One in five deaths worldwide is linked to poor diet(Reference Afshin, Sur and Fay1). Food consumed out of the home in restaurants, cafes and takeaways tends to be high in calories(Reference Roberts, Das and Suen2), saturated fat(Reference Davies, Blackham and Jaworowska3) and salt(Reference Jaworowska, Blackham and Stevenson4), with more regular consumption linked to an increased risk of higher body weight(Reference Nago, Lachat and Dossa5). A 2022 survey conducted by the Food Standards Agency in England, Wales and Northern Ireland found that 53 % of respondents had eaten in a restaurant, 50 % had ordered takeaway food and 38 % had consumed food from a fast-food outlet in the previous 4 weeks(Reference Armstrong, King and Clifford6). Evidence from high-income countries also suggests fast-food outlets are more common in deprived communities compared with more affluent areas(Reference Fleischhacker, Evenson and Rodriguez7–9).
Small businesses dominate the sector, with just over half of the £62 billion of revenue generated from the UK consumer foodservice sector in 2022 coming from small and locally owned restaurants and hot food takeaways(10). Unlike large chain restaurants, small independent businesses usually operate in small premises, with limited staff, equipment and access to suppliers(11), and finite resources to participate in healthy eating interventions specifically(Reference Goffe, Penn and Adams12). Yet in contrast to large, chained businesses, owners of small independent businesses may be more likely to be able to make decisions about whether and how to enact interventions(Reference Ma, Shive and Zhang13).
A study found that independent and small-chain restaurants (under twenty outlets) serve meals with higher energy content than those in larger chain restaurants, with individual meals accounting for 66 % of an adult’s daily energy requirements(Reference Urban, Lichtenstein and Gary14). Nevertheless, existing reviews on healthier eating interventions in restaurants, takeaways and fast-food outlets mostly rely on large chain restaurants(Reference Hillier-Brown, Summerbell and Moore15) or include both chains and non-chains(Reference Espino, Guerrero and Rhoads16,Reference Gittelsohn, Lee-Kwan and Batorsky17) , and policies aiming to support healthier food purchasing in the out-of-home sector (e.g. nutrition labelling) have typically only applied to larger businesses who have greater resources to implement such legislation(18,19) . However, this risks widening health inequalities if the small businesses exempt from them provide less healthy food and are more abundant in predominantly poorer areas. Indeed, evidence from Australia shows that independent takeaways are more common in deprived than affluent areas(Reference Turrell and Giskes20) and studies from the UK describe a high prevalence of independent fast-food and takeaway outlets in disadvantaged areas(Reference Bagwell21,Reference Blow, Gregg and Davies22) . Therefore, separate interventions are needed for small, independent restaurants, which are effective in improving food healthiness whilst being feasible and acceptable to restaurants. For example, in the case of menu labelling, a survey among independent restaurants in Canada found that common worries are the expenses and time effort required to implement such a policy(Reference Mah, Vanderlinden and Mamatis23).
The differences between chained and independent restaurants and takeaways mean that policymakers need specific information on the types of interventions that may be effective in small restaurants and hot food takeaways, including potential challenges and opportunities to developing effective interventions. We aimed to systematically review the extant evidence of interventions to promote healthier food purchasing or consumption in this setting.
The objectives of this review were to:
-
1. Establish the characteristics of healthy eating interventions conducted in small, independent or local restaurants and hot food takeaways (hereafter ‘restaurants and takeaways’).
-
2. Assess the impact these interventions had on food availability and purchasing patterns.
-
3. Identify characteristics of interventions that increased acceptability to small restaurant and takeaway staff and owners.
The findings of this review can inform policymakers on which interventions may be effective and acceptable in small independent restaurants and takeaways and can be implemented at a local level.
Methods
Protocol and registration
The pre-registered protocol is available on PROSPERO (CRD42022341791). This review follows the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 checklist for reporting systematic reviews(Reference Page, McKenzie and Bossuyt24) (Appendix 1).
A review with a wider scope was specified in the protocol including healthier eating interventions in small food stores and restaurants. However, after completing data extraction, we split the review into two papers to focus on each setting in isolation. Instead of using the National Institutes of Health Quality Assessment Tools(25) as pre-specified, we used the Mixed Methods Appraisal Tool (MMAT)(Reference Hong, Fàbregues and Bartlett26), which enabled us to evaluate different study types employing one tool and guidance document.
Data sources and search strategy
The search strategy for this review was tailored to identify interventions in small restaurants and takeaways and food stores. An information specialist (NR) developed the search strategy in consultation with LB after initial scoping searches. We searched the following databases for primary studies from database inception to 15 June 2022: Cochrane Central Register of Controlled Trials (CENTRAL), Issue 5, 2022, the Cochrane Library (Wiley), MEDLINE and Epub Ahead of Print, In-Process & Other Non-Indexed Citations and Daily (OvidSP, 1946–), Embase (OvidSP, 1974–), CINAHL (EBSCOhost, 1982–) and Science Citation Index and Social Science Citation Index (Web of Science Core Collection, 1900–). Search strategies were comprised of keywords and controlled vocabulary terms. We applied no limits on language or publication date. We used the filter designed by the Cochrane EPOC group to identify randomised studies, before and after studies and interrupted time series (https://zenodo.org/record/5106292). We used the Polyglot tool from SR-accelerator (https://sr-accelerator.com/#/polyglot) to adapt the search formatting from MEDLINE to the other databases. All search strategies are provided in Appendix 2A.
