A conceptual framework for the study is given in Fig. 1 illustrating hypothetical associations between feeding control practices and factors such as parental perceptions, children’s characteristics and socio-economic factors.
Study sites and participants
In 2013, two cross-sectional studies were conducted in two Health and Demographic Surveillance Sites, one in DodaLab in Dong Da district in central urban Hanoi, and one in rural FilaBavi in the Ba Vi district (later incorporated into Hanoi). In total, there are approximately 352,000 inhabitants living in Dong Da and the socio-economic characteristics are typical for large urban cities of Vietnam [20]. Ba Vi with a population of 250,000 inhabitants, mainly farmers, is situated about 60 km from the Hanoi city center and is representative of rural districts. In all, 1364 urban children, 725 boys (53.2 %) and 639 girls (46.8 %), and 1313 rural children, 705 boys (53.7 %) and 608 girls (46.3 %), aged from 3 to 6 years participated in the studies. The mothers and fathers were separately interviewed to obtain information about or related to the feeding of the children. In the urban area, 1346 mothers and 1303 fathers completed the interview. For 1299 (95.2 %) of the children, answers from both parents were available. In the rural area, the number of mothers and fathers were 1292 and 1248, and 1242 (94.6 %) children had answers from both parents.
Data collection
Two sources of information about the children and their parents were used; the specific interviews using the CFQ questionnaire (see below) and the information routinely collected in the HDSS, also obtained through interviews. The interview procedure and quality control was the same in both cases. The interviewers were trained to ensure that interviewees understood the questions exactly and to use a suitable language in the interview without changing the meaning. Supervisors in each of the two areas were responsible for supervising the work of the interviewers and for the data quality control. They checked randomly about 3 % of the anthropometric measurements and records. In the case of large differences, the household was revisited by a third person and the information was adjusted accordingly.
The following information was collected:
Perception of responsibility and parental feeding practices
The Child Feeding Questionnaire (CFQ) developed by Birch et al. [10] contains 7 subscales assessing parental attitudes, beliefs and practices related to child feeding. It has been validated in a number of studies involving parents of 2 to 11-year-old children [16, 21, 22]. The questionnaire was translated from English into Vietnamese and piloted to make sure that it was understandable. Four subscales of the CFQ were used to obtain the following information: (i) perceived responsibility (3 items), assessing parental responsibility for feeding the child; (ii) restriction (6 items), measuring the extent to which parents limit children’s access to unhealthy food (e.g. sweets, high-fat food, junk food). Two questions about using food as reward were omitted from the original restriction subscale for logical reasons; (iii) pressure to eat (4 items), measuring the extent to which parents press their children to eat large portions of food at mealtime; and (iv) monitoring (3 items), measuring the degree of parents’ supervision of their children’s food intake. All items were measured using 5-point Likert scales ranging from “disagree” to “agree” or “never” to “always”. High scores indicated more responsibility or greater restriction, pressure to eat and monitoring.
Perception of children’s weight and mothers’ weight
Two questions taken from the CFQ were used: “How would you describe your weight at present” and “How would you describe your child’s weight at present”. The possible answers were “Markedly underweight”, “Underweight”, “Normal”, “Overweight” and “Obesity”.
Children’s diet
Information about diet or children’s level of food consumption was obtained during the same interviews using questions specifically designed for this study. Two variables were defined:
Amount of food
Food intake at each main meal was compared to other children in the same age group and was classified by the parents at one of 3 levels: “less than”, “the same” and “more than”.
Fatty food, sweets and snacks consumption
“Fatty food” includes fatty meat and butter. “Sweets” in the study means candy, chocolate, ice cream. “Snacks” refers to potato chips, popcorn, twisties. Frequencies of fatty food, sweets and snacks intake were categorized by the parents into 6 levels: (1) never or less than once a week, (2) 1–3 times/week, (3) 4–6 times/week, (4) 1 time/day, (5) 2 times/day and (6) 3 times or more/day.
Socio-economic factors
Information about demographic and socio-economic status of the mothers and fathers was obtained from the HDSS databases in 2012. Mothers’ age, level of education and the family economic status were the factors used in the present study.
Mothers’ education was categorized into three levels (1) secondary school or less, (2) high school and (3) higher than high school.
Family economic status: The number of assets in the household was used as an indicator of family economic status. A list of assets included a refrigerator, television, videocassette recorder, gas cooking plate, telephone, mobile phone, personal computer, air conditioner, water heaters, bicycle, motorbike and car.
Anthropometric measurements
Weight (kg) and height (cm) of the children and their parents were measured in their homes by well-trained staff working in pairs. Digital Tanita scales and mobile measurement instruments were used.
BMI (body mass index) (kg/m2) was calculated as weight (in kg) divided by squared height (in m).
Statistical analysis
Conventional statistical methods were used to summarize and describe data in tables and graphs. The key dependent variables were the mean scores over the subscales for restriction, pressure and monitoring. Simple Pearson correlation coefficients were used to describe associations. Correlations between obviously skewed variables were checked using Spearman rank correlations. To compare the questionnaire responses between mothers and fathers taking education differences into account we used a stratification approach.
To address psychometric considerations about reliability and construct validity of the questions we used the Cronbach’s alpha coefficient which measures the correlations between variables within a subscale and can take values between 0 and 1. To demonstrate reliability it shall be reasonably high. Alpha values above 0.70 are conventionally considered acceptable.
A complete assessment of construct validity of the Birch questionnaire would require a confirmatory factor analysis (Structural Equations Modelling). We restricted the validity considerations to the study of the correlation matrices and the results of an exploratory factor analysis with oblique Promax rotation.
Most analyses were carried out separately for the urban and rural sites using the computer software Stata version 12.
Ethical considerations
The field sites in FilaBavi and DodaLab have received ethical approvals from the Ministry of Health of Vietnam as well as from the Scientific and Ethical Committee of Hanoi Medical University. There has been a written approval from authorities of the Dong Da and Ba Vi districts allowing research to be carried out in the sites. People living in the sites have also given their consent to routinely provide background information such as demographic information and household’ socioeconomic information as well as to participate in specific studies implemented in the sites. Our study was an observational study so the Scientific and Ethical Committee of Hanoi Medical University have approved us to work with oral consent from participants.