Evaluating adherence to recommended diets in adults 1991–2015: revised China dietary guidelines index

All data used in this study were derived from the China Health and Nutrition Survey (CHNS) codeveloped by the National Institute for Nutrition and Health of the Chinese Center for Disease Control and Prevention and the University of North Carolina at Chapel Hill in the United States [24]. CHNS was a prospective study begun in 1989 (followed up in 1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011, 2015, and 2019) and was sampled by a multistage, stratified, clustered random method. The study included eight diverse provinces and autonomous regions from 1989 to 1997 and nine from 2000 to 2009. Based on the national administrative regions, a weighted sampling scheme was used to randomly select four counties and two cities in each province. Villages and townships within the counties and urban and suburban neighborhoods within the cities were selected randomly [24]. Townships and urban neighborhoods were grouped into the urban area, and villages and suburban neighborhoods were grouped into the rural area. More detailed sample and design are described elsewhere [24, 25]. Beijing, Chongqing, and Shanghai were added in 2011, and Shannxi, Yunnan, and Zhejiang provinces were added in 2015. The institutional review board of the University of North Carolina at Chapel Hill and the National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention, approved the study protocol (No. 201524). All of the participants signed the informed consents.

The data collected in 2019 are under manipulation. Adults aged 18–64 years with complete dietary, socioeconomic, and demographic data from 1991 to 2015 were involved in this study. Pregnant or lactating women, adults with energy intake less than 1600 kcal per day (kcal/d) or greater than 2400 kcal/d, and those with an implausible CDGI(2019)-A score (less than 1% of total population in each wave) were excluded. A total of 20,975 individuals living in 388 communities participated in this study. All of the participants signed the informed consents.

Consecutive 3d24h dietary recalls (two weekdays and one weekend day) were used in each wave to assess dietary intake at the individual level. Trained investigators interviewed each participant and then recorded the amounts and other details of all foods and drinks consumed except cooking oil and condiments. Household weighing method was used to obtain cooking oil and condiment consumption by all family members during the corresponding 3 days. The consumption of each participant was calculated according to the times cooked at home, meal proportions, and ratio of individual energy intake to household energy intake. The intake of carbohydrate and energy was calculated by means of China Food Composition [26]. The energy supply ratio of carbohydrates (%) = carbohydrates intake (g) × 4 kcal/g ÷ energy intake (kcal) × 100%. Salt intake consisted of NaCl from cooking salt, paste, and sauce.

Demographic information, income, physical activity, nutritional knowledge, and community status were all collected by trained investigators through face-to-face questionnaires. We asked the household annual income and the individual annual income of each participant each year and then calculated the household annual income per capita in different years, which were all inflated to 2015 according to the gross domestic product of each year to make them comparable. We defined whether the participants knew the CDG by their answering yes or no to this question: “Do you know the Chinese Dietary Guidelines?” The physical activity in our questionnaire had five levels: (1) very light; (2) light, meaning working in a standing position (e.g., salesperson, laboratory technician, teacher, etc.); (3) moderate (e.g., student, driver, electrician, metal worker, etc.); (4) heavy (e.g., farmer, dancer, steel worker, athlete, etc.); and (5) very heavy (e.g., loader, logger, miner, stonecutter, etc.). Body weight was measured by standard unified method and equipment. Physical activity level and body weight were used to calculate each participant’s energy requirement.

The urbanization index was established by Jones-Smith and Popkin and was based on twelve multidimensional components reflecting economic, social, demographic, and infrastructural diversity at the community level [27].

To align with the Chinese Dietary Reference Intakes 2013, we regrouped physical activity into three levels: very light and light physical activity were considered level 1, moderate physical activity level 2, and heavy and very heavy physical activity level 3. Energy requirement was calculated by the following formula: energy requirement (kcal/d) = basal energy expenditure (kcal/kg·d) × body weight (kg) × physical activity level values. Table 1 shows basal energy expenditure by age and gender. The physical activity level values were 1.50 for level 1, 1.75 for level 2, and 2.00 for level 3 [28].

Four of the six recommendations of CDG-2016 for the general population concerning food were selected as components of CDGI(2019)-A. CDGI(2019)-A was composed of thirteen food-related components and one nutrient-related component: 1) other cereals and miscellaneous beans, 2) total vegetables, 3) ratio of dark-colored vegetables to all vegetables, 4) fruits, 5) milk and dairy products, 6) soybeans, 7) nuts, 8) seafood, 9) poultry and meat, 10) eggs, 11) cooking oil, 12) salt, 13) alcohol, and 14) energy supply ratio of carbohydrate. “Other cereals” referred to cereals other than rice, flour, and their products. “Miscellaneous beans” referred to beans other than soybeans. “Dark-colored vegetables” referred to vegetables with carotene ≥500 μg/100 g [22]. “Energy supply ratio of carbohydrate” was used to evaluate the intake of cereals and tubers. Equivalent weight continuity scoring was used to assess the diet quality of adults in China.

The components of CDGI(2019)-A can be divided into three categories: (1) adequate intake (other cereals and miscellaneous beans, vegetables, fruits, milk and dairy products, soybeans, and nuts), (2) appropriate intake (seafood, poultry and lean meat, and eggs), and (3) limited intake (cooking oil, salt, and alcohol). For individuals with different energy requirements, neither the subtotal score of each component nor the recommended food intake amounts was equal, as detailed in Table 2. For the adequate intake foods, if intake was 0, the score was also 0; if intake was not less than the recommended amount, the score was full. For the appropriate intake foods, if intake was 0 or more than twice the recommendations, the score was 0; if intake was within the recommended ranges, the score was full. For limited intake foods, if intake was more than twice the recommendations, the score was 0; if intake was less than the recommendations, the score was full. In other cases, the score was calculated proportionally. The maximum score values were 5 for energy supply ratio of carbohydrates, 5 for other cereals and miscellaneous beans, 5 for vegetables, 5 for ratio of dark-colored vegetables to all vegetables, 10 for fruits, 10 for milk and dairy products, 5 for soybeans, 5 for nuts, 10 for seafood, 10 for poultry and lean meat, 10 for eggs, 10 for cooking oil, 10 for salt, and 10 for alcohol. The scoring method has been detailed elsewhere [23]. The total score, the sum of the fourteen components, was no more than 110 and was positive with diet quality.

SAS 9.4 and MLwiN 2.36 software were used in this study. Hierarchy of communities, individuals, and repeated measurements was caused by the sampling, and the CDGI(2019)-A scores (level 1) of individuals (level 2) living in the same community (level 3) for each year were not independent. Therefore, traditional analysis of variance was not applicable, so three-level multilevel models were employed in this study, implemented by MLwiN 2.36 software.

Survey year was variable at level 1. Age, gender, education, income, and awareness of the CDG were variables at individual level (level 2), of which income was in quartile. Urbanization index was variable at community level (level 3). First, a two-level null model and a three-level null model without any independent variables were fitted to analyze the intra-class correlation and variation at different levels. Then, a three-level random intercept growth model including survey years as category variate was fitted to explore the trend of total score. Subsequently, a three-level random intercept growth model, including covariates at both individual and community levels, was fitted to analyze the determinants of CDGI(2019)-A scores. Finally, a three-level random intercept-slope growth model was fitted to analyze the variation of impact of covariates on total score at the community level.

The log likelihood ratio test was used to compare the goodness of fit between models, and the Wald χ² test was used to test coefficients and variances in models. Standardized residual × normal scores plots were used to test the assumption of normal distribution at different levels.