Share of Food Consumed Away from Home of Total Food Consumption% kcals

Summary

Household Consumption and Expenditure Surveys (HCES)—also referred to by a variety of other names including Household Income and Expenditure Surveys (HIES), Household Budget Surveys (HBS), or Living Standards Measurement Surveys (LSMS)—are complex surveys conducted on a nationally representative sample to characterize important aspects of household socioeconomic conditions (Coates et al., 2012). Typically, HCES are conducted every 3-5 years in a range of countries and cover 7,000 to 20,000 households to provide a statistically representative sample (Fiedler et al., 2012). Most HCES are implemented by national statistical agencies, often with technical assistance from the World Bank’s Living Standard Measurement Study (LSMS) group.

The results of HCES have wide-ranging utility. Their primary purpose is to provide information for poverty monitoring, the calculation of national accounts, and as an input for consumer price indices (Smith et al., 2014). However, there is increasing interest in using the food consumption module from HCES as a source of nationally representative data for assessing food security and nutrition. Furthermore, HCES collect a wide range of data on determinants and outcomes (e.g. socioeconomic status, education), potentially enriching food security and nutrition analyses. Based on existing research there is wide consensus that HCES, with carefully designed consumption modules, are a valuable source of data for household-level food security and nutrition measurement (Russell et al., 2018; Zezza et al., 2017).

One of the major drawbacks of using HCES is that the consumption modules are heterogeneous across countries, which means that not all HCES data lend themselves to the same food security and nutrition analyses, and comparisons across countries can be inaccurate. Some of the key ways in which the consumption modules differ across surveys include: 1) the length of the recall period; 2) whether data are collected for acquisition, consumption, or both; 3) whether there is information on the mode of food acquisition (purchases, own production, and in-kind); 4) whether or not information on food consumed away from home is collected and in what form; 5) whether food detail is collected through open recall or a list, and, if a list, how disaggregated and specific the foods and food groups are; and 6) the use of non-standard units without available conversions (Smith et al., 2014). For example, if the food consumption module has a short food list with aggregated items making it difficult to match with a food composition database, excludes food away from home, and has a long recall period (>14 days) then the consumption module may not be adequate for measuring certain food security and nutrition indicators, such as total household-level calorie availability.

While the ‘C’ in HCES stands for ‘consumption’, HCES collect data on acquisition, consumption, or both. While consumption data refers to the food consumed by the household, acquisition data refers to the food acquired through purchases, own-production, and in-kind. Acquisition data serve as a proxy for food consumption, as households may build food stocks or consume food stocks during the reference period, as compared to consumption, which collect data on food consumed in a specified period. This is an important point because some foods (e.g. grains) are not perishable and can be stored, therefore some households may be drawing down stocks acquired to meet current consumption, while other households may be accumulating stocks that will be consumed after that period (Smith et al., 2014). Another type of HCES collects a combination of acquisition and consumption data, wherein households report what they acquired through purchases and what they consumed from own-production and transfers (Smith, 2003). Food consumption estimates generated from acquisition data or a combination of both acquisition and consumption data are typically referred to as 'apparent consumption' in the literature to distinguish from actual consumption (Fiedler & Mwangi, 2016).

The World Bank Microdata Library has the most comprehensive and publicly accessible repository of HCES data. Data also can be accessed—often for a fee—from countries’ National Statistics Office, though each country has its own policies and procedures regarding data sharing. The International Household Survey Network (IHSN) is an informal network to promote data standards and dissemination where additional information (e.g. survey catalogs, guidelines, and software) on existing HCES can be found (IHSN, 2018).

Strengths:

• Typically nationally representative and sometimes representative at provincial and district levels
• Typically collected every 3-5 years, allowing for an examination of trends
• Food consumption data from HCES are an important source of information on food security and nutrition
• Include a wide range of data on determinants and outcomes (e.g. socioeconomic status, education), enabling various analytical options.

