were conducted every 2−4 years. The latest survey, which included 20,914 individuals in 7319 households, was conducted in 2015. A multi-stage random cluster process was used to draw samples of approximately 7200 households, with over 30,000 individuals in 15 provinces and municipal cities. The provinces and cities vary substantially in geography, economic development, public resources, and health indicators. In each survey, a household interview, a health examination for children and adults, a nutrition survey, and a community survey were conducted simultaneously. The health examinations and household surveys were conducted in the participant's home, while the community survey was collected by interviewing the neighborhood officials. The CHNS raw data are accessible through its official website (cpc.unc.edu/projects/china). Detailed information on survey contents and quality control procedures can also be found in the official website. This study focused on data of adult individuals from the latest round of the survey. Inclusion criteria were: (1) adults aged 18 years and over; (2) adults who can answer the questions independently. Exclusion criteria were: (1) households with missing data on health consumption and overall consumption expenditure; (2) respondents with missing data on studied variables. After data cleaning, 12,259 respondents in total were included in this study.

The Dietary Knowledge Score
In the 2015 CHNS survey, 16 questions were preselected to assess the dietary knowledge of respondents. In each question, respondents were asked to choose "strongly disagree", "disagree", "neutral", "agree" and "strongly agree". For easier comparison with future studies, only 12 dietary knowledge items were selected in this study. In regression analysis, a summary dietary knowledge score was constructed by summing up the score of twelve questions. For each answer, we have constructed a score ranging from 0 to 2, with 2 for strongly agree, 1 for agree, and 0 for other answers in positive questions, and 2 for strongly disagree, 1 for disagree, and 0 for other answers in negative questions. The maximum dietary knowledge score is 24. A higher score indicates better knowledge.

Dietary Guideline Awareness
Dietary guideline awareness is a binary variable, indicating whether the respondents know about the Dietary Guidelines for Chinese Residents, which is known as The Chinese Pagoda. The Dietary Guidelines for Chinese Residents is published by the National Health Commission of China. The Dietary Guidelines provides evidence-based food and beverage recommendations for Chinese aged 2 and over. The purpose of these recommendations is to promote population health, prevent the population from developing non-communicable diseases, help people to achieve a healthy weight, etc.

Independent Variables
A great number of variables are available in the CHNS. Referring to previous nutrition studies, some demographic, socioeconomic, residential area and health education intervention variables were associated with nutritional knowledge [9,10,17,18]. Since this study was designed to explore the contribution of socioeconomic factors on dietary knowledge inequality, we constructed a hypothesis that all the socioeconomic variables and potential control variables in the CHNS are all associated with dietary knowledge. After removing redundant variables, four types of independent variables were included in this study to empirically examine the relationship between these variables and the dietary knowledge score/dietary guideline awareness. Demographic characteristics considered in this study were sex and age. Socioeconomic characteristics considered in this study were marital status (unmarried, married, divorced or widowed), educational attainment (illiterate, elementary, middle school, high school, and university), working status, basic medical insurance, and economic status (grouped equally into the poorest, poorer, middle, richer, and the richest quintiles according to the per capita net household expenditure). A set of location variables was included to capture the potential regional heterogeneity. These variables were geographic location (Eastern China, Central China, and Western China), residential area (urban or rural), and birthplace (North China, Northeast, East China, Central China, South China, and Western China). Lastly, the urbanization index, a comprehensive composite variable to measure the degree of urbanization on 12 aspects of urbanization in communities, was also included. The aspects of the urbanization index are population density, economic activity, traditional markets, modern markets, transportation, health infrastructure, sanitation, communication, social services, diversity and housing. Detailed information on this variable is available on the official CHNS website [19][20][21].

Statistical Analysis
Categorical variables are described as proportions (number and percentage), continuous variables are described as means and standard deviation (SD). The chi-square test was performed to compare the difference in dietary guideline awareness among different groups. One-way analysis of variance (ANOVA) and the t-test were used to assess the difference in the dietary knowledge score among groups. The post-hoc multiple comparison procedure was further employed to compare between subgroups.
