The setting for our study was England, and the study period was from November 2019 to March 2022, which coincided with the end of emergency regulations introduced during the COVID-19 pandemic (Table 1). Data were not collected between December 2019 and May 2020, meaning that first available data after November 2019 were from June 2020.

We completed analyses at the postcode district level because this reflects how food outlets registered to accept orders online delineated where they would deliver to in November 2019. Postcode districts are contained in the first half of full postcodes. For example, for the postcode CB2 0QQ, the postcode district is CB2. Postcode districts have an average size of 33 square miles [20] and a median population of 23,610 (IQR 13,320-34,560) [21].

We used postcode district boundary data from 2012, sourced from the United Kingdom Data Service [22], to map postcode districts in a Geographic Information System (GIS; ArcGIS version 10.7.1; ESRI Inc). We considered 2118 postcode districts eligible for inclusion in our analyses, reflecting those with boundaries entirely within or intersecting the border of England.

We modeled our exposure measure (time) based on the frequency of data collection for our outcomes (monthly). We investigated changes in each of our outcomes over time compared with baseline (November 2019).

Our study relied on publicly available data. Research ethics committee approval was not required.

A descriptive summary of online food access in England is shown in Table 2 and Tables S1-S8 in Multimedia Appendix 1.

The summed number of food outlets registered to accept orders online in England increased from 29,232 in November 2019 to 49,752 in March 2022, equating to 70.2% growth (Figure 1).

The median number of food outlets registered to accept orders online per postcode district was 7.0 (IQR 1.0-21.0) in November 2019 and 13.0 (IQR 3.0-34.0) in March 2022. The median percent change from baseline (November 2019) per postcode district was 65.4 (IQR 33.3-100.0) in March 2022. The overall increase in the number of food outlets registered to accept orders was significant at each time point in our adjusted model (Table S10 in Multimedia Appendix 1 provides the IRRs), and there was significant effect modification by deprivation (P<.001). Estimated means calculated from IRRs of our adjusted negative binomial GEE are shown in Figure 2. At each level of deprivation, we observed that the estimated number of outlets registered to accept orders online had initially increased from baseline levels (November 2019); this decreased immediately after June 2020 and was then followed by a more consistent upward trend. The estimated number was consistently highest in the most deprived areas (Q5) and lowest in the least deprived (Q1). Absolute growth over time was also highest in the most deprived areas. For these areas, the estimated number was 39.1 outlets at the end of the study period, compared with 24.6 outlets at baseline, whereas for the least deprived areas, this was 6.6 and 3.5 outlets, respectively. The slope index of inequality between the least and most deprived areas was 5.0 (95% CI 4.5-4.5) outlets at baseline, and 7.8 (95% CI 7.0-8.6) outlets in March 2022.

The median number of food outlets accessible online per postcode district was 63.5 (IQR 16.0-156.0) in November 2019 and 57.0 (IQR 11.0-163.0) in March 2022. The median percent change from baseline per postcode district was 0.0 (IQR −32.0 to 33.3) in March 2022. The overall decrease in the number of food outlets accessible online was significant at each time point in our adjusted model (Table S12 in Multimedia Appendix 1 provides the IRRs), and there was a significant effect modification by deprivation (P<.001). Estimated means calculated from the IRRs of our adjusted negative binomial GEE are shown in Figure 3. At each level of deprivation, we observed that the estimated number of food outlets accessible online had decreased from baseline in June 2020. Although this was followed by an upward trajectory, the estimated number remained lower than baseline in less deprived areas (Q1-Q3) but surpassed baseline in those in the 2 upper Qs of deprivation (Q4 and Q5). For the most deprived areas (Q5), the estimated number increased from 181.9 outlets in November 2019 to 200.0 outlets in March 2022, and this contributed to an increasing gap in online food outlet access between the least and most deprived areas. The slope index of inequality between the least and most deprived areas was 32.0 (95% CI 28.1-35.9) outlets at baseline and 37.3 (95% CI 31.8-42.9) outlets in March 2022.