Weaknesses:

• Due to issues with the structure of some consumption modules (e.g. no information on food consumed away from home), the data may not be useful for certain food security and nutrition analyses
• Some HCES only measure 'apparent consumption' (based on acquisition data), not actual consumption
• The food list is not always designed with the level of detail needed to make exact matches between the food items in the food list and a food composition database
• Recall periods in HCES vary from 1 to 365 days, with long recall periods (>2 weeks) raising concern about reliability and recall bias
• Household-level data from HCES do not allow for measurement of individual-level food security and nutrition indicators
• Many HCES do not capture seasonal variation

Summary

Weighed Food Records (WFR), also called weighed food diaries or simply weighed records, are considered the "gold standard" of individual quantitative dietary assessment methods (Carlsen et al., 2010). WFR require the respondent or enumerator to weigh all foods and beverages at the time of consumption (rather than asking respondents to recall their consumption, as is done in the 24-hour quantitative recall, or 24HR). Any plate waste must also be recorded, as well as a description of the food along with preparation methods and brand names.

Though no dietary assessment methodology can completely prevent measurement error, WFR are often considered the most precise method when it comes to quantifying food intake, since each food is weighed, eliminating issues associated with portion size estimation through recall. As a result, the high degree of accuracy produced by WFR means they are often used as the reference method in validation studies of other dietary assessment methods (e.g. Alemayehu et al., 2011; Nightingale et al., 2016).

If working in a low- or middle-income country with low literacy levels the presence of a trained enumerator in the household is typically required throughout the period being assessed, from the time the first food or beverage is consumed in the morning to when the last one is consumed at night. When using WFR to collect dietary data, enumerators must be carefully trained to standardize measurement, instrument calibration, and interviewing methods in order to reduce measurement error. While enumerator-administered WFR provide very accurate estimates of dietary intake, they can be intrusive and time-consuming, as well as potentially distort the behavior of respondents’ due to the presence of an enumerator throughout the day (Ortega et al., 2015).

Furthermore, administering a WFR can be difficult in populations where people are out of the home for all or most of the day (e.g. urban areas or school age children), as the enumerator would have to accompany respondents throughout the day. If the population of interest is literate, then a food diary or self-administered WFR could be used, in which the respondent weighs and records all of the foods and beverages consumed over a specified period (e.g. 24 hours). Due to the expense and small sample size of most WFR collections, care must be taken to ensure that the sample is representative of the studied population (Wrieden et al., 2003).

Strengths:

• Offer a high degree of accuracy in assessing food and nutrient intake relative
 to other recall-based dietary assessment methods
• Provide quantitative estimates of individual food consumption and nutrient intake
• Take into account preparation methods and the effect on estimated nutrient content
• Applicable to diverse groups with a wide range of dietary patterns

Weaknesses:

• Significant training is required to minimize errors in data collection
• Data are frequently collected from small samples that are not nationally representative
• Enumerator-administered WFR are intrusive and can distort respondent behavior
• Like most surveys, to capture seasonal variation data collection must span the entire year or be repeated in multiple seasons

Summary

The 24-hour Dietary Recall (24HR) method provides comprehensive, quantitative information on individual diets by querying respondents about the type and quantity of all food and beverages consumed during the previous 24-hour period (Gibson & Ferguson, 2008). A standard multiple pass 24HR includes having the respondent iteratively provide increasingly granular data about each food or drink and its preparation method and other attributes, as well as an estimation of the portion size consumed. The multiple pass approach has been validated in many low- and middle-income countries (Gibson et al., 2017).

Enumerator-administered, rather than self-administered, 24HRs are often used in low- and middle-income countries because they are quick, culturally sensitive, and provide quantitative data on both foods and nutrients (Gibson et al., 2017). Data from 24HR can be used to assess dietary patterns, food groups, or nutrient intake. In order to analyze the nutrient content, the food data must be matched with nutrient information from a food composition database. Mean intakes of foods and nutrients can be measured using a single 24HR, while assessing the "usual intake" of a population requires that repeat 24HRs are collected from a sub-sample of the study population (Gibson & Ferguson, 2008).

