Food balance sheets

At a country level, dietary consumption is surrogated using indirect approaches that embed household food purchases, or even food balance sheets (FBS) from which per capita food data availability can be estimated. In FBS, food availability for human consumption is calculated by taking into account production, imports, exports, stock variation and utilisation elements, such as feed, seed, losses and industrial uses^{( 1 )}. FBS are one of the most important sources of data on food availability for human consumption in most countries. However, it remains a challenge to collect accurate data about all the food groups, and the quality of the food data cannot always satisfy high-quality standards. As a result, FBS figures of food consumption are prone to measurement errors⁽ Reference Del Gobbo, Khatibzadeh and Imamura ^{2 )}. Furthermore, FBS neither give any indication of the differences that may exist in the diet consumed by different population groups, such as people of different socio-economic groups, ecological zones or geographical areas within a country⁽ Reference Claro, Jaime and Lock ^{3 )}, nor do they provide information on seasonal variations in the total food supply.

Household expenditure surveys and individual food consumption surveys

Data obtained through household and individual food consumption surveys (IFCS) are often the preferred source of food consumption estimates because they provide more information on food actually consumed than FBS do⁽ Reference Serra-Majem, MacLean and Ribas ^{4 )}. However, it should be emphasised that the food data collected in household consumption and expenditure surveys reflect the quantity of food acquired by a household rather than that consumed by its members. Data on the latter are collected in IFCS, also called dietary intake surveys or diet monitoring surveys. Compared with the household consumption and expenditure survey method, the IFCS method yields a more reliable measure of dietary adequacy and exposure to contaminants at individual and household levels because it is based on actual food consumption rather than food acquisition⁽ Reference Engle-Stone and Brown ^{5 )}. Further, since the data are collected at the individual level, comparisons can be made across age and sex groups, and inequalities in intra-household food distribution can be identified. However, collection and processing costs for these individual data are considerably higher than for the data collected in household consumption and expenditure surveys. For this reason, implementation of the IFCS method at the country level may be challenging for low- and middle-income countries⁽ Reference Pisa, Landais and Margetts ^{6 )}, therefore limiting the use of this approach for cross-country comparisons.

Although dietary monitoring and surveillance is crucial in the development of targeted interventions among population groups at risk of nutritional inadequacies, several countries (particularly those in deprived regions such as Africa) have no representative national IFCS system in place yet⁽ Reference Pisa, Landais and Margetts ^{6 )}. Furthermore, in most countries where food consumption surveys are conducted, the data are not collected routinely, which undermines trend analyses⁽ Reference Micha, Khatibzadeh and Shi ^{7 ,} Reference De Keyzer, Bracke and McNaughton ^{8 )}. Despite the challenges involved, IFCS are performed in several countries and serve as the basis to evaluate compliance with nutrient intake requirements or exposure to potentially harmful dietary contaminants/components. Opportunities and barriers identified in past and ongoing IFCS can serve as a basis to develop and optimise future surveys in regions where IFCS are still lacking. Furthermore, harmonisation of the methodologies of data collection in national surveys, as well as of recruitment, training and supervision of interviewers will increase the comparability of the data collected. A cross-continental comparison of national food consumption survey methods has already been published, including studies reporting on IFCS from 1985 to December 2011⁽ Reference De Keyzer, Bracke and McNaughton ^{8 )}. This previous review described how few countries endeavoured to increase the potential of their national IFCS through implementation in other national health surveys (e.g. health interview surveys or health examination surveys), an approach likely to be cost-effective, especially in resource-constrained environments often reported in low- and middle-income countries⁽ Reference De Keyzer, Bracke and McNaughton ^{8 )}.

In this review, we compare national IFCS and their potential for health research through integration in national health surveillance programmes worldwide. More specifically, we inventoried methods and designs used in national IFCS. We further discuss and compare these methodological properties between nations and cross-continentally and evaluate the potential for embedding food consumption survey in national health surveillance programmes.

