Gender and age are associated with healthy food purchases via grocery voucher redemption

Introduction: Grocery vouchers that specifically target foods associated with reduced cardiovascular disease (CVD) risk result in increased consumption of those foods. In regions with disproportionately high CVD rates, there is little research concerning the impact of vouchers on purchases of risk-reducing foods when there are no restrictions placed on grocery voucher redemption. Since many food assistance programs place few restrictions on type of foods that can be purchased, identifying demographic factors associated with purchasing habits is a prerequisite to promoting healthy eating. The purpose of this study was to determine the associations of age, gender, education and income level with purchasing of healthful foods through the use of a grocery voucher in a rural food desert (poverty rate of ≥20% and ≥33% of residents living >16 km from a large grocery store) with high rates of chronic disease. Methods: The effectiveness of an intervention that included a media campaign, a $5 grocery voucher, local heart healthy food branding and a grocery store event was tested. Brief nutritional articles were published in both local newspapers during four consecutive weeks. These articles explained the physiological actions of healthy foods and listed a health-promoting recipe. During the fourth week of the media campaign, a voucher for a $5 grocery gift card redeemable at one of either community grocery stores was also printed in both local newspapers. In each store, foods that are known to be associated with a reduced risk of CVD were marked with a blue logo. Participants (N=311) completed a questionnaire that assessed demographics and usual servings of fruits, vegetables and grains. Participants received a $5 grocery card and a list of labelled foods. Returned grocery receipts were stapled to the questionnaires to analyse the relationship between demographics and food choices. Results: Participants who bought at least one labelled food item were older (M=48.5, SD=14.7) than those who did not buy any of these items (M=42.3, SD=16.4; p=0.0008). There was a significant association between labelled food purchases and gender, with 47% of male participants purchasing at least one labelled food item compared with 63% of females in the study (p = 0.008). There were no significant associations between purchase of labelled food items and either education or income. The significant


Introduction
Individual and family food preferences are determinants in food purchases, but many other factors, including age, gender, educational level and household income, also influence grocery-buying habits 1,2 .Strategies that promote the purchase of foods associated with reduced cardiovascular disease (CVD) risk should be developed in the context of these factors.Grocery vouchers and educational strategies that specifically target foods associated with reduced CVD risk result in increased consumption of those foods 3 .However, there is little research concerning the impact of grocery vouchers on purchases of these foods when there are no restrictions placed on the types of food that can be purchased.Many food assistance programs place very few restrictions on the types of food that can be purchased.For this reason, identifying demographic factors associated with the types of food that individuals purchase is a prerequisite to developing successful interventions aimed at increasing consumption of healthy foods in regions with disproportionately high rates of CVD.The purpose of this cross-sectional, pilot study was to determine the associations of age, gender, education and income level with purchasing of foods associated with lower CVD risk through the use of a grocery voucher.This pilot study was conducted following a community-based intervention to promote healthy food purchases in a rural food desert with disproportionately high rates of poverty and CVD.A food desert is defined as a lowincome census tract (having either a poverty rate of ≥20%, or a median family income ≥80% of the area's median family income) where ≥33% of the population resides >16 km from a supermarket or large grocery store 4 .

