Obesity is a complex and ongoing public health problem( Reference Flegal, Kit and Orpana 1 ). Persistent daily energy imbalances have contributed to population-level weight gain over the past 40 years( Reference Cutler, Glaeser and Shapiro 2 – Reference Rehm, Matte and Van Wye 4 ). Indeed, the average American diet now includes 200–300 kcal/d (837–1255 kJ/d) more than 30 years ago( Reference Finkelstein, Ruhm and Kosa 5 ). Energy from the consumption of sugar-sweetened beverages (SSB) is the largest contributor to this increase( Reference Finkelstein, Ruhm and Kosa 5 , Reference Kaiser, Shikany and Keating 6 ). It is not surprising then that the consumption of SSB is linked to weight gain( Reference Malik, Pan and Willett 7 – Reference Mozaffarian, Hao and Rimm 9 ).
Fast-food restaurants are a prominent source of SSB energy( Reference Poti, Slining and Popkin 10 , Reference Powell and Nguyen 11 ). This may be in part due to the popularity of fast-food combination meals, which include beverages( Reference White 12 ), and the prominence of soft drinks on fast-food menus( Reference Harris, Schwartz and Brownell 13 ). Moreover, many fast-food restaurants offer free refills which may encourage additional consumption of energy-dense beverages. Some fast-food restaurants have self-serve beverage stations that allow for free beverage refills, while other restaurants offer free refills on beverages but require customers to request a refill from restaurant staff.
There is very little research on consumer characteristics associated with SSB purchases at fast-food restaurants or on how different restaurant policies, such as the availability of refills, affect consumer choice( Reference Wansink 14 ). To our knowledge, the present paper is the first that examines the odds of refilling a beverage in a fast-food restaurant, the contribution of self-serve refill stations to beverage exposure, and the effect of refills on the energy and volume of beverages obtained.
Methods
The data utilized for the present study were collected as part of a larger project evaluating New York City’s (NYC) Sugary Drinks Portion Cap Rule( 15 ). Point-of-purchase surveys and receipts were collected from fast-food customers in the NYC metropolitan region using a customer intercept protocol. Data were collected at NYC and neighbouring Newark and Jersey City, New Jersey locations of the most common restaurant chains in Manhattan (McDonald’s, Burger King, Subway, Wendy’s and KFC) over several months in 2013 and 2014( Reference Elbel, Mijanovich and Dixon 16 ). Restaurants were surveyed on weekdays during lunch (11.30–14.30 hours) and dinner (16.30–19.30 hours) periods. The Institutional Review Board of New York University Medical Center approved the study.
In July 2014, just after data collection was completed, researchers called each surveyed restaurant and asked two questions. The first was ‘Does the restaurant offer free beverage refills of in-store purchases?’ When restaurants reported offering refills on beverages we asked ‘Do customers refill their own drink from a self-serve refill station or are refills provided behind the counter?’ These data were collected later because the focus of the original study was not on the offering of free beverage refills.
The outcome variables for our analyses included: (i) a binary indicator for whether a customer ordered a beverage, including SSB (e.g. soft drinks, low-calorie drinks, sports drinks, lemonade), diet beverages, coffee, bottled water and unsweetened tea; (ii) a binary indicator for whether a customer reported refilling his/her beverage; (iii) estimated energy content of the beverage (‘beverage calories’, i.e. kilocalories; 1 kcal=4·184 kJ); and (vi) estimated volume of the beverage (‘beverage ounces’, i.e. US fluid ounces; 1 US fl. oz=29·5735 ml) (obtained from the receipt). We estimated beverage energy and volume using information listed on each restaurant chain’s website. We doubled these estimates for refilled beverages.
Our analyses focus on the complex relationship of the above outcomes with the type of refill available (self-serve or not) and ordered beverage size. The primary predictor was a categorical variable with three levels: (i) no free refills available; (ii) free refills were available without a self-serve station; or (iii) free refills were available with a self-serve station. We used separate logistic regression models to estimate the odds of ordering any beverage for the full sample and the odds of refilling a beverage (based on responses to the survey question ‘Did you refill your cup while in the restaurant?’) among only the sample who ordered a refillable drink (e.g. non-pre-packaged beverages including soft drinks, tea, lemonade). We estimated meal energy, beverage energy and beverage volume using ordinary least squares for the sample of respondents who ordered a refillable soft drink. The primary treatment variable in the linear regression models was an interaction of the availability of free refills and the presence of self-serve stations in the restaurant. All regression models included controls for consumer and meal characteristics including gender, race/ethnicity, age, education, employment status, meal time (lunch or dinner), meal type (‘to go’ or ‘to stay’; à la carte or combo meal), average fast-food eating frequency per week, restaurant chain, state and survey period. We clustered all standard errors at the restaurant chain level. Analyses were done using the statistical software package Stata version 13.
Results
We surveyed customers at sixty-one fast-food restaurant locations in NYC and thirty-seven in New Jersey.
