We examined the association between food insecurity and total daily energy intakes in American men and women. We estimated the number of daily snacks and meals consumed by individuals in different food security categories. Also, we calculated the energy contribution, energy density, and food group sources of those snacks and meals. Using the 1999–2002 National Health and Nutrition Examination Survey (NHANES), we examined the Food Security Survey Module (FSSM) and dietary information from the 24-h recall. Differences in energy intakes between groups were not significant. Women who were food insecure without hunger (FIWOH) and food insecure with hunger (FIWH) had significantly fewer meals than food secure (FS) women. The energy contribution of each meal and the total energy contributed from snacking were both significantly greater for FIWOH women than for FS women. The number of meals was significantly lower whereas the daily number of snacking occasions and the total energy from snacking were significantly increased for FIWOH men relative to FS men. FIWOH men consumed snack foods that had significantly lower energy density than those consumed by FS men. Among men and women, the major sources of meal energy were the grain group, the meat, poultry, and fish group, and the sugar, sweets, and beverages group whereas the major source of snacking energy was the sugar, sweets, and beverages group. Total energy intakes were not different for FI individuals; however, their meal and snack behaviors were different. Focusing solely on total energy intake would miss important consequences of food insecurity.
Findings from the Current Population Survey indicate that in 2006 11% of households in the United States were food insecure (FI), defined as having “limited or uncertain availability of nutritionally adequate and safe food or limited or uncertain ability to acquire acceptable foods in socially acceptable ways” (1,2). Food insecurity has been associated with obesity, heart disease, diabetes, high blood pressure, and food allergies (3,4,5,6,7).
Despite this vulnerability, very little research attention has been given to the diet of FI individuals. Three studies found that adult women who experienced food insufficiency/insecurity had lower energy intakes than those who reported food sufficiency/security whereas one study found no differences in energy intake (8,9,10,11). Another study that included both men and women found no differences in energy intake between adults from food-insufficient and food-sufficient households (12).
Researchers have found that the majority of women in all levels of food sufficiency reported between two and seven eating occasions per day, and there were no significant differences in number of eating occasions per day or the amount of nutrients obtained from snacks (11). Other researchers have found a reduced meal frequency among low-income and/or food stamp participants (13,14). In fact, skipping meals “because there wasn't enough money” is one construct used to measure food insecurity (2). To our knowledge, evidence of reduced meal frequency has not been documented in FI individuals with dietary assessment techniques, such as a 24-h recall.
Because of the paucity of the literature on this issue, we undertook this study to further the understanding of the dietary behaviors of FI individuals. Differences in dietary behaviors may provide insights regarding the previously reported association between obesity and food insecurity. Using a 24-h recall, we examined the association between food insecurity and total daily energy and macronutrient intakes in American men and women. We estimated the number of daily snacks and meals consumed by individuals in different food security categories. Also, we calculated the energy contribution, energy density, and food group sources of those snacks and meals.
We hypothesized that daily total energy intakes would not be different for FI or food secure (FS) individuals; however there would be considerable differences regarding their meal and snack behaviors. We hypothesized that a reduction in meal frequency among FI individuals would be offset by an increase in energy per meal. We also hypothesized that the energy from snacking would be greater in FI than FS individuals.
Methods and Procedures
National Center for Health Statistics' (NCHS) National Health and Nutrition Examination Survey (NHANES) for 1999–2002 provides information about people's consumption of foods and nutrients, as well as extensive health-related data, and information about Americans' demographic and socioeconomic characteristics. NHANES 1999–2002 is a complex, multistage probability sample of the civilian, noninstitutionalized population of the United States.
Food security measures
The NHANES 1999–2002 contains the 18-item Food Security Survey Module (FSSM), which has been shown to be a stable, robust, and reliable measurement tool (15). The NHANES 1999–2002 data are released in four food security categories: FS, marginally FS (MFS), FI without hunger (FIWOH), and FI with hunger (FIWH). The time frame for the FSSM is the last 12 months. Because we focused on adults for this analysis, we used the ten-item adult measure to ensure that the same measure was used for every individual.
For the 1999–2002 NHANES, individuals' dietary intakes were collected through an interviewer-administered 24-h dietary recall method. Research has shown that food intake data obtained from a 1-day dietary recall are reliable measures of usual intakes of population groups (16). Details regarding the specification of snack and meal occasions have been described elsewhere (17). Nutrient intakes obtained from the NHANES data set do not include nutrients from dietary supplements, medications, or plain drinking water but do include nutrients from beverages other than drinking water. The number of meal occasions and snacking occasions were calculated over the entire 24 h for each individual. The energy contributions per snack and per meal, and the total energy contributions of snacks and meals, were calculated. The relative caloric contribution of food groups defined by US Department of Agriculture (USDA) Food and Nutrient Database for Dietary Studies, 1.0 (2004, Beltsville, MD: Agriculture Research Service, Food Surveys Research Group) was also calculated.
