Estimating spatial patterning of dietary behaviors using grocery transaction data
Bibliographic record
Abstract
ObjectiveTo demonstrate a method for estimating neighborhood foodselection with secondary use of digital marketing data; grocerytransaction records and retail business registry.IntroductionUnhealthy diet is becoming the most important preventablecause of chronic disease burden (1). Dietary patterns vary acrossneighborhoods as a function of policy, marketing, social support,economy, and the commercial food environment (2). Assessmentof community-specific response to these socio-ecological factorsis critical for the development and evaluation policy interventionsand identification of nutrition inequality. Mass administration ofdietary surveys is impractical and prohibitory expensive, and surveystypically fail to address variation of food selection at high geographicresolution. Marketing companies such as the Nielsen cooperationcontinuously collect and centralize scanned grocery transactionrecords from a geographically representative sample of retail foodoutlets to guide product promotions. These data can be harnessed todevelop a model for the demand of specific foods using store andneighborhood attributes, providing a rich and detailed picture of the“foodscape” in an urban environment. In this study, we generated aspatial profile of food selection from estimated sales in food outletsin the Census Metropolitan Area (CMA) of Montreal, Canada,using regular carbonated soft drinks (i.e. non-diet soda) as an initialexample.MethodsFrom the Nielsen cooperation, we obtained weekly grocerytransaction data generated by a sample of 86 grocery stores and 42pharmacies in the Montreal CMA in 2012. Extracted store-specificsoda sales were standardized to a single serving size (240ml) andaveraged across 52 weeks, resulting in 128 data points. Using linearregression, natural log-transformed soda sales were modelled as afunction of store type (grocery vs. pharmacies), chain identificationcode and socio-demographic attributes of store neighborhood, whichare median family income, proportion of individuals who receivedpost-secondary diplomas, and population density as measured by the2011 Canadian Household Survey. Selection of the predictors andfirst-order interaction terms was guided by the minimization of themean squared error using 10-fold cross-validation. The final modelwas applied to all operating chain grocery stores and pharmacies in2012 (n=980) recorded in a comprehensive and commonly availablebusiness establishment database. The resulting predicted store-specific weekly average soda sales was spatially interpolated toprovide a graphical representation of the soda sales (representing anunhealthy foodscape) across the Montreal CMA.ResultsFigure 2 demonstrates the spatial distribution of the predicted sodasales in the Montreal CMA.ConclusionsThe current lack of neighborhood-level dietary surveillanceimpedes effective public health actions aimed at encouraging healthyfood selection and subsequent reduction of chronic illness. Ourmethod leverages existing grocery transaction data and store locationinformation to address the gap in population monitoring of nutritionstatus and urban foodscapes. Future applications of our methodologyto other store types (e.g. convenience stores) and food productsacross multiple time points (e.g. mouths and years) will permit acomprehensive, timely and automated assessment of dietary trends,identification of neighborhoods in special dietary needs, developmentof tailored community health promotions, and the measurement ofneighbourhood-specific response to nutrition policies and unhealthyfood advertising.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.003 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".