The potential influence of the digital food retail environment on health: A systematic scoping review of the literature

may have positive population dietary impacts, whereas online food delivery platforms appear likely to promote unhealthy food purchasing. However, the current evidence base is fragmented, with many knowledge gaps. Further research is required to understand the influence of the digital food retail environment on population diets and how these environments can be designed to support healthy food choices.

Some evidence suggests that online grocery retail and meal kit subscription services may have positive population dietary impacts, whereas online food delivery platforms appear likely to promote unhealthy food purchasing.However, the current evidence base is fragmented, with many knowledge gaps.Further research is required to understand the influence of the digital food retail environment on population diets and how these environments can be designed to support healthy food choices.
food environment, nutrition policy, online food delivery, public health F I G U R E 1 Flowchart to demonstrate relationships between the different strata of the food environment.

| BACKGROUND
"Food environments" are defined as "the collective physical, economic, policy and sociocultural surroundings, opportunities and conditions that influence people's food and beverage choices and nutritional status." 1 There is strong evidence that food environments have a substantial influence on population diets. 2 Food environments that are "obesogenic" are those that promote consumption of energy dense, highly processed foods. 3Unhealthy diets are a leading risk factor in developing non-communicable disease, including type two diabetes and metabolic syndromes such as obesity and high blood pressure. 4,5The global prevalence of obesity has tripled since 1975, with 39% of the world's adult population currently experiencing overweight or obesity. 6igital food environments," including digital food marketing, online food ordering platforms, and food and recipe blogging, are an increasingly prominent component of wider food systems (Figure 1).Digital food retail environments include digital platforms, such as mobile phone apps or websites, which can be used to purchase food and beverages.These digital platforms include, among others, online food delivery platforms (such as Uber Eats and DoorDash), online grocery stores, and "meal kit" subscription services (such as Hello Fresh and Marley Spoon).The use of these platforms is substantial and is increasing over time.For example, globally, DoorDash (an online food delivery platform) has more than 20 million monthly users and generated US$9.9 billion gross for the first quarter of 2021.7 A recent market research report of the Australian Uber Eats online platform found that fast food delivery services doubled their customer base in the 18 months from mid-2018, with 19% of Australians now reporting that they have previously used the service.8 Internationally, the online grocery sector has been experiencing an increase in sales.In the United States, between 2017 and 2022, online grocery sales increased by an average of 17% per year, with the 2022 revenue predicted to be in excess of US$32 billion.9 The Australian online grocery sector has also experienced market growth, with a reported 39% year-on-year growth compared with 2% yearly growth in traditional grocery sales.10 Globally, meal kit subscription service Hello Fresh saw a 42% year-on-year growth in the number of orders between 2018 and 2019, with 10.5 million meals delivered.11 The growth of meal kit subscription services highlights their position as a small, but expanding, component of the digital food retail environment.Although their global revenue (US$17.8 in 2022) 12 is much smaller than that of online grocery retail (US$285 billion in 2021) 13 and online food delivery platforms (US$323 billion in 2022), 14 they have a unique position in the digital food retail environment as an important place of food procurement, though their impact on diet is unclear.Meal kits are mostly purchased entirely through digital means, with limited physical shopfronts or locations for in-store purchasing.Meal kits subscriptions are also designed to fill a niche: to make preparing a home cooked meal as easy and convenient as possible so users do not need to resort to unhealthy takeaways.15 The use of these digital food retail platforms accelerated further during the COVID-19 pandemic, with Australian online grocery sales experiencing a further 5% increase in sales in 2021.16 Australian online food delivery platform Menu Log also reported a 54% increase in orders through their Melbourne (Australia) platform since March 2020, when the first COVID-19 restrictions were introduced.17 Similarly, European online food delivery platform Just Eat generated €3.4 billion in revenue in 2021, a 45% increase in revenue from 2020.18 The penetration rate of online food delivery platforms in China increased from 16.5% in 2015 to 48.8% in 2022, with a user base of approximately 521 million people.19 Additionally, revenue for online sales of food in Mexico rose from $US1,143 million in 2019 to US$4237 million in 2022.20 The digital food retail environment has the potential to influence diet and health in a number of ways.Previous literature has shown that convenience and ease are key determinants of what people choose to eat 21 and that perceived difficulty in preparation is a barrier to healthy eating.22 A recent scoping review by Granheim et al.   broadly examined all aspects of the digital food environment generally, including marketing, regulation, pricing, convenience, desirability, and digital food culture including food blogging and social media communities relating to specific types of food or diets.23 Their review found that there is a relationship between digital and physical food environments and that one likely influences the other, though this influence on health is unclear.A previous narrative review of the Australian online food delivery sector by Bates et al. found that the innovative nature of online food delivery platforms "disrupts" usual food retail methods, taking advantage of the current lack of regulations applying to food retailers in Australia.15 Digital food retail environments can increase convenience by removing barriers to accessing food and drinks, through facilitating ordering using userfriendly apps or websites and by removing the need to travel to purchase.Indeed, the magnitude of convenience will depend on what aspect of the digital food retail environment is being utilized; online grocery retail removes the need to travel to grocery stores and physically purchase items, meal kit subscription services reduce time spent planning meals, and using online food delivery platforms replaces almost all steps involved in food preparation.24 Online food delivery platforms are also theorized to increase sedentary behaviors as consumers no longer need to leave their homes to receive meals, 25 as well as increasing "emotional overeating" behaviors among university students.26 Despite these potentially negative implications for population health, digital food retail environments may also promote healthier diets.For example, a review by Fernadez and Raine suggested the potential for online grocery retail to be used as a way to increase fruit and vegetable intake among those who find physical grocery shopping difficult, such as older adults, people with mobility or health conditions, or those without transport.24 However, there is the potential that these health benefits may not apply equally across population groups because of inequalities in digital access.27 Despite the widespread and increasing use of digital food retail platforms, to date, there has been no synthesis of the literature to understand the potential influence on population diets and health of the digital food retail environment.This scoping review aimed to systematically search the literature in order to understand the potential influence of the digital food retail environment on population diets and health.Our intention was to summarize the evidence in a way that could guide public health policy in this area.

