The purpose of the Barbados Food Consumption and Anthropometric Survey (2000) was to determine the food intake and dietary patterns of the country's adult population, the prevalence of obesity and diagnosed nutrition-related CNCDs, and the relationship between these and dietary patterns. The information obtained through this study will enable NNC to plan and execute various targeted intervention programmes with the aims of increasing the population's levels of awareness of health-related risks and of actually reducing those risks by adopting more appropriate eating and lifestyle patterns.
The specific objectives of the survey were to:
determine the nutritional adequacy of Barbadians' diets;
examine food choices, food purchasing patterns and food sources;
assess the extent of home food production and its contribution to household diets;
determine the prevalence of obesity and diagnosed nutrition-related CNCDs, and their relationship to dietary patterns;
determine whether there are significant pockets of undernutrition in Barbados;
investigate the association between demographic, social and economic characteristics and dietary practices.
This survey was designed to provide national results at a 3 percent tolerance with a 95 percent degree of confidence. The sample design was a systematic random two-stage cluster design.
SAMPLE SELECTION
Selection of households
The first stage involved the selection of 42 Enumeration Districts (EDs) from a total of 547. In the second stage, 13 clusters of three households each were selected systematically within each of the 42 selected EDs. Each cluster consisted of a primary household and two secondary households. The secondary households were the households immediately before and immediately after the primary household. A total of 1 638 households were selected, and 1 051 were in the final sample (see the section on Recommendations for future surveys: difficulties encountered and limitations of the survey). Enumerators reported that several of the houses were vacant, or occupied by tourists on a short-term basis. The occupants of some houses refused to participate.
The sample was selected from the EDs that were used by the Barbados Statistical Service (BSS) for the population census in May 2000. These EDs had been created to include approximately 150 households each - EDs used in previous population censuses were divided in order to ensure this. Each ED thus had an equal probability of selection.
The EDs were first ordered from the northwestern corner of Barbados, moving in a serpentine fashion to the east and south, with the last ED listed being located in the southeastern corner. Only households in these EDs were enumerated. In each selected ED, the 13 clusters of three households were selected systematically with a random start.
Selection of individuals
Two people of at least 18 years of age were interviewed in each household, except in one-person households or in those where only one person was 18 years or more. The head of the household or the person who prepared the meal was interviewed for the household questionnaire and the first individual questionnaire. A second individual questionnaire was administered to a second person who was selected randomly according to the rules set out in the survey manual.
In general, interviewers conducted household visits in pairs to facilitate the weighing and measuring of the respondents, but this practice was not always possible and some households were visited by only one enumerator. Interviewers were advised to leave a callback card at the households of people who were not at home asking them to contact NNC; however none of these prospective respondents called. The interviewer made two follow-up visits, and if contact was not established, "refusal" was indicated on the questionnaire, which was given to the supervisor.
THE FIELD TEAM
The survey team consisted of three supervisors, to each of whom was assigned a group of 12 enumerators at the start of the data collection phase. The enumerators included community nutrition officers, a dietetic technician, a school meals supervisor, nurses, an environmental health officer, agricultural field officers and field and supervisory staff who had recently been employed on other surveys.
It was the responsibility of each supervisor to identify and indicate the 13 clusters of households in each ED to the enumerators. In most cases, two enumerators visited each household, but there were instances when only one interviewer visited in order to ensure that the identified respondents were interviewed as scheduled. Interviews were conducted in the evenings and at weekends, as well as during regular working hours.
STANDARDIZATION AND TRAINING
Three standardization and training workshops were conducted for interviewers, who were taught skills for conducting interviews and for collecting anthropometric and dietary intake information. Each received a copy of the Interview manual for Barbados, which was the reference document during the sessions. Consultants from FAO and the Caribbean Food and Nutrition Institute (CFNI), and technical staff from BSS and NNC were the resource persons at the standardization workshops. Data collection commenced on 5 September 2000.
Interviewers were trained to:
ask each question in the same way;
estimate the weights of preportioned cooked foods, before they were informed of the actual weights. For fruits and vegetables, edible portions were used before the "as purchased" weight was divulged. During the interviews, the interviewers used common household measuring equipment such as measuring cups, pot-spoons and measuring spoons to assist respondents in recalling food intake. Where possible, the weight of food reported by the respondent was recorded. All food portions were converted to the measurements stated in the food composition tables;
probe in order to determine the method of food preparation and the addition of fat, sugar, salt and other condiments.