Because of methodological differences in the treatment of beverages, it has been recommended that energy density values be calculated using solely food items (18). Thus using the USDA Food and Nutrient Database for Dietary Studies, 1.0 to beverages, we calculated the energy density (kcal/g) of food items (specific USDA food codes used are available from the authors). The contributions from beverages were excluded only from the energy density values. All other calculations included contribution from beverages.
The analytical sample included adults who completed the adult-level FSSM. Individuals were screened into the FSSM using the USDA food adequacy indicator and/or income. We excluded women who were pregnant and/or breastfeeding. Because researchers have reported associations between food insecurity and obesity among adults, we included individuals ≥18 years of age (7,19,20,21,22,23,24,25). To avoid including older individuals, many of whom have low energy intakes, we excluded respondents >60 years of age (17). Because prior researchers have found differences among FI men and women, we stratified by gender (women n = 2,707; men n = 2,933) (ref. 3).
Data analysis strategy
To examine the relationship between food security status and our dietary outcomes, we used multivariate linear regression. Possible confounders, including age, race-ethnicity, education, and income, were included in our analysis. Age was categorized as 18 to ≤30 and >30 to ≤60 years. Three race/ethnicity categories were examined: non-Hispanic white, non-Hispanic black, and Hispanic. Level of education was defined as less than high school degree or more high school degree. Household income was defined as a ratio; each household's income was compared with appropriate poverty thresholds. In all models, FS individuals were the referent group.
To account for mobile examination center sample design and sampling weights, we used STATA (Version 8; STATA, College Station, TX) to estimate all descriptive and inferential statistics. To examine whether age, race/ethnicity, and education were differentially distributed among the levels of food security and thus possible confounders, we used χ2-tests of independence. A P value ≤0.05 was considered statistically significant. This study was approved by the Institutional Review Board, Office of Human Subjects Research.
Demographic characteristics are presented by food security status for men and women in Table 1. Differences across levels of food security for men and women were found for every characteristic: age, race-ethnicity, education, and income.
Table 1. Characteristics of adults (18–60 years) who responded to the Food Security Survey Module (FSSM) in the 1999–2002 National Health and Nutrition Examination Survey (NHANES)a
Adjusted mean values for daily energy and macronutrient intakes are reported in Table 2 (model coefficients are presented in Supplementary Tables S1 and S2 online). Differences in energy intake were not observed among men or women. MFS women had significantly higher intakes of protein than FS women, whereas FIWH women had significantly higher intakes of total and saturated fat than FS women.
Table 2. Energy and macronutrient intakes of women and men (18–60 years) by level of food security in the 1999–2002 National Health and Nutrition Examination Survey (NHANES)a
The adjusted means for meal and snack occasions and their energy contributions are presented in Table 3. FIWOH and FIWH women had significantly fewer meals than FS women. Mean energy contribution of each meal and the total energy contributed from snacking were both significantly greater for FIWOH women than for FS women. Among FIWOH men, the daily number of meals was significantly decreased whereas the daily number of snacks and the total energy from snacking were significantly higher than for FS. FIWOH men consumed snack foods that had a significantly lower energy density than FS men.
Table 3. Meals and snacks and their energy contribution and energy density values among women's and men's diets (18–60 years) in the 1999–2002 National Health and Nutrition Examination Survey (NHANES)a
The major sources of energy during meal occasions were similar for all levels of food security and for both women and men. For men and women, the grain group was the major source of energy during meal occasions (ranging from 30% among FIWH men to 34% among FIWH women), followed by meat, poultry, fish, egg, and mixtures (ranging from 22% among FS women to 27% among MFS men). The third largest energy source was the sugars, sweets, and beverages group for men and women, ranging from 14% in MFS and FIWH women to 21% in FIWH men (results regarding the caloric contribution of all food groups are available from the authors).
For men and women, the major source of energy from snacking was the sugar, sweets, and beverages group. Among women, sugar, sweets, and beverages contributed between 34% among FS and 39% among MFS. Among FIWH men, the sugar, sweets, and beverages group contributed more than half their snacking energy (Figure 1). Among women, grain products were the next largest energy source and contributed between 33% among MFS and FIWOH and 26% among FIWH. The dairy group was the third leading energy source for women and contributed between 15% among FS and 9% among FIWOH.
Our results show that an increase in meal size and the energy obtained by snacking appears to compensate for a reduced meal frequency among FI individuals. Our results provide a possible mechanism in which FI individuals who skip meals may actually be able to consume sufficient or more than sufficient amounts energy. The increase in snacking energy may to be attributed to an increase in snacking frequency among FIWOH men. Although the energy per snacking occasion was not significantly different in FIWH women, the modest increase likely becomes amplified when considering the frequency of snacking. Other researchers have also shown that snacking is associated with higher total energy intake among adults (26,27,28).