| METHODS
This systematic scoping review followed protocols outlined by Arksey and O'Malley, and the Joanna Brigg's Institute (JBI) Manual for Evidence Synthesis, [28][29][30] and reported according to the PRISMA guidelines for scoping reviews. 31A systematic scoping review methodology was selected to allow for a broad research question, while maintaining the rigor of a systematic review.

| Inclusion criteria and exclusion criteria
Included studies were those that examined any aspect of the digital food retail environment (including, but not limited to, online food delivery platforms, online grocery stores, and meal kit subscription services) where an aim or rationale of the study was related to health.
Because this review was intended to inform public health policy recommendations, all included studies were required to have a public health focus.This focus on health could be included as part of the study design (e.g., classifying foods or purchases by an indicator of healthiness and measuring participant health indicators such as body mass index [BMI]) or referred to by the authors in the rationale for the study or aims.
We considered all aspects and types of digital food retail environments.Populations for included studies could include human (adults, children) users/customers or digital food retailers, companies, or products.No restrictions were placed on country studied or year of publication.All study designs were considered, including quantitative, qualitative, observational, and experimental.Editorials were screened against the inclusion criteria.
We excluded intervention studies that targeted a specified population with pre-existing health conditions that require therapeutic diet, for example, low carbohydrate diets for those with diabetes.Studies that primarily examined alcohol delivery services were excluded, as this was beyond the scope of this review.Studies examining online marketing external to a digital food retail environment were excluded.Systematic reviews and conference abstracts were excluded.

| Search strategy
The search strategy aimed to locate both published studies and grey literature to identify non-empirical research.An initial limited search of Medline was undertaken to identify key articles based on the research aim, which was used to guide the search strategy.
The search was conducted in September 2023.Three overarching search terms were used ("food and beverages," "digital," and "purchasing"), combined with the operator "AND."Within these broader search terms, more specific terms were combined with the operator "OR."Databases that were searched included MEDLINE Complete, EBSCO Business Source Complete, EBSCO Communication and Mass Media Complete, and Medline ProQuest.A full list of search strings is available in Table S1.
The Medline specific search terms were food and drink (food* OR drink* OR beverage* OR grocer* OR diet*) and digital (online OR digital OR app OR apps) and purchasing (prevalence OR purchas* OR use*).Boolean search operators were modified for each database searched, and search results were limited to English.Reference lists of all included articles were screened for additional literature.Forward and backward citation screening of included studies was also undertaken to identify further articles.
Grey literature was sourced using systematic keyword searches of Google, Google Scholar, Proquest Dissertations & Theses Global, and the Preprints database.The first 100 search results from each of these databases were screened against the inclusion criteria.
Following the search, all identified citations were collated and uploaded into Covidence, 32 which is a web-based software platform designed to streamline the research review process.Duplicate search results were removed.Titles and abstracts were screened by the first author (RB), with a sample of articles (n = 1583, approximately 7% of total papers) screened independently by two authors (RB and CD) to determine the degree of consistency in exclusion of irrelevant articles.
This produced an agreement rate of 90%, similar to other scoping reviews. 23Subsequent full text screening against inclusion and exclusion criteria was carried out by two authors independently (RB, MK and CD).Reasons for exclusion at full text were recorded using Covidence.Disagreements between these two reviewers regarding inclusion were resolved through discussion and, if needed, inclusion of a third reviewer to reach consensus.

| Reporting of findings
Data were extracted from included studies.Using a standardized data extraction template, the following data were included: country, population, study aims, methods, study type, digital food retail environment category, and key findings.
Pilot searching of the literature identified three key digital food retail environments: (i) online food delivery platforms, (ii) online grocery retail, and (iii) meal kit subscription services.Findings were therefore narratively synthesized, according to these three identified domains of the digital food retail environment (we were also open to other domains emerging as we synthesized the data, but no further domains were identified).Reporting was further divided into studies that reported on (i) characteristics of the digital food retail environment, (ii) potential influence on healthiness of purchasing behavior, or (iii) other contextual factors relevant for health.

