In 2019, 13·7 million households in the USA experienced food insecurity(1). Among households with children, 2·4 million had children who were food insecure(1). High prevalence of food insecurity was observed in households with children (13·6 %) and those headed by single mothers or single fathers (28·7 % and 15·4 %, respectively). Additionally, non-Hispanic Black households, Hispanic households and low-income households below 185 % of the poverty line were also observed to have high prevalence of food insecurity (19·1 %, 15·6 % and 27·6 %, respectively)(1). The prevalence of household food insecurity was highest for those living in the Southern USA (11·2 %) and in the principal cities of metropolitan areas (12·4 %)(1).
Food insecurity is associated with poorer physical health, cognitive development, academic achievement and behaviour in children and has been linked to higher odds of childhood obesity(Reference Shankar, Chung and Frank2–Reference Kral, Chittams and Moore4). Risk of childhood obesity is also associated with physical activity and sedentary behaviours(5,Reference Hong, Coker-Bolt and Anderson6) . Few studies have investigated the connections between food insecurity and physical activity in children, however, and those that have explored this link demonstrate mixed associations, with some finding negative associations with moderate-to-vigorous physical activity, null or negative associations with sedentary behaviour and null associations with total physical activity, defined as the sum of light, moderate and vigorous intensity physical activity(Reference Fram, Ritchie and Rosen7–Reference To, Frongillo and Gallegos9). It is hypothesised that food insecure children may be less likely to participate in physical activity because food insecurity can have physical (e.g. hunger, fatigue and illness) and emotional (e.g. stress, anxiety and depression) manifestations(Reference Ke and Ford-Jones10,11) . Thus, because both food insecurity and inadequate physical activity can result in serious health consequences and interventions aimed at decreasing food insecurity may positively impact physical activity, it is critical to examine this relationship in children. This research will help clarify the relationship between food insecurity and physical activity, contributing meaningful insights into the implications of food insecurity in children.
The primary aim of this study was to examine whether household food insecurity is associated with children’s physical activity and sedentary behaviour. Prior studies have indicated that factors influencing physical activity may operate inconsistently across certain subgroups, such as age, sex and race/ethnicity. For example, trends in physical activity behaviour among US youth reveal that children aged 12–15 years tend to be less physically active than children aged 6–11 years, high-school girls tend to be less physically active than high-school boys and there are reported differences in activity by race/ethnicity(12). Thus, this study will also explore whether the relationship between food insecurity and physical activity differs within specific subpopulations, including age, sex and, race/ethnicity.
Methods
Study design and participants
The Healthy Communities Study is a cross-sectional observational study of 5138 US children designed to assess the impact of community programmes and policies addressing childhood obesity on child diet, physical activity and weight outcomes. Participants included those 4–15 years in grades K-8 and were chosen from a diverse sample of 130 communities, defined as high-school catchment areas, across the country between 2013 and 2015. Community selection was based on hybrid sampling that included a national probability-based sample of 102 communities as well as a sample of twenty-eight ‘certainty’ communities known for having programmes and policies that target childhood obesity prevention. Data collectors interviewed children and families in their homes, gathering comprehensive information on demographics, nutritional status, anthropometry, physical activity and the surrounding environment. The Battelle Memorial Institute Institutional Review Board approved the study protocol, and parents provided written informed consent for child participation. A more detailed explanation of participant recruitment, study design and data collection are included in papers by Sagatov et al., Strass et al. and John et al. (Reference Sagatov, John and Gregoriou13–Reference John, Gregoriou and Pate15).
Variables of interest
Demographic information was provided by the parent or guardian and included items on child age, child race/ethnicity, geographic region, urbanicity, country of origin, household income, parental education attainment, marital status and employment status. Household food insecurity was indicated by two items from the US Department of Agriculture’s 18-item US Household Food Security Survey Module that assess uncertainty that the household food budget or food supply is sufficient to meet basic needs(Reference Bickel, Nord and Price16). The indicator from these two items has high sensitivity (97 %) and specificity (83 %) in predicting responses to the full module and are associated with poor or fair child health, hospitalisations and developmental risk(Reference Hager, Quigg and Black17). Furthermore, any affirmation of household food insecurity is associated with children’s academic performance, weight gain and social skills(Reference Jyoti, Frongillo and Jones18). Administered by an interviewer, respondents were asked to assess how often the following occurred within the past 12 months: (1) ‘We were worried whether our food would run out before we got money to buy more’ and (2) ‘The food we bought just didn’t last and we didn’t have money to get more.’ Participants could respond with very often, often, sometimes, rarely or never. Those who responded with very often, often or sometimes for at least one of the items were categorised as having household food insecurity. For all households, the adult was the respondent.