The median number of food outlets registered to accept orders online as a percentage of the number of food outlets in the physical food environment per postcode district was 14.3 (IQR 3.8-26.0) in November 2019 and 24.0 (IQR 6.2-43.5) in March 2022. The median percent change from baseline per postcode district was 55.3 (IQR 23.0-96.1) in March 2022. The overall increase in the percentage of food outlets registered to accept orders online was significant at each time point in our adjusted model (Table S14 in Multimedia Appendix 1 provides the coefficients), and there was a significant effect modification by deprivation (P<.001). Estimated means calculated from the coefficients of our adjusted GEE are shown in Figure 4. We observed an initial increase from baseline in June 2020 that was followed by a slight decline, a second increase that equaled or surpassed previous levels, and then another decline before a more stable increase. Although this trend was evident across all levels of deprivation and the estimated percentage was significantly increased at each level of deprivation by the end of the study period, the magnitude varied. Nevertheless, the estimated mean percentage was the highest in the most deprived areas (40.0% in March 2022 compared with 27.4% in November 2019), with these areas also having a higher growth over time in absolute terms, compared with other areas. The slope index of inequality between the least and most deprived areas was 4.5% (95% CI 4.1-4.9) at baseline and 5.9% (95% CI 5.4-6.5) in March 2022.

To the best of our knowledge, our study was the first to examine changes in levels of online food access across an entire country during the COVID-19 pandemic. We identified an increase in the number of food outlets registered to accept orders online, reaching approximately 50,000 in England by March 2022. There was a parallel increase in the percentage of food outlets in the physical food environment registered to accept orders online. In contrast, the number of food outlets accessible online, that is, those that would deliver to a given population living in a postcode district, was on average lower in March 2022 than before the COVID-19 pandemic. The magnitude of change for all measures of online food access varied by level of deprivation. The most deprived postcode districts consistently had the highest number and percentage of food outlets registered to accept orders online. Importantly, contrary to the trend for England as a whole, only in the most deprived postcode districts was the number of food outlets accessible online higher in March 2022 than at baseline. We did not observe this for postcode districts at any other level of deprivation, with the number decreasing in less deprived areas. As a result, there is some evidence that existing socioeconomic inequalities in the opportunity to use online food delivery services to purchase food prepared out of home widened over time.

Our finding that an increased number of food outlets had registered to accept orders online in the first 2 years of the COVID-19 pandemic aligns with the reports presented in business and news media from major online food delivery services operating in England [35,36]. Moreover, the decrease in July 2020 accords with contemporaneous reports of food outlet closures and decreased order volume through online food delivery services [37]. For the most part, after July 2020, we observed an increase in the number of food outlets registered to accept orders online. As opportunities for on-premises food consumption were limited by national stay-at-home orders imposed in March 2020, followed by periods of time with restrictions on the capacity for on-premises dining, food outlet owners perhaps made a strategic business decision to register to accept orders online as a way to generate revenue.

For England as a whole, we identified an overall decline in the number of food outlets accessible online (ie, the number that will deliver to a given population living in a postcode district) between November 2019 and March 2022. This is in contrast to the increased number of food outlets registered to accept orders over the same period (irrespective of whom they would deliver to). The decline we observed was particularly pronounced in the early stages of the COVID-19 pandemic, which might have reflected a period of transition among those that had only recently registered to accept orders online. Food outlets unaccustomed to food delivery plausibly operated with a more limited delivery radius to ensure they could fulfill customer orders. Moreover, at this time, there was broader workforce capacity instability owing to self-isolation rules, unclear restrictions on maximum travel distances legally allowed, and concerns for online food delivery service courier safety [38]. All of which might have contributed to the implementation of limited delivery areas.

