Evaluation of Bangladesh Healthy Eating Index (BD-HEI) | BMC Nutrition

Conduction of community nutrition survey and dietary assessment

Study design, sample size, and sampling

Following a two-stage cluster sampling procedure, this cross-sectional survey covered 1080 reproductive-aged women from different rural and urban areas of Bangladesh. The whole country was categorized into four distinct regions based on agricultural productivity and composite scores: low, moderately low, moderately high, and highly productive regions [11, 12]. From each region, one district was selected randomly. Then, one urban and two rural areas were randomly selected from each district. Each selected area was divided into three Enumeration areas (EA). Therefore, in each district, there were nine EAs (six from rural areas and three from urban areas). Hence, a total of 36 EAs were obtained from four districts. A total of 30 households were randomly selected from each of the 36 EAs. Thus, the survey covered 1080 households from different rural and urban areas of Bangladesh. From each household, one reproductive-aged woman was interviewed who is primarily responsible for household food preparation. In the Bangladesh context, this group is primarily responsible for household food preparation, child care, and they are also somewhat nutritionally vulnerable than others [13,14,15]. Figure 1 elicits the details of the sampling procedure, which has been elaborated by Jubayer et al. 2024 [10].

Our entire sample of 1,080 households was found to have sufficient power to identify variations in important food security and nutritional indicators across geographic and agricultural productivity strata in order to ensure statistical reliability. In particular, the sample enables the identification of a 10% regional variation in binary outcomes (such as the prevalence of food insecurity), assuming 80% power, 95% confidence level, design effect of 2 (to account for clustering). This is in line with the methodology of Lemeshow et al. (1990) [16], which offers recommendations for figuring out the minimum sample sizes in health surveys that use cluster sampling.

Additionally, practical considerations and previous experiences with large-scale national surveys, like the Bangladesh Demographic and Health Survey (BDHS) and Multiple Indicator Cluster Surveys (MICS), which frequently use 25–30 samples per cluster, influenced the decision to choose 30 households per Enumeration Area (EA). This figure strikes a compromise between logistical viability and statistical efficiency, permitting adequate within-cluster variability while preserving manageable fieldwork procedures.

Data collection

A structured questionnaire was developed for recording dietary and associated socio-demographic information. We referred to earlier national surveys, including the National Micronutrient Survey, the Bangladesh Integrated Household Survey, and other small-scale surveys carried out in comparable contexts. In addition, when drafting the questionnaire, we conferred with a number of nutrition specialists from the Institute of Nutrition and Food Science, University of Dhaka. The English version of the questionnaire is attached in Supplementary File 2. To ensure the validity of the questionnaire, a pilot survey was carried out before the final data collection. Upon receiving a two-day-long training session, nutrition graduates and supervisors were appointed to collect the data. Dietary data were collected using a quantitative multiple-pass 24-h dietary recall method. 24 h dietary data of two consecutive days were collected from 355 respondents (almost one-third of the total sample). Reproductive-aged women responsible for household food preparation reported the weight or portion size of foods consumed, both at home and outside. Enumerators used measuring cups and kitchen scales to assist with accurate portion estimation.

Dietary data analysis

Grouping of individual food items into a food group

Food items consumed by the respondents were categorized into different food groups according to the national food-based dietary guidelines: Cereal, Roots and tuber, Pulse, Fish/meat/egg, Milk and milk products, Fats and oil, Fruits, Leafy Vegetables, Non- leafy Vegetables, Sugar, and Salt.

Calculation of energy and nutrient intake

To match the food composition table, raw weights of consumed foods were required. For items like rice, raw weight was calculated using a cooked-to-raw conversion factor. For foods eaten directly, like fruits and vegetables, the edible coefficient was applied to determine the final raw weight. Based on food intake data, usual energy, and nutrient intake were calculated using the food composition table for Bangladesh [17]. The current study utilized the Multiple source method (MSM) to calculate the usual food and nutrient intake distribution [18]. It is a web-based application accessible via the MSM website (https://nugo.dife.de/msm/).

Development of BD-HEI

Identification of index component

BD-HEI component was identified and categorized into adequacy, optimum, or moderation components based on the healthier options provided in the FBDG, prepared by the Ministry of Food and the Ministry of Health and Family Welfare of the Government of Bangladesh. Table 1 provides a summary of the 11 components of the BD-HEI. Each component comprises different food items and has distinct scoring criteria [10].

Components of each food group

Each food group comprises different food items, as given in Table 2. Food items were selected based on FBDG as stated by Jubayer et al., 2024 [10].

Weighing, standard, and scoring of the index component

Every index component had an equal weight, and scores ranged from 0 to 5 or 10. Leafy and non-leafy vegetables were separated into two sub-components, which equally divided the overall score of 10. Similarly, cereals, roots, and tubers were grouped, with their score of 10 distributed proportionally. The remaining nine components each received a score of 10, resulting in a total BD-HEI score between 0 and 90, —with higher scores indicating better diet quality [10]. Three rating systems were used for adequacy, moderation, and optimum components, as shown in Fig. 2 and Table 1.

Graphical presentation of BD-HEI for (a) adequacy category, b moderation category, and c optimum category. This figure is adapted from Looman et al., 2017 [20]

Scoring for the adequacy component

Participants scored 0 for no consumption, while the maximum score was given for meeting or exceeding the recommended servings — 10 for fruits and 5 for each vegetable sub-component (leafy and non-leafy). A formula (Eq. 1) adapted from Bekele et al., 2019 [19] calculated scores for intakes below the recommended portions.

Scoring for the optimum component

These dietary components should be consumed within an optimal range. The maximum score (5 or 10) was given for intake within the recommended range, while a score of 0 was assigned for no consumption or intake exceeding the upper limit (Threshold value). Equation 2a was used when consumption was less than the lower limit of the recommended range, and Eq. 2b exceeded the upper limit of the recommended range.

Scoring for the moderation component

Unhealthy food items are placed into this category; thus, less consumption carries a higher score. A maximum score (ten) was assigned if consumption was less than the recommended servings. Zero was taken when consumption exceeded the highest consumption limit (threshold value). When participants consumed more than the recommended servings, scores were calculated using the following equation:

Analytical approach for measuring validity and reliability of the index

Table 3 summarizes the strategies used to evaluate the construct validity and reliability of BD-HEI. Jubayer et al., 2024 [10] suggested to follow the same approaches. Before conducting statistical analysis, normality of the total HEI score was investigated using statistical test- Kolmogorov–Smirnov and Shapiro–Wilk test as well as visual inspection of the histogram, Q-Q plot. It has been found that the HEI score was approximately normally distributed.

Study variables

Table 4 lists the major study background variables. The details of these variable measures are described in the Supplementary File 1.