Physical activity was measured using self-report in the 7-d Physical Activity Behavior Recall (PABR-7) instrument. The tool was designed to detect forms of physical activity and sedentary behaviour that are typically targeted in community interventions to reduce childhood obesity. Through a computer-assisted interview, respondents specified whether the child participated in each activity over the past 7 d, the days the activity was performed and the average intensity of the activity (i.e. light, moderate, hard and very hard) using sex- and age-appropriate visual aids. For children aged 4–8 years, their parent/guardian responded, and the child was there to assist. For children aged 9–15 years, the child responded, and their parent/guardian was present to assist(Reference Pate, McIver and Colabianchi19). In this study, physical activity terms refer to the frequency and intensity of specific forms of activity and are therefore used in a manner that is specific to the methodology of the study.
Four measures were used to assess a child’s physical activity: total physical activity, vigorous physical activity, moderate-to-vigorous physical activity and sedentary behaviours. These variables were converted into index scores by multiplying the number of reported activities for each activity intensity level by the frequency of participation in those activities. Greater scores indicate more physical activity and sedentary behaviour. The total physical activity index score included fifteen activities: PE class, recess/free play, dance, activity breaks, school sports, non-school sports, pick-up sports, physical activity in after-school, physically active games, swim/water activities, outdoor/adventure activities, walk/bike to school, walk/bike to store/friend’s house/etc., walk/bike for fun/exercise and active video games. The vigorous physical activity index score included four activities: school sports, non-school sports, pick-up sports and outdoor/adventure activities. The moderate-to-vigorous physical activity index score included eleven activities: PE class, recess/free play, dance, school sports, non-school sports, pick-up sports, physical activity in after-school, physically active games, swim/water activities, outdoor/adventure activities and walk/bike for fun/exercise. The sedentary behaviours index score included four activities: used a computer for games or playing on the internet (not for schoolwork or social networks), used a computer or phone for social networking, watched TV and played non-active video games(Reference Pate, McIver and Colabianchi19).
Data analysis
Secondary data analysis was conducted on the Healthy Communities Study data set. Participant demographics were analysed by household food insecurity status using t-tests for continuous variables and χ 2 tests of independence for categorical variables. Multilevel linear regression models were generated to relate food insecurity with physical activity, adjusting for child- and community-level covariates and clustering by school and community levels. Multiple imputation by chained equations was used to account for missing data due to participant nonresponse whereby twenty imputed data sets were created and then combined(Reference van Buuren and Groothuis-Oudshoorn20). Approximately a quarter of cases (24·5 %) were missing at least one variable. The least absolute shrinkage and selection operator (LASSO) technique was used to select a subset of covariates from the large pool of demographic variables available, increasing accuracy and interpretability of models(Reference Tibshirani21). The covariates included in the models were maximum parental education, maximum parental employment status, child age, child age squared, child race/ethnicity, seasonality, US region, income classification, community percentage of African Americans, community percentage of high school graduates and community unemployment rate for those in the labour force 16 years and older.
Additional, exploratory multilevel linear regression models were generated to assess for interaction with child age, sex and race/ethnicity. Child sex was coded as female or male, based on the identification by the data collector. The race/ethnicity variable had four options: Non-Hispanic White, Non-Hispanic African American, Non-Hispanic other and Hispanic or Latino. Based on a review of the relevant literature assessing how physical activity varies by age, age was divided into 4–11 years and 12–15 years(12). All analyses were conducted in SAS 9.4 (SAS Institute). Reported P-values are two-sided, and significance was set at P < 0·05.