| Quality of included studies
The quality of included studies was assessed using the Joanna Brigg's Institute critical appraisal tools. 33The appropriate tool was selected based on study design.These appraisal checklists evaluate methodology by assessing the inclusion criteria of the sample, the descriptions given of study population, the appropriateness of the analysis methods, and other aspects of the study design as a binary "yes/no" question. 33llowing methods described by Shi et al., studies were reported as "good" quality when they had only "yes" or "not applicable" ratings, "fair" if they had one or two "no" or "unclear" ratings, and "poor" quality if they had three or more "no" or "unclear" ratings. 34| RESULTS

| Search results
The results of the search are reported in a PRISMA flow diagram 35 (Figure 2).
The literature search identified 21,382 studies, of which 762 were duplicates.All studies were identified through database and grey literature searching.We did not identify any further relevant articles through forward or backwards citation screening; 19,812 articles were excluded due to irrelevant titles, and a further 681 studies were excluded for irrelevant abstracts.
A total of 126 articles were reviewed in full, with 57 articles meeting the inclusion criteria.

| Overview of included studies
A summary of all included studies is shown in Table S2 Table 1: Of the 57 included studies, 11 were classified as "good quality."Thirty-two were classified as "fair" quality, and 14 classified as "poor." Common methodological weaknesses included participant inclusion criteria not being adequately described, study subjects and settings not being described in detail, confounding variables not being clearly identified, and not addressing the influence of the researcher on the research for qualitative studies.

| Online grocery retail
Thirty of the included studies examined online grocery retail, from the United States (n = 18), the United Kingdom (n = 7), China (n = 1), Canada (n = 1), Australia (n = 1), and the Netherlands (n = 1).One study included multiple high-income European countries.Twenty-two studies examined the items purchased by online grocery retail users, and the remaining eight studies examined the websites or some aspect of the online grocery retail platform itself, including the availability of nutritional information.Ten studies examined users of online grocery by an indicator of socioeconomic position.Generally, it was reported that shopping online for groceries enabled healthier purchasing compared with in-store shopping, despite studies also reporting that nutritional information for products was inconsistently available.

| Characteristics of the online grocery retail environment
The online grocery retail environment was generally characterized as having variably available nutritional information.A study by Lee et al.   analyzed the availability of nutritional information on eight Canadian supermarket websites.The researchers found that although all of the 555 items examined had some product images available, most (81%) of these images were illegible even after zooming in to increase image size. 53Of the products that were required to provide nutritional information per regulations (i.e., all pre-packaged foods), this information was only available for 61% of those products. 53A similar study by Stone et al. examining nutritional labeling across five UK grocery retailers' websites found that nutrition information was available for every product analyzed in the product's "description page," but this showing the search strategy.
T A B L E 1 Summary of the examined health related outcomes or measures for included studies.information was not available on the "results" page after searching for a product. 58Similarly, Wallis and Moore reported that of price promoted products available on three major UK supermarket websites, front of pack nutritional information was often unavailable on webpages (70% available) and images (52% available). 64A 2021 cross sectional study by Bhatnagar et al. compared the price, promotions, and nutrition information availability of products from supermarkets in store and online. 39The study found that the median price for the same products in store and online was similar (£1.87 and £1.85, respectively) and that price promotions were more common in store (32%) compared with online (24%). 39No difference was found in the nutrient composition information displayed across the two different shopping mediums. 39

| Potential influence of the online grocery retail environment on the healthiness of purchasing behavior
It was generally reported that shopping for groceries online led to healthier food and drink purchases compared with shopping in store.

