Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jun 1.
Published in final edited form as: Appetite. 2021 Jan 26;161:105128. doi: 10.1016/j.appet.2021.105128

Persistence and transience of food insecurity and predictors among residents of two disadvantaged communities in South Carolina

Angela D Liese 1, Patricia A Sharpe 2, Bethany A Bell 3, Brent Hutto 2, Jessica Stucker 2, Sara Wilcox 2,4
PMCID: PMC7987735  NIHMSID: NIHMS1666860  PMID: 33513414

Abstract

Background:

Little is known about patterns of household food insecurity (HFI) across more than two time points in adults in the United States, the frequency predictors of different trajectories. The distinctions between persistent and transient food insecurity trajectories may be crucial to developing effective interventions.

Objective:

To characterize dominant trajectories of food security status over three time points between 2013 and 2016 and identify demographic, socioeconomic and health-related predictors of persistent and transient HFI.

Design:

Cohort study in disadvantaged communities in South Carolina.

Setting and subjects:

397 middle-aged participants, predominantly female, African American, living in USDA-designated food deserts.

Main outcome measure:

Household food insecurity over time using the 18-item USDA’s Household Food Security Survey Module.

Statistical analyses performed:

Descriptive analyses of food security trajectories and multinomial regression analyses.

Results:

At baseline (2013–2014), 61% of households reported HFI during the previous 12 months, which decreased to 54% in 2015 and to 51% in 2016. Only 27% of households were persistently food secure, 36% experienced transient and 37% persistent food insecurity. Female sex (OR 2.7, 95%CI 1.2–5.9), being married or living with a partner (OR 2.4, 95CI% 1.1–5.3) and fair health status (OR 4.4, 95%CI 2.2–8.8) were associated with increased odds of persistent food insecurity. Fair health was also a significant predictor of transient food insecurity.

Conclusions:

These findings suggest that future research should focus on persistent versus transient trajectories separately and that tailored interventions may be needed to make progress on alleviating food insecurity among disadvantaged communities.

Keywords: Food security, poverty, cohort

Introduction

Food insecurity is defined as the ‘limited or uncertain availability of nutritionally adequate and safe foods or limited or uncertain ability to acquire acceptable foods in socially acceptable ways’ (Bickel, Nord, Price, Hamilton, & Cook, 2000) and has been linked to increased risk of chronic disease, obesity and lifelong mental health effects (Jyoti, Frongillo, & Jones, 2005; Laraia, 2013; Seligman, Laraia, & Kushel, 2010). The United States Department of Agriculture (USDA) monitors household food security annually with a series of questions about conditions, experiences, or behaviors related to food insecurity occurring at any time in the past 12 months (Bickel et al., 2000). After a historical peak of 14.9% in 2011, the prevalence of household food insecurity has slowly declined to 12.3% in 2016 (Coleman-Jensen, Rabbitt, Gregory, & Singha, 2017a). However, state-level variation is significant, ranging from 8.7% food insecure households in Hawaii to a high of 18.7% in Mississippi in 2014–16. US-wide, non-Hispanic Black populations have the highest burden of household food insecurity at 23% in 2016, compared to 19% in Hispanics and 9% in non-Hispanic whites. (Coleman-Jensen et al., 2017a) Overall, the Southern US, which is characterized by a higher proportion of minorities, also exhibits the highest food insecurity levels compared to other regions, and South Carolina (13% household food insecurity in 2014–2016) ranks as the 22nd highest state (Coleman-Jensen et al., 2017a).

To estimate the duration of food insecurity, the USDA uses indirect evidence by posing supplemental questions about the frequency of occurrence over the past year, whether food insecurity occurred in the past 30 days, and if so, on how many days. Based on these data, the USDA estimates that for 2016 “For about one-fourth of food-insecure households and one-third of those with very low food security, occurrence of the associated conditions was frequent or chronic. That is, the conditions occurred often, or in almost every month” (Coleman-Jensen, Rabbitt, Gregory, & Singha, 2017b). These estimates stand in stark contrast to the USDA’s statement that for the nation as a whole, “when food insecurity occurs in US households, it is usually recurrent but not chronic”.(6) However, these data do not shed light on persistence of food insecurity over a time period longer than a year.

