Dietary assessment methods used for low-income populations in food consumption surveys: a literature review

Abstract

In order to assess nutritional adequacy, reliable estimates of nutrient intake are required. Specifically, the EURopean micronutrient RECommendations Aligned (EURRECA) Network of Excellence needs clear guidelines on methods suitable for assessing micronutrient intakes among vulnerable population groups. The objective of this project was to collect, evaluate data and review the suitability of methods used to assess dietary intake of low-income groups across Europe. A comprehensive literature review methodology was employed, which involved structured search MEDLINE, from 1990 to 2008, on (nutrient intake methods) and (validity terms) and (human studies) and finally (low-income terms). In addition, manual searches were conducted for published books and national studies. Seven studies satisfied the selection criteria for inclusion in the review in which four dietary intake methods had been described and validated. Three other studies found in the grey literature used similar methods. Only one study tested a variety of methods to assess reliability and the method preferred by those having low incomes. Preferred methods were 24 h recalls and a FFQ which, compared with the weighed inventory, also yielded higher estimates of energy and nutrient intakes. Many of the methods used in low-income populations have not been subjected to evaluation and consequently may not demonstrate sensitivity and/or specificity when used in this population. Based on one study only, four multiple-pass 24 h recalls are recommended as the most appropriate method for the evaluation of nutritional adequacy in low-income households.

Adequate and reliable assessment of dietary intake in nutritional research is key when attempting to understand dietary profiles linked to living conditions such as low income. There are many retrospective (24 h recalls, dietary history method and FFQ) and prospective (food records over 1, 2, 3, 4 and 7 d, duplicate portions) intake methods⁽¹⁾. However, none is suitable in and of itself to give sufficiently accurate dietary information for all purposes.

Dietary assessment methods including food diaries, dietary recalls and list-type methods, such as FFQ, are subject to considerable biases and errors, and none can be considered as a gold standard⁽²⁾. Moreover, there has been relatively little research on the comparability of these instruments across diverse ethnic and socio-economic groups. It is important to know whether and by how much the measurement characteristics of dietary assessment instruments differ, especially in research designed to address dietary quality, food inequality or diet as part of health inequalities in at-risk populations compared to the rest of the population.

Much of the evidence of diets among low-income households until now has been based on national surveys in which the numbers of low-income households are few, cooperation rates are low and classification of households is based on receiving social benefit or on employment status⁽³⁾. The limitation, or outright lack, of published data and information on food intake and nutritional outcomes of those living on low incomes thus seems to be due to the difficulties faced when engaging such groups to participate in research projects for the purpose of measuring their food and nutrient intakes. As a result their circumstances are not well reflected in the literature, resulting in a knowledge gap in Europe.

Definitions of low income are often different depending on the country, the study purpose or even on the authors of the publications. Poverty, materially deprived, low socio-economic status (SES) or position, low social class and food insecurity are the most commonly used terms in research projects, reports and papers reporting on the studies of low-income populations. SES in scientific studies is often represented by multiple indicators including: income, educational level, occupation, profession and cultural aspects, or indices where several of these indicators are used (for example, disability-adjusted life year or DALY, a composite index used by WHO). All may operate independently or interact in ways that influence food choices. For households with limited resources, income and the cost of food (both being determinants of diet composition) are two important factors influencing food choice, which also interact with each other⁽⁴⁾. Based on Euro Stat data (2007), for the European union-25 as a whole, an average of 16 % of the population was currently at risk of being impoverished and had an equivalised household disposable income below 60 % of the national median for the country in which they lived⁽⁵⁾. The percentage of income spent on food may indicate how severe the lack of food is likely to be, but this would depend on how income is categorised – either as personal income, household income or defined in another way. Low-income families have the highest percentage (23 %) of income expenditure designated for food acquisition in comparison to the highest income families in each country (13 %)⁽⁵⁾. According to the indicators of income inequality and objective poverty, there was a divide among former European union member states (European union 15), with the UK, Ireland and South European countries having higher and the Continental and Nordic countries lower indicators of inequality and poverty. Among new member states, the Baltic countries and Poland had the highest and Slovenia and the Czech Republic showed the lowest indicators of inequality and poverty⁽⁶⁾. The data for 2002–2007 in Serbia defined low-income households as those who have monthly incomes of 8883 dinars (RSD) per equivalent scale (approximately 100 €)⁽⁷⁾.

