Introduction
Food prepared away-from-home is typically energy-dense and nutrient-poor [
1,
2]. Over the last two decades, the number of food outlets selling such food has increased across the world [
3‐
6]. Whilst purchasing food prepared away-from-home is likely to be determined by many factors [
7], the influence of the number of food outlets physically accessible in one’s neighbourhood has been researched extensively [
8‐
10]. Evidence from cross-sectional and longitudinal studies has shown that individuals living in areas with greater neighbourhood food outlet access consume food prepared away-from-home more frequently, and live with higher bodyweight and obesity [
11,
12], and that these associations are stronger for those of lower socioeconomic position [
13]. In the past, individuals may have been limited to purchasing food prepared away-from-home from outlets that were physically accessible [
14]. However, changing social-norms [
15], technological advances [
16], widespread internet availability [
17], and a desire for greater convenience [
18], have influenced the emergence of alternative purchasing formats.
Online food delivery services provide online access to food outlets selling food prepared away-from-home. Based on their location, customers receive aggregated information about food outlets that will deliver to them. Customers select a food outlet and place their order through the online food delivery service platform. Orders are then forwarded to the food outlet where meals are prepared. When ready, meals are delivered by couriers who work for either the online food delivery service or independently for the food outlet. In 2018, around one in six adults in the UK had used an online food delivery service in the previous week [
19]. These customers tended to have higher levels of education, were younger, male, or living with children.
Socio-ecological models propose that among other factors, an interplay between physical food outlet access and individual-level characteristics influences the purchase of food prepared away-from-home [
9,
20‐
22]. However, little is known about the determinants of online food delivery service use. Based on findings from research investigating the role of physical food outlet access, it is reasonable to suggest that greater access to food outlets via online food delivery services is associated with more frequent use, and that the influence of this exposure varies according to customer characteristics. Food outlet characteristics may also be important. Online food delivery services facilitate access to different types of food outlets, including restaurants and takeaways, often selling a range of cuisines [
23]. As taste preferences contribute to food choices [
24], another possible determinant of online food delivery service use is the type of cuisine sold by accessible food outlets. Finally, food available through online food delivery services is typically energy-dense and nutrient-poor [
23,
25]. Since consumption of such food has been associated with weight gain over time [
26,
27], it is plausible that greater online food outlet access is associated with living with a higher bodyweight.
In this study, we investigated the association between the number of food outlets and unique types of cuisine accessible online and online food delivery service use, and in the presence of an association, whether it differed according to sociodemographic characteristics of online food delivery service customers. In secondary analyses, we investigated the association between the number of food outlets accessible online and bodyweight.
Methods
We used Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines to report the Methods of our study [
28] (Checklist S1).
Study design
In this cross-sectional analysis, we used an automated approach to collect data on the number of food outlets and unique types of cuisine accessible through an online food delivery service. We linked this with data on the use of online food delivery services and bodyweight amongst adults in the UK, collected via an online survey.
Study population
The International Food Policy Study (IFPS) is an ongoing annual repeat cross-sectional survey conducted in Australia, Canada, Mexico, the UK, and USA. We used data collected between November–December 2019 from respondents in the UK. Data collection methods for the IFPS have been described elsewhere [
29]. Briefly, respondents were recruited through Nielsen Consumer Insights Global Panel and their partners’ panels. Email invitations with links to an online survey were sent to a random sample of eligible panellists aged between 18 and 100 years. Informed consent was obtained from all respondents prior to survey completion. The IFPS received ethics clearance through a University of Waterloo Research Ethics Committee (ORE# 21460). Data collection in the UK was approved by the University of Cambridge Humanities and Social Science research ethics committee (case 19/225), and all methods were carried out in accordance with relevant guidelines and regulations.
Online food outlet and unique type of cuisine access
The online food delivery service
Just Eat has been available in the UK since 2006, and in 2020 was the market leader in terms of the number of food outlets registered to accept orders (around 35,000) and annual order volume (over 120 million) [
30,
31]. Unlike competitors, which focus on the largest UK cities,
Just Eat reports that it has outlets accessible across all parts of the UK [
32]. Through pilot work in one region of England, we identified that 95% of food outlets registered to accept orders through a competitor (
Deliveroo) were also registered with
Just Eat (Additional file
1). Therefore, we used data from
Just Eat (referred to as ‘the online food delivery service’ hereafter) as a proxy for all online food outlet access.