Results
In this study sample, children were on average 9 years old and half were female (Table 1). Nearly half of children were Hispanic or Latino (44·7 %). The majority (72·9 %) had at least one parent working full-time; less than half of parents (42·9 %) had a maximum education of high school or less. Geographically, 41·6 % of participants were from the Southern USA. At the community level, nearly 19·7 % of children lived in a community with a census tract minority classification of African American and 34·8 % with an income classification of low income(Reference Ritchie, Woodward-Lopez and Au22,Reference Pate, Frongillo and McIver23) .
Socio-demographic characteristics of the children in the Healthy Communities Study by household food insecurity status (n 5138)
* Variables selected using the LASSO technique.
† Statistical significance was set at P < 0·05.
‡ Results from χ 2 tests of independence for categorical variables or t-tests for continuous variables. Multiple imputation by chained equations employed to account for missing values. Approximately a quarter (24·5 %) of cases were missing at least one variable.
§ Child race/ethnicity ‘other’ category includes Asian, Native Hawaiian/Pacific Islander, Non-Hispanic American Indian/Alaska Native, more than one race including African American and more than one race not including African American.
|| Unemployed includes those unemployed and looking for work; other includes those who are only temporarily laid off, on sick leave, or maternity leave, disabled, keeping house, retired, student and other.
¶ Graduate degree includes Master’s degree, professional degree and doctoral degree.
** Physical activity index scores represent the multiplication of the number of reported activities for each activity intensity level within the past 7 d by the frequency of participation in those activities. Greater scores indicate more physical activity and sedentary behaviours.
Close to half of children were classified as residing in a household with food insecurity, with the highest prevalence seen among Hispanic or Latino children (58·1 %) followed by non-Hispanic African American children (21·5 %) (Table 1). A high prevalence of household food insecurity was also observed in children whose parents had a maximum educational attainment of some college or Associate degree or less (87·9 %). Additionally, those living in the Southern USA (43·7 %) had higher prevalence of household food insecurity. On the community level, communities with lower income classifications, greater proportions of African American children, higher unemployment rates and fewer high school graduates also had higher prevalence of household food insecurity.
Children on average had a total physical activity index score of 17·9, a vigorous physical activity index score of 3·0, a moderate-to-vigorous physical activity index score of 13·5 and a sedentary activity index score of 11·8 (Table 1). Children in households with food insecurity had less moderate-to-vigorous physical activity than children in households without food insecurity (P < 0·001).
No associations were observed between household food insecurity and physical activity (Table 2). A significant interaction between food insecurity and child sex at the sedentary level was found (P = 0·03) (Table 3). When stratified to explore individual effects by sex, associations between food insecurity and physical activity were NS for either females or males (data not shown). There were no other associations by race/ethnicity or age (data not shown).
Associations of household food insecurity and physical activity of the children in the Healthy Communities Study as measured by the 7-d Physical Activity Behaviour Recall (PABR-7) instrument (n 5138)
* Binary predictor of interest was having household food insecurity (1) compared with not having (0) household food insecurity.
† Statistical significance was set at P < 0·05.
‡ Multilevel mixed modelling adjusted for maximum parental education, maximum parental employment status, child age, child age squared, child race/ethnicity, seasonality, US region, income classification, community percentage of African Americans, community percentage of high school graduates and community unemployment rate for those in the labour force 16 years and older. Standard errors are clustered at the community and school levels. Multiple imputation by chained equations employed to account for missing values. Approximately a quarter (24·5 %) of cases were missing at least one variable.
§ Total physical activity index score is the number of reported activities (out of 15) multiplied by the frequency of participation in those activities in the past 7 d.
|| Vigorous physical activity index score is the number of reported activities (out of 4) multiplied by the frequency of participation in those activities in the past 7 d.
¶ Moderate-to-vigorous physical activity index score is the number of reported activities (out of 11) multiplied by the frequency of participation in those activities in the past 7 d.
** Sedentary activity index score is the number of reported activities (out of 4) multiplied by the frequency of participation in those activities in the past 7 d.
Moderating effect of child sex on the association between household food insecurity and physical activity of the children in the Healthy Communities Study as measured by 7-d Physical Activity Behaviour Recall (PABR-7) instrument (n 5138)
* Statistical significance was set at P < 0·05.