People from low-income households typically have less nutritionally adequate diets, especially those who live for long periods of time on limited incomes. There is growing evidence to suggest that poor nutritional status is a specific cause of higher disease rates in low-income households^(8–10). Health inequalities revealed striking differences in many countries between social groups showing that lower socio-economic groups have greater incidences of premature babies, low-birth weight babies, heart disease, stroke and some cancers, in addition to risk factors like smoking, physical inactivity, obesity, hypertension, the lack of breast feeding and poor diet^{(11, 12)}. In addition, among other factors, lower literacy, numerical and language skills, physical disabilities and mental health problems are more common in low-income groups, as well as low motivation, and as such, constitute obstacles when identifying and assessing this population's food and nutrient intakes.

The above-mentioned evidence indicates the need to identify different intake methods that are appropriate for low-income groups. Thus, the aim of the present paper was to examine dietary intake methods available for use by researchers in order to screen or assess nutrient intake in low-income populations in Europe. Moreover, the extent to which validity, specificity and acceptability of these methods have been addressed was also evaluated. By examining these issues, the understanding of the science of dietary intake assessment and recommended measures to improve food and nutrient data reproducibility and validity in low-income populations will therefore be enhanced.

Methods

Search strategy

The electronic database MEDLINE was utilised to identify dietary validation studies addressing nutrient intake from 1990 to March 2008 using text terms with appropriate truncations and relevant indexing terms. The reference lists in the articles, reviews and textbooks retrieved were also to be referred. The search terms were as follows:

1. 1 Block A: Descriptors of low income: Medical subject heading (MESH) term ‘low income’ (includes indigency, indigents and indigent); poverty; material deprived; low SES; social class; low socio-economic position.

2. 2 Block B: Descriptors of dietary intake: MESH term ‘nutrition assessment’ (includes nutrition survey and diet survey) and ‘diet assessment’; MESH term ‘dietary habits’ (includes food habits and diet modification); MESH term ‘diet records’ (includes food diary, 24 h dietary recall and FFQ); MESH term ‘feeding behaviour’ (includes feeding pattern, eating behaviour), ‘food consumption’, ‘nutritional status’ and ‘nutrient intake’.

3. 3 Block C: Descriptors of validity: MESH term ‘validity and reproducibility’; ‘validation study’; ‘replication study’; ‘correlation coefficient’; ‘correlation study’.

Two independent reviewers (V. V. and R. N.) ran the electronic searches and screened all titles and abstracts for studies that might have met the following inclusion/exclusion criteria. The articles were excluded when both reviewers determined that titles/abstracts met the exclusion criteria. When a title/abstract could not be rejected with certainty, the full text of the article was obtained and further evaluated. Reviewers excluded studies that were not conducted in Europe, studies exclusively focused on diseased or institutionalised persons, as well as studies exclusively focused on energy and/or macronutrients. Studies evaluating the physiological effects of foods, nutrients and those assessing their relationship to genetic determinants, calibration studies and those discussing statistical methods, studies in animals and those without abstracts in PubMed were also excluded.

The full text of all articles collected was screened for definitive exclusion by the first reviewer, with independent duplicate assessment of a random sample of 50 % by a second reviewer. Where the two reviewers disagreed, the study was discussed and a consensus decision reached where possible. If this was not possible then a third reviewer was asked to consider the results. Data from each study were extracted onto an Excel database file.

This review includes cross-sectional or validation studies published in English language journals presenting validated results of dietary intake methods in low-income adults, assessing data on their micronutrient intake.