Cohort studies provide a unique opportunity to characterize dynamics of temporal patterns observed in the same group of people over time, including the persistence and the transience of food insecurity over time (Grineski, Morales, Collins, & Rubio, 2018; Hernandez & Jacknowitz, 2009; Ip et al., 2015; Jansen et al., 2017; Jones & Frongillo, 2006; Nord, Andrews, & Winicki, 2002; Whitaker & Sarin, 2007; Wilde, Nord, & Zager, 2010). To date, the majority of cohort studies have utilized only two time points and characterized four mutually exclusive states: persistent food insecurity, persistent food security, becoming food insecure, and becoming food secure. Using this approach, Whitaker et al. reported that about 15% of the mothers in the Fragile Families and Wellbeing Study experienced persistent food insecurity over a two-year period between 2001 and 2005, 26% transitioned between states, and 59% persistent food security (Whitaker & Sarin, 2007). An analysis of the Early Childhood Longitudinal Study (ECLS) Birth Cohort reported only 3% persistent food insecurity, 11% transient food insecurity, and 86% persistent food security in the course of one year (Hernandez & Jacknowitz, 2009).

Several cohort studies have used more advanced statistical models to capitalize on three or more consecutive assessments of food insecurity (Hernandez & Jacknowitz, 2009; Ip et al., 2015; Jansen et al., 2017; Ryu & Bartfeld, 2012; Wilde et al., 2010) . A study covering Kindergarten through 8th grade in the ECLS study found only 1% of the sample was food insecure in four consecutive years and 3% in three years, whereas 79% had persistent food security (Ryu & Bartfeld, 2012). Similarly, Wilde et al. analyzed the Survey of Program Dynamics (SPD) panel, which evaluated how US households experience food insecurity between 1998 and 2001 (Wilde et al., 2010). They concluded that only about 1% of the US population experienced persistent food insecurity for four or five consecutive years in the nationally representative SPD sample (Wilde et al., 2010). Thus, past cohort studies seem to suggest that food insecurity is typically episodic or transient. However, only a single published cohort in the US has addressed the persistence or transience in a particularly disadvantaged population, households of Head Start pre-schoolers in Michigan, and reported 54% persistent food insecurity, 36% transient food insecurity and 10% persistent food security (Jansen et al., 2017).

Cohort studies are also well positioned to evaluate predictors of different food insecurity trajectories, but the literature to date is quite finite with disparate foci because of the dominance of the static view offered by cross-sectional research. Wilde et al. focused on household composition, demographic and socioeconomic characteristics, and disabilities,(Wilde et al., 2010) whereas Hernandez et al. evaluated household composition, maternal characteristics, socioeconomic indicators, and geographic attributes (Hernandez & Jacknowitz, 2009). In contrast, the study by Ip et al. focused on migration, household immigration status, and seasons given the interest in female farmworkers (Ip et al., 2015). Compared to the prospective studies predicting a particular outcome state at a single timepoint (often assumed to be representative of all future states), studies of predictors of outcome trajectories offer more nuanced and possibly more accurate temporal inferences informed by empirical data on outcome states at multiple timepoints.

We had an opportunity to study food insecurity trajectories in a sample of residents of two disadvantaged communities in South Carolina who were interviewed three times between November 2013 and August 2016 as part of an evaluation of the impact of a food hub initiative on fruit and vegetable intake, diet quality, perceptions of the environment, body mass index, and shopping behaviors. (Sharpe et al., 2020) For the present analyses, we aimed to quantify the persistence and transience of household food insecurity and identify predictors of trajectories of food insecurity.