Although the initial decline in the number of food outlets accessible online was followed by an increase at all levels of deprivation, it was only among more deprived areas, and particularly the most deprived, that the number eventually surpassed prepandemic levels over the period of our study. The number of food outlets accessible online was already the highest in the most deprived areas in England before the pandemic [12]. However, we found evidence of an increasing divergence between the most and least deprived areas, suggesting that inequality in online food access widened during the period of our study. A perceived lack of demand in less deprived areas might have meant that registered food outlets chose not to include them in their delivery areas. Moreover, the delivery of food sold through online food delivery services is mostly completed by couriers on bicycles. As such, there is a natural limit in the distance that can be traveled while maintaining food quality. In rural, less deprived areas, this challenge might be insurmountable. Regardless, given that the number of food outlets accessible online is positively associated with online food delivery service use [11], it is possible that food-purchasing practices, diet, and health were negatively influenced during the COVID-19 pandemic, especially among populations living in more deprived areas. Although adults living in the United Kingdom self-reported a decrease in their diet quality during the COVID-19 pandemic [39], evidence on changes in the frequency of having food prepared out of home delivered is inconsistent [40]. Further research is required to understand the extent to which the changes in online food outlet access we observed were associated with changes in online food delivery service use during the same period and subsequent implications for diet and health. This future research could incorporate further measures of food access, for example, accounting for the affordability of food sold, which is recognized by online food delivery service customers as being an important consideration preceding use [41]. It is plausible that it is not financially viable for populations living in the most deprived areas to use online food delivery services despite the increase in the number of food outlets accessible, especially owing to the increase in the cost of living in the United Kingdom and elsewhere [42].

The number of food outlets registered to accept orders online as a percentage of the number of food outlets in the physical food environment had increased by the end of the study period. However, this increase was only apparent after a period of instability. Owners of food outlets with customer-facing premises in the physical food environment reported that although being registered to accept orders online was a way for customers to access their food when on-premises food consumption and travel was restricted, simultaneously managing orders placed in-person and online was difficult [38]. If food outlets that did not previously accept orders online only did so out of necessity when on-premises food consumption was restricted, it is plausible that they subsequently deregistered when restrictions ended. This scenario would partly explain the successive increases and decreases that coincided with the start and end of restrictions introduced during the COVID-19 pandemic (Figure 4).

Although the relative change over time with respect to the percentage of food outlets registered to accept orders online was consistent across all levels of deprivation, the absolute change was highest in the most deprived areas, which led to a widening of absolute inequality in online access to food prepared out of home. Food sold through online food delivery services is typically prepared in the kitchen facilities of food outlets located in the physical food environment [9]. Our finding is likely a reflection of the existing urban form in the most deprived areas in England, which is characterized by a high density and concentration of food outlets [43,44]. Although we cannot be certain, if more food outlets were operating within emergency regulations introduced during the COVID-19 pandemic, more outlets may have also registered to accept orders online. Our findings therefore provide evidence to suggest that the introduction of emergency regulations at least partly contributed to widening inequality in this exposure.

There remains considerable scope for the percentage of food outlets in the physical food environment that are registered to accept orders online to increase. This emphasizes the coexisting and overlapping nature of digital and physical food environments, which together provide multiple opportunities to purchase energy-dense, nutrient-poor food [45]. However, there are likely to be natural limits to growth because all food outlets do not necessarily need or want to register to accept orders online. In fact, the relative stability from June 2021 onward suggests that a plateau might have already occurred. An important limitation of this paper is that we do not have additional prepandemic data that would allow us to account for any existing trends in analyses. Nevertheless, our findings and associated data can contribute to future surveillance of longer-term trends.