Dillahunt et al. studied the effect of an intervention providing grocery
delivery to 20 adults with limited transport access.The researchers found that participants who were in the intervention group purchased a higher percentage of "green" (healthier; 55%) than "red" (less healthy; 22%) classified foods, when compared with the non-intervention group (46% "green" and 25% "red"). 60A 2017 study by Huyghe et al. also compared the purchasing of "vice" food products (including salty snacks, chocolate, chips, etc.) across online and instore shopping events. 49This study was conducted using purchasing data from 4313 shoppers at large European grocery retailer. 49The mean relative expenditure per shopping trip on "vice" products was significantly lower when shopping online compared with in-store (0.056 vs. 0.111, respectively), and fewer "vice" items were purchased online compared to in-store. 49This was despite customers spending more overall when shopping online compared to in-store (€167 and €120, respectively). 49On the other hand, Lacko et al. analyzed Neilson Homescan data and found that those who shopped for groceries online purchased 73 calories per person per day, whereas in-store shoppers purchased 686 calories per person per day. 52However, the authors highlighted that the number of online grocery shoppers in the study sample was low (i.e., only 0.8% of total volume sales reported were from online grocery retail stores). 52 8-week randomized control trial by Gorin et al. found that participants who were overweight and were assigned to the intervention group, which instructed participants to order their groceries online instead of shopping in store, perceived that the grocery delivery helped them to "decrease impulse purchases" and "make healthier choices" compared with shopping in-store. 45Although, there was no significant difference in weight loss between the intervention and control groups (7.5 ± 8.5 vs. 9.0 ± 6.7 lbs, respectively), a nonsignificant trend was reported for the number of grocery deliveries participants completed and weight loss after 8 weeks (r = 0.71, p = 0.8). 45In contrast, a thesis by Thomas found that when using a mobile phone app for grocery purchases, 68 university student reported lower "health fluency" scores due to reduced access to external reference points such as nutrition labeling, and other products at hand for easier comparisons, compared with if they shopped in store (mean health fluency score = 5.97 vs. 5.28, respectively). 61wever, the study did not find that this reduction in health fluency was associated with the healthiness of products purchased. 612011 Australian study by Sacks et al. found that introducing greater detail of nutrition information on a supermarket website in the form of a "traffic light" warning system for amounts of fat, saturated fat, sugar, and sodium produced no significant changes in the real-world online sales of products that were labeled "red" when compared with sales from a website that did not provide traffic light labeling (control).57 A 2022 study by Stuber et al., analyzing sales data from 11,775 shoppers in an online Dutch supermarket, found that there was no overall significant effects from four different conditions (information nudge, positing nudge, both nudges, or control).59 The information nudge included labeling being applied to designated healthier options, including "convenient," "tasty," or "popular."Shoppers from deprived areas that were allocated to the information nudge intervention purchased 2.4% (95% CI: 0.8, 4.0) more healthy products compared with the control group.59 A randomized controlled trial conducted in conjunction with a major UK supermarket found that shoppers randomly assigned to the intervention arm (where they saw healthier top of page banners, healthier recipe ingredient lists, and reordered search results for yoghurts and ice creams by healthiness) found no significant differences in overall energy purchased between trial arms, despite intervention participants purchasing greater quantities of selected healthy recipes ingredients (spaghetti bolognese recipe [B = 0.89, SE = 0.22, 95% CI (0.45, 1.33), p < 0.001] and fajita recipe [B = 1.72,SE = 0.17, 95% CI (1.28, 2.05) p < 0.001]). 41

| Other contextual factors relevant for health
Eleven studies reported on use of online grocery retail across socioeconomic strata.Of these studies, six examined recipients involved in SNAP, 36,54,56,60,62,63 two studies examined low income populations, 37,65 one study examined seniors with low incomes, 43 one study examined a food insecure population, 42 and one study examined SNAP WIC participants. 50nerally, participants in the SNAP or WIC programs were reluctant to shop online due to their concerns over the quality and freshness of groceries, particularly fruit, vegetables, and meat products.
The mixed methods study by Martinez highlighted perceived barriers to using online grocery services, 54 including that participants in the United States felt that online grocery services were more expensive, only for affluent people or for those who had limited mobility and could not shop independently. 54They were also concerned that by using online grocery retail stores, they were surrendering some "control" over the freshness and quality of what they were purchasing as they were not there in person to select items using their own sense of smell and sight. 54Similar concerns were raised by SNAP participants in a focus group study by Rogus et al., where participants were hesitant to shop for perishable foods online because of a perceived risk of receiving spoiled items as part of their grocery order, particularly for meat, dairy products, and fruits and vegetables. 56Trude et al. used mixed methods to examine online grocery purchasing behaviors among SNAP recipients in Maryland. 62Participants self-reported that they less frequently bought fresh produce (OR = 0.34, p < 0.001), meat and seafood (OR = 0.29, p < 0.001), and sweets (OR = 0.54, p = 0.005) when shopping online compared with in-store. 62Jilcott Pitts et al. examined perceived advantages and disadvantages of online grocery shopping among WIC participants from North Carolina, USA, and found that those interviewed perceived prices would be higher than in store, and as with the other studies, there were concerns about a lack of control surrounding the quality and freshness of products. 50We did not identify any studies examining if high socioeconomic position groups share similar sentiments regarding online grocery shopping.
In contrast to findings from studies using general population groups, Jilcott Pitts et al. found that WIC participants generally perceived that they made more impulse food and beverage purchases when shopping online compared with in store. 50Despite concerns about freshness, being able to use their SNAP or WIC allowances for payment online was a motivator for these groups to use online grocery retail stores. 54Crowell et al. examined online grocery orders placed by seniors with low incomes in New Jersey, USA, as part of a program aiming to improve their nutrition while living in a "food desert." 43Analyzing receipts from 151 orders placed over 4 months by 28 participants, they found that 45% of items ordered by the seniors had "no nutritional value" according to the Food Finder classification, but 36% of items ordered had the highest rated nutritional value. 43No comparison was made with in-store grocery purchases. 43

| Online food delivery platforms
Twenty two studies examined online food delivery platforms, from Australia (n = 4), the United Kingdom (n = 4), Brazil (n = 3), Canada (n = 2), the United States (n = 2), Malaysia (n = 1), Saudi Arabia (n = 1), China (n = 1), New Zealand (n = 1), Singapore (n = 1), one from multiple high income countries, and one from multiple Latin American countries.Seventeen studies used a cross sectional study design, two used longitudinal methods, and two used mixed methods, and one was a qualitative study.
82,83,86,87 The studies that examined the users of online food delivery platforms were based in the United Kingdom, 79,81 Australia, 70,84 Malaysia, 72 Saudi Arabia, 67 Singapore, 85 the United States, 40 and multiple high income countries. 78Two studies examined the influence of online food delivery platforms on dietary habits during COVID-19 lockdown restrictions, 67,85 one study examined the in-app advertising of online food delivery platforms during the COVID-19 pandemic, 75 and one study examined what food items were promoted on online food delivery platforms during the COVID-19 pandemic. 77Studies generally reported that the offerings available through online delivery platforms were mostly unhealthy and that frequent users of these platforms were highly educated or had high BMI.