All references were downloaded to Endnote 20(27) before being transferred to Covidence(28). In November 2022, we additionally conducted forward and backward searches of eligible studies and six reviews(Reference Hillier-Brown, Summerbell and Moore15,Reference Espino, Guerrero and Rhoads16,Reference Liberato, Bailie and Brimblecombe29–Reference Gittelsohn, Rowan and Gadhoke32) using Citation Chaser(Reference Haddaway, Grainger and Gray33), with results subsequently imported into Covidence(28). We excluded two more restaurant and takeaway papers(Reference Schuldt, Levings and Kahn-Marshall34,Reference Schneider, Markovinovic and Mata35) after the citation tracking due to them not meeting our setting or outcome requirements.
Eligibility criteria
Eligibility criteria were determined following the population, intervention, comparison, outcome and study design (PICOS) framework(36) and are outlined in Table 1. We included primary studies where the study authors described the restaurant or takeaway as small, independent, community-based or local and where there was no evidence that the restaurants or takeaways were part of a chain. A community-based restaurant was defined as a place for local people to come together to eat freshly cooked food.
Eligibility criteria based on population, intervention, comparison, outcome and study design (PICOS)(36)
* Exclusion of theses, dissertations, protocols, trial registrations and conference abstracts as well as national-level interventions was decided after the review had begun.
Study selection and data extraction
After the exclusion of duplicates, abstract and full-text screening was completed independently and in duplicate by two reviewers (LB and CJ) using Covidence(28). Any conflicts were discussed by the two reviewers, and a third reviewer arbitrated if needed.
Data extraction was conducted by one reviewer, with reviewers subsequently checking the data extraction forms completed by the other reviewer. Again, any conflicts were discussed and resolved between the two reviewers. Data extracted included first author name and year, country and location, stakeholders involved, study aim, methods (design, start and end date, targeted population, co-design and stakeholder involvement, if applicable), setting type, sample size, recruitment methods, inclusion criteria, intervention characteristics (name, description, duration, comparator/control), outcomes, measured data, statistical or analysis methods, main findings, barriers and facilitators to working with businesses and recommendations for future studies.
Study quality assessment
Study quality was rated independently by one reviewer and verified by a second reviewer (LB or CJ) using the MMAT(Reference Hong, Fàbregues and Bartlett26). Each of the included studies was first categorised into one of five groups based on study design: (i) qualitative research; (ii) randomised controlled trials; (iii) non-randomised studies; (iv) quantitative descriptive studies; and (v) mixed methods studies. Studies were then assessed for quality using the category-specific criteria and presented in full, as recommended, rather than being adapted into a single score(Reference Hong, Fàbregues and Bartlett26).
Synthesis of results
Results were synthesised narratively(Reference Mays, Pope and Popay37). The main characteristics and outcomes of interventions were summarised in tables, and the patterns were identified. Additionally, our analysis was guided by the Nuffield intervention ladder(38), which categorises interventions according to their intrusiveness (i.e. their impact on individual freedom) (Table 2). Briefly, higher steps on the ladder represent more intrusive interventions, with eliminating choice being the highest step (i.e. most intrusive intervention)(38). Each element of the included interventions was grouped depending on whether it was designed to impact consumer behaviour (customer-level interventions) or the business’s behaviour (business-level intervention).
The Nuffield intervention ladder(38), used for categorising included interventions
Results
Searches retrieved a total of 7455 records, and after removing duplicates, 4624 records were screened (Fig. 1). We assessed 287 full-text records for eligibility, resulting in the inclusion of 12 studies reporting on 13 interventions.
PRISMA flow diagram showing the study selection process(Reference Page, McKenzie and Bossuyt24).
Quality assessment
Of the twelve studies included, most used a mixed methods(Reference Hillier-Brown, Lloyd and Muhammad39–Reference Goffe, Hillier-Brown and Hildred43) (n 5) or a quantitative, non-randomised(Reference Ma, Shive and Zhang13,Reference Lee-Kwan, Bleich and Kim44–Reference Fitzgerald, Kannan and Sheldon46) (n 4) study design (Table 3). Two were randomised controlled trials(Reference Ayala, Castro and Pickrel47,Reference McNally, Anzman-Frasca and Bowman49) , and one study was quantitative descriptive(Reference Chen, Carrillo and Kapp48). No qualitative studies were identified, potentially due to our outcomes of interest being geared towards quantitative measurements.
Study quality assessment of included studies using MMAT(Reference Hong, Fàbregues and Bartlett26)
MMAT, Mixed Methods Appraisal Tool; N, no; Y, yes; C, can’t tell. MMAT study design group: 2 = randomised controlled trial; 3 = quantitative non-randomised study; 4 = quantitative descriptive study; 5 = mixed methods study. Questions for each study design were as follows in Appendix 2B.
Neither of the randomised controlled trials(Reference Ayala, Castro and Pickrel47,Reference McNally, Anzman-Frasca and Bowman49) met all of the MMAT’s(Reference Hong, Fàbregues and Bartlett26) quality criteria. Randomisation was either not appropriately performed(Reference Ayala, Castro and Pickrel47) or insufficiently described(Reference McNally, Anzman-Frasca and Bowman49), and neither study reported whether outcome assessors were blinded to the intervention, which limited their quality assessment scores.
Only one of the four quantitative, non-randomised studies provided sufficient detail to be appraised and met all five criteria(Reference Lee-Kwan, Bleich and Kim44). Other studies provided insufficient detail on the population’s representativeness(Reference Fitzgerald, Kannan and Sheldon46), whether there were complete outcome data(Reference Ma, Shive and Zhang13,Reference Lee-Kwan, Goedkoop and Yong45) , whether confounders had been accounted for(Reference Ma, Shive and Zhang13,Reference Fitzgerald, Kannan and Sheldon46) and whether the intervention was implemented as intended(Reference Fitzgerald, Kannan and Sheldon46).