† Analyses used multilevel modelling to identify moderating effects of sex on the association between household food insecurity with physical and sedentary activity. Models included an interaction term for food security status (with v. without) by child sex (female v. male). Covariates included maximum parental education, maximum parental employment status, child age, child age squared, child race/ethnicity, seasonality, US region, income classification, community percentage of African Americans, community percentage of high school graduates and community unemployment rate for those in the labor force 16 years and older. Standard errors are clustered at the community and school levels. Multiple imputation by chained equations employed to account for missing values. Approximately a quarter (24·5 %) of cases were missing at least one variable.
‡ Total physical activity index score is the number of reported activities (out of 15) multiplied by the frequency of participation in those activities in the past 7 d.
§ Vigorous physical activity index score is the number of reported activities (out of 4) multiplied by the frequency of participation in those activities in the past 7 d.
|| Moderate-to-vigorous physical activity index score is the number of reported activities (out of 11) multiplied by the frequency of participation in those activities in the past 7 d.
¶ Sedentary activity index score is the number of reported activities (out of 4) multiplied by the frequency of participation in those activities in the past 7 d.
Discussion
The results from the exploratory analysis indicate that sex may modify the relationship between household food insecurity and physical activity in children. Girls in households with food insecurity engaged in more sedentary behaviours than did girls in households without food insecurity. In contrast, boys in households with food insecurity engaged in less sedentary behaviours compared with boys in households without food insecurity. Food insecurity in the USA impacts school-age girls more than boys and is associated with poorer academic performance and higher BMI(Reference Jyoti, Frongillo and Jones18,Reference Burke, Frongillo and Jones24,Reference Au, Zhu and Nhan25) . Additionally, participation in the Supplemental Nutrition Assistance Program is associated with better academic performance of girls more than boys, suggesting that girls may be more susceptible to some consequences of food insecurity(Reference Frongillo, Jyoti and Jones26). Since stratification by sex did not yield significant results, additional studies are needed to better understand the influence of sex, physical activity and food insecurity.
One potential explanation for the lack of association in the main analysis lies with the food insecurity indicator used. In this study, household food insecurity was assessed by two items from the US Department of Agriculture’s eighteen-item measure that identifies households experiencing uncertainty about food access. A previous study demonstrated that the two-item tool was sensitive, specific and valid when tested in a sample of 30 098 families with young children of which 23 % were food insecure(Reference Hager, Quigg and Black17), but these two items may not capture the experience of all individuals in the household, particularly children. Although parents attempt to protect their children from food insecurity, children are often aware of food insecurity and take responsibility for it(Reference Frongillo, Fram and Escobar-Alegría27). Children’s experiences of food insecurity are often not known by parents(Reference Frongillo, Fram and Escobar-Alegría27). Many children have greater awareness of household food insecurity than parents assume, and child report of their food insecurity is more strongly associated with outcomes than parent report(Reference Fram, Frongillo and Jones28). Perhaps there is a relationship between food insecurity and physical activity in children, but the use of this two-item household food insecurity indicator was unable to detect this association by not assessing child experiences of food insecurity.
Previous studies that assessed the relationship between food insecurity and physical activity among children have also not used food insecurity measures that specifically assess children, such as the five-item questionnaire used in the US National Health and Examination Survey(Reference Asfour, Natale and Uhlhorn8,29) and the US Department of Agriculture’s eighteen- or six-item measures, which may have contributed to the mixed findings(Reference To, Frongillo and Gallegos9,30) . Using measures that specifically assess child food insecurity may be able to more accurately capture this relationship. The US Child Food Security Survey Module is one validated child-report measure for older children aged 12 to 17 years(Reference Connell, Nord and Lofton31), but it is not recommended for use in younger children. Additionally, because it was adapted from selected items in the US Household Food Security Survey Module for adults, some researchers suggest that it reflects adult experiences of food insecurity that some children may not share(Reference Fram, Ritchie and Rosen7,Reference Fram, Frongillo and Draper32,Reference Fram, Bernal and Frongillo33) . The Child Food Security Assessment is another self-reported measure that is validated for use in children as young as 6 years of age. Unlike the US Household and Child Food Security Survey Modules, the Child Food Security Assessment was developed based on interviews with children and assesses multiple subdomains of child food insecurity including cognitive awareness, emotional awareness, physical awareness, participation in parent food strategies, initiation of child food management strategies and resource generation(Reference Fram, Frongillo and Draper32).