Urban planning has been used by over half of local authorities in England to promote the creation of healthier physical food environments, specifically by preventing new takeaway food outlets from opening [46]. To our knowledge, restrictions targeted specifically at online food delivery services are not in place. Although not yet fully clear, similar to the physical food environment, interventions to restrict access to food prepared out of home through these services might be increasingly necessary in the future [47,48]. At this point, public health interventions that do not necessarily attempt to restrict online food access but instead attempt to mitigate the potential public health burden posed by online food delivery service use possibly represent the most appropriate route to regulation. For example, changing the nutritional composition of food sold inside outlets would be expected to have a crossover effect and also change food sold online. Additionally, in England, in 2022, mandatory calorie labeling of non–prepacked food and soft drinks was introduced for retailers that operate with more than 250 employees, which meant that this information had to be displayed both inside food outlets and on the platforms of online food delivery services [49,50]. The introduction of this regulation recognizes the role of the digital food environment in the purchase of food prepared out of home. Research in the future could seek to further understand if and how regulations in the physical food environment can be extended to the digital food environment.

Furthermore, there has been evolution in the preparation location for the food sold through online food delivery services. This is best demonstrated by the development of facilities known as dark kitchens [51]. These facilities allow food businesses to register to accept orders through online food delivery services and prepare meals for delivery without the financial costs of having a customer-facing premises. To date, there is only limited evidence on the dark kitchen business model, which may not be generalizable beyond the data collection context of a single London borough [52]. Nevertheless, the development of these facilities plausibly influenced our findings related to the percentage of food outlets registered to accept orders online. Future research might seek to understand the public health implications of the dark kitchen business model, including how it uniquely contributes to the number of food outlets accessible online.

Our data collection spanned 2 years and allowed us to closely monitor trends in important metrics of online food outlet access. In doing so, we present new baseline levels for future assessment. Nevertheless, our study is not without limitations. Our analysis was observational, and we cannot definitively conclude that the changes we observed were due to the COVID-19 pandemic. We also do not have sufficient prepandemic data that would allow underlying or existing trends to be accounted for in analyses. This is particularly the case for changes between November 2019 and June 2020 when we had no data, and we could not determine whether changes coincided with the COVID-19 pandemic.

We completed data collection at a single time point on a monthly basis. If food outlets were only registered to accept orders or accessible in the intervening period, they might not have been returned in our searches. However, the changes in the number of food outlets accessible online that we observed, especially the initial decrease in June 2020, is similar to reports of food outlet closures and decreased order volume at this time [37].

We did not track individual food outlets over the study period. Instead, we studied the total number of food outlets registered to accept orders at any given time point. This means that we are unable to comment on the number that were newly registered during the COVID-19 pandemic. It would be interesting for future work to investigate the extent to which food outlets that registered only during this period remained registered.

Our findings demonstrate potential online food outlet access. For this to be realized, any given individual must be an online food delivery service customer. As such, our findings do not necessarily translate into realized individual-level online food outlet access [53].

We used postcode districts as our unit of analysis and acknowledge the possibility that our findings are subject to the modifiable areal unit problem. Although the spatial unit that we adopted for analyses has the potential to introduce bias [54], our approach was data driven to allow consistency with our previous research [12]. Relatedly, using postcode districts as our unit of analysis also meant that we were limited to using boundary data from 2012. These data were not temporally aligned with further covariate data that were the most recently published or collected.

We investigated changes in multiple measures of food access through online food delivery services during the first 2 years of the COVID-19 pandemic. We identified that the number of food outlets in England that were registered to accept orders online increased. In parallel, the number of food outlets registered to accept orders online as a percentage of the number of food outlets in the physical food environment increased. Although the number of food outlets that could be accessed decreased for the whole of England, trends differed by the level of deprivation. The number of outlets decreased compared with prepandemic levels in the least deprived areas and increased to surpass prepandemic levels only in the most deprived.

Overall, then, we identified considerable changes in measures of online food accessibility during a period when on-premises food consumption was often restricted. Our data from March 2022 represent a new baseline to which future changes in measures of online food accessibility can be investigated. Future research might attempt to understand the extent to which the changes we identified were associated with the changes in online food delivery service use and, in turn, the implications for diet quality and health.