One study was a quantitative descriptive study(Reference Chen, Carrillo and Kapp48). It met the criteria for sampling strategy and statistical analysis but provided insufficient information or did not meet the criteria for representativeness of the sample, appropriateness of variables and measurements and the risk of non-response bias.
Five studies were mixed methods(Reference Hillier-Brown, Lloyd and Muhammad39–Reference Goffe, Hillier-Brown and Hildred43). Whilst all were strong on integrating qualitative and quantitative components of their research questions, they did not provide sufficient information on or failed to address the divergences and inconsistencies between quantitative and qualitative findings. Some also failed to meet the criteria of providing an adequate rationale for using mixed methods(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Nothwehr, Snetselaar and Dawson40,Reference Goffe, Hillier-Brown and Hildred43) , integrating quantitative and qualitative interpretation(Reference Nevarez, Lafleur and Schwarte41) and adhering to the quality criteria of the different methods involved(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Nothwehr, Snetselaar and Dawson40,Reference Goffe, Hillier-Brown and Hildred43) .
Settings and stakeholders involved in the interventions
Twelve studies reporting on thirteen interventions were included, with two papers assessing the same intervention(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) and two papers testing two interventions each(Reference Ayala, Castro and Pickrel47,Reference McNally, Anzman-Frasca and Bowman49) . Eight interventions involved small restaurants(Reference Nothwehr, Snetselaar and Dawson40,Reference Nevarez, Lafleur and Schwarte41,Reference Fitzgerald, Kannan and Sheldon46–Reference McNally, Anzman-Frasca and Bowman49) , three focused on takeaway outlets(Reference Ma, Shive and Zhang13,Reference Hillier-Brown, Lloyd and Muhammad39,Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) and two included both(Reference Bagwell42,Reference Goffe, Hillier-Brown and Hildred43) . The number of businesses involved varied, ranging from one(Reference Chen, Carrillo and Kapp48,Reference McNally, Anzman-Frasca and Bowman49) to 206(Reference Ma, Shive and Zhang13) (Table 4).
Intervention name, location, study design and stakeholders involved in included studies
RCT, randomised controlled trial; WA, Washington; CA, California; MD, Maryland; PA, Pennsylvania.
Ten interventions were conducted in the USA(Reference Ma, Shive and Zhang13,Reference Nothwehr, Snetselaar and Dawson40,Reference Nevarez, Lafleur and Schwarte41,Reference Lee-Kwan, Bleich and Kim44–Reference McNally, Anzman-Frasca and Bowman49) and three in the UK(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42,Reference Goffe, Hillier-Brown and Hildred43) . Most took place in cities(Reference Ma, Shive and Zhang13,Reference Nevarez, Lafleur and Schwarte41,Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45,Reference Chen, Carrillo and Kapp48) , highly populated counties(Reference Ayala, Castro and Pickrel47,Reference McNally, Anzman-Frasca and Bowman49) , boroughs(Reference Bagwell42) or suburban areas(Reference Fitzgerald, Kannan and Sheldon46). One intervention was set in rural small-town settings(Reference Nothwehr, Snetselaar and Dawson40), and one included both urban and rural settings(Reference Goffe, Hillier-Brown and Hildred43). Although not all studies provided this information, several targeted low-income areas(Reference Ma, Shive and Zhang13,Reference Nevarez, Lafleur and Schwarte41,Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) , and others spanned areas with various levels of deprivation(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42,Reference Goffe, Hillier-Brown and Hildred43) . All but three interventions(Reference Nothwehr, Snetselaar and Dawson40,Reference Ayala, Castro and Pickrel47) engaged a wider range of stakeholders other than businesses and academic researchers, commonly from the local authority (n 6; e.g. health teams, environmental health officers)(Reference Ma, Shive and Zhang13,Reference Hillier-Brown, Lloyd and Muhammad39,Reference Nevarez, Lafleur and Schwarte41,Reference Bagwell42,Reference Fitzgerald, Kannan and Sheldon46,Reference Chen, Carrillo and Kapp48) or local community organisations or non-governmental organisations (n 3)(Reference Ma, Shive and Zhang13,Reference Nevarez, Lafleur and Schwarte41,Reference Chen, Carrillo and Kapp48) .
Some interventions focused on specific cuisines, such as American(Reference Nothwehr, Snetselaar and Dawson40,Reference Ayala, Castro and Pickrel47) , Latino(Reference Ayala, Castro and Pickrel47), Chinese(Reference Ma, Shive and Zhang13) or British ‘Fish & Chip’ shops(Reference Goffe, Hillier-Brown and Hildred43). Two interventions had inclusion criteria relating to business owner ethnicity, targeting African-American or Korean-American takeaway owners(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) or Chinese American restaurant owners or chefs(Reference Ma, Shive and Zhang13). Several interventions conducted in the USA were set in areas with a high or growing proportion of residents identifying as Latino or Hispanic(Reference Nevarez, Lafleur and Schwarte41,Reference Ayala, Castro and Pickrel47–Reference McNally, Anzman-Frasca and Bowman49) , African American(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) or areas with a high proportion of ethnic minority residents(Reference Ma, Shive and Zhang13).
