In this cross-sectional study, we used 2019–21 data from the National Survey of Children’s Health (NSCH).¹⁹ The NSCH is funded and directed by the Maternal and Child Health Bureau. It is a nationally representative survey conducted annually in all 50 states and the District of Columbia on a systematic random sample of households with children ages 0 to 17 years. Survey respondents are the child’s parent or guardian (hereafter referred to as “caregiver”), and caregivers complete the survey online, on paper, or over the phone. Due to the nature of some of the questions included in our measure of family resilience and connection, we included only families with children ages 6 to 17 years in analyses. This study was reviewed and considered exempt by the University of North Carolina’s Institutional Review Board.

Our exposure was household food hardship during the past 12 months, defined as struggling to afford sufficient quality and/or quantity of food. In the NSCH, household food hardship is measured with a single question, “Which of these statements best describes your household’s ability to afford the food you needed during the past 12 months?” with response options including, “We could always afford to eat good nutritious meals,” “We could always afford enough to eat but not always the kinds of food we should eat,” “Sometimes we could not afford enough to eat,” and “Often we could not afford to eat.” The United States Department of Agriculture (USDA) defines food insecurity as limited or uncertain availability of or ability to access nutritionally adequate and safe food and is typically measured by an 18- or 6-item questionnaire to assess experiences in the past 12 months or past 30 days.²⁰ Food insecurity captures reductions in diet quality and variety, anxiety about food running out, and changes in food intake. The USDA defines food insufficiency, a related but narrower construct, as a household not having enough to eat, sometimes or often, in the past 7 days.²⁰ In contrast to food insecurity, food insufficiency focuses solely on whether a household has enough food to eat and is intended to assess rapid changes in food availability.²¹ The question and response options included on the NSCH reflect aspects of both food insufficiency (not having enough food to eat) and food insecurity (not being able to afford sufficient quality and quantity of food and a 12-month timeframe). Thus, we considered response options of “We could always afford enough to eat but not always the kinds of food we should eat,” “Sometimes we could not afford enough to eat,” or “Often we could not afford to eat” to capture a broad construct of food hardship, or that a household experienced challenges affording quality food (not being able to afford the kinds of food they should eat) and/or sufficient quantity of food (not being able to afford enough to eat). We also conducted sensitivity analyses with a measure of food insufficiency as the exposure, defined as response options of “Sometimes we could not afford enough to eat,” and “Often we could not afford to eat.”²⁰

Our outcome was family resilience and connection, conceptualized as interpersonal functioning within families and measured by the family resilience and connection index (FRCI). The FRCI is an established measure created using 6 questions available from the NSCH.¹¹ The FRCI questions capture key aspects of family resilience and connection commonly agreed upon in the literature, such as beliefs that facilitate coping, adherence to routines while allowing for deviations when necessary, effective communication skills, and adaptive problem solving, and reflect characteristics that have the potential to support child wellbeing.¹¹ This index has been used in prior studies examining family resilience and connection.¹⁸ The first 4 items in the FRCI index ask, “When your family faces problems how likely are you to do each of the following: 1) talk together about what to do; 2) work together to solve our problems; 3) know we have strengths to draw on; and 4) stay hopeful even in difficult times.” Response options for each of the 4 items include “all of the time,” “most of the time,” “some of the time,” and “none of the time.” Consistent with prior research, we assigned 1 point to responses of “all of the time” for each of the 4 items.¹¹ Two additional items ask, “How well can you and this child share ideas or talk about things that really matter?” and “How well do you think you are handling the day to day demands of raising children?”. Response options include “very well,” “somewhat well,” “not very well,” and “not at all.” Consistent with prior research, we assigned 1 point to responses of “very well” to both items.¹¹ We summed points assigned to each of these 6 items, with totals ranging from 0 to 6 and 6 representing the highest family resilience and connection score. Consistent with prior research, we categorized the FRCI into “high” (4–6 points) as compared to low or moderate (0–3 points) family resilience and connection.¹¹

Our modifier was household SNAP participation in the past 12 months. In the NSCH, SNAP participation is captured with the question, “At any time during the past 12 months, even for 1 month, did anyone in your family receive Food Stamps or Supplemental Nutrition Assistance Program (SNAP) benefits?”. Response options were “yes” or “no.”