Background and significance
Grocery store marketing strategies influence food purchases 5 .Point-of-purchase marketing, degree of variety of fruits and vegetables, display stands in prominent locations and staff promotion of healthful foods are all successful strategies to increase purchasing of foods associated with decreased chronic disease risk [6][7][8][9] .Financial incentives (ie food vouchers) provided to low-income households that can only be used to purchase specific grocery items (eg fruits, vegetables, fruit juices) result in greater purchases of those items [10][11][12][13] .In contrast, financial incentives without concurrent promotion of healthful foods result in an overall rise in food expenditure but not an increase in purchases of specific food items known to decrease the risk of chronic disease 14 .Financial incentives were combined with a location-based marketing community intervention to promote the purchase of healthful foods but grocery voucher redemption was not restricted to certain food items.
The current US Department of Agriculture Dietary Guidelines include recommendations to make half of meals fruits and vegetables, choose whole grains for half of all consumed grains, and drink low-fat or skim milk 14 .Guideline-based interventions may be ineffective if socioeconomic factors and population characteristics that influence dietary habits are not accounted for during development of these interventions.Older adults and women are more likely to have higher quality diets than younger adults and men 15 .There is also evidence that men and women respond differently to marketing strategies aimed at increasing nutritional knowledge and purchases of foods associated with reduced CVD risk 16 .It is important to evaluate the participant characteristics associated with the effectiveness of marketing strategies because a healthy diet has a greater impact on decreasing mortality rates in men than in women 17 .
The ability to recognise the health-related impact of dietary choices, degree of motivation to improve diet and physiological changes in appetite all differ throughout the lifespan.Advancing age is associated with healthier dietary habits.Older adults are more likely to report healthy behaviors, including adherence to a diet of foods associated with lower CVD risk, than younger adults [18][19][20][21] .In spite of this, overweight and obesity in older adults is steadily rising 22 .Older adults who adhere to the Dietary Guidelines for Americans are also less likely to experience disabilities in activities of daily living (eg dressing oneself, getting in and out of bed), instrumental activities of daily living (eg doing household chores, preparing meals), leisure and social activities (ie attending social events and performing leisure activities), and physical activities (eg walking, kneeling, lifting) than older adults who do not adhere to the dietary guidelines 23 .Given this reality, age-appropriate interventions aimed at improving dietary behaviors are more likely to be successful than interventions that do not account for agespecific responses to lifestyle behavior modification strategies.
Interventions that aim to increase consumption of foods associated with reduced CVD risk should also include genderspecific strategies 24 .There are gender-related differences in body composition, appetite, satiety and diet quality throughout the lifespan 25,26 .Women tend to have better quality diets, experience lower hunger sensations and higher post-meal satiety than men 15,18,26,27 .There are also genderspecific differences in the brain's response to external stimuli (ie pictures of food) related to appetite and food 28 .Female gender is also associated with a greater likelihood of recognising food as a means to influence health outcomes 29 .College-educated individuals tend to eat healthier and low parental education levels are associated with poor-quality diet in children 30,31 .However, the relationship between educational level and diet quality is often mediated by income level and attitude towards healthy eating 32 .Measures of selfefficacy and healthy lifestyle planning to improve diet are similar across educational levels, suggesting these lifestyle behavioral goals are not limited to college-educated individuals 33 .
Food cost is a barrier to healthy eating, particularly for those living in lower income households.Poverty is generally associated with poor dietary habits and, consequently, increased obesity and chronic disease risk  . Howeer, some contrasting studies have shown only marginal or no significant relationship between income level and consumption of foods associated with lower CVD risk 18,37 .Rising food cost can be an even greater barrier to healthy eating for individuals with low-incomes living in rural food deserts 38 .The increasing cost of feeding families is most concerning in areas where a significant percentage of people receive Supplemental Nutrition Assistance Program (SNAP) funds, a federally funded, state-administered nutrition support program.Nearly 47 million low-income Americans depend on SNAP funds to meet their nutritional needs.Barriers to healthy eating among low-income households include limited access to healthy foods, inadequate SNAP benefits, the cost of nutrient-rich foods and environmental factors such as rurality.
A Central Appalachian food desert, Breathitt County, Kentucky, which has disproportionately high rates of poverty and chronic disease, was chosen as the location to test this pilot community intervention.Breathitt is one of 50 Kentucky counties where many families consistently experience food insecurity (ie the state or risk of being unable to provide food for oneself or one's family).Food insecurity in Appalachia is associated with higher rates of chronic diseases, obesity, poor management of health conditions and depression [39][40][41] .The poverty rate in Breathitt County, which is one of the nation's top 25 poorest counties, is 33.8%, compared to 16.9% for the state of Kentucky and the national rate of 15.1% 42 .According to the 2010 census, 54.6% of Breathitt County households have less than $35,000 annual income (Table 1).The 2006-2010 average household income in Breathitt County was $19,906.In December 2011, more than 6000 of the county's 13 878 adult residents received Supplemental Nutritional Assistance Program or Kentucky Transitional Assistance Program funds 43 .While cost can be a barrier to healthy eating, 26% of individuals in Breathitt County with limited access to healthy foods have middle to high incomes, indicating that other factors beyond financial resources may inhibit health-promoting dietary habits 44 .Breathitt ranks as one of the least-healthy Kentucky counties, with a cardiovascular mortality rate of 540 per 100 000 population, compared to 409 for the entire state and a national rate of 326 45 .
In rural Appalachia, the traditional diet consists of a high amount of calorie-dense meats and starches that are often fried or have added animal fats.A mixed methods convergent parallel design study was conducted prior to development of this community marketing event aimed at promoting purchases of CVD risk-reducing foods.During this earlier study, a nominal group process was used to determine factors that affected adherence to a risk-reducing diet in individuals living in this rural food desert.Food cost and limited access to healthy foods were factors affecting the potential to adhere to a CVD risk-reducing diet.Younger participants and those with lower incomes were more likely to identify food cost as a barrier to adherence than those who were older with higher incomes.Women were more likely than men to report an understanding of the health impact of certain foods as motivation to adopt and adhere to a risk-reducing diet 46 .Based on these findings, a media campaign/financial incentive was developed to promote the purchase of CVD riskreducing foods.