Interventions based on their classification on the Nuffield intervention ladder
Almost all interventions had components classed as ‘customer-focused’ as well as ‘business-focused’(Reference Ma, Shive and Zhang13,Reference Hillier-Brown, Lloyd and Muhammad39,Reference Nevarez, Lafleur and Schwarte41–Reference McNally, Anzman-Frasca and Bowman49) , with one intervention solely aimed at the customer level(Reference Nothwehr, Snetselaar and Dawson40) (Table 5). All but three interventions operated on more than one rung of the Nuffield ladder(Reference Nothwehr, Snetselaar and Dawson40,Reference Ayala, Castro and Pickrel47,Reference McNally, Anzman-Frasca and Bowman49) . The highest rung used was restricting choice on the business level(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) . The lower Nuffield ladder classifications that ‘provide information’ and ‘enable choice’ were most commonly used, aimed at both customers (e.g. menu labelling) and business owners and staff (e.g. cooking guidelines for chefs) (Table 5).
Included interventions coded by the Nuffield intervention ladder(38)
Business-level intervention components and outcome measures
Five interventions measured business-level outcomes(Reference Ma, Shive and Zhang13,Reference Hillier-Brown, Lloyd and Muhammad39,Reference Nevarez, Lafleur and Schwarte41–Reference Goffe, Hillier-Brown and Hildred43) (Table 6). Three studies used the number of businesses meeting certain criteria as an outcome measure(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42,Reference Goffe, Hillier-Brown and Hildred43) , and three studies measured the nutrient content or weight of dishes sold(Reference Ma, Shive and Zhang13,Reference Nevarez, Lafleur and Schwarte41,Reference Goffe, Hillier-Brown and Hildred43) , with one also describing self-reported changes to cooking habits(Reference Nevarez, Lafleur and Schwarte41). Four studies only provided descriptive evaluations(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Nevarez, Lafleur and Schwarte41–Reference Goffe, Hillier-Brown and Hildred43) .
Summary of intervention characteristics, outcome measures and main findings
Four interventions resulted in small increases in the number of businesses complying with criteria(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42,Reference Goffe, Hillier-Brown and Hildred43) , reduced weight of sold meals(Reference Goffe, Hillier-Brown and Hildred43) or sodium content of dishes(Reference Ma, Shive and Zhang13). One intervention described staff reporting positive changes to cooking habits(Reference Nevarez, Lafleur and Schwarte41).
Restrict choice
Three interventions(Reference Ma, Shive and Zhang13,Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) aimed to reduce the sugar, fat and salt content of foods, for example, by changing cooking practices (e.g. cooking oil usage) or switching to healthier products. Two used the number of businesses and number of criteria met as outcome measures and reported small positive effects(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) , whilst another recorded lower sodium content of dishes(Reference Ma, Shive and Zhang13). However, only one conducted statistical testing(Reference Ma, Shive and Zhang13).
The Healthy Catering Commitment in London is a series of criteria relating to cooking, serving and selling practices; businesses are expected to meet eight out of twenty-two criteria before being awarded a Healthy Catering Commitment Award by their local authority(Reference Bagwell42). Seventy-seven businesses were surveyed, each having to make an average of 2·5 criteria-related changes to secure the award(Reference Bagwell42). More businesses (n 26) signed up for ‘provision of information’ (e.g. promotion of healthy eating by staff) compared with ‘enabling choice’ criteria (n 1–15, depending on change) (e.g. offering fresh fruit, smaller portion sizes) due to cost and potentially reduced revenue associated with the latter. Criteria to ‘eliminate choice’ that were cheap and perceived as not interfering with customer preferences (e.g. cooking oil practices) were readily implemented; however, there was more hesitancy for changes visible to customers (e.g. thick-cut chips).
Similarly, the Takeaway Masterclass intervention asked businesses to commit to health-promoting practices and provided interactive training(Reference Hillier-Brown, Lloyd and Muhammad39). Businesses committed to a median of 4 goals/criteria (range 1–7) and achieved a median of 3 goals, including increasing vegetables in meals and grilling and poaching instead of frying(Reference Hillier-Brown, Lloyd and Muhammad39).
The Healthy Chinese Take-out Initiative included a media campaign and low-sodium training, with takeaways adopting sodium-reduction techniques such as lowering the amount of soy sauce used(Reference Ma, Shive and Zhang13). A significant and sustained reduction in the sodium content of three target dishes was found, with relative reductions of 36 % for Dish 1 (5·5–3·5 mg/g), 28 % for Dish 2 (5·7–4·1 mg/g) and 19 % for Dish 3 (5·9–4·8 mg/g), although all three dishes remained above the local authority’s recommended sodium intake per meal(Reference Ma, Shive and Zhang13).
Guide choice through incentives
The Fish and Chip Wholesaler Study combined public pledging, provision of smaller-sized packaging and an information and engagement session(Reference Goffe, Hillier-Brown and Hildred43). Although only reporting descriptive statistics, the number of venues offering smaller portion meals increased from 6 at baseline to 8 at 6-week post-intervention, and the weight of fish and chip meals sold decreased by a mean of 37 g for regular meals and 27 g for small meals(Reference Goffe, Hillier-Brown and Hildred43).
Enable choice
Salud Tiene Sabor, a menu labelling intervention, reported restaurant staff declaring they employ healthier cooking practices as a result of the intervention, including using more vegetables(Reference Nevarez, Lafleur and Schwarte41). They also tested meals served for their calorie content and found that post-intervention, 58 % of main meals and 59 % of side dishes remained above the local authority’s recommended calorie content per meal, although there was no baseline or control for comparison(Reference Nevarez, Lafleur and Schwarte41).