| Characteristics of the online food delivery platform retail environment
Studies that examined online food delivery platform characteristics examined the access of food outlets to a specified delivery location and also the healthiness of menu items, 68,82 the healthiness of menu items only, 69,71,73,76,83,87 or access to food outlets from different locations, 80 and the availability of calorie labeling or nutritional information, 74,86 and others examined the health aspects of in-app promotional marketing strategies. 75,77udies that examined the healthiness of menu items found that they were mostly unhealthy.A 2021 study by Brar and Minaker examined the distance from food outlets to delivery location, as well as the healthiness of foods available through Canadian online food delivery platforms. 68They found a high correlation between city population size and the total number of food retailers available through the online food delivery platform (r = 0.71) and a positive correlation between population density and the total number of retailers (r = 0.51). 68This study also examined the menus of the most popular three food outlets across four online food delivery platforms.Most (n = 9 of 12) of these menus scored fewer than 36 points out of 100 when scored against the Healthy Eating Index-2015. 68A score of 36 corresponds to an "F" score, the lowest possible. 93ng et al. and Partridge et al. examined the menus of outlets available through the Uber Eats online delivery platform in Australia 87 and Australia and New Zealand. 83 Uber Eats were "discretionary," with the "cereal based mixed meal" category (pizzas, pastas, burgers, etc.) comprising 42.3% of total discretionary menu items. 87Wang et al. also found that discretionary menu items available from online delivery platforms were more likely than healthy menu items to be offered as part of an value bundle (9.6% and 1.6%) and include a product photo (30.7% and 25.1%). 87milar findings were found in Brazil where Horta et al. also found that the foods most promoted and available through Brazilian online food delivery platforms were "ultra-processed" foods such as sugar sweetened beverages, ice creams, and salty packaged snack foods. 76Additionally, these ultra-processed foods compromised 70% of menu offerings, with the remaining 30% made up of healthier traditional, vegetable-based meals. 76However, a study by Cong et al. examining   the healthiness of meals available through online food delivery platforms in China found that more than 60% of menu items analyzed were classified as "healthy," with these most commonly being classified as "noodle and dumpling meals." 69eas with greatest deprivation were found to have the greatest access to unhealthy food outlets through online delivery platforms.A 2021 study by Keeble et al. aimed to examine associations between neighborhood postcode district deprivation level and access to takeaway food outlets through online food delivery platforms. 80Across 2118 postcodes, they found that the most deprived areas had significantly greater access to takeaway food outlets through online food delivery platforms, compared with the least deprived areas (106.1 outlets vs. 70.4outlets, respectively). 80study by Goffe et al. engaged with 20 expert stakeholders to create a scoring system ("Health Rating score") to objectively assess the healthiness of outlets offered through the UK online food delivery platform "Just Eat." 73 Following creation of this scoring framework, 149 menus from different outlets were assessed for healthiness by the expert stakeholders.Metrics including the availability of water or sugar sweetened beverages as a drink choice and the presence of deep fried potato products like chips were also used to assist with rating the healthiness of a food outlet. 73The rating score ranged from 0 (least healthy) to 5 (most healthy).The most frequently awarded Health Rating score was 1 with no outlets achieving a score of 5. 73 Studies that examined the prevalence of nutritional information available on online food delivery platforms found that it varied across platforms and across locations.A study by Vanderlee et   86 Similarly, an American study found calorie labeling was available for 27% of menus on DoorDash, 19% of menus on Uber Eats, and 6% of menus on Grubhub. 74