After the first workshop, the interviewers pilot-tested the instrument in parishes that BSS identified as having similar demographic profiles to those in which the survey would be conducted. Following the pilot tests, the consultants and other resource persons made adjustments to the instrument as required.
A second pilot test was carried out, and discussions were held with the consultants and interviewers before final changes were made to the instrument.
The following quality control measures were implemented during data collection for the survey:
One week after data collection, all enumerators were invited to a meeting at which to discuss issues that had emerged and that were relevant to the carrying out of interviews and the identification of foods. Information was summarized and distributed to the enumerators as a complement to the information in the interview manual.
The supervisors attended the discussions and performed spot checks to assess the enumerators' interview techniques, as well as to ensure that the questionnaires were being administered as planned.
The coordinator held regular meetings with the three supervisors to review the progress of fieldwork and make adjustments as required.
Each enumerator was required to insert the food code from the food composition tables into each item of the 24-hour recall period.
The supervisors checked the questionnaires to ensure that all items and questions were completed.
THE SURVEY QUESTIONNAIRES
The following two questionnaires were used to gather the information for the survey:
A household-level questionnaire was administered to the head of household and included questions on:
- household composition;
- income and food expenditure information;
- food production activities;
- shopping habits;
- food preparation practices.
An individual-level questionnaire was administered to the head of household and to one other household member (see section on Sample selection) and included questions on:
- demographic and socio-economic factors (age, sex, education, occupation);
- diagnosis and family history of CNCDs;
- healthy lifestyle practices (exercise, dieting, food preparation practices, frequency of purchase of ready-to-eat meals);
- perception of own nutritional status;
- use of micronutrient and dietary supplements.
In addition, anthropometric measurements were taken. The interviewers were trained to perform standard anthropometric assessment. Each individual was weighed and measured twice. When the difference between the two measurements was within pre-set limits, the second measurement was recorded. Body weight was recorded to the nearest half pound (converted later to kilograms) with an electronic scale, and height was measured to the nearest centimetre with a metal measuring tape and a stadiometer.
A 24-hour recall period was used to determine each individual's food intake in the 24 hours prior to the interview, using a standardized four-step protocol: 1) a complete list of all foods and beverages consumed during the previous 24-hour period, or preceding day, was obtained; 2) detailed descriptions of all the foods and beverages consumed, including cooking methods and brand names (if possible), were recorded; 3) estimates of the amounts of all foods and beverages consumed were obtained, using household measurement equipment such as kitchen spoons, cups and kitchen scales; and 4) the interviewer reviewed the recall with the respondent to ensure that all items had been recorded correctly.
In addition, a semi-quantitative food frequency questionnaire with a list of 84 foods was administered. Frequency was coded as daily, weekly, monthly or none, and the number of predefined measuring units consumed was recorded. A measuring unit, approximated by staff from NNC in collaboration with the CERES[5] consultant, served as a reference portion from which the enumerator determined the amount that the respondent stated he or she had consumed, according to the frequency of intake specified. Portions were also recorded in grams, as required for foods that were recorded by weight.
A pilot study was conducted to validate these methods. Interviewers were provided with a bag containing household equipment such as a kitchen scale, measuring cups and spoons to help respondents recall the quantity of foods eaten. For food items such as carbonated drinks and juices packed in boxes, a chart was prepared showing the amounts in each container of various beverages.
DATA CLEANING AND ENTRY
Supervisors and a team of nutritionists checked the 24-hour recalls and the semi-quantitative food frequency questionnaires before two trained data entry clerks entered the data. As a quality control measure, the CERES consultant randomly selected and entered 200 of the completed questionnaires in order to compare this with the data entry carried out by the clerks. Household measures were converted to grams and used to record quantities of cooked or raw food, as appropriate. A preliminary data analysis (frequencies and histograms) was used to identify errors and outliers, which were then corrected or recoded as missing values. The Statistical Package for the Social Sciences (SPSS) Version 8 was used for all data analysis.
FOOD AND NUTRITIONAL DATA HANDLING
The CERES computer program was used to determine the nutritional value and nutrient adequacy of the diet, using Food composition tables for use in the English-speaking Caribbean (FCT) and Recommended dietary allowances for the Caribbean (RDA) (CFNI, 1993 and 1998).
The nutritional value and adequacy of the diet was calculated from the 24-hour recalls. The semi-quantitative food frequency questionnaire was used to examine dietary diversity and to identify the best food sources for the nutrients.