Customer-related intervention components and outcomes
Ten interventions measured customer-related outcomes(Reference Nothwehr, Snetselaar and Dawson40,Reference Nevarez, Lafleur and Schwarte41,Reference Goffe, Hillier-Brown and Hildred43–Reference McNally, Anzman-Frasca and Bowman49) (Table 6). Nine interventions used value sales and/or order data to measure intervention impact on customer-related behaviour change(Reference Nothwehr, Snetselaar and Dawson40,Reference Goffe, Hillier-Brown and Hildred43–Reference McNally, Anzman-Frasca and Bowman49) , and six reported customer-interview data(Reference Nothwehr, Snetselaar and Dawson40,Reference Nevarez, Lafleur and Schwarte41,Reference Goffe, Hillier-Brown and Hildred43,Reference Lee-Kwan, Goedkoop and Yong45,Reference McNally, Anzman-Frasca and Bowman49) . Studies that used sales data reported challenges with data collection. There was heterogeneity in registers/tills across restaurants that made the data hard to process(Reference Ayala, Castro and Pickrel47). Not all intervention restaurants and takeaways provided data(Reference Goffe, Hillier-Brown and Hildred43), and it was laborious to manually input paper order slips(Reference Lee-Kwan, Bleich and Kim44).
One intervention reported a positive effect on smaller portion size orders(Reference Goffe, Hillier-Brown and Hildred43), although only evaluated descriptively. Three interventions found mixed results (positive and no changes) on healthy foods sold(Reference Lee-Kwan, Bleich and Kim44,Reference McNally, Anzman-Frasca and Bowman49) . Two interventions recorded no significant increases in healthy item orders(Reference Nothwehr, Snetselaar and Dawson40,Reference Fitzgerald, Kannan and Sheldon46) . Three interventions reported orders occurred from the new healthier menu, but it was unclear if these replaced or supplemented orders from the existing menu(Reference Ayala, Castro and Pickrel47,Reference Chen, Carrillo and Kapp48) . One intervention reported that nutrition information influenced the purchase decisions of one-third of customers, but there was no baseline comparison(Reference Nevarez, Lafleur and Schwarte41).
Guide choice through incentives
Two interventions provided financial incentives for healthy meal choices through price promotions(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) and donations to local causes(Reference McNally, Anzman-Frasca and Bowman49), finding mixed effects (positive and no changes) on sales of healthier items depending on the item targeted(Reference Lee-Kwan, Bleich and Kim44) or comparison time period(Reference McNally, Anzman-Frasca and Bowman49).
The Baltimore Healthy Carry-out intervention recorded a statistically significant interaction between groups for healthier sides and beverages sold in two of three intervention phases, but not for healthier entrees or healthy items overall(Reference Lee-Kwan, Bleich and Kim44). The greatest increase was seen in phase 3, where a price promotion (incentive) was implemented alongside healthier cooking methods(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) . Although the effect of intervention phases cannot be isolated due to each new phase building on the previous one, this could suggest that intervention elements higher up the ladder may have been more successful within this study. In the process evaluation, 42·6 and 65·3 % of surveyed customers reported choosing an option due to the BHC leaf logo or photos on the menu, respectively(Reference Lee-Kwan, Goedkoop and Yong45). The Fundraising Healthy Eating Scheme intervention made higher financial donations to local schools contingent on orders of healthier menu items and found a higher percentage of healthier menu items ordered during the intervention than in the post-intervention period, but not in the pre-intervention period(Reference McNally, Anzman-Frasca and Bowman49). Six of the surveyed customers (20·7 %) said they chose their meal option due to the incentive(Reference McNally, Anzman-Frasca and Bowman49). There was no significant difference in orders of healthy items between the intervention arms (other arm reported under provide information)(Reference McNally, Anzman-Frasca and Bowman49); suggesting adding a higher ladder level component (incentive) did not provide additional benefit in this instance.
Enable choice
Five interventions enabled choice by adding new healthier meals or sides to menus(Reference Ayala, Castro and Pickrel47,Reference Chen, Carrillo and Kapp48) or marking healthy options on menus(Reference Nevarez, Lafleur and Schwarte41,Reference Lee-Kwan, Bleich and Kim44–Reference Ayala, Castro and Pickrel47) . Three recorded orders of new healthy items, but it remained unclear how order numbers from existing menus were affected(Reference Ayala, Castro and Pickrel47,Reference Chen, Carrillo and Kapp48) . One intervention found no changes in healthy item orders(Reference Fitzgerald, Kannan and Sheldon46). In one study, a third of clients stated their order was influenced by nutrition information, but there was no comparison(Reference Nevarez, Lafleur and Schwarte41).
The Kids Choice Restaurant Program created new healthier menus in both interventions, with one intervention additionally employing marketing and employee training(Reference Ayala, Castro and Pickrel47). Both interventions recorded increased sales of new healthier menu items immediately after implementation, with the proportion of healthier items making up 23 % (menu plus) and 17 % (menu only) of all children’s menu items in the first four intervention weeks(Reference Ayala, Castro and Pickrel47). However, sales of pre-existing menu items did not differ between the two conditions during the intervention, and the difference from baseline was not assessed statistically(Reference Ayala, Castro and Pickrel47). The Galerias Restaurant intervention found that after 6 weeks, 11·6 % of item orders were from the new intervention menu, but there was no comparator, and it is unclear whether sales of less healthy items decreased(Reference Chen, Carrillo and Kapp48). The Healthy Dining Program labelled and promoted healthy menu items and found no change in targeted healthy menu orders from pre-intervention to 6-week post-intervention(Reference Fitzgerald, Kannan and Sheldon46). Salud Tiene Sabor found that one-third of customers stated their purchases were influenced by the point-of-sale nutrition information that was displayed during the intervention, but there was no pre-intervention comparator(Reference Nevarez, Lafleur and Schwarte41).