Methods
This community-based pilot study tested the effectiveness of a culturally appropriate, community-based campaign aimed at promoting purchases of food items associated with reduced CVD risk.A partnership was established with local US Department of Agriculture Cooperative Extension Office Family and Consumer Science (FCS) specialists who delivered this intervention, which involved a print media campaign, a $5 grocery store voucher, local heart-healthy food branding and a community grocery store event.Brief nutrition articles were published in both local newspapers during four consecutive weeks in July 2012.The newspapers have a combined reach of 7000 readers.These articles, written at a literacy level appropriate for readership, explained the physiological actions of healthy foods and listed a healthpromoting, minimal-ingredient recipe using foods readily available in local grocery stores.A 'Healthy Eating for Men' article that specifically described the impact of diet on men's health was written in each week's newspaper.During the fourth week of the media campaign, a voucher redeemable for a $5 grocery gift card was also printed in both local newspapers.Individuals were instructed to bring the grocery voucher to either of the county's two grocery stores on a single day in August 2012.One of the grocery stores is a national, full-service chain and the other is a national discount chain with their own exclusive and national brands.
In each store, foods that are known to be associated with a reduced risk of CVD were marked with a blue logo.These labelled items included whole grains; fresh and frozen fruits and vegetables; canned and dried legumes; skim dairy; omega-3 rich fresh, frozen and canned fish; unsalted almonds and walnuts and olive oils (not including oil blends).At both stores, brand name and generic options were available for all food product categories, with the exception of fresh produce.A table with a registered nurse and two other research personnel was set up near the entrance/exit of each of the two grocery stores, at locations that were also near the stores' produce sections.When individuals brought their grocery voucher to either store, they were asked to complete a brief questionnaire that assessed demographics and usual servings of fruits, vegetables and grains.After completing the questionnaire, participants received a $5 grocery card that could be used to purchase groceries only on the day of the study.Participants were also given a list of logo-labelled foods and a reusable grocery bag printed with the logo when they entered the store.The labels were affixed as permanent displays in the stores and the reusable bags served as prompts to purchase these foods during future visits to the grocery stores.Participants were instructed that if they returned their grocery receipts to the table after check-out, they would receive free 'healthy eating' items (ie two food preparation tools and a cutting mat).Returned grocery receipts were stapled to the questionnaires to analyse the relationship between demographics and food choices.A second table was staffed by a Family and Consumer Science specialist from the local Cooperative Extension Office, who provided prepared samples of some of the labelled CVD-risk reducing foods free to all customers.A well-known local community member was also available in each store to assist individuals in locating labelled food items and to ensure adequate stocking of these food items.All respondents (N=311) returned their grocery receipts following check-out.

Measures
Sociodemographic factors: These factors included age in years, gender, education and income.Education was assessed by participant response to the highest level of education completed with the possible response options of 'Grade school', 'High school', 'Some college', 'College degree' and 'Masters or doctoral degree'.Household income per year was assessed by asking participants to select one of these options: <$20,000, $20,000-30,000, $30,001-40,000, $40,001-50,000, $50,001-60,000, $60,001-70,000 and >$70,000.
Purchase of food items associated with CVD riskreduction: Each labelled purchased item from any of the food categories (ie whole grains; fresh and frozen fruits and vegetables; canned and dried legumes; skim dairy; omega-3 rich fresh, frozen and canned fish; unsalted almonds and walnuts and olive oils) was recorded as a single purchase using the USDA grocery taxonomy.The number of labelled, CVD risk-reducing food items purchased ranged from 0 to 6. Due to almost half of the respondents purchasing zero designated food items (42%), the responses were dichotomised to: 'No purchase of the risk-reducing food items' and 'Purchase of at least one of the risk-reducing food items.'