Methods

This manuscript presents results of a secondary analysis of a study conducted in South Carolina among two low-income communities, described previously (Liese, Ma, Hutto, & Et, 2017; Ma, Liese, Hibbert, & Et, 2017; Sharpe et al., 2018) . The parent study evaluated the impact of a healthy food access initiative, a food hub intervention to increase healthy food access with a longitudinal, quasi-experimental design. The target population of this study comprised: (a) residents of high-poverty areas (i.e., a poverty level greater than or equal to the state’s level of 16%) of the food hub’s service area (Location 1); and (b) residents of areas with similar poverty levels in another part of the state in which no food hub was present or planned (Location 2). Overall, the two populations were very similar and group differences were small, as previously described. (Sharpe et al., 2020) Recruitment focused on seven census tracts (i.e., four tracts in Location 1 and three tracts in Location 2) in South Carolina, six of which were USDA-designated urban food deserts (tracts of low income and low access)(Dutko, Ploeg, & Farrigan, 2012; Ploeg et al., 2009; USDA, 2017). We refer to these communities as disadvantaged due to the high poverty and food desert status of the areas. To accommodate local community neighborhood definitions, which extend beyond the geographic boundaries of census tracts, we extended the boundaries for participant eligibility to 1 mile past the seven recruitment tract boundaries into adjacent tracts, but only if the adjacent tract had a poverty level greater than or equal to that of the state (≥16%, range, 16.7% to 62.3%; (USDA, 2017). This yielded an expanded participant eligibility area that included residents of 19 tracts, of which 12 were food deserts. In total, 82.5% of the study sample lived in urban food deserts.

Data were collected at time 1 (baseline) between November 2013 and May 2014, time 2 (first follow-up) between April and August 2015, and time 3 (second follow-up) between May and August 2016. The targeted enrollment goal of 560 at baseline for the evaluation study was determined based on pre-specified hypotheses, statistical power analysis, and assumptions about retention rates over time. Using purchased address lists from a survey sampling firm, letters addressed to the ‘family food shopper’ were mailed to all eligible addresses (n=6,136) with an invitation to call for information about the study. In-person, printed, and electronic recruitment strategies followed this initial letter. Eligibility screening occurred by phone or in-person at community locations. Interviews were conducted with 527 participants at time 1 prior to the food hub’s opening, with 429 participants at time 2 (after the food hub had opened), and 408 at time 3. Analysis was restricted to 397 participants with complete food security data for all descriptive analyses. For analyses of predictors, 12 individuals with missing data on select predictor variables (e.g. marital status, income, food assistance, self-reported health) were excluded, resulting in a sample of 385 for multivariate analysis.

Participants were asked their age, race/ethnicity, gender, marital status, having children in the household for whom they shopped, their total household income in the past year, highest level of education, employment status and whether the household received Supplemental Nutrition Assistance Program (SNAP) benefits during the past 12 months. Additionally, self-reported health status was queried in five levels (excellent, very good, good, fair, and poor). Household food security was assessed using the 18-item USDA US-Household Food Security Survey Module (HFSSM), which queries the past 12 months (Bickel et al., 2000). We refer to the food security assessments by the year of data collection (i.e. 2013–14, 2015, 2016), recognizing that the questions refer retrospectively to the previous 12 months, a period that varies by participant, depending on the participants’ dates of data collection (interviews).

This study was reviewed and approved by the Institutional Review Board of the University of South Carolina. Verbal informed consent was obtained from all participants and was witnessed and formally recorded.

Statistical Analysis

Household food security (possible responses range from 0–10 affirmations for households without children and 0–18 affirmations for households with children) was classified into four categories: high food security (zero affirmations), marginal (1–2 affirmations), low (3–7 affirmations among households with children, 3–5 among households without children), or very low food security (≥8 affirmations in households with children, ≥ 6 affirmations in households without children (Bickel et al., 2000; Coleman-Jensen et al., 2017a). Subsequently we created a dichotomized household food security variable, combining high and marginal into the food secure group and low or very low into the food insecure group (Bickel et al., 2000).