In November 2019, we used a web-browser extension to collect data about food outlets in England, Wales and Scotland that were accessible through the online food delivery service. First, on one weekday, we identified all food outlets registered to accept orders. Second, within 72-h, we visited the profile of each outlet on the online food delivery service website and collected data on their physical location, the types of cuisine sold (a maximum of two classifications self-determined by outlet owners), and their delivery area, which is a list of all postcode districts to which they delivered. In the UK, the first half of a full postcode is the postcode district, for example, in the
postcode ‘CB2 0QQ’, the
postcode district is ‘CB2’. Postcode districts are predominantly used for mail and delivery routing purposes and vary in size. Based on data from the 2011 census, the median postcode district population was 23,610 (IQR; 13,320-34,560) [
33].
When completing the IFPS survey, respondents in the UK reported their residential postcode. From this, we extracted the postcode district and identified the number of food outlets accessible online by summing the number of food outlets that listed the same postcode district in their delivery area. From the accessible food outlets, we summed the number of unique types of cuisine.
Online food delivery service use
In the IFPS survey, respondents were asked “During the past 7 days, how many meals did you get that were prepared away-from-home in places such as restaurants, fast food or takeaway places, food stands, or from vending machines?”. Respondents who had purchased at least one meal prepared away-from-home in the previous week were then asked to report the number of meals that were “Ordered using a food delivery service (e.g., Uber Eats, Just Eat, Deliveroo) and delivered”. We used answers to the second question for our primary outcome, and dichotomised respondents into those who reported use of an online food delivery service in the previous week, and those who did not.
Body mass index and weight status
We used self-reported height and weight data for our secondary outcomes. We calculated body mass index (BMI; kg/m
2, continuous), and used World Health Organization BMI cut-offs to classify respondents as being: ‘not overweight’ (BMI < 25 kg/m
2); ‘overweight’ (BMI
> 25–29.9 kg/m
2); or ‘obese’ (BMI
> 30 kg/m
2). We included respondents in a ‘not available’ category when we were unable to calculate BMI due to bodyweight non-report, which is a possible reflection of social-desirability bias [
34], or when calculated BMI was < 14 kg/m
2 or > 48 kg/m
2.
In accordance with findings from research investigating the relationship between neighbourhood food outlet access and bodyweight [
12,
26], it is plausible that greater online food outlet access is positively associated with BMI and weight status. However, the focus of our study was on the association between online food outlet access and the more proximal outcome of online food delivery service use, which is a relationship less susceptible to bias from unmeasured confounding [
10,
35]. Therefore, we report the findings for our secondary outcomes in additional file
2.
Potential confounders
In the IFPS survey, respondents reported sociodemographic information. We included potential confounders, decided
a priori, based on previous findings that they were positively associated with online food delivery service use, or purchasing food prepared away-from-home [
19,
36]. Sex at birth was reported as male or female and treated as a binary variable in analysis. Age was reported in years. Due to a possible non-linear influence on food purchasing, we grouped respondents into four age categories for analysis: 18–29 years, 30–44 years, 45–60 years, > 60 years. Ethnicity was reported as the group that best described racial or ethnic background. We grouped respondents into a binary variable for analysis: ‘majority’ (white alone) or ‘minority’ (all other responses), which reflects that the majority of IFPS respondents in the UK were white. We used education and perceived income adequacy as markers of socioeconomic position [
37]. Education was reported as the highest qualification completed. We categorised respondents as having: ‘low’ (high school completion or lower), ‘medium’ (some post-high school qualifications), or ‘high’ (university degree or higher) education for analysis. Perceived income adequacy was reported based on how well total monthly incomes allowed a respondent’s needs to be met. We dichotomised respondents as finding it ‘not easy’ (don’t know, refuse to answer, very difficult, difficult, or neither easy nor difficult responses), or ‘easy’ (easy or very easy responses) to make ends meet. Living with children under the age of 18 years, and smoking status in the past 30 days were reported as binary (yes/no) measures.