Provide information
Three interventions provided information only, promoting healthier products using marketing materials such as table tents(Reference Nothwehr, Snetselaar and Dawson40,Reference Goffe, Hillier-Brown and Hildred43,Reference McNally, Anzman-Frasca and Bowman49) or providing point-of-purchase nutrition information(Reference McNally, Anzman-Frasca and Bowman49). One intervention described slightly increased small-portion orders(Reference Goffe, Hillier-Brown and Hildred43), however not using any statistical tests. One intervention found mixed results (positive and no effect)(Reference McNally, Anzman-Frasca and Bowman49), and one had no effects on healthier orders(Reference Nothwehr, Snetselaar and Dawson40).
The Fish and Chip Wholesaler Study encouraging fish and chip shop owners to offer and promote smaller portion sizes found increases in the number of small-portion meal orders from 14·2 % of total Fish & Chip orders before the intervention to 21·2 % post-intervention, with 20 % of surveyed customers indicating they had tried a smaller portion meal(Reference Goffe, Hillier-Brown and Hildred43). The Signposting to Healthy Meals did not have a comparator group and found no significant changes in order slips, although 34 % of customers who were aware of the signs said that these had impacted their order(Reference Nothwehr, Snetselaar and Dawson40). One intervention arm from the Fundraising Healthy Eating Scheme provided information on healthier items and a 15 % donation of the total bill value to the school wellness programme and recorded significantly increased healthy item orders compared with follow-up, but not the baseline period(Reference McNally, Anzman-Frasca and Bowman49). Only four surveyed customers (10·8 %) said they selected their option due to the promotion materials(Reference McNally, Anzman-Frasca and Bowman49).
Intervention barriers and facilitators
Recruitment of restaurants and takeaways can be challenging, with recruitment rates for businesses varying from 10(Reference Hillier-Brown, Lloyd and Muhammad39) to 100 %(Reference Nothwehr, Snetselaar and Dawson40) of those approached to participate in the evaluation. Four studies did not report recruitment rates(Reference Ma, Shive and Zhang13,Reference Nevarez, Lafleur and Schwarte41,Reference Fitzgerald, Kannan and Sheldon46,Reference McNally, Anzman-Frasca and Bowman49) . One research team was approached by a business owner wanting to conduct an intervention(Reference Chen, Carrillo and Kapp48). Identifying and visiting potential restaurants and takeaways several times before recruitment was reported as a strategy for successful recruitment(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) . One other study reported that a local restaurant association played a strategic role in recruiting restaurants(Reference Ma, Shive and Zhang13).
Five studies reported intervention fidelity, all achieving moderate to high fidelity(Reference Nothwehr, Snetselaar and Dawson40,Reference Goffe, Hillier-Brown and Hildred43,Reference Lee-Kwan, Goedkoop and Yong45,Reference Ayala, Castro and Pickrel47,Reference McNally, Anzman-Frasca and Bowman49) . Three studies reported barriers relating to difficulties engaging busy restaurant and takeaway staff with the training(Reference Ayala, Castro and Pickrel47), high turnover rates(Reference Ma, Shive and Zhang13) and trusting that staff would correctly deliver smaller portion sizes as intended(Reference Goffe, Hillier-Brown and Hildred43). Framing the intervention as ‘good customer service’ was reported to be potentially beneficial to serving staff implementing the intervention as intended(Reference Goffe, Hillier-Brown and Hildred43). Motivated staff, especially owners and managers, were key to keeping businesses engaged with and implementing the intervention(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Lee-Kwan, Goedkoop and Yong45,Reference Chen, Carrillo and Kapp48,Reference McNally, Anzman-Frasca and Bowman49) . Building good relationships with owners and involving them in decisions(Reference Lee-Kwan, Goedkoop and Yong45) as well as building strong partnerships(Reference Ma, Shive and Zhang13), with, for example, support from community groups(Reference Nevarez, Lafleur and Schwarte41,Reference Chen, Carrillo and Kapp48) or working with a wholesaler(Reference Goffe, Hillier-Brown and Hildred43), were also mentioned as facilitators.
Two studies reported that businesses are better engaged with intervention elements that were cheap, easy to implement and perceived as acceptable or less noticeable to clients, which included easy-to-implement intrusive interventions (e.g. changing cooking oil used, categorised as restrict choice)(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) . One study reported that a two-phase intervention where low-cost, low-burden intervention elements are implemented first whilst building a stronger rapport with business owners and managers before introducing higher-burden intervention elements was effective at keeping businesses engaged(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) . Conversely, worries about customer satisfaction, a lack of demand for healthier products and associated costs were common barriers(Reference Hillier-Brown, Lloyd and Muhammad39–Reference Bagwell42,Reference Lee-Kwan, Goedkoop and Yong45) . Six studies reported that an intervention’s economic impact is an important factor for business owners when considering whether to engage with interventions, primarily because of small restaurants’ and takeaways’ small profit margins and susceptibility to economic fluctuation(Reference Nothwehr, Snetselaar and Dawson40–Reference Lee-Kwan, Goedkoop and Yong45,Reference Chen, Carrillo and Kapp48) . Businesses were reported to be motivated by the potential (financial) benefit of an intervention(Reference Goffe, Hillier-Brown and Hildred43,Reference McNally, Anzman-Frasca and Bowman49) , positive feedback from clients(Reference Hillier-Brown, Lloyd and Muhammad39) and financial incentives (e.g. supplies, covering first stock)(Reference Lee-Kwan, Goedkoop and Yong45).