Eight trajectories were defined: (1) Persistent food security; (2) recent change from secure to insecure; (3) persistent change from secure to insecure; (4) two changes ending in secure; (5) two changes ending in insecure; (6) persistent change from insecure to secure; (7) recent change from insecure to secure; and (8) persistent food insecurity. Proportions and 95% confidence intervals were calculated. This grouping represents the conceptual extension of the categorization of two data points on food security utilized by previous investigators (i.e. “persistently food secure,” “persistently food insecure,” “becoming food secure,” “becoming food insecure”). (Hernandez & Jacknowitz, 2009; Jansen et al., 2017; Jones & Frongillo, 2006; Whitaker & Sarin, 2007)

For the analysis of predictors of food insecurity we used multinomial logistic regression. The outcomes of interest were derived from the trajectories above: persistent food security (group 1, reference); persistent food insecurity (group 8); in addition to two groups that entailed transient food insecurity, i.e. the predominantly food insecure (i.e. food insecure at two of three time points, groups 3, 5 and 7); and the food insecure at a single time point (groups 2, 4 and 6). Our motivation was that we wished to characterize the degree of persistence, similar to Wilde et al. (Wilde et al., 2010) and because the transient versus food secure may differ in their predictors.

Predictors included demographic and socioeconomic variables, receipt of food assistance and self-reported health status at baseline. Variables with more than two levels were categorized as follows. Marital status was dichotomized as married or living together with a partner vs. not married (reference group, which included never married, divorced, widowed, or separated individuals). Education and income were coded as three-level variables: Having completed high school or a GED, having completed some college or more education than high school/GED, versus not having completed high school (reference group); Annual household income > $20,000, $10,000-$19,999, less than $10,000 (reference group). Self-rated health status was dichotomized as fair health (including poor, fair, and good health) versus very good or excellent (reference group) because research has shown that self-reported health varies with sex, age, income and education in that women, older individuals, those with lower income and those with less education tend to over-rate their health status in interview situations.(Layes, Asada, & Kepart, 2012)

Because no meaningful differences in food security changes were observed between the two communities, all results are presented for the overall sample. Analyses were performed using SAS version 9.4. (SAS. Statistical Analysis System, 2013).

Results

As shown in Table 1, the study sample was predominantly female (81%) and African American (94%) and had an average age of 54 years (standard deviation, SD, 13.6 years); 17% were married or living with a partner. Most households had two or more adults (53.4%) and 68.5% had no children in the household. A large percentage participated in SNAP in the last 12 months (64%), had completed high school or a lower level of education (69%), and had an annual household income less than $20,000 (77%).

Table 1.

Demographic characteristics of 397 household food shoppers from two disadvantaged communities in South Carolina (2013 – 2014)

Characteristics Total sample
Age (years), mean, SD 54.1 13.6
Adults in household, mean, SD 1.7 0.8
Women, % 81.1
African American, % 94.0
Marital status, %
  Married or unmarried couple living together 16.9
  Never married, divorced, widowed, or separated 83.1
Households without children, % 68.5
Education, %
  Less than high school 30.0
  High school/GED* 39.3
  Some college or more education 30.7
Income, %
  $0–9,999 44.0
  $10,000–19,999 32.9
  >$20,000 23.1
Participation in SNAP**, past year % 63.5
Self-reported health, %
  Very good or excellent 21.5
  Poor, fair, or good 78.5
Open in a new tab
*

General Education Diploma

**

Supplemental Nutrition Assistance Program (SNAP)

Household food security characteristics at all three data collection time points are presented in Table 2. The percent of food insecure households (i.e. either having low or very low food security) decreased from 61% at baseline to 54% in 2015 to 51% in 2016. This trend was consistently observed in both study communities. The duration between data collection at baseline (2013–2014) and 2015 interviews was on average 464 days (SD 44 days) and 366 days between the 2015 and 2016 interviews (SD 19 days).

Table 2.

Household food security characteristics of 397 households in two disadvantaged communities in South Carolina assessed at three time periods

Household Food Security Assessment Period
2013 – 2014* 2015 2016
Household food security status n % n % n %
Food secure
High food security 72 18.1 84 21.1 122 30.7
Marginal food security 81 20.4 98 24.7 73 18.4
Food insecure
Low food security 126 31.8 127 32.0 125 31.5
Very low food security 118 29.7 88 22.2 77 19.4
Open in a new tab

Food secure: High or marginal household food security

Food insecure: Low or very low household food security

*

The heading 2013–2014 refers to the first data collection period of November 2013 through May 2014 with 20% of the total interviews occurring in the last two months of 2013.