Food sold through online food delivery services is typically prepared in in the kitchens of existing food outlets. Therefore, online food outlet access might be a function of neighbourhood food outlet access. We used Ordnance Survey Points of Interest (OS POI) data from June 2019 to account for this. This commercial data contains information about food outlets from over 170 suppliers, is one of the most complete sources of food outlet location data available for the UK, and has been used in previous research investigating neighbourhood food outlet access [
38‐
40]. We extracted information for the following categories as they include food outlets predominantly registered to accept orders through online food delivery services: ‘Fast food and takeaway outlets’ (food outlets serving food for consumption away from the premises), ‘Fast food delivery services’ (food outlets serving food for delivery, not through online food delivery services), ‘Fish and Chip shops’ (food outlets predominantly serving a specific type of cuisine for consumption away from the premises) and ‘Restaurants’ (food outlets serving food for consumption inside the premises) [
41]. We mapped food outlets in a geographic information system (GIS) (ArcGIS version 10.7.1) using coordinates supplied in OS POI data, which have a stated accuracy of 1 m [
42]. We obtained coordinates for the postcodes of IFPS respondents through Doogal (a free web-based resource), or when this was not successful, the GeoConvert tool (maintained by the UK Data Service), and mapped them in our GIS. We counted the number of food outlets listed in OS POI data within a 1600 m (1-mile) Euclidean (straight-line) radius of respondents’ postcodes to determine neighbourhood food outlet access. This distance has been shown to reflect the spatial extent of an individual’s typical shopping behaviour, and could reasonably be walked by an adult in 15–20 min [
43].
Exclusion criteria
Data were available for 4139 IFPS respondents in the UK. We removed 732 (17.7%) respondents as they had either missing postcode, covariate (except BMI and perceived income adequacy), or outcome data; lived in Northern Ireland (not covered by OS POI data); or when the total number of meals purchased away-from-home in the past 7 days exceeded 21 (as this was not considered plausible). The final analytical sample included 3067 (74.1%) respondents.
Statistical analysis
We used Stata (version 16.1) to complete statistical analysis with a significance threshold of
p < 0.05 throughout. To reduce non-response and selection bias, we applied post-stratification sample weights constructed based on population estimates of age, sex, ethnicity and education from the 2011 UK census [
29]. We rescaled sample weights to our analytic sample, and unless specified, report weighted findings.
Residuals for all models were not normally distributed. Therefore, we modelled exposures (the number of food outlets and the number of unique types of cuisine accessible online) as quarters (Q). Q1 was the quarter with the lowest number and used as the reference category throughout statistical analyses. We used binomial logistic regression models to estimate the association between each exposure and online food delivery service use in the previous week. We completed unadjusted analyses and analyses adjusted for potential confounders of neighbourhood food outlet access, sex, age, education, perceived income adequacy, living with children and ethnicity. Where the exposure was the number of unique types of cuisine accessible online, we additionally adjusted for the number of food outlets accessible online, which was positively related.
Where the exposure was the number of food outlets accessible online, we added multiplicative interactions to our adjusted binomial logistic regression model to investigate if the association with online food delivery service use varied according to respondent education level, age and sex, and whether they lived with children. We used post-estimation Wald Tests to determine interaction significance and completed analyses stratified by the respective sociodemographic characteristic when there was evidence of a significant interaction.
Sensitivity analyses
In sensitivity analyses, we wanted to test our assumption that we had appropriately defined neighbourhood food outlet access. We used our adjusted model for all sensitivity analyses. When constructing our variable to control for broader neighbourhood food outlet access, we first included the number of supermarkets accessible in the neighbourhood (as these outlets provide access to food sold ready-to-eat). Second, we included additional types of food outlet. Alongside the four OS POI food outlet categories initially included (‘Fast food and takeaway outlets’, ‘Fast food delivery services’ (not through online food delivery services), ‘Fish and Chip shops’ and ‘Restaurants’) we also included: ‘Cafés, snack bars and tea rooms’, ‘Convenience stores’, ‘Supermarkets’, ‘Bakeries’ and ‘Delicatessens’.
Secondary outcomes
We used linear regression models to investigate the association between online food outlet access and BMI, and multinomial logistic regression models to investigate the association between online food outlet access and weight status. We completed unadjusted analyses and adjusted analyses that included the aforementioned potential confounders in addition to smoking status, which is negatively related to bodyweight [
44]. Due to the exploratory nature of these analyses, we did not complete sensitivity analyses or explore interactions.
Conclusions
More frequent online food delivery service use could increase the consumption of energy-dense and nutrient-poor food that is typically sold away-from-home, which has known implications for health. Our study is the first to investigate the association between the number of food outlets or unique types of cuisine accessible online and the use of online food delivery services. After adjusting for a range of potential confounders, adults in the UK with the greatest number of accessible food outlets had 71% greater odds of online food delivery service use in the previous week compared to those with the lowest number. This association was particularly evident in adults who were highly educated, younger, living with children or female. We did not find evidence that the number of unique types of cuisine accessible online was associated with online food delivery service use, and the number of food outlets accessible online was not associated with bodyweight. Future research might further explore reasons for using online food delivery services and the implications for public health resulting from use of this purchasing format.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.