Key results
Our study provides a comprehensive exploration of population-level patterns of protein purchased and consumed in older adults in the UK. We find a decline in the percentage of energy derived from protein both consumed and purchased in the oldest age groups, and similarly fewer adults in the two oldest age groups meeting recommended protein intake levels. A higher proportion of protein was reported in the consumption data compared to the purchasing data. Protein purchasing and consumption was largest for meat and poultry products, but was also high for dairy and bread/bakery items as well. There was low purchasing and consumption of plant proteins.
Interpretation
Our findings reveal a nuanced association between age group and protein behaviours. Protein consumption declined with increasing age group for males, but not for females (Fig. 1). In contrast, declining protein purchasing from 60 years onwards was consistent by sex (Fig. 2). The proportion of individuals meeting protein recommendations declined linearly with age group for females, but not for males (Table 1). Declines in protein behaviours by age group were not consistent across all sources of protein (Table 2 and Fig. 4). Designing interventions around protein and age group may help to elucidate these complexities.
Evidence of lower levels of protein consumption and purchasing in the oldest age group represents an important area for policy makers. Protein intake is associated with the degree of muscle decline in ageing adults [16], and therefore represents an important determinant of sarcopenia. Dietary interventions might be more feasible and less intrusive than strength training in older adults [6]. However, focusing purely on diet ignores the complex reasons for this pattern including lower energy needs in older adults, difficulty in preparing meals, underlying comorbidities, changing preferences, and changes in dentition that impact the ability to consume certain foods [28]. Protein consumption is linked to satiety and therefore any intervention would need to offset potential declines to energy consumption. Furthermore, since sarcopenia is promoted by food choices over many years, to be most effective interventions need to target behavioural change much earlier in the life course [26].
The lower percentage of protein found in the purchasing data compared to the consumption data present novel findings. Few studies have previously compared these data types due to a lack of availability of loyalty card data [27]. It was not clear why the difference in consumption and purchasing data might exist. A variety of reasons include individuals underreporting what they consume (e.g. underreporting consumption of purchased non-protein sources) [3, 20], different sample years or the varying sociodemographic makeup of individuals within each dataset. Our purchasing data was closer than our consumption data to reported values in the US (~ 15%) [6]. They may also reflect broader consumer dynamics, such as individual’s purchasing protein from alternative retail outlets than just the high street supermarket (e.g. local butchers). People purchasing foods may not be buying foods that they will necessarily consume; household composition matters (data may be more accurate for single or dual person households than families). Finally, differences may reflect food wastage patterns where high value meat is prioritised over bread, fruit and vegetables [25].
We find evidence that not all adults meet a range of protein recommendations, with patterns declining with age group as well. While our findings contrast to evidence from the UK Biobank [7], UK Biobank is not a representative survey and our findings match results from other contexts [15, 23]. There are currently no UK guidelines on recommended protein intake that take into account ageing [21]. Improving the clarity of messaging around protein might help encourage adequate consumption of protein in older adults. However, defining the expected level of protein is difficult. While the international recommended dietary allowance of 0.8 g/kg is set at two standard deviations above the minimum amount of protein to maintain body protein and loss of nitrogen [28], this figure does not necessarily represent optimal intake or reflect evidence that older adults would benefit from greater amounts of protein intake (not solely for muscle maintenance, but also broader health benefits) [6, 11, 16, 19, 29]. Focusing on the more stringent protein guidelines we examined might be more beneficial, particularly given that a minority of individuals currently meet such recommendations.
Timing of protein was skewed towards evening meal, and to a lesser extent lunch meals. This finding follows previous research [10]. On average, individuals were not meeting the recommended 25-30 g of protein required to maintain muscle mass and function at either of these periods [21, 24]. Protein intake during breakfast was particularly low, suggesting an opportunity to target this meal in interventions [10]. The low levels of protein in-between main meal times also present opportunities for interventions. Smaller but more regular protein snacks throughout the day may help individuals increase their protein consumption [21]. While many protein-rich snacks exist, they tend to be focused towards high performance and athletes, which limits their general appeal [26].
Finally, our study provides novel insight into how supermarket loyalty card data may aid our understanding of protein-related behaviours. There has been a lot of excitement about the promise of big data, however there have been few applications in nutrition-related studies so far [27]. Our study reveals nuances in purchasing behaviours that change across age groups, which both support and extend the observations in the NDNS. However, the loyalty card data cannot answer all questions surrounding protein; we were unable to assess whether protein purchased was consumed (or who it was purchased for), when it was consumed or how purchasing related to recommended levels of intake. Purchase patterns are likely to be driven by a main shopper, who may act as a gatekeeper for a household. It is clear that these new forms of data can only supplement traditional data and such data will remain important in answering future research questions.
Limitations
There are several limitations of our study. A non-disclosure agreement was signed for use of the in-store purchase data and we are restricted in the details that we can report here. We were unable to report key sample information including sample size (which was in the order of millions) or basic demographic characteristics of loyalty card holders (which were not representative of the UK population). Not being able to report the representativeness of a sample can limit the ability to scrutinise the quality of the data, which is important when we are making comparisons across data sets. Loyalty card usage varies by age group and sex introducing bias into estimates as well [31].
Loyalty cards may be shared between individuals (e.g. individuals purchasing for friends and family unable to visit supermarkets themselves) and within households reducing their applicability for studying individual-level patterns in purchasing behaviours. This restricts the ability to draw comparisons to the NDNS data and may limit our conclusions. It represents a broader concern with the applications of supermarket loyalty card data and how they might supplement other forms of data [20].
The focus of our paper on protein restricts the wider conclusions we can draw on the use of supermarket loyalty card records as a dietary assessment tool. Although this was not the aim of our study, validating how purchasing patterns for a greater range of macro- and micro-nutrients relate to consumption patterns is needed to evaluate the usefulness of supermarket loyalty card data in diet-related research. Future studies should also consider how patterns vary to other dietary assessment methods and surveys, link loyalty cards to individual-level surveys to allow for direct comparisons, and collect data from multiple loyalty cards to cover all potential transactions.
Our analyses were mainly descriptive and cross-sectional. We are limited in our ability to draw associations between dietary behaviours and age. There are few studies that have examined longitudinal associations of protein intake [21]. Extending our analyses across the life course might help to shed light on reasons why protein consumption changes across age groups. Longitudinal and repeated cross-sectional data would also help to examine trends in dietary behaviours and the context to our data. Examining the existence of cohort effects is also important to determine how protein consumption and purchasing changes across age groups.