The concentration index and the related concentration curve have become the most frequently used measures to quantify income-related inequalities in the health domain [22,23]. The concentration curve is displayed graphically, with the cumulative percentage of the population in income on the horizontal axis and the cumulative percentage of the health sector variable on vertical axis (for this study, the health sector variables are mentioned in the variables section). If the health sector variable has higher (lower) values among poorer people, the concentration curve will lie above (below) the line of equality. In the case that the health sector variables are distributed equally among all income groups, the concentration index would coincide with the 45-degree line [24,25]. The concentration index was employed to measure inequality in the health sector variables over the distribution of income. The concentration index is defined as twice the area between the concentration curve and the line of equality (the 45-degree line) [26]. The concentration index value is bound between −1 and +1.
The negative concentration index indicated that the health sector variables were disproportionately concentrated among the poor, and vice versa [27]. The concentration index formula is as follows: where c is the concentration index, y i is the health sector variable of individual i, R i is the fractional ranks of income, and µ is the mean of y i . The health concentration index can be decomposed into the contributions of each determinant to observed income-related health inequality-in which, each contribution is the product of the sensitivity of health with respect to that determinant and the degree of income-related inequality in that determinant [28]. Health economists normally put forward the different decomposition methods in the literature to decompose the concentration index. In this study, we employed the decomposition approach from the Wagstaff. In the case that the health sector variables are continuous variables, the decomposition approach based on ordinary least squares regression was commonly adopted [29]. The OLS regression equation between y and x can be written as: where y i is the health sector for i, β k is the coefficient for kth variable, x k is the kth independent variable, and ε i is the residual error.
The concentration index for y, can be written as: where c is the overall concentration index of health sector variable, x k and c k are the mean and the concentration index of x k , and GC ε is the generalized concentration index for ε. The equation shows that the concentration index is made up of two components: the explained component and residual variation (unexplained) component that was not explained by any of independent variables. In the health sector, outcome variables are seldom continuous and are often binary. In that case, a marginal effect based on logistic regression could be employed to approximate the decomposition analysis. The logistic regression can be written as follows: where logit(P) is a linear function of independent variable, β k is the coefficient for kth variable, x k is the kth independent variable, and ε i is the residual error.
A linear approximation of the non-linear estimations using a marginal effect obtained from logistic regression can be written as follows: where β k m is the marginal effect for kth variable and u i is the error generated by the linear approximation used to obtain the marginal effects. Marginal or partial effects have been analyzed in the analysis of health sector inequalities in non-linear settings. In this case, the concentration index also can be decomposed into two parts, the explained component and the residual variation component More details on decomposition methods when the dependent variable is binary can be found in previous studies [26,30].
All statistical analyses were performed with SAS version 9.4 and Stata/MP version 14.0. Table 1 shows the characteristics of respondents. A total of 12,208 adults, including 5746 (47.07%) men and 6462 (52.93%) women, participated in the study, with a mean (standard deviation, SD) age of 52.59 (15.31) years, and most of them were married. Over 33% of respondents had middle school education. In total, 97.31% of respondents were covered by basic medical insurance and 39.93% lived in urban areas.   Table 2 shows the means of the dietary knowledge score and dietary guideline awareness by study variables. Univariate analyses showed that there was a statistically significant difference in the dietary knowledge score and dietary guideline awareness among age, income, educational attainment, marital status, working status, residential areas, and geographic region. There was no significant difference in the dietary knowledge score and dietary guideline awareness between women and men. Further post-hoc comparison analysis showed that there was a statistical difference in the dietary knowledge score between Central China and Western China. However, no difference was found in dietary guideline awareness.   