Results

The literature search yielded 1216 references. Figure 1 depicts a flow chart of studies assessed and excluded at various stages of the review. Out of 1216 studies, twenty-seven abstracts were retrieved for more detailed evaluation. The reasons for exclusion were: studies were not conducted in Europe, studies focused on patients or on children, studies did not examine micronutrients or did not have validated intake methods. Three abstracts could not be traced, but twenty-four full papers were evaluated. After examination, only seven studies (Table 1) met the inclusion criteria. Most of the examined studies were either applied nutritional studies or observational studies, with the exception of a comparison (validation) study.

Fig. 1 Summary of search strategy process and results.

Table 1 Characteristics of studies on intake methods in low-income populations

NA, not applicable.

The characteristics of studies involved in this review are presented in Table 1. The comparison study published by Holmes et al. ⁽³⁾ compared three 4 d dietary survey methods (multiple-pass 24 h recall, FFQ and semi-weighed method) against a 4 d weighed inventory. The present study provided useful insight into the assessment of validity for different dietary intake methods applied to low-income populations. It suggested that in terms of acceptability and effectiveness of the methods:

1. 1 The 24 h recall was the method most preferred by interviewers for reporting diet.

2. 2 The food frequency was the method most preferred by respondents for reporting diet.

3. 3 The semi-weighed method was least liked for reporting diet by both interviewers and respondents.

4. 4 The greatest number of difficulties reported by the interviewers was with the weighed inventory.

5. 5 The 24 h recall and FFQ yielded higher estimates of energy and nutrient intakes more often than either the semi-weighed or weighed inventory.

6. 6 The 24 h recall was more likely to have a significantly higher number of food items reported.

7. 7 The 24 h recall provided the most consistent results across all age and sex groups compared with the other methods.

Besides the validation study presented earlier, in another study the same authors compared a 4 d assessment of diet by 24 h recall to a FFQ, semi-weighed method and weighed inventory over 10–12 weeks. The obtained results confirmed the 24 h recall as the most reliable dietary assessment method⁽¹³⁾. Freisling et al. ⁽¹⁴⁾ assessed diet of pregnant women by a single 24 h recall plus interview. A trained master's student of nutritional sciences interviewed women individually about their food and beverage intakes from the previous day. The field work covered all days of the week, including weekends. All portion sizes were described in household measures and for manufactured food items by the manufacturer's information. The diet records were coded by the same master's student using specifically designed food composition software⁽¹⁴⁾. Two other studies used FFQ: a self-administered, semi-quantitative FFQ, where participants were asked about consumption of 100 food items and their serving sizes, and then converted into daily energy, nutrient and alcohol intakes⁽¹⁵⁾ and an interview-based FFQ estimated by a variety of visual aids, e.g. a display of a series of dishes (slices of meat, cakes, cups and glasses) of recognisable weights and volumes⁽¹⁶⁾. Haste et al. ⁽¹⁷⁾ applied 7 d weighed records to low-income pregnant women, while Andrie et al. ⁽⁸⁾ chose estimated 7 d diet records aided by a photographic manual of portion sizes (Table 1).

Another important part of each study was the way SES was defined and how this was used in sampling. Detailed sampling methods used for the evaluation of social status are presented in Table 2. As can be seen in Table 2, social class was categorised in different ways in these studies, suggesting that there is no consistent method for identifying low-income groups in Europe.

Table 2 Sampling methods and determination of material deprivation

Discussion

Evaluating intake methods in low-income populations require that the following two key factors are taken into consideration: determination or definition of low income and choosing the most appropriate method of intake assessment to apply.

