A comparison of national dietary surveys (Scottish Health Survey 2021 and Dietary Intake in Scotland’s children 2024) with Worldpanel by Numerator household purchase data

Collecting accurate information about dietary intake is vital for public health surveillance and policy. However, self-reported diet surveys, which ask people to recall what they ate and drank over a period of time, come with the well-known challenge of misreporting of energy intake. A study using the doubly labelled water (DLW) method as part of the UK Diet and Nutrition Survey (NDNS), in 2019-20 and 2020-22, found that people reported eating 30% fewer calories in 24-hour dietary intake recalls than their actual energy use measured by the DLW method. DLW measures how much energy the body uses and is often used as a proxy for energy intake, however, it cannot show which types of food or drink are mis-reported, or whether specific nutrients are under- or over- reported. Understanding which foods and drinks are under-reported, and by how much, helps us see how misreporting affects estimates of dietary intake in national surveys.

This report aims to investigate the types of food and drink that may be misreported by comparing food and drink intake data from the self-administered 24-hour dietary recall method used in two national diet surveys - The 2021 Scottish Health Survey (SHeS) and the 2024 Dietary Intake in Scotland’s children (DISH) survey, with retail purchase data from Worldpanel by Numerator (‘Worldpanel data provided to FSS’). It also explores how misreporting may affect estimates of energy intake and nutrients such as fat, saturated fat, carbohydrates, fibre, total sugars and sodium. The report concluded by identifying ways to improve dietary reporting using the self-administered 24-hour dietary recall method and recommending further research to improve understanding of misreporting.   

A review of previous research was carried out to look at misreporting of specific types of foods and drinks in 24-hour dietary recalls. Thirteen relevant studies were identified from published research and evaluated. These studies showed inconsistent findings with no consistent pattern of which foods and drinks were over- or under- reported. Several possible reasons for misreporting were identified, such as forgetting foods and drinks that had been consumed, and difficulty in estimating portion sizes.

The data from the SHeS and DISH surveys were adjusted so they could be combined and analysed together. This adjustment followed official guidance from the NDNS (UK Data Archive Study (2020)). Worldpanel by Numerator data used in the report was from a subset of the Worldpanel by Numerator take home panel including data for Scotland only, for 52 w/e 26 Dec 21 and is referred to as ‘Worldpanel data provided to FSS’.

Only certain food groups could be compared, as some food groups in the ‘Worldpanel data provided to FSS’ did not correspond closely to those in the SHeS or DISH survey data. For the food groups that could be compared, new food group categories were created, and the ‘Worldpanel data provided to FSS’ and the SHeS/DISH food groups were matched to these new categories. The new groups were rated as ‘high’, ‘medium’ and ‘low’ confidence, depending on how well the foods included in the groups matched. Those classified as ‘high’ had largely like-for-like comparisons; ‘medium’ had small caveats; and ‘low’ had larger caveats. Together, these new food groups represented 71% of total reported energy intake in SHeS/DISH, and similar proportions for fat, total sugars and sodium.

To compare data from SHeS/DISH with ‘Worldpanel data provided to FSS’, first the average daily calories, fat, saturated fat, carbohydrates, fibre, total sugars, and sodium from each food group for different age and sex groups were calculated, using only foods and drinks bought from retailers for SHeS/DISH data (takeaway and out-of-home consumption were excluded). Then the averages were scaled up using Scottish population census data to estimate total annual intake for the population to match the ‘Worldpanel data provided to FSS’, for comparison.

The analysis showed consistently lower intake of energy and nutrients in SHeS/DISH  compared to the ‘Worldpanel data provided to FSS’. Almost all high confidence and all medium confidence food groups had lower reported energy intake in SHeS/DISH when compared with ‘Worldpanel data provided to FSS’. For fat intake, most matched food groups showed lower intakes in SHeS/DISH than in the ‘Worldpanel data provided to FSS’; 86% of high-confidence food groups and 80% of medium-confidence food groups. A similar pattern was seen for saturated fat, carbohydrates, fibre, total sugars and sodium.

Alcohol, biscuits, confectionery, cakes and pastries were among the foods that had the largest discrepancy between SHeS/DISH and ‘Worldpanel data provided to FSS’ for energy, fat, saturated fat, carbohydrates, and total sugars. Fibre from some of these foods was also lower in SHeS/DISH. Milk and cream were also lower in SHeS/DISH, which matches findings from a previous study in Aberdeenshire. However, some of the differences might be due to food waste rather than reporting errors. 

Bread, yoghurt, fruit and eggs showed strong agreement between the data sources for energy and nutrients. However, because SHeS/DISH does not include bread used for homemade sandwiches, bread can't be compared with high confidence, although results suggest it is reported more accurately than other food groups. Eggs had slightly larger differences in energy than bread, yoghurt or fruit. This might be partly because the egg category created for comparison only included certain foods. For example, plain omelettes were included, but those with meat or vegetables were not. 

The report helps to identify food groups that may be mis-reported by participants using the self-administered 24-hour dietary recall method, however it has some limitations. For example, differences caused by reporting errors could not fully be separated from methodological differences such as differing underlying nutrient databases, and the entire diet could not be compared due to differences in some food groups. Also, ‘Worldpanel data provided to FSS’ is household rather than individual level data so it could not be determined whether differences between SHeS/DISH and ‘Worldpanel data provided to FSS’ were due to smaller portions or missing foods. It was also not possible to account for foods wasted in the household, which would apply to the food purchase but not the diet recall data. 

Key recommendations for further research and improvements to monitoring of diets in Scotland include:  

• Future analyses comparing household food expenditure data in single occupancy households to triangulate findings.
• Additional research on the accuracy of reporting of milk and cream.
• The software used in the self-administered 24-hour dietary recall method, Intake24, could consider integrating multi-pass questions relating to alcohol intake and snack foods, and review and update portion size estimation images for biscuits, confectionery, cakes and pastries.