In Table 1, we report characteristics for all survey respondents, for the sub-sample of respondents who ordered a refillable drink and for the sub-sample of respondents who reported refilling their drink. We collected receipts from 11795 adults, of whom 3541 (30 %) ordered a drink. Respondents were evenly distributed between males (53 %) and females (47 %). African American (43 %) was the most frequently race/ethnicity reported, followed by Hispanic (31 %), white (15 %) and other race/ethnicities (10 %). The age distribution across respondents was fairly uniform. Approximately 58 % of the sample had a high-school degree or less, almost 65 % were employed and 80 % were surveyed during lunch time. Approximately 35 % ordered their meal to stay in the restaurant and 18 % purchased a combination meal.
Table 1 Descriptive statistics for fast-food consumers from ninety-eight fast-food restaurants in New York City, and Newark and Jersey City, New Jersey, 2013–2014
The 30 % of respondents who ordered a refillable soft drink were different from the rest of the sample. Proportionally, fewer females than males, and more whites and fewer respondents reporting other races compared with blacks and Hispanics, ordered a soft drink. Soft drink purchasers were more likely to consume their meal in the restaurant. More lunch-time purchases included a refillable drink than during dinner time. Meals with a beverage were more likely to have been a combination meal compared with all meals. Proportionally more purchases included a refillable drink at Burger King, KFC and Wendy’s compared with McDonald’s and Subway.
There were also differences between survey respondents who did and did not refill their soft drink. Fewer blacks refilled their soft drinks than Hispanics and respondents of other races. Unemployed respondents were more likely to get a refill. Respondents who refilled their beverage were more likely to eat in the restaurant. A greater portion of respondents who ordered a child or value size soft drink reported getting a refill compared with respondents who had larger drinks. Respondents at fast-food restaurants in New Jersey were more likely to get refills compared with respondents in NYC. Lastly, there was a difference in the distribution of refills by restaurant chain; customers at Burger King and KFC were more likely to get refills, while McDonald’s and Wendy’s consumers were less likely to refill their beverage, compared with Subway customers. There were no statistically significant differences in beverage refills between groups within gender, education or the time of day the purchase was made. Refilled beverages were mostly SSB (90 %). The remaining refilled beverages include diet beverages (9 %) and other (1 %; i.e. juice drinks, unsweetened tea).
Results from the multivariable logistic regression models are shown in Table 2. Among the full sample, African Americans (adjusted OR (aOR)=0·68; 95 % CI 0·59, 0·78) and respondents reporting other race (aOR=0·61; 95 % CI 0·57, 0·64) had lower odds of ordering any beverage relative to whites. Respondents aged 65 years or older had increased odds of ordering a beverage (aOR=1·56; 95 % CI 1·17, 2·09) relative to respondents aged 18–24 years. Among respondents who had a soft drink, there were no differences by age or race in the odds of getting refills. Customers who ordered a combination meal had considerably higher odds of having a beverage (aOR=28·39; 95 % CI 8·84, 91·14), but not of refilling their soft drink. Further, respondents who ate at the restaurant had higher odds of both obtaining a beverage (aOR=2·61; 95 % CI 2·10, 3·24) and of refilling the beverage (aOR=4·45; 95 % CI 2·48, 7·99) compared with those who took their meals to go. Although self-serve refill kiosks were not associated with the customer ordering any beverage, we did find among respondents with a beverage that the presence of a self-serve refill kiosk was highly associated with reporting having a refill (aOR=7·37; 95 % CI 3·91, 13·89). Size of beverage was related to the odds of having a refill. In particular, relative to customers who ordered the smallest soft drink size, respondents who ordered small (aOR=0·46; 95 % CI 0·38, 0·57), medium (aOR=0·34; 95 % CI 0·17, 0·71) and large (aOR=0·39; 95 % CI 0·16, 0·95) beverages all had lower odds of having a refill. In contrast, those aged 25–39 years (aOR=0·77; 95 % CI 0·60, 0·98), 50–64 years (aOR=0·63; 95 % CI 0·43, 0·92) and 65 years or older (aOR=0·40; 95 % CI 0·23, 0·68) had lower odds of refilling a soft drink, compared with those aged 18–24 years. Finally, relative to customers at Burger King, customers at KFC (aOR=2·18; 95 % CI 1·76, 2·71) had higher odds of getting a refill while Wendy’s customers had lower odds of obtaining a refill (aOR=0·41; 95 % CI 0·36, 0·45).
Table 2 Adjusted odds of obtaining a beverage or beverage refill at ninety-eight fast-food restaurants in New York City, and Newark and Jersey City, New Jersey, 2013–2014Footnote *
aOR, adjusted OR; ref., reference category.
* Please note that the odds ratios presented in each column are estimated using different samples and are thus not directly comparable.
Based on our estimates, free soft drink refills were associated with increased exposure to beverage and total meal energy. Even after controlling for customer demographic and meal characteristics, we found that soft drink refills were associated with a customer obtaining an average additional 29 beverage ounces (858 ml) and 250 beverage calories (1046 kJ; Table 3). We found that free refills were associated with 330 additional total meal calories (1381 kJ; data not shown). This suggests customers who refilled their beverages did not offset the additional beverage energy with lower-energy food orders.