The nutrient compositions of some frequently consumed local recipes were calculated and added to the food composition table in the CERES program. Local foods were also grouped into the "generic" classification and installed into the CERES program. The calculation of the 38 generic foods (see Annex IV) was based on the quantities used in FCT.
RDA was used as the standard against which the adequacies of energy and selected nutrient intakes for each respondent were assessed. Four age groups were used in analysing the results: 18 to 29 years (<30 years); 30 to 49 years; 50 to 64 years; and ³ 65 years.
Using the information collected from the 24-hour recall and the CERES program, the following information was obtained:
intakes of energy and nutrients;
percentages of adequacy of energy and nutrient intakes, based on age- and sex-specific RDAs;
contributions to total energy from proteins, fats and carbohydrates;
per capita consumption of the 38 generic foods.
Based on World Health Organization (WHO) standards, percentages of adequacy of respondents' energy and nutrient intakes were categorized for the analysis into the following categories: £ 70 percent (very inadequate), 70.1 to 90 percent (inadequate), 90.1 to 110 percent (adequate) and > 110 percent (excessive). Adequate dietary protein, fats and carbohydrate intakes as percentages of total energy were determined and categorized as follows:
proteins, 10 to 15 percent of dietary energy;
total fats 15 to 20 percent of dietary energy;
carbohydrates 55 to 60 percent of dietary energy.
The nutrient goal for fruits and vegetables was set at 370 g per day. Two scores were developed: 1) the micronutrient adequacy score was calculated from the levels of adequacy of a range of micronutrients, using data from the 24-hour recalls (see the subsection on Micronutrient adequacy in Chapter 3); and 2) the dietary diversity score was calculated from the number of items consumed and the frequency of their consumption, using data from the food frequency questionnaire (see the subsection on Dietary diversity in Chapter 3).
Body mass index (BMI) was calculated from each respondent's weight and height according to the following equation:
BMI = weight (kg)/height (m)2
The standard WHO/FAO classification was used to define underweight, normal weight, overweight (also called pre-obesity), and obesity classes I to III (see Table 3.11).
DATA ANALYSIS
Statistical analysis was performed using SPSS Version 8. Frequency distributions, medians, means and standard deviations were used to describe the data. Cross-tabulations and chi-squared tests were performed to evaluate the strength of association between categorical variables, especially to examine differences across age groups and sexes. For continuous variables (such as BMI and dietary intake data), t-tests and one-way analyses of variance were used to investigate differences between age groups and sexes. In all cases, p < 0.05 was used as the cut-off for statistical significance.
Vulnerability profiles were developed using a number of definitions of vulnerability (see Chapter 4). This was done either by performing multiple regression analyses to determine the contributions of a range of variables to the variation of certain key indicators (BMI, micronutrient adequacy scores and dietary diversity scores), or by using t-tests, one-way analyses of variance and chi-square tests to examine other potentially vulnerable conditions (poverty, old age, and female-headed households). The results of in-depth analyses of three key issues - iron deficiency anaemia, healthy lifestyles and younger adults - are presented in Chapter 5.
Throughout this report, the results from the Barbadian survey have been put into context by comparing them with earlier Barbadian data (to examine trends), with data from other Caribbean countries and, in some instances, with data from the United Kingdom or the United States. Chapters 3, 4 and 5 focus on significant findings, rather than presenting the results of all the analyses undertaken. Chapter 6 presents the policy and programmatic implications, as well as recommendations based on the survey results.
A national survey such as this one requires considerable forward planning. It demands a substantial commitment of time and resources, and should therefore be undertaken only when adequate time and resources exist and when the survey results will be put to good use in guiding policy construction and programme planning. It is useful to examine the planning and implementation of this survey in order to extract lessons learned that might benefit future surveys. The following recommendations are based on the experience of a survey that was successfully executed by NNC. Many of the lessons learned, difficulties experienced and limitations of the data are by no means unique to this survey, and can be found in textbooks on survey methodology, especially those on dietary methodology and nutritional epidemiology. However, it is helpful to consider them in the context of this survey.
PLANNING AND IMPLEMENTING THE SURVEY
A national food consumption and nutrition survey cannot be undertaken by a relatively small nutrition unit or centre (such as NNC) alone; it needs the support, approval and collaboration of at least one government ministry (in this case, the Ministries of Health and of Agriculture and Rural Development were involved), as well as of other government departments (BSS for this survey). Securing such support, approval and commitment is the essential first step in planning the survey. However, political commitment to support the nutrition centre and oversee all stages of the survey is equally essential.