Data analysis
Descriptive analysis, including means and standard deviations or frequency distributions, was used to summarise the study variables and check for missing or out-of-range values.Due to more than half of the participants completing either in 'Grade School' or 'High School' education (51.8%), responses were dichotomised to 'At most high school' and 'At least some post-secondary education'.Two-group comparisons of study variables by purchase of one or more designated food items were accomplished using χ 2 tests of association, two-sample t-test or Mann-Whitney U-test, as appropriate.Multiple logistic regression was used to assess whether sociodemographic factors predicted the purchase of at least one of the designated food items.Variance inflation factors were used to assess whether multicollinearity was present in the model.Data analysis was conducted using Statistical Analysis Software v9.3 (SAS Institute, www.sas.com); an alpha level of 0.05 was used throughout.

Ethics approval
The study was approved by the University of Kentucky Medical Institutional Review Board Approval #12-0362-P2H.

Results
The average age of the 311 participants was 45.8 (SD=15.7); the range of ages was 16-89 years.Thirty percent of the respondents were male and 52% had at most high school education.Forty-three percent had annual household income of <$20,000, which is similar to county demographics.One hundred and thirty-six (46.4%) respondents spent less than $10.Sociodemographic and total amount spent data are presented in Table 2.As shown in Table 3, there was a significant bivariate association between purchase of the labelled food items and age: participants who purchased at least one labelled food item were older (M=48.5,SD=14.7)than those who did not buy any of these items (M=42.3,SD=16.4;p=0.0008).Also there was a significant association between labelled food purchases and gender, with 47% of male participants purchasing at least one of labelled food items compared with 63% of females in the study (p=0.008).There were no significant associations between purchase of labelled food items with either education or income.
Sociodemographic factors were assessed as potential predictors on purchase of at least one of the labeled food items using multiple logistic regression (Table 4).Consistent with the bivariate analysis, the significant predictors in this model were age (p=0.003) and gender (p=0.01).The results of the model suggest that for every 10 year increase in age, there was a 29% increase in the likelihood that at least one designated food item would be purchased.Male participants were 48% less likely than female participants to purchase at least one designated food item.The variance inflation factors for this model were all less than 1.4, suggesting little evidence of multicollinearity due to correlations among the predictor variables.
There was a significant relationship between spending less than $10 (yes or no) and purchasing at least one labelled item (p<0.0001).Of those respondents whose total amount spent was less than $10, 46% purchased at least one labelled food item and 54% didn't purchase any of the labelled food items.
Of those respondents who spent $10 or more, 75% purchased at least one labelled food item and 25% didn't purchase any labelled food items.