Table 2 shows the means of the dietary knowledge score and dietary guideline awareness by study variables. Univariate analyses showed that there was a statistically significant difference in the dietary knowledge score and dietary guideline awareness among age, income, educational attainment, marital status, working status, residential areas, and geographic region. There was no significant difference in the dietary knowledge score and dietary guideline awareness between women and men. Further post-hoc comparison analysis showed that there was a statistical difference in the dietary knowledge score between Central China and Western China. However, no difference was found in dietary guideline awareness. Table 3 shows the proportion of respondents who correctly answered dietary knowledge by each question. Urban residents had a higher proportion of correctly answered questions compared to rural residents, except two. "Choosing a diet with a lot of staple foods is not good for one's health" was the question with lowest score. Only 2077 (42.61%) of urban residents and 3006 (40.99%) of rural residents answered the question correctly. Table 4 shows the association between determinants and the summary dietary knowledge score/dietary guideline awareness from OLS regression and logistic regression, respectively. Income and educational attainment had a positive association with the dietary knowledge score and dietary guideline awareness. An increase in income and education attainment significantly increased the dietary knowledge score and dietary guideline awareness. The higher the urbanization index, the higher the dietary knowledge score and dietary guideline awareness. Compared to urban residents, rural residents had a lower dietary knowledge score and dietary guideline awareness. Compared to respondents who lived in Eastern China, respondents who lived in Western and Central China had a lower dietary knowledge score and lower dietary guideline awareness.      Figure 2 shows the concentration curve and the concentration index on the dietary knowledge score and dietary guideline awareness. Both curves lie under the 45-degree line, and the corresponding concentration index was 0.054 ± 0.002 and 0.274 ± 0.008 for the dietary knowledge score and dietary guideline awareness, respectively. That means that there was pro-rich inequality in the dietary knowledge score and dietary guideline awareness. A higher dietary knowledge score and better dietary guideline awareness were more concentrated on the rich.   Table 5. Since the health sector variables of interest were concentrated among the rich, positive contribution means that these variables increase the degree of pro-rich inequality. That is, the rich were more likely to have a higher dietary knowledge score and better dietary guideline awareness. Results showed that the largest and the second largest absolute contributions were from income and educational attainment for the observed inequality of the dietary knowledge score and dietary guideline awareness. These positive contributions were slightly offset by negative contributions from marital status. Quantifying the contributions expressed as the percentage of each independent variable, urbanization index showed 11.868% and 9.135% contribution to the dietary knowledge score and dietary guideline awareness, respectively. Living in Western China contributed 19.46% and 8.16% to the total observed inequality in the dietary knowledge score and dietary guideline awareness, respectively.  Note: C, the concentration index; C K , the concentration index for the k th independent variable; ref, reference group.

Discussion
Increasing the dietary knowledge of the general population is an important way to improve their dietary habits and lifestyle while reducing incidences of obesity-related non-communicable diseases throughout the whole lifespan [31]. Using the large-scale CHNS data, this study measured the proportion of residents who were aware of dietary guidelines and who correctly answered the dietary knowledge questions. The study found that compared to rural residents, urban residents had a higher proportion of correctly answered dietary knowledge questions in nearly every question. Among all of the questions, "Choosing a diet with a lot of staple foods is not good for one's health" had the lowest proportion of residents with correct answers for both urban and rural areas. In China's history, the Chinese suffered from a long period of hard times and lack of food approximately half a century ago. In these hard times, staple food played a prominent role in supplying a large fraction of energy needs to people. Although China's economic growth has increased rapidly since the open and reform policy, the attitudes and beliefs have not changed much among Chinese people. Chinese people still maintain the traditional habit of eating a lot of staple food, such as rice and noodles. The main function of staple food is compensatory energy. Some staple food can be good for residents in moderation; however, excessive consumption may negatively affect general health. White rice and refined wheat contribute approximately 30% of energy to diets in China. Highly processed carbohydrates, such as white rice, may lead to a quick rise, high peak and fast fall in blood glucose, resulting in high glycemic index and glycemic load values [32]. Too much sugar contained in staple food is a predicator for type 2 diabetes [33]. Currently, most residents, especially in remote rural areas, are still not aware of this. Further actions need to be taken in future public health plans to educate residents to understand the risk of consuming excessive staple food. Despite most people having a clear understanding that physical activities have innumerable physical and mental health benefits, more than half of them do not realize that excessive intense physical activities can be harmful to their health, such as increasing cardiac risks [34]. Slightly larger gaps were observed between urban and rural residents in term of answering questions related to sugar consumption and fat consumption. Compared to other questions, these two questions showed a significant difference, with urban residents providing more correct answers than rural residents. Recent studies found that various fats influence health differently, e.g., unsaturated fats are considered to play a number of beneficial roles in health, while other fats, such as trans fats, are considered to do the opposite. There is still a lot that can be done to improve people knowledge regarding this topic, especially in rural areas.
