Online Grocery Swap Prompts Improve Diet Quality, Cut Carbon
Key Takeaways
- A randomized online grocery trial in 1,201 US adults found that climate, health, and combined swap prompts were each associated with more healthful selections than control.
- Only climate and combined prompts lowered the carbon footprint of selected foods, by about 39% and 33%, while health-only prompts did not.
- Participants received an average of 2.3 to 3.6 swap offers per intervention visit and accepted about one-quarter of them.
- Most participants approved of the swaps and labels, with approval ranging from 77% to 87%.
In the PLOS Medicine randomized online grocery swap trial, 1,201 US adults completed three weekly visits in a naturalistic online grocery store, with visit 1 as a no-intervention baseline and visits 2 and 3 delivering assigned interventions after 1:1:1:1 randomization to climate swaps, health swaps, climate plus health swaps, or control. Participants were not masked to assignment. The 2 primary endpoints were average healthfulness, measured with the Office of Communications (Ofcom) Nutrient Profiling Model on a 1 to 100 scale, and average carbon footprint, expressed as kilograms of carbon dioxide equivalents (CO2-eq) per kilogram of selected products. Intent-to-treat analyses pooled visits 2 and 3 in linear mixed models to estimate change from baseline relative to control.
Compared with control, average healthfulness improved with climate swaps at B=2.4 (95% CI 1.1 to 3.7), health swaps at B=4.7 (95% CI 3.3 to 6.0), and combined swaps at B=4.0 (95% CI 2.7 to 5.4), with corrected p<0.001 for all three comparisons. Carbon footprint fell with climate swaps at B=-3.3 kg CO2-eq/kg (95% CI -4.6 to -2.1) and with combined swaps at B=-2.8 (95% CI -4.1 to -1.6), whereas health swaps were not significant at B=-0.4 (95% CI -1.7 to 0.8; corrected p=0.51). Health-only and combined prompts outperformed climate-only prompts on healthfulness, while climate-only and combined prompts outperformed health-only prompts on carbon footprint.
Behavioral uptake was measurable but modest. Participants received mean offers of 2.4, 2.3, and 3.6 swaps in the climate, health, and combined arms and accepted 25%, 27%, and 27%, respectively. Across secondary outcomes, all three interventions lowered energy density, saturated fat density, and total spending; health and combined swaps also lowered sugar density; only health swaps increased fiber density; and sodium and protein density did not differ. Domain-specific thinking shifted in the expected direction, negative emotions increased, and positive emotions and social-norm measures did not change. Participants also viewed the swaps and labels favorably, aligning observed use with broad acceptability.
Interpretation remains bounded by the authors' stated limits. The sample differed somewhat from the broader US population, the store was simulated, and participants did not ultimately pay for or receive the groceries they selected. The findings also may not extend to brick-and-mortar retail settings, and labels and swap recommendations were delivered together so their independent effects could not be separated. Some food groups were not represented, and recommended alternatives were not designed to match original products on price. These results therefore describe shopping selections within an online retail environment rather than long-term dietary intake, clinical outcomes, or in-store behavior.
The authors interpreted the combined climate-plus-health strategy as improving both primary outcomes without an apparent loss of effect relative to the stronger single-focus arm in each domain. The trial also supports the feasibility and acceptability of point-of-selection swap recommendations in online grocery settings.
Clinician Questions
How were climate and health food swaps triggered in the online grocery trial?
Swap offers appeared when US adult shoppers in the simulated online grocery trial tried to add products graded C, D, or F. Climate swaps displayed climate grade labels and suggested lower-carbon alternatives, health swaps displayed Ofcom-based health grade labels and suggested more healthful alternatives, and the combined arm displayed both labels while offering products that improved at least one dimension, were at least as good on the other, and were never graded C, D, or F on either dimension.
What do the online grocery swap findings apply to, and what settings remain uncertain?
The findings apply to food selections made by US adults during two intervention shopping visits in a simulated naturalistic online grocery store. Participants did not ultimately pay for or receive the groceries they selected, the sample differed somewhat from the broader US population, and generalizability to brick-and-mortar grocery settings remains uncertain.
Did age, health consciousness, or environmental consciousness change response to climate or health swap prompts?
The investigators did not observe meaningful differences in the two primary outcomes by age group, health consciousness, or environmental consciousness. That pattern suggests the online grocery swap effects were not clearly concentrated within only one age- or attitude-defined subgroup in this trial.