Table 3 Regression-adjusted mean beverage volume and energy content, for the sub-sample of respondents who obtained a refillable beverage, at ninety-eight fast-food restaurants in New York City, and Newark and Jersey City, New Jersey, 2013–2014
Adjusted means show the expected number of beverage ounces and beverage calories obtained for each combination of size and refill, with the assumption that the entire sample either refilled or did not refill a beverage of each available size. Results calculated using ordinary least-square regression models with the following covariates: presence of a self-serve station in the restaurant, the offering of free refills at the restaurant, respondent gender, race, age, education level, employment status, whether meal was had for lunch or dinner, the location the meal was had, whether a combination meal was ordered, the size of the beverage, the restaurant chain, state where the survey was collected, survey period and frequency of fast-food visits.
* Estimated volume of the beverage (US fluid ounces; 1 US fl. oz=29·5735 ml).
† Estimated energy content of the beverage (kilocalories; 1 kcal=4·184 kJ).
Discussion
Our findings suggest that demographic, consumer and restaurant characteristics are all associated with fast-food restaurant customers obtaining soft drink refills. In particular, the availability of refills from self-serve beverage stations was associated with significantly larger odds of a customer refilling his/her drink. While this is true, only a small percentage (8·7 %) of fast-food customers with a soft drink reported refilling their beverage. An even smaller percentage (2·6 %) of all fast-food customers refilled their beverage. Yet, getting a refill was associated with customers obtaining substantially more energy from SSB even after controlling for the size of the beverage.
The current study has several limitations. First, we do not know how representative this sample is of fast-food consumers because our street intercept sampling strategy is subject to non-random selection. Unfortunately, we do not have the response rate for the survey. A previous study using street intercept surveys reported a 60 % response rate( Reference Dumanovsky, Huang and Nonas 17 ). Second, we surveyed only customers arriving on foot at restaurants within three adjacent Northeast cities. We recognize that, outside this region, many fast-food eaters obtain their meal from drive-through windows. Regional differences in customer characteristics mean our results may not generalize to non-urban consumers. Third, we offered survey respondents a $US 2 incentive for participation, which could have affected the participants’ purchase decisions. Fourth, we do not have an objective measure for how much of each beverage was actually consumed. Thus, we limit our conclusions to obtained, rather than consumed, beverages. Fifth, we assume that a beverage was completely empty when refilled and that the refill completely filled the cup. These assumptions could have resulted in an overestimation of exposure to fast-food beverage energy and volume if the beverage was only partially consumed or refilled. Alternatively, our results may underestimate the true association if the beverage was refilled multiple times. Results from a sensitivity analysis in which we assumed that customers refilled only half of their beverage still found that the beverage refills contributed a substantial and statistically significant number of beverage calories (134 or 561 kJ more) and ounces (14 or 414 ml more; data not shown). Sixth, we did not collect information on stores’ refill policies until after data collection occurred. We think it is unlikely that stores changed their policy during the interim period. Status quo bias, financial costs and space constraints of changing beverage dispensing systems reduce the likelihood of restaurants changing their beverage refill policy during the year and a half between the first data collection period and when we re-contacted stores. Regardless, it is possible that some stores may have changed their refill policy or that the employees we spoke with inaccurately reported the availability of beverage refills at their restaurant. Thus, our estimates could be subject to measurement error bias.
Our findings suggest that one novel opportunity to reduce SSB consumption is to restrict the availability of self-serve beverage stations at fast-food restaurants. France recently passed such a proposal in fast-food chains and restaurants( Reference O’Connor 18 , Reference Farand 19 ). Based on our findings a similar domestic restriction could lead to reductions in the energy, volume and grams of sugar obtained from beverages among fast-food customers who order refillable beverages.
Acknowledgements
Acknowledgements: The authors thank Olivia Martinez BS, of New York University School of Medicine, for informal feedback. Financial support: This research was funded by the National Institute of Health/National Institute of Diabetes and Digestive and Kidney (grant number R01 099241); the National Institute of Health/National Heart, Lung, and Blood Institute (grant number R01 HL095935); the New York State Health Foundation (grant number 12-01682); and the Robert Wood Johnson Foundation (grant number 70823). The funding sources played no role in the study design, collection, analysis or interpretation of data, in the writing of the manuscript, or in the decision to submit the manuscript for publication. Conflict of interest: None. Authorship: A.B. contributed to the analysis, interpretation of results, writing, and conceptualization of the study and research design. J.H.C. contributed to the analysis, interpretation of results, writing, and conceptualization of the study and research design. B.E. obtained financial support for the study and contributed to conceptualizing study design, interpretation of results and writing; and had full access to all of the study data and takes responsibility for the integrity of the data and the accuracy of the data analysis. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the New York University Medical Center’s Institutional Review Board. Verbal informed consent was obtained from all subjects/patients.