Food security trajectories over time are presented in Table 3. Persistent food insecurity at all three time periods was the dominant experience (37%, 95% CI 32.3–41.7%), 36% of the sample experienced transient food insecurity but only 27% (95%CI 22.6–31.4%) experienced persistent food security. Only 9.3% of the households (n=37) experienced two changes in food security status (i.e. from food secure to insecure to secure, or from food insecure to secure to insecure). Among those households that were food insecure at baseline, 60% remained food insecure at both subsequent assessments. For those households with very low food security at baseline, 78% remained in the persistent food insecurity group (data not shown). Additionally, Table 3 addresses the question of continuous food insecurity over consecutive assessment periods. Whereas overall 61% of households were classified as food insecure in the first observation period (2013–2014), 46% were food insecure in both the first and second (2015 and 2016), and 37% food insecure in all three assessment periods.

Table 3.

Percentage of households experiencing different household food security trajectories over time (2013 – 2016)

Household Food Security Assessment Period
Definitions of Trajectories 2013 – 2014 2015 2016 Households N=397
n % 95% CI
Persistent food security at all time points S S S 107 27.0 22.6–31.4
Recent change from secure to insecure S S I 13 3.3 1.5–5.1
Persistent change from secure to insecure S I I 19 4.8 43.1–52.9
Two changes ending in secure S I S 14 3.5 1.7–5.3
Two changes ending in insecure I S I 23 5.8 3.5–8.1
Persistent change from insecure to secure I S S 39 9.8 6.9–12.7
Recent change from insecure to secure I I S 35 8.8 6.0–11.6
Persistent food insecurity at all time points I I I 147 37.0 32.3–41.7
Open in a new tab

Food secure (S): High or marginal food security

Food insecure (I): Low or very low food security

Table 4 presents results of the evaluation of predictors of persistent and transient food insecurity. Female sex, married or living together with a partner, receiving SNAP benefits and fair health status were significantly associated with increased odds of persistent food insecurity. Whereas the point estimates for higher education and income were suggestive of inverse associations, they did not reach statistical significance. The only predictors of the transient food insecurity states were income greater than $20,000 (protective) compared to less than $10,000 and fair health for the predominantly food insecure trajectory and only fair health for the trajectory that entailed only a single food insecure time point. Fair self-reported health was the one variable consistently associated with all three food insecurity trajectories (at a single timepoint: OR 2.6, 95% CI 1.2–5.8); being predominantly food insecure: OR 2.2, 95% CI 1.1–4.5; and persistent food insecurity: OR 4.4, 95% CI 2.2–8.8).

Table 4.

Association of baseline characteristics with food security trajectories (2013–2016) (n=385)

Persistent food security Transient food insecurity Persistent food insecurity
Food insecure at a single timepoint Predominantly food insecure
Characteristics OR (95% CI) OR (95% CI) OR (95% CI)
Age (years) Reference 0.99 (0.96, 1.02) 0.99 (0.97, 1.02) 1.01 (0.98, 1.03)
Female vs. male Reference 0.85 (0.38, 1.91) 0.91 (0.41, 2.02) 2.67 (1.23, 5.83)
African American vs. other Reference 0.87 (0.22, 3.46) 1.20 (0.31, 4.72) 0.82 (0.25, 2.67)
Married or living together vs. not married†† Reference 0.82 (0.30, 2.25) 1.16 (0.47, 2.89) 2.44 (1.13, 5.27)
Children in household vs. no children Reference 0.66 (0.28, 1.54) 0.47 (0.21, 1.04) 0.74 (0.36, 1.52)
High school/GED vs. less than high school Reference 1.03 (0.46, 2.31) 0.93 (0.41, 2.08) 1.15 (0.57, 2.34)
Some college or more education vs. less than high school Reference 0.48 (0.20, 1.17) 0.76 (0.33, 1.74) 0.48 (0.22, 1.01)
$10,000–19,999 vs. $0–9,999 income Reference 1.14 (0.52, 2.52) 0.87 (0.41, 1.86) 0.84 (0.42, 1.66)
>$20,000 vs. $0–9,999 income Reference 0.72 (0.28, 1.85) 0.37 (0.14, 0.96) 0.56 (0.24, 1.27)
SNAP receipt vs. not Reference 1.47 (0.66, 3.25) 1.18 (0.55, 2.56) 3.83 (1.87, 7.84)
Fair, good or poor vs. very good or excellent self-reported health Reference 2.65 (1.22, 5.75) 2.17 (1.05, 4.48) 4.39 (2.18, 8.83)
Open in a new tab
†