Presently, there is no consensus about how to define low-income populations in Europe. A variety of definitions of low income have been described, including low-income families (whose income is at the level of income support), households below average income (usually expressed as numbers of households with incomes < 40 or 50 % of the national average household income) and quantitative indices of material and social deprivation. Deprivation factors in the deprivation index for identifying low-income/‘deprived’ households were used in an analysis of the Health Survey in England data and proved to be the strongest household-level predictor of poor general health⁽¹³⁾. Furthermore, Nelson et al. ⁽¹³⁾ state that low-income families have ‘low cost but acceptable’ diets. The low cost but acceptable food budget standard identifies a basket of foods and corresponding menus, which provides (for a given household composition) a palatable diet that is consistent with prevailing cultural norms, and that satisfies existing criteria for health in relation to dietary reference values, food-based dietary guidelines and safe levels of alcohol consumption⁽¹³⁾. The level of poverty in Britain was defined as < 60 % of the contemporary mean, e.g. the threshold for a two-adult household, i.e. 180 pounds/week⁽¹⁸⁾. In France, low socio-economic groups or food-insufficient individuals are those whose estimated cost of the mean diet is below 2·5 €/d. This represents a population facing very severe food choice restrictions because of economic constraints⁽¹⁹⁾. The poverty in Serbia is defined as households that have monthly income of 8883 dinars per equivalent scale (approximately 100 €)⁽⁷⁾.

A number of studies in Europe^(20–23) have clearly identified differences between socio-economic subgroups of the population on the level of foods, food groups and nutrients, pointing – in general – to a healthier diet in persons with higher levels of education. Several European studies^{(24, 25)} have also shown the relationship between lower SES and poorer micronutrient density in the diet. A large body of epidemiological data shows that diet quality follows a socio-economic gradient, higher-quality diets are associated with greater affluence, whereas energy-dense diets that are nutrient poor are preferentially consumed by persons of lower SES and of more limited economic means⁽²⁶⁾. The Pan-European Survey of Consumer Attitudes to Food, Nutrition and Health found that the top five influences on food choice in fifteen European Union countries are quality/freshness (74 %), price (43 %), taste (38 %), healthy eating (32 %) and family preferences (29 %)⁽²⁷⁾.

The consumption of whole grains, lean meats, fish, low-fat dairy products and fresh vegetables and fruit (containing considerable amounts of vitamins, minerals and fibre) was consistently associated with higher SES groups, whereas the consumption of fatty meats, refined grains and added fats was associated with lower SES groups⁽⁸⁾. Low-SES groups also had the lowest consumption of vitamin C, vitamin D, carotene, folate, vitamin E, plant-based polyphenols, as well as low Fe, Ca and K intakes. Studies of plasma biomarkers of dietary exposure provide additional evidence that SES affects diet quality^{(8, 26)}. Increased health inequalities and micronutrient deficits are also a result of poverty and social inequalities⁽¹¹⁾.

There are also many difficulties in dietary assessments for low-income population: literacy – level of education of respondents, biases, respondents not informed and not accessible, low motivation and the need to design special tools for nutrient intake assessment, especially in those whose primary source of nutrition is based on social and welfare systems (i.e. public kitchens). Variations in the level of language and numeric skills across the whole sample could mediate against comparable record keeping in all households⁽²⁸⁾. Those with the lowest skills may have difficulty in completing dietary records unless assistance is provided by either interviewers or other household members. Physical problems of record keeping exist among some subgroups, especially the elderly and disabled. Poor eyesight, impaired hearing and other physical disabilities such as being chair- or bed-ridden make weighing and recording of diet difficult or impossible⁽³⁾. A higher incidence of mental health problems also exists among individuals living on limited incomes, particularly the elderly. Dementia creates adverse consequences for data collection, particularly when using retrospective methods of dietary assessment that are reliant on memory⁽²⁹⁾. Drug and alcohol abuse may create similar problems, impacting not only on the quality of the data but also on the safety and welfare of the interviewers⁽³⁾. All these difficulties may affect the application of intake assessment methods as well as their results.