Individual-level quantitative dietary data can also be used to develop a better understanding of typical household food preparation, cooking methods, and brand names of foods consumed within the household. Furthermore, if individual-level dietary data are collected in conjunction with information on socioeconomic status, education, and health, the data can be used to examine linkages between income levels and dietary choices, as well as dietary patterns and health outcomes.

While 24HRs offer a higher degree of accuracy in assessing food and nutrient intake relative to Food Frequency Questionnaires (FFQ) or estimates derived from Household Consumption and Expenditure Surveys (HCES), they are collected infrequently on nationally representative samples, and when they are collected, these data are typically not publicly available (Pisa et al., 2018). However, there is increasing interest in, and demand for, individual-level quantitative dietary data, particularly in light of the nutrition transition and rapid food system changes (Coates et al., 2017). For example, the Food and Agriculture Organization and World Health Organization Global Individual Food consumption data Tool (FAO/WHO GIFT) aims to make publicly available existing quantitative individual food consumption data from countries all over the world. In addition, the Global Dietary Database (GDD) provides information on dietary intakes of foods and nutrients for children and adults by age, sex, pregnancy/nursing status, rural/urban, and level of education.

Strengths:

• 24HRs provide quantitative estimates of individual food consumption and nutrient intake
• Can structure survey—and analyses—to include information on food sources and preparation methods
• Can account for foods consumed together that may enhance or inhibit micronutrient absorption
• Offer a higher degree of accuracy in assessing food and nutrient intake relative to FFQs or estimates derived from HCES

Weaknesses:

• Given relative complexity of 24HR surveys, significant training of enumerators is required to minimize errors in data collection
• Accurately recalling the quantity consumed can be challenging for respondents and a relatively large source of error in 24HRs compared to enumerator-administered Weighed Food Records (WFR)
• Data are frequently collected from small samples that are not nationally representative
• As with other surveys that rely on memory and are administered by an enumerator, there is potential for recall bias and interviewer bias
• Like most surveys, to capture seasonal variation data collection must span the entire year or be repeated in multiple seasons

Summary

Food Frequency Questionnaires (FFQ) are a type of dietary assessment instrument that attempts to capture an individual’s usual food consumption by querying the frequency at which the respondent consumed food items based on a predefined food list. Given that food lists are culturally specific, FFQs need to be adapted and validated for use in different contexts (Thompson & Subar, 2013).

FFQs are a common method for measuring dietary patterns in large epidemiological studies of diet and health. FFQs are often limited to the food items that are a source of nutrients related to the particular dietary exposures under study, for example, fruit and vegetable consumption or foods with high levels of saturated fat. Dietary diversity scores are a type of metric that are often calculated from a simplified FFQ (see the description of Dietary Diversity metrics to learn more). Food consumption modules of Household Consumption and Expenditure Surveys (HCES) that use a food list and an extended recall period can also be considered a type of FFQ.

In general, FFQs rely on a longer recall period in order to capture foods that are not consumed every day but are still part of the individual’s typical diet. FFQ recall periods vary greatly, but typically range from 7 to 30 days (although some are as long as one year). A drawback is that recall bias may increase with longer periods of recall (Coates et al., 2012). However, these measures of ‘usual intake’ are a more valid indicator of the relationship between diet and health outcomes than those capturing only a single 24-hour snapshot of the diet (24-hour Dietary Recalls can only provide information on usual intake if data are collected from respondents on multiple non-consecutive days). Longer FFQs can better assess total diets, but shorter FFQs have higher response rates and lower respondent burdens (Thompson & Subar, 2013).

FFQs typically collect information on the frequency of consumption but not necessarily on the quantity consumed. When FFQs do include questions about quantity consumed it is typically based on standard portion sizes, rather than direct weight or use of household utensils. Therefore, FFQs are not as accurate as other quantitative dietary assessment methods (e.g. 24-hour Dietary Recall) (Coates et al., 2012). Additional measurement error is introduced when food lists are not specific to the studied population, when questionnaires use inconsistent or imprecise portion sizes (Shim et al., 2014), or when the food lists are not granular enough to make an accurate match to a food composition table for deriving nutrient content of the diet. Because food lists are developed with a specific population in mind, it can be difficult to accurately compare results across populations (cultures or countries) with different dietary patterns.