One main barrier to implementation was a lack of availability of healthier products from suppliers, either at all or at a comparative price point to regular versions(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) . This may be more common in more rural areas and outside of large cities(Reference Hillier-Brown, Lloyd and Muhammad39), and both businesses and customers in more affluent areas may be more willing to pay the extra costs involved in healthier options(Reference Bagwell42).
A web-based tool kit was a useful tool for dissemination of lessons learned and for potential participating businesses to learn more about the intervention(Reference Nevarez, Lafleur and Schwarte41).
Discussion
Summary of findings
Interventions to encourage healthy eating in small, independent restaurants and takeaways were mostly a complex mix of initiatives integrating business-level elements and consumer-focused components. Study quality was poor with limited quantitative outcome data, and it was not possible to conduct a meta-regression to identify effective components. Nonetheless, we found some narrative themes. Interventions focused at the customer level were mostly at the lower rungs of the Nuffield ladder. Enabling choice through introducing new and healthier menu items resulted in healthier items being ordered, with take-up varying from 11(Reference Chen, Carrillo and Kapp48) to 23 %(Reference Ayala, Castro and Pickrel47) of orders, but it was less clear whether these items substituted or supplemented other less healthy items. There was also a lack of evidence on whether the uplift in sales when new menu items were introduced could be sustained. Providing incentives (at the mid-point of the ladder) also resulted in a mix of positive results and no effect, with impact varying across product categories or comparison periods. Price promotions appeared to have some effect at least in the short term to boost sales of healthy products(Reference Lee-Kwan, Bleich and Kim44) but may not be a sustainable option for small businesses with tight margins. Most business-level interventions were classified as operating at mid-to-high rungs of the Nuffield ladder. Few interventions evaluated business-level outcomes, but almost all reported some positive effect including greater adherence to nutritional criteria or reduced salt content or weight of dishes, though quantitative evidence of effectiveness was scarce.
Strengths and limitations
We comprehensively searched relevant academic databases, including through multiple screeners and new software (e.g. Citation Chaser(Reference Haddaway, Grainger and Gray33)), building confidence in the scope and accuracy of our review. Our synthesis of studies provides the first overview to identify characteristics that are important for successful intervention design and implementation to improve food healthiness in small restaurants. We used the Nuffield intervention ladder to categorise intervention components; two included studies similarly used the Nuffield ladder to characterise intervention components(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) , whilst another study reflected on their results using the Nuffield ladder(Reference Goffe, Hillier-Brown and Hildred43), highlighting the relevance of this framework. Although we risk excluding studies by not conducting additional grey literature searches, higher-quality studies would likely be published in peer-reviewed journals.
The small number of heterogeneous and relatively low quality studies identified in the review is in itself a finding of interest but limits the potential generalisability of these results. Few studies had a randomised design, and it was not possible to directly compare interventions due to the heterogeneity of intervention components, study designs and settings. Furthermore, the narrow geographic range (urban areas in the UK and USA) of studies included means that findings may not translate to other food cultures (e.g. informal food economies). Additionally, our review may be limited by publication bias(Reference DeVito and Goldacre50), particularly considering most interventions described at least some positive effects.
Interpretation and comparison to existing literature
To the best of our knowledge, this is the first review that has focused specifically on small, independent restaurants and takeaways. The poor quality of available evidence and lack of impact evaluations in the out-of-home field have been reported in previous reviews(Reference Hillier-Brown, Summerbell and Moore15–Reference Gittelsohn, Lee-Kwan and Batorsky17,Reference Hillier-Brown, Summerbell and Moore51) . For example, in a review summarising interventions in food outlets in England, only twenty-one out of seventy-five interventions included evaluations of the impact or outcome of the intervention, and such evaluations were done to aid service delivery rather than research-led initiatives(Reference Hillier-Brown, Summerbell and Moore51). Challenges with data collection as reported by many studies in this review may impede rigorous evaluation.
Previous reviews also found that ‘simple’ environmental changes such as information provision and promoting existing healthy options are particularly common intervention strategies in community-based restaurants(Reference Espino, Guerrero and Rhoads16,Reference Gittelsohn, Lee-Kwan and Batorsky17) , consistent with our finding that easily implemented and cheap interventions are most acceptable to businesses. One reason the provision of information appeared to have mixed effects across studies in our review may be that customers often arrive at food outlets with pre-established order intentions; therefore, material to highlight new menu options and point-of-sale nutrition information may have a limited effect(Reference Ayala, Castro and Pickrel47). Additionally, nutrition labels may be ignored if the main eating motivation is hedonic and quick decisions are required(Reference Sanjari, Jahn and Boztug52). Indeed, research shows that taste is valued more strongly than health for restaurant meals(Reference Biermann and Rau53). Therefore, intervening to encourage healthier eating may be particularly challenging in these settings.