To our knowledge, this is the first report where a 24-h recall was used to demonstrate a reduced meal frequency among FI individuals. Our findings provide validation that the responses to the FSSM questionnaire are reflected in the dietary behaviors observed in the 24-h recall. Other research has shown that skipping meals is a food choice coping strategy (14). Competing demands for individuals' time and energy, such as an increased need to work more than one job, may contribute to fewer meals.
Another item used to measure food insecurity, cutting back on portion size, was not evident in our results. However, “cutting back” is difficult to assess and this behavior would result in lower energy values per eating occasion if the energy density of the consumed foods at that occasion remained constant. FIWOH women had greater energy per meal values and FIWOH men had greater energy per snack values than those of FS groups. A larger meal or snack is another mechanism in which those who skip meals can achieve sufficient or more than sufficient daily energy intakes.
Our findings, that FI and FS daily energy intakes did not differ, suggest that skipping meals might not indicate an energy shortage. Two other reports found no differences in energy intakes of adults from food insufficient/insecure and food sufficient/secure families (10,12). However, three other studies found women from food insufficient/insecure households to have lower energy intakes than those from food sufficient/secure households (8,9,11).
Other research has demonstrated a link between obesity and food insecurity, which raises several questions. Researchers have explored the relationship of rising obesity levels and the technical change that has made food less expensive and physical activity more expensive (29). These researchers propose that higher incomes would result in an increased demand for food and the desire to be thin. Changes in the relative costs of food and physical activity would have different effects depending on an individuals' income.
We cannot definitely state that dietary behaviors alone promote obesity. However, in a study examining women's socioeconomic status and weight control practices, Jeffery and French concluded that dietary behaviors are the most likely proximal cause of socioeconomic status differences in obesity (30). Several mechanisms involving dietary behaviors have been proposed to explain the obesity and food insecurity link (6,19,31,32). One mechanism proposes that FI women over-consume one macronutrient. For example, Dietz concluded an increased fat consumption was responsible for the obesity/insecurity association (31). We found FIWH women had higher fat intakes than FS women; however, this did not translate to higher energy intakes. Previous research has shown that FIWH was not related to BMI among women (7). Researchers did find overweight to be associated with MFS and FIWOH among women (7). Our results showed that MFS women had higher protein intakes than FS women but this did not translate to higher energy intakes. We found that a considerable portion of snack energy was provided by the sugars, sweets, and beverages group among men. Previous research has shown that the highest consumers of sugar in all its forms are younger males; however this subpopulation is not linked to a greater risk of obesity (33).
Another mechanism suggests that low-income families maintain their energy intakes by purchasing low-cost items that are high-energy dense foods which may promote overconsumption (19,20,21). We found that the energy density values for the FI (MFS, FIWH, FIWOH) were not greater than FS. In fact FIWOH men had a lower energy density value for snack foods than FS men. We observed the energy density of snack foods to be higher than meal foods in all groups and thus, the passive overconsumption of snack foods may be a concern.
Another possible mechanism for the food insecurity-obesity association is the “boom and bust” cycle of food consumption linked to food stamps benefits (34). Researchers found food stamp recipients who shopped for food infrequently consumed less energy over the food stamp cycle; however, this difference was not extreme (34). In the same study, frequent shoppers who received food stamps showed no significant differences in energy intakes. Two other studies that examined repeated 24-h recalls found no significant differences in energy intakes over time (8,35).
Our results provide insights regarding the food stamps-obesity association (36,37). Food stamps may alleviate the most stringent deprivations; however, they do not increase the likelihood that a FI household will become completely FS (37). Our results show that the dietary behaviors of women at every level (FIWH, FIWOH, and MFS) were different from FS women. Thus food stamps may not allow recipients, particularly those in the most FI categories, to achieve the dietary behaviors of FS individuals.
One important consideration is that the dietary intake data we examined are self-reported, and self-reported information may yield biased results (38,39,40). Underreporting has been inversely related to socioeconomic status and other evidence suggests that overweight individuals underreport intakes (41,42,43,44). Another consideration is that the food security measurements encompass the past 12 months while the dietary information encompasses one 24-h recall. When individuals are classified as FI in the past 12 months, one cannot be certain that the episodes of hardship occurred during the 24-h period assessed by the dietary recall.
The results from this study demonstrate that skipping meals can be associated with diets that are adequate and possibly more than adequate in energy. We found an increase in meal size, and the energy obtained by snacking appears to compensate for a reduced meal frequency. Thus focusing solely on total energy intake is likely to miss important consequences of food insecurity. We therefore recommend that nutrition interventions aimed at FI audiences should also target meal and snack behaviors.
This study was supported by the US Department of Agriculture, Economic Research Service, Research, Innovation, and Development Grants in Economics (RIDGE) Grants and the Alabama Agricultural Experiment Station (ALA013-020). Mark Lino and Catherine Wernette provided comments on drafts of the article. We are grateful to Yanling Ma for her programming assistance and Caroline Brawley for support in administrative matters.