To identify potential confounders, we created a conceptual diagram (Supplemental Figure 1) based on an extensive literature review and author subject matter expertise. We adjusted analyses for potential confounding variables including family income (0%–99%, 100%–199%, 200%–399%, or ≥400% FPL), highest education level achieved of caregiver(s) within the household (high school or less, some college, or college or higher), child race and ethnicity (Hispanic, White non-Hispanic, Black non-Hispanic, or Multiracial/non-Hispanic/Other), household participation in the Special Supplemental Nutrition Program for Women, Infant, and Children during the past 12 months, household participation in SNAP during the past 12 months, household participation in free or reduced lunch in the past 12 months, family size (≤4 or ≥5 family members), immigration history (child of immigrant or US-born parents), and caregiver employment (unemployed or employed). Further description of confounders is available in Supplemental Table 1.

We used multivariable linear regression with a log link and binomial distribution to calculate risk ratios (RRs) and corresponding 95% confidence intervals (CIs). We adjusted all models for the above noted confounders and accounted for the complex sampling design of NSCH. We first calculated RRs comparing the likelihood of high family resilience and connection among households experiencing food hardship to those not experiencing food hardship. Then, among households with incomes < 200% FPL (ie, likely to be eligible for SNAP benefits), we calculated RRs and 95% CIs comparing the likelihood of high family resilience among households experiencing food hardship to those not experiencing food hardship separately for households participating in SNAP and households not participating in SNAP. We conducted analyses in Stata SE 18 (StataCorp LLC).

This study has several limitations. First, the cross-sectional design prevents establishing temporality between food hardship, SNAP participation, and family resilience. Future longitudinal studies are needed. Second, food hardship was measured with a single NSCH question rather than USDA’s validated scales, potentially missing nuances in household experiences of food hardship. Data on SNAP participation specifics, such as the duration of participation and benefit amount received, were unavailable. This may obscure potential differences in the impact of shorter- versus longer-term SNAP participation and higher versus lower SNAP benefits on family resilience and connection. Third, the Family Resilience and Connection Index is a valid and reliable measure,¹¹ but may not fully capture all relevant aspects of family resilience and connection. It is also possible that participants underreported levels of food hardship or overreported family resilience and connection due to social desirability. Further, we dichotomized the FRCI into a binary outcome, grouping low and moderate scores together. This approach was necessary to ensure sufficient statistical power for our multivariable analyses but may potentially limit the granularity of our findings. Future research could maintain 3 FRCI categories to capture a fuller understanding of family resiliency and connection in the context of food hardship and SNAP participation. Fourth, in some states, the cutoff for SNAP eligibility is < 200% FPL. However, we took a conservative approach by using 200% FPL recognizing that household income can fluctuate over time and may not always be reported with complete accuracy. It is possible that we may have overestimated the number of households eligible for SNAP, potentially biasing results towards the null. Fifth, we adjusted analyses for a comprehensive set of potential confounders, however, there is always potential for residual or unmeasured confounding.

Last, the data we used in this study are situated primarily within the COVID-19 pandemic context, during which time many families received increased benefits from social support programs such as SNAP emergency allotments.³⁹ While our study contributes to the understanding of family resilience during this challenging time, future research is needed to explore the relationship between food hardship and family resiliency and connection post-COVID. Further, given that SNAP participation can vary by state, additional studies could examine whether family resilience and connection differ by US region. Qualitative research could also shed light on potential pathways between food hardship and family resilience and connection and strategies to mitigate this association.