General survey characteristics

National IFCS systems are available in most of the high-income countries, while such surveys are scarce in low- and middle-income countries (such as Africa, Eastern Europe and several Asian countries). The IFCS were cross-sectional, with only few having a continuing character (Table 1). Nineteen surveys were identified as being conducted erratically (Ethiopia^{( 12 )}, Nigeria⁽ Reference Maziya-Dixon, Akinyele and Oguntona ^{13 )}, South Africa⁽ Reference Shisana, Rehle and Simbayi ^{14 )}, Bahrain⁽ Reference Wahab, Gharib and Al Sairafi ^{15 )}, Belgium⁽ Reference Bel, Van den Abeele and Lebacq ^{16 )}, Bulgaria⁽ Reference Petrova and Angelova ^{17 )}, Czech Republic⁽ Reference Ruprich, Dofkova and Rehurkova ^{18 )}, Estonia⁽ Reference Pudule, Grinberga and Kadziauskiene ^{19 ,} Reference Pomerleau, McKee and Robertson ^{20 )}, Finland⁽ Reference Reinivuo, Hirvonen and Ovaskainen ^{21 )}, France⁽ Reference Dubuisson, Lioret and Touvier ^{22 )}, Hungary⁽ Reference Szeitz-Szabó, Bíró and Bíró ^{23 )}, Ireland⁽ Reference Harrington, Robson and Kiely ^{24 )}, Latvia⁽ Reference Merten, Ferrari and Bakker ^{10 )}, Poland⁽ Reference Sekula, Nelson and Figurska ^{25 )}, Slovakia⁽ Reference Merten, Ferrari and Bakker ^{10 )}, Slovenia⁽ Reference Merten, Ferrari and Bakker ^{10 )}, Brazil⁽ Reference Sichieri, Pereira and Martins ^{26 )}, Chili⁽ Reference Hugo, Patricia and Marcelo ^{27 )}, New Zealand^{( 28 )}); fifteen on a regular basis (at a fixed frequency; e.g. annually, biennial; India^{( 29 )}, Israel⁽ Reference Keinan-Boker, Noyman and Chinich ^{30 , 31 )}, Japan^{( 32 )}, South Korea⁽ Reference Kim ^{33 )}, Malaysia^{( 34 )}, Philippines^{( 35 )}, Austria⁽ Reference Elmadfa, Freisling and Nowak ^{36 )}, Denmark⁽ Reference Pedersen, Christensen and Matthiessen ^{37 )}, Germany⁽ Reference Heuer, Krems and Moon ^{38 )}, Italy⁽ Reference Leclercq, Arcella and Piccinelli ^{39 ,} Reference Sette, Le Donne and Piccinelli ^{40 )}, The Netherlands⁽ Reference Rossum, Fransen and Verkaik-Kloosterman ^{41 )}, Spain⁽ Reference Ortega, Lopez-Sobaler and Ballesteros ^{42 )}, Sweden⁽ Reference Becker and Pearson ^{43 )}, Canada^{( 44 )}, Mexico⁽ Reference Barquera, Hernandez-Barrera and Campos-Nonato ^{45 )}), while five were continuous (ceaselessly from one year to another) surveys over years (China⁽ Reference Zhang, Zhai and Du ^{46 , 47 )}, the UK⁽ Reference Bates, Lennox and Prentice ^{48 )}, the USA^{( 49 )}, Australia^{( 50 )}, Venezuela^{( 51 )}). Japan was one of the first countries to start collecting IFCS routinely with data collected throughout years since 1946. Other countries conducted IFCS regularly: India, Israel, Japan, South Korea, Malaysia, Philippines, Austria, Denmark, Germany, Italy, The Netherlands, Spain, Sweden, Canada and Mexico. For the USA and China, participants are included in a cohort for tracking over time. Countries which have integrated IFCS in their health survey are: South Africa, Australia, China, Japan, South Korea, Malaysia, Canada, Mexico and the USA. All of the IFCS integrated in health surveys, besides South Africa, have the advantage of being continuous (Australia, the USA and China) or regular (Canada, Japan, South Korea, Malaysia and Mexico), while most of the other IFCS are rather erratic.

Target population, survey design and sampling method

Supplementary Table S1 summarises the study design aspects and methods of the selected surveys. The age of the target populations ranged from <1 year to over 80 years (in surveys that do not include an age <18 years limit). In all surveys, both sexes were included except for the Nigerian IFCS that included solely women. Different sampling designs were used to select study participants with a majority applying a multistage and stratified sampling. The sampling frames used for the selection of sampling units were based either on census data (e.g. the USA, Mexico, Brazil, South Korea and New Zealand), a combination of frames like healthcare registries and labour force data (Canada), strata from counties (China) or other sampling frames, such as enumeration blocks, which are geographical areas artificially created to have several living quarters (Malaysia).

Numbers of participants and participation rates

In Table 1 and in Supplementary Table S1, recruitment aspects of all selected surveys are listed. Sample sizes of single surveys ranged from about 400 (Spain and Slovenia) to over 30 000 (Canada and Brazil). Participation rates were above 90 % in Malaysia, Ethiopia and South Africa; between 80·0 and 89·9 % in the USA (National Health and Nutrition Examination Survey (NHANES) 2001, 2005), Mexico (National Health and Nutrition Survey 1999), Brazil, China and South Korea; between 70·0 and 79·9 % in Canada, the USA (NHANES 2003, 2007 and 2009) and Australia (for the FFQ); and below 70 % in Japan, Australia (for 24-HDR) and New Zealand. Overall, participation rates were lowest in European countries (between 23 and 80 %, except for Poland and Slovakia where response rates were 96 %). The integrated surveys often seem to have higher response rates, a random sampling design and are therefore often more representative for the general population. Although the higher response rates in these integrated surveys may be due to other factors that are out of the scope of this inventory, these numbers are demonstrating the feasibility of integrating an IFCS in other health surveys as the extra respondent burden seems not to reduce the response rate.