Strengths:

• Better at estimating ‘usual diet’ due to longer recall period than the 24-hour Dietary Recall or 24-hour Weighed Food Records
• Captures individual-level dietary patterns
• FFQs can be easier and less time-consuming to implement than a 24-hour Dietary Recall, if the food list is relatively short (e.g. <100 items)

Weaknesses:

• FFQs require substantial up-front investment to develop and validate the instrument (food list and quantities) for a given context or country.
• Usual frequency of intake is prone to measurement error, particularly with recall periods longer than seven days (and usual portion size questions are prone to measurement error)
• If the FFQ is too long it can be more time consuming to administer than a standard 24-hour Dietary Recall and cause respondent fatigue
• Like most surveys, to capture seasonal variation data collection must span the entire year or be repeated in multiple seasons

Summary

Food Composition Databases (FCDB)—sometimes also referred to as Food Composition Tables (FCT) if in printed or PDF format—are collections of data on the nutritional content of foods. They are derived from quantitative analyses of representative samples of foods (Gibson, 2005). FCDBs have multiple uses, including for nutrient analysis of foods from dietary consumption surveys, nutrition labeling, and to inform nutrition-sensitive agricultural policies (Charrondiere et al., 2011). The Food and Agriculture Organization (FAO) is the global coordinator of the International Network of Food Data Systems (INFOODS), through which it compiles a directory of national, regional, and international FCDBs that can be useful to those interested in analyzing food and nutrient availability and consumption, food fortification or supplementation programs (Greenfield & Southgate, 2003). In addition, FAO/INFOODS produces guidelines for developing FCDBs, food matching, and converting food data (FAO/INFOODS, 2018).

FCDBs are sometimes available online, but not always. FAO/INFOODS provides contact information and links to country and regional FCDBs. In addition, the ILSI Research Foundation has created the World Nutrient Databases for Dietary Studies (WNDDS), which catalogues 90 electronically available FCDBs and provides detailed information about each one. Electronic access to these data means that the information can be quickly updated, they contain a greater volume of material, and they are readily available for users with internet access (Greenfield & Southgate, 2003). Additionally, online FCDBs make it easier for the information to be reformulated according to the needs of various users.

While some low- and middle-income countries have national FCDBs, they often contain data that is several decades old and/or rely on information that is from another country’s FCDB. This is due to the expense and time-intensive nature of analyzing or gathering nutrient composition data (Greenfield & Southgate, 2003). Frequently, borrowed food composition data comes from the USDA and EU FCDBs, or from other countries in the region and regional FCDBs (Coates et al., 2017). However, nutrient contents of foods can vary due to environmental factors, production, and processing and thus can differ from one country to the next and even within countries (Greenfield & Southgate, 2003). Rapid food processing advancements and globalized food systems can challenge the task of keeping FCDBs both up-to-date and specific to the locale (Thompson & Subar, 2013). These various issues may result in decreased precision when it comes to identifying the nutrient content in a given food.

Given the complexity of FCDBs, adequate training on food composition data use is recommended, to comprehend and use the data. One recommended resource is the FAO e-learning course on food composition data (Food Composition Data E-learning Course, 2013).

Strengths:

• Food composition data have a wide variety of uses, including matching foods with nutrients from dietary assessment data in order to conduct analyses, nutrition labeling, policy making, and nutrition-sensitive agriculture
• Well-developed national FCDBs can offer a picture of the types of food available and consumed
• When FCDBs are paired with dietary consumption data, researchers are able to answer questions about nutrient adequacy in a population

Weaknesses:

• Differences in the development of FCDBs (e.g. nutrient calculations and sampling) can reduce comparability of nutrient data for specific foods across databases
• Adequate training on food composition data use is recommended to comprehend and use the data in FCDBs
• Food composition data that are outdated or from other countries are sometimes relied upon to update the national FCDBs of low- and middle-income countries