Danish receipt data create an unusually granular cohort for diet and disease research
Denmark now has a cohort built from something most epidemiology studies never see in usable form: the receipt itself. The Health, Food, Purchases and Lifestyle (SMIL) cohort links smartphone-captured consumer purchases with nationwide administrative health and social registries, creating a dynamic open cohort that moves beyond the weak recall of conventional dietary questionnaires. By the end of 2022, the database contained 11,214 participants who had consented to share their unique identification number for registry linkage. That linkage is the core scientific strength of the platform, because it allows purchase behavior to be studied alongside real-world health history rather than in isolation.
The cohort’s baseline structure already signals wide clinical relevance. Men made up 62% of participants, and 24% were aged 45 to 55 in 2022. The median number of unique shopping trips was 63, with an interquartile range of 26 to 116, a pattern that suggests repeated exposure measurement rather than a one-off snapshot. Within that population, 21% had a history of cardiovascular disease and 8% had diabetes before donating receipts. Those comorbidities matter because they place the cohort squarely in the zone where diet is likely to influence future vascular and metabolic outcomes, whether through excess energy intake, dietary fat quality, added sugar exposure, or broader purchase patterns that track with household habits.
One sentence captures the practical appeal of this platform: it turns routine consumer behavior into analyzable exposure data.
Shopping trips, not memory, define the exposure signal
The most consequential advance in SMIL is methodological rather than therapeutic. Investigators demonstrated that consumer purchase data can be translated into operational food groups and merged with registry data at scale. That step may sound technical, but it is what separates a promising digital dataset from a usable epidemiologic instrument. Once purchases can be coded into food groups, the cohort becomes capable of addressing questions that are notoriously difficult with self-report: what people actually bought, how spending shifts across demographic groups, and whether those patterns precede disease trajectories in linked registries.
The cohort also revealed marginal differences in the monetary percentage spent on food groups by sex, age, and hour of the day. Those timing patterns are more than a curiosity. Buying behavior by hour may function as a proxy for work schedules, household structure, impulsivity, routine meal timing, or late-day convenience purchases, all of which can influence diet quality in ways that standard questionnaires rarely capture. The exposure is not just what is purchased, but when it enters the home and how consistently it appears across repeated transactions.
A second sentence stands on its own: sugary drinks occupied a larger share of spending among young men than among young women.
Sex and age differences appeared across food spending patterns
The clearest example in the cohort came from sugary drink expenditure. Men under 30 spent 8.2% of total reported expenditure on sugary drinks, compared with 6.5% for women under 30. Among participants over 30, men spent 4.3% and women 3.9%. The absolute differences are not enormous, yet they are clinically meaningful because they point to demographic gradients in sugar exposure that may accumulate across years. Even modest shifts in beverage purchasing can matter when they are sustained over time and layered onto other dietary behaviors, especially in populations already carrying cardiovascular or metabolic disease.
The age contrast is just as relevant. Younger participants spent a larger share on sugary drinks than older adults, which fits with broader behavioral patterns seen in diet research but is often hard to measure accurately. In a cohort linked to registry data, those differences can be tested against hard outcomes rather than inferred from survey responses. A future analysis could examine whether high-spending categories cluster with incident diabetes, weight-related diagnoses, or downstream cardiovascular events, although this paper itself stops at cohort description and baseline purchasing patterns.
Because the study is open and continuously updated, the exposure history is not frozen at enrollment. That structure creates a living database that can absorb new participants over time while continuing to extend follow-up for earlier entrants. For dietary epidemiology, this is especially valuable. Food habits shift with season, income, age, household change, and health status, and a static cohort often misses those transitions. SMIL was built to keep them in view.
Registry linkage gives the cohort its long-term value
The strongest scientific promise of the cohort lies in what happens after the purchase data are captured. Denmark’s nationwide registries allow linkage across health and social outcomes, giving investigators the chance to examine associations between food purchasing and disease without relying on self-reported diet histories alone. That is important because recall bias, social desirability, and selective underreporting routinely weaken dietary studies. Purchase data do not eliminate confounding, but they offer a cleaner exposure source and a repeatable method for measuring change over time.
This matters most for chronic disease research. In participants who already had cardiovascular disease or diabetes at baseline, purchase patterns may reflect treatment adherence, health consciousness, household support, or attempts at dietary change after diagnosis. In those without prior disease, the same patterns may identify upstream risk behavior long before clinical events appear in registries. The same dataset can therefore support both prevention research and prognosis research, depending on the analytic question.
The cohort began with data spanning 2018 to 2022, and collection is set to continue for many years. Investigators also plan initiatives to expand the cohort. Growth will matter because larger samples will improve power for subgroup analyses and allow better separation of stable dietary behavior from short-term fluctuation. With repeated shopping data and registry linkage, the SMIL cohort is positioned to examine whether purchase behavior tracks with later disease, healthcare use, or social outcomes in ways that standard diet studies simply cannot resolve.
Questions that now depend on follow-up, not theory
The baseline paper establishes feasibility, not causation. It shows that receipt data can be collected, coded, and linked, and it documents real differences in spending by sex, age, and time of day. It does not yet answer whether those differences translate into measurable disease risk, whether they persist after adjustment for social and health factors, or which food categories best predict later outcomes. Those are the questions that will determine whether SMIL becomes a high-value platform for preventive medicine or remains primarily a descriptive cohort.
For now, the value is concrete: a large Danish population with repeated purchase records, registry linkage, and enough baseline clinical diversity to support future analyses in cardiometabolic disease. The remaining test is whether the cohort’s purchase signals can predict outcomes strongly enough to justify using consumer data as a serious epidemiologic exposure measure rather than a convenient proxy for self-reported diet.