Dietary intake assessment methods

Most surveys used 24-HDR as the principal DIA method (Table 3). Multiple recalls for all participants were available in China (three recalls) and in most of the European countries. In some countries, duplicate recalls were available in a subsample only (Canada, South Korea, Australia and New Zealand). Computer-assisted personal interviewing was performed in the USA, Malaysia, Brazil and New Zealand. In the surveys from China and Australia, the 24-HDR was performed with paper and pencil in a face-to-face interview. In South Korea, a face-to-face interview was performed, no interview software was reported, and in Mexico, the method of administration of the 24-HDR was also not reported in the study report. A prospective DIA method was only used in Brazil and Japan (2-d estimated dietary records and 1- or 3-d semi-weighed dietary records, respectively). Finally, Mexico (Mexican Health and Nutrition Survey 2006) and South Africa used only a semi-quantitative FFQ to report on frequencies of intake during the past 7 d. An FFQ (formerly called food propensity questionnaire) was also used in addition to a principal DIA method to identify frequencies of consumption and non-consumers of various food groups in Canada, the USA, Japan, South Korea, Malaysia, Australia and New Zealand (National Nutrition Survey 1997) and several European countries. For particular population groups such as children, the 24-HDR interview is often complemented with a food diary, or food diaries can also be used as stand-alone method. This is the case in most European countries and allows food consumption that occurred when the children were in the presence of care givers who differ from the adult being interviewed (e.g. care givers of kindergarten) to be captured⁽ Reference Freisling, Ocke and Casagrande ^{52 ,} Reference Ocke, Brants and Dofkova ^{53 )}.

Additional assessments and measurements

Almost all IFCS collected additional information in addition to the dietary intake data. The most common extra parameters included in the IFCS were anthropometrics (measured or estimated), physical activity questions/measurements and socio-demographical information (Supplementary Table 3). Anthropometrical data were always collected through measurements performed by trained researchers in the IFCS integrated in health surveys, while in other dietary surveys, these were often estimated (respondent recall/report).

Fieldwork characteristics and data controls

In Table 2 and Supplementary Table S1, the fieldwork aspects of the nutrition surveys are presented. Interviews were conducted either face-to-face or by telephone. Places other than the homes of the participants for administrating the DIA were examination centres (e.g. South Korea and several European countries and mobile examination centres in the USA). The time-span of the fieldwork was at least 1 year (all seasons) in almost all European countries, Canada, the USA, Brazil, South Korea (Korean NHANES 2008 and onwards), Malaysia, Australia and New Zealand. In the African countries running an IFCS, only few seasons/months were included in the fieldwork period.

Food linking and statistical analysis

Supplementary Table S4 summarises features related to data analyses of the dietary intake data. Surveys using multiple measures of intake allow for correcting of within-person variability. Most surveys used the Nusser method⁽ Reference Nusser, Carriquiry and Jensen ^{54 )} (using Software for Intake Distribution Estimation SIDE or C-SIDE) developed at the Iowa State University to calculate distributions of usual intake (Canada, the USA NHANES 2003, Brazil, South Korea and New Zealand and few European countries). For the USA, from NHANES 2005 and onwards, the NCI method developed by the National Cancer Institute was used (https://epi.grants.cancer.gov/diet/usualintakes/method.html). Finally, in the Australian survey, an equation developed by the US National Academy of Science was used to adjust for within-person variance⁽ Reference Mackerras and Rutishauser ^{55 )}. Misreporting of energy intake was assessed using either the Goldberg method⁽ Reference Goldberg, Black and Jebb ^{56 )} (the USA, Malaysia, Australia and most European countries) or the equations by Black and Cole⁽ Reference Black and Cole ^{57 )} (Canada). Whether the Goldberg method was applied at the individual or group level was not always specified. Two surveys indicated that no calculation of misreporting was performed (South Korea and New Zealand), while several surveys did not report whether misreporting was analysed and if any misreporting adjustments were done.

Recruitment and training of field staff

In Supplementary Table S2, recruitment and training of the interviewers and field staff in the nutrition surveys are described. In twenty-three IFCS, the interviewers were nutritionists or dietitians. In some countries, interviews were performed by trained interviewers (sometimes in addition to the nutritionists) who are familiar with local food customs (New Zealand) or professional interviewers working on a variety of surveys (Canada). For interviewers in the USA, a high school diploma was considered to be the minimum education requirement, as this is necessary for government jobs. However, also dietitians/nutritionists were available to perform the IFCS in the USA. Training was provided on a variety of topics such as interviewing and probing skills (Canada, the USA, Brazil, China, South Korea, Malaysia, Australia and New Zealand), training on contacting participants and software training. The duration of these training sessions ranged from 3 d (China) to 3 weeks (India). The average duration of reported training programmes for interviewers was approximately 7 d.