Not married group includes never married, divorced, widowed, separated

Values marked in bold are statistically significant at p<0.05

Discussion

Similar to national trends (Coleman-Jensen et al., 2017a), the percent of households that were food insecure in the previous 12 months decreased markedly between 2013 and 2016, from 61% to 51% in the two disadvantaged communities. Improvements in the economy, including an unemployment rate that returned to levels prior to the 2007–2009 recession, are likely contributors to this trend (Bauer & Schanzenbach, 2018). The decrease was substantially larger in this study than the 2% decrease nationally and 1.1% in South Carolina (Coleman-Jensen et al., 2017a). This difference was to be expected given the higher than average prevalence of household food insecurity in the disadvantaged communities, compared to high-risk groups described nationally, which includes households nationally with incomes below 185% of the federal poverty threshold (31.6% food insecure) or households headed by non-Hispanic Blacks (22.5%)(Coleman-Jensen et al., 2017a). Small-area or community-level data can thus provide a very different perspective than national or state-wide data, which may mask more localized patterns.

Food assistance benefits are unlikely to explain the overarching improving trend in food security, as benefits were cut in South Carolina in fall 2013 due to the expiration of a recession-triggered boost to SNAP and cut again in spring 2016 when they were restricted to 3 months every 3 years accompanying the reinstatement of the work requirement for able-bodied adults without dependents (Concannon, 2013). It is possible that reliance on emergency, supplemental, or subsidized food sources, which was very common in this sample, buffered the cuts to benefits (Sharpe et al., 2018). However, food insecurity levels were still high in the summer of 2016 with 51% of the sample affected.

Moreover, more than one-third (37%) of the households experienced persistent food insecurity at all three time points, 46% were food insecure at the first and second assessment period, 36% transitioned between food security states and only 27% had persistent food security. To the best of our knowledge, this magnitude of food insecurity over time has only been reported in the US among households of children enrolled in the Head Start in Michigan which reported 54% persistent food insecurity over a one-year period, 36% transient food insecurity and 10% persistent food security.(Jansen et al., 2017) Estimates of other studies are substantially lower and range from 1% to 15% for persistent food insecurity (Hernandez & Jacknowitz, 2009; Whitaker & Sarin, 2007; Wilde et al., 2010).

It is important to remember that the HFSSM queries whether the food security-related experience or behavior ever occurred in the past 12 months; thus, no direct inferences can be made related to frequency of occurrence or duration of the food insecurity experience or behavior within that 12-month window. Thus, we do not imply that those who experienced persistent food insecurity were food insecure every day or every month of the year. However, it is likely that this group was food insecure in the majority of the time period, because analyses of supplemental data from the Current Population Survey Food Security Supplement (CPS-FSS) have shown that among food insecure households, food insecurity was experienced in seven months during the year (Coleman-Jensen et al., 2017a).

Similar to Wilde et al. (Wilde et al., 2010), we observed that the proportion of households that experienced persistent food insecurity decreased over the observation period, from 61% of households being classified as food insecure in the first observation period, 46% food insecure in first and second observation period and 37% food insecure in all three observation periods. However, the level and rate at which this occurred was very different compared to the national data (Wilde et al., 2010). In South Carolina, the proportion with persistent food insecurity decreased from 61% to only about half of that (37%). This contrasts with the CPS’s results, showing that the persistent group decreased to less than a quarter of its initial level of 8.6%. Thus, it seems that the conclusion of the 2016 USDA household food security report that “…when food insecurity occurs in US households it is usually recurrent but not chronic” (Coleman-Jensen et al., 2017a) may not be tenable for this study population. Given that 60% of those who were food insecure at baseline remained so throughout the study, and this proportion was even higher at 78% for those who had very low food security at baseline (data not shown), we instead conclude that in the disadvantaged communities studied in South Carolina, food insecurity was predominantly a chronic condition (Nord et al., 2002).