Further, given that the physical activity tool, PABR-7, was based on self-report, it may be subject to recall error. The inclusion of objective measures of physical activity, such as data collected from wearable activity tracking devices, was available but limited to 10 % of the full study sample due to the original study’s budgetary constraints(Reference Pate, McIver and Colabianchi19). The tool has not yet been validated against accelerometry. Additionally, although the PABR-7 focused on activities targeted by community programmes and policies, which could exclude other important forms of physical activity, it was structured similar to other physical activity self-report instruments that have been widely used in previous studies with children(Reference Pate, McIver and Colabianchi19). The tool reflects the composite of the child’s participation in multiple types of physical activity that are often offered in communities that are implementing obesity prevention programmes for children and youth. Therefore, the instrument and composite index should be reflective of the child’s engagement in activities that are often conceptualised as supporting prevention of obesity.
The Healthy Communities Study used a large, diverse sample of children and households from across the United States(1) with a high prevalence of household food insecurity (44·6 %) and childhood obesity (24·8 %)(Reference Plimier, Hewawitharana and Webb34) compared with national estimates of food insecurity in households with children (19·5 %(Reference Coleman-Jensen, Gregory and Singh35) and 16·6 %(Reference Coleman-Jensen, Rabbitt and Gregory36)) and childhood obesity (17·2 % and 18·5 %)(Reference Fryar, Carroll and Afful37) in 2013 and 2015, respectively. The study oversampled communities with certain criteria, including high proportions of African American, Hispanic and low-income households, to lend statistical power and relevance to populations experiencing health disparities(Reference Strauss, Sroka and Frongillo38). While the study was conducted between 2013 and 2015 and has not been repeated since, the findings are relevant to today’s public health context as trends in food insecurity and childhood obesity in the United States remain high, especially in low-income and racially diverse populations. The coronavirus pandemic has increased the prevalence of child food insecurity to a projected 17·9 % in 2021(Reference Hake, Dewey and Engelhard39), and recent data suggest that food insecurity in households with children is close to a combined 40·0 % among African American and Hispanic households and 37·1 % among households with an income-to-poverty ratio under 1·0(Reference Coleman-Jensen, Rabbitt and Gregory40). Additionally, childhood obesity has risen in 2017–2018 to 19·3 % and is estimated to be a combined 49·8 % among African American and Hispanic children(Reference Fryar, Carroll and Afful37) and close to 38·8 % in households with incomes less than or equal to 350 % of the federal poverty level(Reference Ogden, Carroll and Fakhouri41). Given its observational design, the study is unable to determine causal relationships.
Conclusion
This study suggests that sex may modify the relationship between household food insecurity and physical activity in children, though additional research is needed to clarify this relationship. Results from this study indicate that use of a more detailed food insecurity measure that better represents child experience may be warranted to assess how severity of child food insecurity is contemporaneously related to physical activity. Future studies may also explore more sub-group analyses by sex or design qualitative studies to better understand this relationship from the perspective of children.
Acknowledgements
Acknowledgements: The authors would like to acknowledge Kaela Plank and Sridharshi Hewawitharana for their assistance with data analysis and Laurel Moffat and Isabel Thompson for their feedback in manuscript development. Financial support: Research reported in this publication was supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number K01HL131630. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Conflict of interest: There are no conflicts of interest. Authorship: S.N.: data analysis and manuscript drafting; M.T.: data analysis support and manuscript revising; L.R. and L.A.: conception and design of research and manuscript revising; E.F., B.L. and R.P.: interpretation of data and manuscript revising. All authors had responsibility for final content and read and approved the final manuscript. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were approved by the Battelle Memorial Institute Institutional Review Board. Written informed consent was obtained from all participants.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980021002536
Open access