| Other contextual factors relevant to health
The frequency of online food delivery platform use ranged from 28% in the prior month among Australian adults (ordering at least once in the prior month) 70 to 40.3% among Malaysian university students (who reported using the platforms 1-3 times per month). 72Moreover, 15% of 19,378 respondents from Australia, Canada, Mexico, the United Kingdom, and the United States reported using an online food delivery platform at least once in the previous week. 79aracteristics of online food delivery platform users were mostly reported as being highly educated or having higher BMI than nonusers.The multi-country analysis (including Australia, Canada, Mexico, the United Kingdom, and the United States) found that respondents who used online food delivery platforms were most likely to be male, White, highly educated, or living with children aged under 18 years. 78proximately 40% were living with overweight or obesity, and the median age was 33 years. 78In another study of 2010 Australian adults (aged 18+ years), multivariate logistic regression modeling found that a younger age, higher BMI, higher education, and higher income were positively associated with likelihood of using online delivery platforms at least once per month. 70The likelihood of using these platforms was also higher for groups that also reported consuming other fast food and sugar sweetened beverages regularly. 70Eu and Sameeha reported that of 290 Malaysian university students surveyed, 85.9% intended to purchase healthy food through online delivery platforms. 72Despite this, most had negative perceptions of the healthy food available for purchase (76.9%), so ordered unhealthy foods most often (77.6%). 72In interviews with 22 UK online food delivery platforms users, Keeble et al. found that participants reported that they did not use online food delivery platforms for the purpose of ordering healthy food. 81A longitudinal Australian study found that online food delivery platform use at baseline, relative to non-use, was associated with greater assessed urges to overeat, though these were small in magnitude. 84A study of young adults aged 18-25 by Buettner et al. found that experiencing food insecurity was significantly associated with greater use of online food delivery platforms (β = 0.32, p < 0.001). 40ere was some suggestion that the COVID-19 pandemic influenced the type and amount of advertising by food delivery platforms.
Horta et al. examined the Brazilian digital food retail environment in the context of the pandemic, but their cross sectional methodology did not allow for comparisons with before the pandemic. 75They described the digital food retail environment in Brazil during the pandemic to be "obesogenic" due to the high prevalence of unhealthy foods and beverages being promoted by in-app promotional marketing in online food delivery platforms. 75Promotions for free delivery were most often used for ice creams, candies, salted snacks, and pizza rather than traditional meals or vegetables ( p < 0.01), and photos were most often used for unhealthy menu items such as ice creams, candies, or salted snacks rather than traditional meals or water ( p = 0.03). 75In another study by Horta et al. conducted during the COVID-19 pandemic, it was found that a Brazilian online food delivery platform most promoted messaging relating to the tastiness and "value for money" of food offerings, rather than their healthiness. 77ditionally, "traditional meals" (unprocessed or minimally processed Brazilian foods), pasta, ultra-processed beverages, and sandwiches were the most commonly promoted food items, with promotions for water, natural juices, fruits, and vegetables being least common (20-25% of promotions vs. <5%, respectively). 77

| Meal kit subscription services
Five studies reported on meal kit subscription services, with three conducted in Australia 88,89,92 and two in Denmark. 90,91The years of data collection ranged from 2017 to 2020.Two studies 89,92 analyzed the nutrition quality of the meals provided through subscription meal kits, and three 88 aimed to understand the motivations of shoppers for using these subscriptions.The nutritional composition of meal kits was generally reported to be healthy, with users reporting that they helped to alleviate stress surrounding family meal planning.

| Characteristics of the meal kit food retail environment
Overall, it was found that meal kit subscriptions provided mostly FoodWorks software. 89,92Both studies reported that the time of preparation across the different meal kit brands was between 23 and 37 min, most recipes instructed to season with salt (86% and 100%, respectively), and nutrition information was provided to consumers. 89,92Moores et al. also identified that many aspects of using meal kit subscription services align with the Conceptual Model of Healthy Cooking for Chronic Disease Prevention, that is, cooking at home, cooking from basic ingredients, accurately measuring ingredients, adding unprocessed vegetables, using olive oil in cooking, and using herbs, spices, and citrus to add flavor. 92th studies highlighted the high sodium content of the meals, with Gibson et al. reporting that all meals analyzed were above 30% of the Suggested Dietary Target (SDT) for sodium, 89 and Moores et al.
finding that median sodium content of HelloFresh meals was 42% of the SDT and 8% of recipes exceeded the total daily recommended intake of sodium. 92However, both studies also found that the meals provided a sizeable proportion of the recommended five serves of vegetables per day, with 3 92 and 2.3-3 89 vegetables per serve per meal on average.Moores et al. also reported that although meal kits were cheaper than eating at restaurants, the cost of purchasing meal kits was up to three times more per serve than traditional grocery purchasing. 92

| Other contextual factors relevant for health
Three further studies analyzed meal kit subscription services, the motivations for using them, and their role in families with children aged under 18 years. 88,90,91In an Australian study of 16 interviews with primary family meal providers, Fraser et al. found that reasons for using meal kit subscription services varied.Many participants explicitly detailed the ways that the meal kits helped to ease the burden of their "mental load" surrounding shopping, planning, organizing, and preparing family meals. 88Participants discussed how meal kits added variety to their weekly meal plan, increased the range of foods the family was willing to eat, and allowed other family members besides the primary provider to be involved in food preparation. 88e authors reported that almost all participants felt that the meals provided through meal kit subscription services were high quality, fresh, healthy, and aligned with nutritional guidelines.
In the two Danish studies, 90,91 interviews were also conducted with 13 primary family meal providers.Both studies used the same dataset of individual interviews, observations of participant food preparation, and focus group responses.Halkier found that meal kit subscriptions presented a new hybrid model of food preparation, a combination of convenience with the health benefits of cooking homemade meals.Participants also discussed how they believed meal kits to be a "huge help" by outsourcing some of the labor of meal preparation and grocery provisioning to the meal kit suppliers. 90milarly, the study by Hertz and Halkier found that participants reported feeling less stressed about preparing family dinners and felt that meal kits provided relief from the daily planning of meals and grocery shopping. 91Participants also reported that they found the meals from the meal kits added variety to their diet and helped their family to eat more vegetables. 91Participants also discussed using meal boxes as a way to provide care for their children and socializing with them during meal preparation. 91The researchers reported that it was difficult to recruit low income participants due to the relatively high cost of meal kit subscriptions compared with usual grocery prices. 91