The evaluation of predictors of food insecurity trajectories in the sample revealed that female sex, being married or living together with a partner, and fair self-reported health status were significantly associated with increased odds of persistent food insecurity over time. That women are more likely to be food insecure has been shown in many studies (Coleman-Jensen et al., 2017a; Ribar & Hamrick, 2003), but this study adds that this association extends to the experience of persistent food insecurity for which the women were at 160% higher risk compared to men. The findings with respect to income and education were also very consistent with the literature (Coleman-Jensen et al., 2017a; Heflin & Butler, 2013; Kennedy, Fitch, Warren, & Rivera Drew, 2013; Wilde et al., 2010). Bearing in mind that all income categories were in the very low income range, the inverse association of higher income was statistically significant for the predominantly food insecure group (i.e. repeated exposure to food insecurity) and point estimates were in a similar direction (i.e. << 1.0) but not statistically significant for the persistent food insecurity category and the group experiencing food insecurity in a single year. Likewise, having some college education was also inversely associated with the different food insecurity trajectories though did not reach statistical significance. Lastly, receiving SNAP benefits was associated with persistent food insecurity, a finding typical in observational studies, in that although food assistance alleviates the burden of food insecurity is it often the most food insecure households that enroll for federal food assistance programs (Fox, Hamilton, & Lin, 2004).

One unexpected findings was that being married or living together with a partner was associated with increased risk of persistent food insecurity, which contrasts with previous work showing a protective effect of the presence of a married couple in a household (Wilde et al., 2010). Recent work has shown that among women, intimate partner violence is significantly associated with more severe food insecurity (Chilton & Booth, 2007; Hernandez, Marshall, & Mineo, 2014; Lown, Schmidt, & Wiley, 2006; Ricks, Cochran, Arah, Williams, & Seeman, 2016). Moreover, unmarried women living with a partner had higher risks of experiencing violence than married women but this difference was affected by severity of food insecurity (Ricks et al., 2016). Another potential explanation for this finding is that in disadvantaged communities such as ours, unemployment, disabilities, or chronic health needs of a woman’s partner may contribute to additional financial strain, thereby increasing the likelihood of persistent food insecurity in partnered women (Drucker et al., 2019; Jackson & Vaughn, 2017). It is therefore conceivable that the effect attributed to marital status may be due to unmeasured confounding by these partner-related attributes.

Finding self-reported health to be the predictor most strongly associated with persistent food insecurity merits discussion as this measure has rarely been used in food security research. Instead, disabilities have been the focus of previous research, demonstrating very consistent associations with increased risk of food insecurity (Schwartz, Buliung, & Wilson, 2019). Wilde et al. found the presence of a person with a disability in the household to be associated with increased odds of persistent food insecurity (Wilde et al., 2010). The significant association of fair self-reported health with all three food insecurity trajectories, with odds ratios ranging from 2.2 to 4.4, suggests that health states that are likely not as severe as a documented disability can have similar adverse effects on a household’s food security status over time. Fair self-rated health has been shown to be associated with increased mortality, with somewhat stronger associations in whites than African Americans which have not been explained to date (Assari, Lankarani, & Burgard, 2016; Mossey & Shapiro, 1982). Thus, it is noteworthy that in the overarchingly African American sample a strong and consistent association of fair self-reported health and all food insecurity trajectories emerged.

Several cohort studies have used more advanced statistical models to capitalize on three or more consecutive assessments of food insecurity. These include a study of farmworkers that reported that households had a 29% probability of staying in the same, least food secure state over two consecutive quarters, whereas households in the most secure state has a 85% probability of remaining in this state (Ip et al., 2015; Wilde et al., 2010). A study covering Kindergarten through 8th grade in the ECLS study found only 1% of the sample was food insecure in four consecutive years and 3% in three years, with 79% persistent food security (Ryu & Bartfeld, 2012). Similarly, Wilde et al. analyzed the Survey of Program Dynamics (SPD) panel, which evaluated how US households experience food insecurity between 1998 and 2001 (Wilde et al., 2010). They concluded that only about 1% of the US population had persistent food insecurity for four or five consecutive years in the nationally representative SPD sample (Wilde et al., 2010). Thus, past research on cohort studies seems to suggest that food insecurity is typically more of an episodic or transient phenomenon. However, the one published study on a particularly disadvantaged population, a Head Start sample, which found 54% persistent food insecurity, suggests that there may be need for more in depth study (Jansen et al., 2017).