Some key factors were identified as the determinants of the dietary knowledge score and dietary guideline awareness in this study. As expected, educational attainment was associated with the dietary knowledge score and dietary guideline awareness. People with a higher level of education are more likely to be health literate, take the initiative to seek dietary information for health considerations, achieve a better understanding of nutritional information from mass media, and achieve better informed judgment ability to distinguish between correct and incorrect commercial promotions. In this study, as a continuous variable, the urbanization index was used to explore the association between the dietary knowledge score and dietary guideline awareness. The study found that there were positive relationships between the dietary knowledge score/dietary guideline awareness and the urbanization index. A feasible reason for this is that residents who live in communities with a higher urbanization level are likely to have more nutritional education from school, and more opportunity to access to health knowledge correctly. Our unpublished data show that 38.65% of respondents who lived in communities with a higher urbanization index proactively looked for nutrition information, while this figure was only 15.55% in communities with a lower urbanization index. Less developed communities commonly use TV and radio as sources for nutritional knowledge. However, mass media is full of fast food advertisements, which usually has lower nutritional benefits. Combined with relatively poor analysis skills and judgment ability, residents who lived in lower urbanization communities are prone to lower dietary knowledge. Geographic region is another factor associated with dietary knowledge. Due to geographical conditions, residents in Central and Western China may lack the resources for better dietary knowledge, limiting the possibility to acquire nutrition knowledge from media, and less opportunity to visit dietitians. Regarding marital status, unlike previous studies, individuals who were married had a higher score on dietary knowledge and better dietary guideline awareness. That may be due to the fact that people who are married are most likely to share their nutrition information with each other [17]. This study confirmed the findings from previous studies that women had a higher dietary knowledge than men. Men are less interested in nutritional knowledge than women. Previous studies showed that more than 25% of men were unwilling to increase their knowledge about food and nutrition [35].
Previous studies on nutritional knowledge found that nutritional knowledge was not distributed equally among residents with different incomes [4]. The study also observed a dietary knowledge difference among income groups, and further measured the magnitude of this inequality. This study found that people with a higher income were more likely to have higher dietary knowledge. The implementation of dietary education may motivate residents to change their dietary behaviors and provide them with the knowledge and skills to make healthy food choices in the context of their lifestyles and economic resources. This study recommends future dietary education and promotion programs to provide different strategies for individuals with different educational levels. Furthermore, these programs should be made more available to disadvantaged people. The contents of education should be adapted to local dietary habits. This would make it easier for local residents to understand and adopt these practices. This study further explored the source of this observed inequality. As expected, income and educational attainment showed the largest proportion of contribution on pro-rich inequality in dietary knowledge sore and dietary guideline awareness. However, the contribution of sub-educational attainment groups was heterogeneous. Elementary and middle school education showed a negative contribution to this pro-rich inequality, whereas high school and university education showed a positive contribution to this kind of inequality. That is because the poor are more likely to have had elementary or middle school attainment, whereas the rich are more likely to have had high school and university education. The study showed that geographic region contributed a large proportion to observed inequality. The possible explanation for this is as follows. Compared to residents living in Eastern China, residents living in Central and Western China had a lower score of dietary knowledge and a lower likelihood of awareness of dietary guidelines. This may be due to the fact that Central and Western China are relatively less prosperous compared to the Eastern China. Increasing the economic investments in Central and Western China, which may lead to a reduced gap of income, could be a feasible way to decrease this kind of inequality.
This study has some limitations. Firstly, due to the cross-sectional study design, causal inference between dietary knowledge and independent variables cannot be established. Secondly, since all of the data were self-reported, respondents may not fully remember past events or experiences accurately or omit details. Therefore, recall bias is unavoidable. To limit recall bias, different recall periods were used in the CHNS. Thirdly, the determinants of the dietary knowledge score and dietary guideline awareness considered in the study were from the pre-specified questions in the survey. Other unobserved confounders were not considered in the regression model.

Conclusions
Using a national-scale survey, this study measured the dietary knowledge of Chinese adults. The study found that the proportion of respondents who correctly answered dietary knowledge questions was significantly low for some questions. Income, educational attainment, the urbanization index, and geographic region were the main determinants for both the dietary knowledge score and dietary guideline awareness. There was pro-rich inequality in the dietary knowledge score and dietary guideline awareness. Income and educational attainment appeared to be the first and second largest proportion of contrition on this pro-rich inequality. In addition, the urbanization index and geographic region also showed large contributions to the observed inequality. This study recommends that future dietary education incorporates different strategies for individuals with different educational levels, with more focus on disadvantaged people. It would be beneficial to consider local dietary habits in developing education materials. This approach could greatly assist local residents in understanding and adopting something they are familiar with to improve their knowledge and to embrace healthy diets and lifestyles.