| DISCUSSION
This is the first systematic scoping review to synthesize literature on multiple aspects of the digital food retail environment and the potential ways that they may influence population diets and therefore health.Our synthesis of 57 studies published between 2007 and 2023 shows that existing literature on this topic is fragmented with diverse potential impacts on health and many knowledge gaps.This is likely due in part to these digital food retail environments, and therefore research in this area, being relatively novel.Evidence suggests that online grocery retail may reduce the purchase of unhealthy food and beverages compared with shopping in-store, despite nutritional information commonly being incomplete for online products.Similarly, meal kit subscription services may have positive population dietary impacts, whereas online food delivery platforms appear to sell and promote predominantly unhealthy foods and beverages, using promotional strategies such as price discounts and marketing images.
3][14] Nevertheless, while we only found two studies describing the nutritional quality of online meal kit subscription services, these were consistent in the finding that online meal kit subscriptions provide mostly healthy meals containing a substantial contribution towards recommended intakes of vegetables, though sodium content was found to be higher than recommended limits. 89,92,94ese findings align with evidence that cooking food at home leads to higher diet quality compared with regularly eating foods that are prepared away from home. 95Another potential benefit of meal kit subscriptions is the opportunity they may provide adults to socialize with their children. 91There have been concerns that children are not learning cooking skills as they grow up, due partly to the increased time pressure faced by working parents. 96Cooking in childhood can lead to greater confidence with food preparation as an adult and associated increases in diet quality. 97Meal kit subscription services may provide a convenient option that increases food literacy, confidence with food preparation, and self-determination skills by making cooking feel more achievable by using pre-measured ingredients and photos for each step of the recipe.
For online food delivery platforms, our review highlights the increased accessibility to unhealthy food that these platforms provide.
The "neighbourhood food environment," previously been defined as 1-km radius of an individual's home or workplace, is much greater when considering the digital food retail environment. 98,99For example, Partridge et al. found that 90% of food deliveries available from online food delivery platforms originated outside of a 1-km radius. 83milarly, Keeble et al. noted that the "neighbourhood" food environment must be thought of as broader than what is in a consumer's immediate vicinity due to the increasing ubiquity of online food delivery platforms. 79The term "food swamp" has been used to describe a food environment where unhealthy food and drink retailers overwhelm healthy offerings and make it difficult for consumers to choose healthy foods. 100The unbalanced healthiness of foods promoted and available through online food delivery platforms is creating a "digital" food swamp.When combined with heavy promotion of unhealthy options on online food delivery platforms, it is not surprising that users most frequently purchase unhealthy offerings. 72r review found that the main users of digital food retail environments have a higher socioeconomic position.Low-income shoppers were found to be wary of using online food retail to purchase perishable groceries such as meats or vegetables.These results are supported by studies conducted in-store, where low-income shoppers have reported to place high importance on being able to carefully examine perishable groceries such as fruit for freshness and value having the ability to personally inspect their food. 101The risk of purchasing spoiled or poor quality produce and needing to repurchase is larger for low compared with high income groups, because food is a large proportion of their household budget. 102While convenience is a motivator to use online grocery retail for some shoppers, 103 the perceived risk of inferior quality groceries may outweigh a desire for convenience in lower income groups. 104Low income shoppers are also more price sensitive and may be deterred from shopping online for groceries because of delivery fess and the possibility of higher prices than in-store. 54Nevertheless, some studies have found that the price of foods available through online grocery stores and traditional instore supermarkets are similar. 105In contrast, our review also suggested that those living in areas of greatest deprivation also have greater access to food outlets through online food delivery platforms, which we suggest are mostly unhealthy. 80Given those with a low socioeconomic position generally have poorer diets and diet-related health compared with those with a higher socioeconomic position, [106][107][108] it will be important to understand the health equity implications of the digital food retail environment as it evolves in the future.
It is important to note that many studies included in our review were conducted prior to the COVID-19 pandemic, and it is unclear if the digital food retail environment has changed due to the increased consumer demand associated with the pandemic.0][111][112] It is possible that these increases in demand may drive changes in the digital food retail environments; however, it is difficult to ascertain or how has occurred without regular monitoring.Standardized monitoring frameworks for these three different digital food retail domains are required, with monitoring conducted at regular intervals.There are some emerging frameworks for online grocery retail monitoring, 113 which could be extended and adapted to monitor meal kits and online food delivery platforms.
5][116] Given the increasing use of these digital food retail environments, it is important that they promote public health.As a first step, recommendations and actions that are currently applied to the physical food environment could be extended to also include the digital food retail environment.
An example of doing so was highlighted recently in the United Kingdom where the government introduced legislation to restrict high fat, salt, sugar (HFSS) foods and beverages from being price promoted (using volume based price promotions, e.g., "2 for the price of 1") and from being placed in prominent store locations (e.g., end of aisle displays or checkouts). 117Importantly, the policy explicitly includes "online marketplaces" (as well as in-store supermarkets), to prohibit promotion of HFSS foods and beverages with volume based price promotions or by positioning them in high traffic areas, such as "home" or "checkout" webpages when shopping online. 117Other offline actions that could be applied online include nutritional labeling (such as kilojoule labelling) of menus, as is currently mandated for large chain out-of-home food stores in Australia and the United Kingdom.However, it is important to note that the digital food retail environment is rapidly evolving, and their unique features are also likely to require unique policy provision.
