The UK-CGIAR’s FRESH-EN UP project has released a new preprint, shedding light on the use of household surveys to inform food and nutrition policies.
An increasing number of research initiatives draw on household consumption and expenditure surveys (HCES) for dietary information, with insights used by national and international stakeholders to guide agriculture and nutrition strategies as well as interventions. But new evidence highlights why we need to use and interpret these data with caution, and how adjustments to survey tools could strengthen the quality of these data for future agri-food systems and nutrition research.
“Our findings show that HCES data, even when carefully analysed, may systematically underestimate food intake—particularly in larger households,” said Fanny Sandalinas, Research Fellow at the London School of Hygiene & Tropical Medicine (LSHTM) and lead author of the paper. “This may lead to overestimation of nutrient deficiency risks, especially when applied to population-level planning.”
The research was conducted by the UK-CGIAR Centre’s commissioned project Fruit and Vegetables for Sustainable Healthy Diets (FRESH)-Expanded Network with UK Partners (FRESH-EN UP), led by the London School of Hygiene & Tropical Medicine, International Food Policy Research Institute (IFPRI), and Sokoine University of Agriculture (SUA). The project’s aim is to increase fruit and vegetable intake, especially among women and youth, and production of vegetables in Tanzania.
“This project brought some new partners together to define the study objectives and to build their partnership more generally, which contributed to an increased understanding of what it takes to build strong partnerships that leverage institutions respective strengths,” said Deanna Olney, joint lead on the FRESH-EN UP project, and a co-author of the paper. “The findings are useful as we consider what type of evidence is needed to inform well-designed diet-related programmes and policies in Tanzania and elsewhere.”
A pre-print version of the paper can be read on medRyix.
The use of household surveys for nutrition surveillance
Constrained and shrinking budgets for nutrition surveillance mean many countries are turning to existing, multi-purpose household surveys for dietary data on populations. HCES are routinely conducted by national governments to generate economic and social indicators of development and wellbeing. These surveys typically involve interviewing thousands of households and are designed to be representative of populations at national, regional or district scales. In the absence of large-scale, individual-level dietary data, researchers have often turned to these data to draw insights on population diets, estimate nutrient gaps, guide national food fortification strategies, and track progress of diet quality goals.
While HCES were not originally designed with nutrition research in mind, food consumption data are usually collected within the survey. Typically, household members responsible for acquiring and preparing food are asked by researchers to recall all food consumed by household members over the past seven days, using a fixed food item list.
The obvious problem with this approach is that it can yield large recall errors, (i.e. over or under-estimates of food item consumption). It is difficult enough recalling one’s own food consumption in the past 24 hours, let alone food consumed by all household members over the past week. Furthermore, depending on the use-case, the household-level dietary data may need to be ‘individualised’ (i.e., converted to consumption quantity per capita or per adult equivalent) to normalise intake estimates across households with diverse demographic compositions.
When methods diverge
This paper used data from 2,600 rural households in the Arusha and Kilimanjaro regions. Researchers compared estimates of food group consumption from individual-level dietary recall (24hR) data from adult women with household-level food consumption data collected in the same household. These data were generated by the CGIAR FRESH Initiative, which developed and applied an end-to-end approach to increasing fruit and vegetable intake in Tanzania and other target countries. FRESH’s work is now part of the CGIAR Science Program on Better Diets and Nutrition.
Additionally, researchers compared food group consumption estimates with household-level food consumption data from the Tanzania National Panel Survey (TNPS, wave 4). The TNPS and FRESH household survey used almost identical methods and survey tools, although the FRESH survey used a more detailed food item list for fruit and vegetables. The TNPS and FRESH household survey data were ‘individualised’ using the adult female equivalent (AFE) method, to generate an estimate of food group consumption per adult woman, enabling comparability with the individual-level 24hR data.
Overall, dietary patterns were broadly similar between the individual-level and household survey methods, including the low consumption of nutrient-dense foods such as fruit, vegetables, and animal-source foods.
However, some critical differences emerged. Energy intakes were substantially lower when estimated from HCES data—by as much as 42% in large households. Also, trends in food group consumption by household wealth differed between the two methods for some food groups, including cereals and milk.
“If you rely solely on household data, you might misrepresent the scale and nature of the nutritional challenges facing populations. Also, you can miss intra-household differences in access to nutritious foods,” said Lily Bliznashka, Research Fellow at IFPRI and a co-author on the paper. She added, “Individual-level dietary data remain essential for a more complete understanding of who is being left behind.”
Interpreting HCES data with care
Despite these challenges, the authors emphasize that HCES remain a valuable tool for identifying nutritional issues and informing strategies to improve diets—including in Tanzania, which like many countries does not have individual-level dietary data from national-scale surveys.
“We need to increase fruit and vegetable consumption in Tanzania to address nutrient deficiencies and prevent non-communicable diseases”, saidJoyce Kinabo, a Professor of Human Nutrition at SUA and a co-author on the paper. “Using household survey data adds an important contribution to understanding the level of consumption at household level, how this varies across the country, and how this relates to household factors. The information could be used to make important national decisions in terms of allocation of resources for production of various foods, such as fruits and vegetables, that are limiting nutrient consumption profiles in the household. Nevertheless, individual data are necessary for interventions at household level.”
A smarter way forward
To improve the value of HCES as a proxy source of individual-level dietary data, the authors propose a combination of practical survey adaptations and analytical innovations.
Small changes to surveys could make a big difference. For example, questions could be added on how food is typically shared, what proportion of meals each household member consumed at home over the past week, and whether some foods are reserved for specific household members.
Methods for analysing results could be refined to adjust for household size effects or to use relevant individual-level dietary data to inform assumptions on within-household food distribution.
“Household surveys can provide valuable data on what people eat, including at sub-national resolution and with equity insights, informing the development of appropriate, context specific agri-food system interventions to improve nutrition”, said Edward Joy, a co-author on the study and joint lead on the FRESH-EN UP grant. “Simple adaptations to survey methods and tools can result in significant improvements to the robustness and value of household survey dietary data.”
Household surveys could be periodically redesigned to collect individual-level dietary data, improving the calibration of HCES-based intake estimates and enhancing their usefulness for nutrition programme and policy design. Pilot efforts to integrate individual-level modules into household surveys are already underway in Tanzania.
As dedicated nutrition data systems face challenges, HCES are filling critical gaps. Improving their ability to reflect diets and inform agri-food system interventions is both a strategic opportunity and a necessity.
Acknowledgements
This material has been funded by UK International Development from the UK government; however, the views expressed do not necessarily reflect the UK government’s official policies.
The paper referenced in this blog used data collected and prepared with funding from the CGIAR Research Initiative on Fruit and Vegetables for Sustainable Healthy Diets (FRESH) was implemented by CGIAR researchers from IFPRI, CIMMYT, the Alliance of Bioversity International and CIAT, IWMI, and CIP, in close partnership with the World Vegetable Center, Applied Horticultural Research, the University of Sydney, the Institute of Development Studies, Wageningen University & Research, the University of California, Davis, Sokoine University of Agriculture, Wayamba University of Sri Lanka, and the Philippines Department of Science and Technology–Food and Nutrition Research Institute, along with other partners from 2022-2024. FRESH’s work is now part of the CGIAR Science Program on Better Diets and Nutrition (BDN), which aims to deliver consumer-oriented solutions to ensure sustainable healthy diets for all. Thus, we would also like to thank all funders who supported this research through their contributions to the CGIAR Trust Fund.
Main image credit: The World Food Vegetable Center