The studies that reported intervention fidelity found compliance to be moderate to high. However, engagement varied both between intervention venues and different intervention components, highlighting the need for a tailored approach. The relatively flexible format of some interventions – for example, where restaurants were given some liberty to choose which changes they would like to implement or adapt – meant that restaurants were able to select changes that best fit their context. Studies also reported that it is easier to engage participating businesses with interventions that are low-cost, low-effort and unlikely to be rejected or noticed by customers(Reference Hillier-Brown, Lloyd and Muhammad39,Reference Bagwell42) , and one intervention reported success using a staggered approach that slowly introduced more intrusive components(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) . Whilst rare, some higher-level interventions were identified, demonstrating these can be implemented. However, higher-level interventions requiring structural changes may be beyond the financial resources of small restaurants and takeaways. One strategy could be creating greater and equal opportunity for small restaurants and takeaways to access and serve healthier foods, which are often more expensive or only come in large package sizes unsuitable (and unaffordable) for small businesses(11), for example, through the provision of wholesaler subsidies for healthy foods for small restaurants and takeaways. An intervention providing discounts on healthy foods for small stores at wholesalers found this led to increased availability of healthier options(Reference Budd, Jeffries and Jones-Smith54).
Economic incentives or perceived economic viability of interventions was a main facilitator for engagement. Additionally, establishing rapport with owners may benefit recruitment(Reference Lee-Kwan, Bleich and Kim44,Reference Lee-Kwan, Goedkoop and Yong45) , a finding corroborated by previous evidence stressing the need for community outreach(11). Although studies reported which stakeholders were involved in the intervention in their backgrounds and methods, there was a lack of discussion and identification of the roles and benefits that other stakeholders played. Greater information about motivations and barriers to stakeholder involvement could improve the design and delivery of interventions in the future.
Implications for policy and research
Most studies relied on descriptive statistics and short follow-up periods and had no control or comparator sites, likely partly due to resource constraints and recruitment difficulties. More high-quality studies of interventions are needed, evaluating the longer-term impacts and sustainability of interventions using objective measures of outcomes (e.g. sales data). Investing in a new data system, or training staff on how to input data so that it is usable for the study, is advised if possible within resources available – improved sales data may also help inform businesses’ strategies, as well as be beneficial for researchers. Additionally, none of the interventions evaluated cost-effectiveness (see also (Reference Hillier-Brown, Summerbell and Moore15)). Making the best use of available resources is crucial considering the economic constraints of many small restaurants and takeaways. Whilst none of the included studies mentioned that any of the included restaurants and takeaways offered online food deliveries, if online deliveries are offered, this could limit the exposure to some intervention components, particularly marketing materials in-store. Given the growing size of the online food delivery sector(55), future interventions and research should consider the interaction between in-store and the growing online food delivery market.
Policymakers who want to work with small restaurants and takeaways should be mindful of potential resource constraints and adopt flexible approaches with scope for restaurants to tailor interventions to their needs. Partnering with other stakeholders such as local business associations as well as building rapport with restaurant owners can facilitate recruitment. In addition to drawing on the findings from our review that has systematically appraised the evidence base of interventions in small restaurants, policymakers who want to work with small businesses to make healthier changes should consider recommendations from existing toolkits on how to work with small, independent restaurants and takeaways(11,56) .
The majority of the interventions included in this review were conducted in areas broadly described as low-income or spanning multiple areas of deprivation. However, there was very little reporting on the impact the interventions may have had on reducing health inequalities. Most intervention elements were classed as belonging to the lower levels of the Nuffield ladder that are seen as less intrusive and require more agency and therefore are less likely to reduce health inequalities. In the future, researchers should consider reporting on neighbourhood levels of deprivation or collecting consumer demographic information to better assess how the interventions of interest might impact health inequities.
Conclusion
Interventions to encourage healthy eating in small, independent or local restaurants and hot food takeaways report mostly limited positive effects. The thirteen included interventions reflect a narrow set of countries (conducted in the USA or the UK) and over the past 20 years (published between 2004 and 2020). Most interventions used less intrusive strategies (e.g. providing information, enabling choice), although we found that more intrusive interventions can be acceptable to business owners if they are inexpensive, low-effort and not perceived as threatening customer satisfaction. Almost all interventions targeted the behaviour of both customers (e.g. menu labelling) and restaurant staff (e.g. cooking practices). However, the small number and poor quality of included studies hinder inference. More high-quality studies of interventions with objective purchase and consumption measures are needed to inform substantive policy-led actions.
Supplementary material
For supplementary material accompanying this paper, visit https://doi.org/10.1017/S1368980025000035.
Acknowledgements
None.
Authorship
L.B.: conceptualisation, project administration, supervision, methodology, investigation, formal analysis, writing – original draft, writing – review and editing; C.J.: methodology, investigation, formal analysis, writing – original draft, writing – review and editing; H.F.: writing – review and editing; N.R.: methodology; S.A.J.: conceptualisation, methodology, writing – review and editing; R.P.: methodology, writing – review and editing.
Financial support
L.B. is funded by the National Institute for Health and Care Research (NIHR) Applied Research Collaborations Oxford. R.P. is funded by a Royal Society and Wellcome Trust Sir Henry Dale fellowship (222 566/Z/21/Z) that also supports C.J. C.J. receives additional support from Livestock, Environment and People (205 212/Z/16/Z). H.F. is funded by the COPPER project (NIHR Public Health Research Programme (NIHR133887)). S.A.J. is funded by NIHR Applied Research Collaborations Oxford and by the NIHR Oxford Biomedical Research Centre. R.P. is also supported by NIHR Oxford Health Biomedical Research Centre. The funders had no role in the study design, data collection, analysis or interpretation. The views expressed are those of the author(s) and not necessarily those of the funders. For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.
Competing interests
There are no conflicts of interest.
Ethics of human subject participation
Not applicable.
Open access