This review identified several key gaps in the literature on the potential effect of the digital food retail environment on diets and health.First, as mentioned previously, given the rapidly evolving nature of the digital food retail environment, it will be important that these are monitored over time.Standardized tools for doing so will be important to enable compatibility across jurisdictions and across time.
Second, to ensure our digital food retail environments promote public health, we need a better understanding of the unique aspects of these environments that influence health most and how these can be changed to better support healthy purchases.For example, algorithmic amplification of certain products and prices on landing pages and targeted marketing techniques towards individuals are possible in digital, but not in-store, food retail environments.The digital food retail environment may therefore require unique regulatory conditions.There is also currently little insight into the relationship between ways that marketing and purchasing cues for unhealthy food and beverages in the built food environment are augmented through additional purchasing influences as part of the digital food environment.Additionally, there is limited understanding of the ways in which aspects of the digital food environment (including digital food retail, pricing, and marketing, among others) interact with each other to influence food purchasing behaviors.Although beyond the scope of this review, these interactions are likely important in understanding the effects of the digital food environment on public health and could be explored further in subsequent research.
Our review had several limitations.Overall, the quality of the studies was low, with only 11 of the 57 studies considered "good" quality.Methodological issues affecting study quality decreased their informativeness, with common weaknesses including participant inclusion criteria not being adequately described and study subjects and settings not being described in detail.Further, the literature to date is very heterogeneous in nature, making it difficult to draw conclusions on how healthy the digital food retail environment is and how it could be improved to better promote health.Our decision to limit the inclusion criteria to focus only on health-related studies may also mean that some relevant literature is not captured in this review.For example, while we report on the types of people using online food delivery platforms (i.e., those with a high BMI and those with higher educational attainment), our review lacks data about potential user information for meal kits.These data are available in the broader literature and show that in multi-person households, meal kit users are most likely to be female, between 30 and 39 years old, and highly educated. 118Our study was also limited in that we did not consider all aspects of the digital food environment, instead choosing to contain the scope of the review to the retail aspects, but we acknowledge that there are many interrelated aspects of the digital food environment that influence shoppers' experiences with food and therefore their diets, such as food marketing through social media platforms.Further, the screening of titles and abstracts for this review was primarily conducted by one author, with approximately 7% of titles and abstracts screened by two authors to check consistency.This may lead to misclassification of articles at this early stage of the review process, meaning some potentially relevant articles may have been missed for inclusion in our review.However, in the early screening process, we took a conservative approach where articles were only excluded based on title and abstract if they were clearly irrelevant to the research question, with all other articles progressing to the next stage of review.We are therefore confident that any misclassification is . The date of publication for included studies ranged from 2007 to 2023.The majority of studies were quantitative and used a cross sectional (n = 33), longitudinal (n = 5), or randomized control trial (n = 6) study design.Seven studies used a mixed methods approach, and six were qualitative studies.Thirty studies examined online grocery retail, 22 examined online food delivery platforms, and five examined meal kit subscription services.Forty-eight studies were conducted in high income countries, including the United States (n = 20), the United Kingdom (n = 11), Australia (n = 7), Canada (n = 3), Denmark (n = 2), the Netherlands (n = 1), New Zealand (n = 1), Saudi Arabia (n = 1), Singapore (n = 1), and multiple high income countries (n = 3).Seven studies were conducted in upper middle income countries, including Brazil (n = 3), China (n = 2), Malaysia (n = 1), and multiple Latin American counties (n = 1).The year of data collection ranged from 2007 to 2022.Thirty-three studies examined how adults engage with the digital food retail environment, and 24 studies examined the contents or an aspect of the digital food retail platform themselves.Eleven studies had a focus on a particular socioeconomic group, with six studies examining Supplemental Nutrition Assistance Program (SNAP) recipients, two studies examining low income population groups, one study examining seniors with low incomes, one study examining a food insecure population, and one study examining participants of the SNAP for Women, Infants, and Children (WIC).Three studies examined the way that the COVID-19 pandemic has affected the digital food retail environment.Health-related outcomes measured by each study are reported in healthy meals.The studies by Gibson et al. and Moores et al. both had similar methodology, with dieticians analyzing the macronutrient composition of the prescribed meals supplied by different meal kit subscription services.Moores et al. analyzed 12 months of HelloFresh recipes, and Gibson et al. physically prepared 60 meals across five different meal kit subscription services (Dinnerly, HelloFresh, Marley-Spoon, Pepper Leaf, and Thomas Farms Kitchen), weighed the supplied ingredients, then analyzed the nutritional composition using Other food environment measures (e.g., in-app marketing and food environment scoring tools) Health outcomes associated with use of online food retail environments