If the results of the survey are to be used in policy and programme planning, potential users need to be consulted at an early stage in the planning phase. Users need to specify what information they require and in what form. Ideally, a survey advisory committee should be formed, including users, at least one statistician and a nutritionist. In this survey, a committee was formed comprising two senior medical officers of health, NNC staff, statisticians from BSS and FAO, and the director of the Chronic Disease Research Centre.
Adequate human resources need to be found and committed for all stages of the survey (planning and supervision, data collection, data entry, data analysis and preparation of the report). Survey staff must be dedicated full-time to the survey from an early stage, and should not attempt to carry out other jobs and duties at the same time. A statistician needs to be consulted and involved from the start of survey planning. It is recommended that large surveys be coordinated by a department that is oriented to survey methodology and has access to suitable, well-trained enumerators.
The survey budget must be realistic. The sample size of the survey needs to be calculated ahead so that a realistic estimate of the time and staff needed for data collection can be made, bearing in mind that much interviewing will have to be conducted after working hours and on weekends. Where applicable, allowance needs to be made for the costs of travel and meals, equipment, and repeat visits to ensure the inclusion of as many selected sample households as possible.
A public information campaign is needed well ahead of survey implementation to alert the public that they may be called on to participate in the survey, and to urge their compliance. A high refusal rate can produce biased results and causes much time wasting.
The size of the survey team needs to be carefully planned. On the one hand, a team that is too large can pose problems regarding adequate supervision and quality control. On the other hand, if enumerators are asked to conduct too many interviews, boredom can set in and affect the quality of the work.
A work plan needs to be developed early during survey planning, with the timing of key events clearly marked. This requires a realistic assessment of the time needed for fieldwork. Training should be conducted close to the start of data gathering, and refresher training must also be scheduled. In this survey, delays in training and in the availability (and set-up) of computers, as well as the imminent start of the national census, resulted in the start of the survey being postponed from January to September 2000.
There are a number of ways of collecting dietary intake data. Each has its limitations, and these need to be recognized when choosing the most appropriate methodology. Methods range from the "quick and dirty" food frequency questionnaire to the more intensive (and intrusive) "gold standard" of weighed intakes, for which survey respondents are asked to weigh and record all the foods they consume. The food frequency method provides limited quantitative information on dietary intakes (many nutritional epidemiologists would argue that it should only be used to provide qualitative information on food patterns), but can be administered easily to a large sample. The weighed intake method, however, can only be used with a small sample of highly motivated and literate respondents. This survey chose the 24-hour dietary recall method - which is perhaps the most common method of assessing dietary intakes for large surveys - supplemented by data from a semi-quantitative food frequency questionnaire. The limitations of these methods are described later in this section.
The training of enumerators is crucial, regardless of whether they have a nutrition background. Some enumerators may "fail" the training, so more enumerators than needed for the survey should be trained, and the weakest trainees placed in reserve. In this survey, some enumerators were absent from some training sessions, which limited the adequacy of the training they received, despite efforts to compensate for absences by providing written instructions. If the fieldwork is to be of long duration, refresher training may be needed. Training for a food consumption and anthropometry survey must include the following:
- standardization of anthropometric techniques, including assessing enumerators' abilities by calculating the technical errors of measurement made by each enumerator, and providing ample opportunity for repeated practice sessions;
- measuring and weighing each person twice, and assessing the adequacy of the measurements by comparing the values;
-instruction in the care and standardization of all equipment;
-standardization of interview techniques, and achieving familiarity with the questionnaire and the 24-hour recall method. Role play is a useful training tool for this.
A schedule for supervising fieldwork is essential to ensure quality control throughout the fieldwork. This must include random, but regular, observation of enumerators, frequent meetings to discuss problems, and routine daily checking of completed questionnaires. Regular meetings between supervisors and enumerators are especially important, and time and opportunity must be allowed for these.
A system for gathering in questionnaires at the end of each day also needs to be established. For this survey, it was decided to collect questionnaires weekly, because several respondents were absent at the first visit and revisits were required. In preparation for data entry, supervisors must collect and check the completed questionnaires and correct errors in coding immediately. Assistance from a statistician who is knowledgeable about food consumption and anthropometry would facilitate the planning of quality control during fieldwork.
Data analysis must be planned as part of questionnaire development, with the full involvement of the survey statistician. Thinking ahead to how the data are to be analysed, presented and disaggregated, as well as to which statistical tests are to be used, will help determine the coding of the questionnaire. For example, in this survey, income data were obtained as actual amounts, but food expenditure data were obtained only as categorical information. It was therefore not possible to determine the proportion of a household's income spent on food.