Discussion
This study examined the associations of age, gender, education and income level with purchasing at least one healthy food item following an intervention that included a grocery voucher distributed via newspapers in a rural food desert with high rates of poverty and CVD.In this study, the majority of participants (58%) purchased at least one of the labelled foods associated with CVD risk reduction.Age and gender were associated with purchases of these food items, with older adults and women being more likely to purchase labelled food items with the grocery voucher than younger adults and men.Educational level and income were not associated with purchasing of labelled foods in this study.This lack of association may be due to a relatively homogenous sample from this county, one of the nation's poorest, with a lower percentage of high school graduates when compared to the overall population in Kentucky (65% vs 82% 42 ).
Age and gender should be considered when developing strategies to improve dietary habits of individuals living in the Central Appalachian rural food deserts.Young adults tend to be more technologically proficient and are more likely to respond to healthy eating strategies that use social media (eg Twitter and Facebook) and electronic devices (eg PDAs, mobile phones and cameras) rather than print media 47,48 .
Young adults are also more like to alter poor dietary habits after receiving education about the health effects of foods known to contribute to chronic disease and social normative messages about healthy eating 49 .The media campaign consisted of only print media, and future interventions should be broadened to include both social media and more technologically advanced modes of delivery.Information was also disseminated about the impact of foods associated with positive health outcomes but did not address the impact of those foods associated with negative health outcomes.This region has disproportionately high rates of CVD, type 2 diabetes mellitus and metabolic syndrome.The impact of foods that are traditional staples in the Appalachian diet (eg preserved or processed meats, refined carbohydrates and animals fats) should be included in future nutrition educational interventions.
These findings are also noteworthy because of the current national and state-level debate of whether restrictions should be placed on the types of foods that can be purchased with SNAP funds 50 .Individuals living in lower-income households may be more apt to shop at discount grocery stores but their attitudes towards healthy eating do not differ from those who shop at higher-cost supermarkets 51 .The intervention of a brief media campaign, grocery voucher and point-of-purchase information resulted in a majority of participants purchasing at least one healthy food item, irrespective of income or educational status.
The recent increase in food costs has resulted in the realisation that SNAP benefits may be inadequate to meet those needs and there is current discussion that these benefits should be increased 52 .However, concern about the negative impact of unhealthy food choices has raised the question as to whether SNAP-funded purchases of high-calorie, poornutrient foods and beverages should be restricted.While most agree that additional SNAP funding is necessary to meet the needs of low-income families, public support is mixed for policies that restrict the purchase of food products known to contribute to chronic illnesses 53,54 .The rationale for restricting SNAP purchases to only certain foods in the effort to decrease the risk of chronic disease is predicated upon two unfounded assumptions: (1) that SNAP participation is associated with risk of obesity; and (2) that SNAP participants will consume less of the foods low in nutritional quality if purchasing restrictions are in place 55,56 .Results of this study shows that income was not a factor in whether participants purchased healthy or unhealthy foods with a grocery voucher.Many of the counties located in rural Appalachia are among the poorest in the nation and consequently many rural Appalachians consume poor-quality diets 57 .Poverty-related food insecurity in low-income regions like Appalachia has also resulted in a tradition of food preservation by canning, drying, brining or sugar curing, techniques that increase the glycemic index of foods 58,59 .This predominant tradition influences dietary choice even in the current age of supermarkets and fast food.
A diet that consists of foods associated with reduced risk of CVD is important across the lifespan.Because age and gender were associated with purchasing of healthy foods items via a free grocery voucher, future programs in this impoverished region should be tailored to include gender-and age-related marketing strategies.Children who are taught to make healthy food choices are more likely to maintain those healthy eating habits into adulthood and, as they age, have better health outcomes than individuals who consume more foods associated with poor health outcomes [60][61][62][63] .Although changes in the diet can occur at any point in the lifespan, there is less readiness to change once eating patterns have been established for decades 64 .Parental modelling of lifestyle behaviors predicts healthy eating in children and adolescents 65,66 .Therefore, it is vital that young adults, particularly those with children, understand the health implications of diet and are empowered to make healthy food choices for themselves and their families.Trends over the past 30 years show that men are now spending more time grocery shopping but food preparation and consumption times have decreased for both men and women 67 .Grocery stores may be ideal locations for 'healthy eating' education and interventions.Marketing and financial strategies that specifically target men's nutrition-related behaviors should be developed and tested in vulnerable populations, particularly since diet affects the health of men more than women.

Limitations
Limitations to this study include required access to a copy of the local newspaper or shopper paper.Both grocery stores had newspapers available for sale at $0.50 and a small group of participants did purchase papers at the stores in order to access the grocery voucher questionnaire.Also, physical activity level was not assessed.However, in a cohort study of 1246 participants, physical activity, BMI and diet/weight concern was not associated with appetite 16 .Race/ethnicity was not assessed due to the county's demographics, with non-Hispanic whites comprising 97.5% of the population 42 .Also not assessed was whether participants intended to buy logo-labelled food items prior to receiving the grocery voucher.However, spending less than $10 was associated with purchasing at least one labelled item, indicating that grocery voucher redemption was likely the motivation to come into the store during the grocery event.Additional research using a control group is needed to determine whether financial incentives can prompt individuals to purchase health-promoting foods in addition to or instead of foods that contribute to chronic disease risk.

Conclusions
Younger adults and men may be less responsive than older adults and women to media-based educational strategies, heart-healthy food labelling and grocery vouchers to defray the cost of healthy eating.In this study, income was not associated with heart-healthy food purchases.Point-ofpurchase strategies in community grocery stores may be effective in increasing consumption of foods associated with reduced risk of CVD.Previous studies show that concerns about cost and availability of foods associated with reduced CVD risk are greater factors in the decision to purchase and consume these foods than demographic characteristics 68 .However, age and gender are associated with the likelihood of using vouchers for the purchase of healthful foods.Younger adults may be more responsive to grocery vouchers and coupons via other types of media campaigns (eg text, email and social media).Additional research is needed to determine whether different educational strategies paired with food labelling and grocery vouchers may be successful strategies to promote purchase of heart-healthy foods, particularly for men and younger adults.

Table 1 :
Breathitt County annual family income levels (2010 US Census Bureau)

Table 3 :
Designated food items purchased comparisons of demographic and personal characteristics (N=311) † Group comparison done using two-sample t-test.¶ Group comparison using the χ 2 test of association.§ Mann-Whitney U-test was used for this comparison due to limited range of possible values.

Table 4 :
Multivariate logistic regression with purchase of at least one designated food items as the outcome (N=280)

Table 5 :
Designated food items purchased comparison with total amount spent (N=311)