In the papers reviewed in the present study, one of the commonly used methods for dietary assessment was the weighed food record, which was also used as a gold standard in two studies that evaluated three other methods^{(3, 13)}. However, they differed in the length of time of assessment, for example 4 d v. 7 d^{(3, 8, 13, 17)}, as well as adjunct administration of 24 h recalls or FFQ⁽¹³⁾. These results are similar to published data for the general population. Maillot et al. ⁽³⁰⁾, for example, used an estimated 7 d record aided by a photographic manual of portion sizes in French adults. Andersen et al. ⁽³¹⁾ applied a pre-coded food diary with lists of 277 drinks, food items and dishes for overweight and obese Norwegian schoolchildren, who indicated an eating event by filling in how many units they ate of each food item in the correct time span. Some studies show that there were no striking differences between different socio-economic groups in the proportion of underreporters⁽³²⁾, but the Whitehall II Study shows that the proportion of low-energy reporters was approximately four times higher in the lowest compared to the highest socio-economic group⁽³³⁾. Moreover, the person's attitude towards consumption was found to be a good predictor of reported consumption⁽³⁴⁾.

Methods for measuring food intake are not standardised across Europe and data in low-income populations are generally inadequate. Based on the results of one study, the dietary survey method preferred by the interviewers of low-income populations was the 24 h recall. Most respondents preferred the FFQ. The FFQ constitutes another method that was one of the most utilised by studies reviewed in the present paper. However, compared with the weighed inventory, repeat 24 h recalls and FFQ yielded higher estimates of energy and nutrient intakes. The semi-weighed method was least liked and yielded the lowest estimates of intake. Other intake methods appear not to have been subjected to specific and/or reliability testing, but are used in many other population groups⁽³⁾. Thus, four multiple-pass 24 h recalls were shown to be the most appropriate method for a study of diet and nutrition in low-income households. However, problems with this method include errors in recall, poor estimations of serving size and poor generalisability from a specific day to overall dietary habits⁽¹⁾. Therefore, a single 24 h recall is not appropriate to analyse associations between nutrition intakes and status biomarkers or health endpoint and several measurements are required.

Moreover, there is a lack of data for these kinds of studies in Central and Eastern Europe⁽³⁵⁾. However, data presented as grey literature sources (national studies, monograph, theses, etc.) could be an important and additional source of valuable information. Food consumption of low-income groups in Poland and Belgium (grey literature, data not shown) applied two intake methods: 24 h recalls administered by trained interviewers using an album of food portions and a forty-two item FFQ representing all food groups where the respondent is asked about usual frequency of food and drink consumption⁽¹²⁾. Another study consists of the Low Income Diet and Nutrition Survey in the United Kingdom (n 3728 individuals aged >2 years) in which four 24 h recalls were administered on random days (including at least one weekend day) within a 10 d period by trained interviewers and nurses in a face-to-face interview as well as self-administered questionnaires⁽³⁶⁾. In a large national study in Serbia, 7 d food records were used as the method of determining nutrient intake in schoolchildren and their families (age 10, n 1318 and age 15, n 960)^{(37, 38)}. Even though the present study was not specified for low-income groups, Serbia was classified by the World Bank definition⁽³⁹⁾, based on gross national income per capita, as a low-income European country. The paucity of literature highlights the difficulties in conducting intervention studies that show how economic constraints orient food choices due to intentional or non-intentional bias/misreporting. This is a clear knowledge gap that requires further investigation into the development of optimal tools for dietary assessment for this population.

Conclusions

The present paper centres on the use of various methods for obtaining dietary information applied in evaluating diets of low-income populations. Based on the previously cited validation study, four multiple-pass 24 h recalls were shown to be the most appropriate method for a study of diet and nutrition in low-income households. Compared with the weighed inventory, repeat 24 h recalls and food frequency checklists yielded higher estimates of energy and nutrient intakes. The semi-weighed method was least liked and yielded the lowest estimates of intake. Other intake methods appear not to have been subjected to specific and/or reliability testing, but are used in many other populations groups. The decision to use a particular method must therefore be considered carefully. In general, the findings of this literature review will help in defining the optimal methods of dietary assessment to be used for collecting nutrient data in future surveys targeting low-income populations. Future research studies should include an examination of validity, specificity and acceptability; furthermore, they should also shed light on recommendations for best practice in harmonisation, obtaining and using dietary intake data to compare intakes in low-income populations from different regions throughout Europe.