DATA ENTRY AND ANALYSIS
Quality control of data entry can be ensured either by double data entry of all questionnaires or by double data entry of a sample of questionnaires. Double data entry entails entering data from a questionnaire twice, then comparing the records, and correcting if needed. This survey employed double data entry for a sample (200 questionnaires), the second data entry being done by the CERES consultant.
Entry of the 24-hour recall data is best done by a nutritionist, because judgements regarding dietary data need to be made by someone knowledgeable in the subject. If it is not possible to employ a nutritionist for the data entry, the dietary information needs first to be coded by a nutritionist and then entered by a data entry clerk, under the close supervision of a nutritionist.
The coding of dietary data includes the food item code (which is related to the food analysis program's database, and hence to whether cooked or raw food is referred to) and the quantity of food consumed (in grams). This entails converting household measures into grams and assessing the composition of composite dishes and pre-prepared foods. It also entails making a separate list of foods (such as composite dishes or purchased pre-prepared foods) that are not included in the database, so that these may be included in the data analysis. Decisions regarding the coding of dietary data, quantities and compositions of non-database foods are best made through discussion with nutritionists.
The entry of dietary data is time-consuming, and this needs to be allowed for. In this survey, it would have been better to check the dietary data as the questionnaires came in, rather than after the fieldwork had been completed. To ensure the accurate coding of dietary data, a system needs to be established in which a second person recodes a sample of dietary recalls, which should then be compared with the original coded data. This will highlight issues and problems that can be discussed at meetings. In this survey, the food codes recorded by enumerators were reviewed by nutritionists, but after a considerable lapse of time. The 24-hour recalls of respondents with very high or very low energy intakes were checked.
The 24-hour recall method presents the following difficulties and limitations, which were experienced in this survey:
- It relies on the respondent's memory, hence the need for adequate probing by the interviewer.
- Interviewers experience difficulties in recording quantities consumed and assessing portion sizes. Respondents may also fail to quantify accurately the amounts of food consumed.
- The nutrient content database (and hence the CERES database used in this survey) is incomplete: nutrient content information is not available for every food item in FCT. For example, the zinc and cholesterol contents of many foods are missing. An apparently low zinc intake (as found in this survey) may therefore be the result of the incomplete database, rather than reflecting a real dietary zinc deficiency. Unfortunately, the CERES program used to analyse this survey's dietary data does not indicate cases of incomplete information.
- The nutrient values provided in FCT (and hence CERES) were extracted from other tables, and not obtained from the analysis of Caribbean foods. This may have introduced errors in the moisture content of staple foods and the vitamin content of processed foods and fresh fruit and vegetables.
- Foods and composite dishes can be added to the CERES database. Local recipes from a Caribbean cookery book were used to enter composite dishes, and there may have been errors in the nutrient values of these dishes. Furthermore, the nutrient compositions of many foods that are pre-prepared and processed in Barbados were not available from the food companies concerned, so estimates based on similar foods may have been inaccurate for some nutrients.
The food frequency method used in this survey entailed the gathering of information on the frequency of consumption (monthly, weekly, daily) of 84 foods, and the number of portions of predetermined size consumed. It is recommended that in future no effort be made to quantify the information (through portions consumed), but that a more detailed breakdown of frequency of consumption be obtained. Although the food frequency method is often described as "quick and dirty", it demands the same level of training in good technique as any other method. It provides valuable information on dietary patterns that are not captured by 24-hour recalls.
All scales and measuring tapes used in this survey were calibrated in both metric and imperial units, but enumerators were instructed to record in the unit of their choice. As both data entry and data checking took place long after the fieldwork was complete, it was necessary to return to some homes to re-measure respondents because it was not clear which units had been used for the original measurements. This experience emphasizes the importance of early data entry and checking.
Data analysis must not be carried out by a statistician alone, but must be a collaborative effort between an experienced nutritionist (preferably with a good understanding of statistics) and a statistician (preferably with experience in nutrition surveys). The plan of the analysis and the structure of the survey report need to be agreed before data analysis begins. Input from potential users at this stage would be valuable.
[5] CERES is the software
package that was used to analyse the food intake data. It was developed by the
Instituto de Nutrición e Higiene de los Alimentos in Havana, with funding
from FAO, and incorporates the Caribbean food composition tables. |