Our study has several limitations. Our study did not ascertain information on health insurance status, employment at baseline, housing, transportation, mental health of the adults, intimate partner violence or other partner-related attributes such as employment, disabilities and chronic conditions, all of which may have an impact on food security (Fleischer et al., 2018). There was variability in the time interval between the three assessments; thus, we cannot exclude the possibility that changes in food security occasionally occurred outside of the 12-month timeframe queried. For the definition of trajectories, we opted for a dichotomous definition of food insecurity at each time point, recognizing that using more categories might have revealed more subtle patterns but would also have substantially complicated the analysis (Grineski et al., 2018; Ip et al., 2015). Given that the food hub evaluation study focused on information from the household’s primary food shopper, the present analysis needs to be understood as a cohort analysis of a select group of households, not a representative sample of all households in the study areas. Thus, the sample was heavily skewed toward women who traditionally take on more of the role of food shoppers. Comparison of select demographic characteristics of this sample (94% African American, 68.5% households without children) with the Census 2010-based tract characteristics (90% African American, 62% households without children (Bureau, 2010), revealed substantial similarities on these additional demographic attributes, which supports the notion that this sample may be a reasonable representation of female food shoppers in disadvantaged communities in the South (United States Census Bureau, 2010) .

Among the strengths of this study are longitudinal design, which moves beyond the predominantly static view of food insecurity in cross-sectional studies, and the very high overall retention of 77.4% from baseline to end of study, and retention of 83.3% in the first interval and 92.9% in the second interval, which speak to the internal validity of our study. Last but not least, we evaluated a significant number of demographic and socio-economic characteristics and self-reported health as potential predictors of food insecurity trajectories.

Conclusion

In this sample, a substantial proportion (37%) of households remained food insecure across the three-year time span, despite improvements in the economy observed in South Carolina and nationally, while 36% experienced transient food insecurity. Thus, we feel that in these disadvantaged communities, food insecurity is clearly a chronic condition. Of the various predictors of food insecurity trajectories evaluated, self-rated health emerged as the key risk factor for persistent and transient food insecurity trajectories. Our study has several implications. Given that we found differences between the predictors of transient versus persistent food insecurity, it stands to reason that there may be specific causal pathways that distinguish between these trajectories which should be explored in future longitudinal research. Because self-rated health was a significant predictor of all food insecure trajectories, this construct merits further investigation to unpack its causal underpinnings, which may include combinations of mental and physical health challenges and yet-to-be discovered factors. Given that the proportion with transient food insecurity over time was notable at 36% and this group, by definition, transitions in and out of food insecurity, tailored interventions may have a higher likelihood of success. Finally, the extent of persistent food insecurity in our study was quite high at 37% and indicative of likely chronicity, suggesting that much more strenuous efforts are needed to identify interventions to aid this particularly disadvantaged population.

Supplementary Material

1

Acknowledgments:

The authors appreciate the contributions from the university and local project staff and students in recruitment, retention and data collection, and assistance from the Hub City Farmers’ Market, Butterfly Foundation, Northside Development Group, Via College of Osteopathic Medicine, the Soulfully Fit Committee, and many community leaders and community-based and governmental organizations.

This work was supported by the National Cancer Institute (NCI) under Grant Number R01CA180336 (PA Sharpe, S Wilcox, Principal Investigators). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views or policies of the National Cancer Institute or the National Institutes of Health. NCI had no role in the design, analysis or writing of this article.

Financial Support

This work was funded by the National Cancer Institute (NCI) under Award Number R01CA180336 (PA Sharpe, S Wilcox, Principal Investigators). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views or policies of the National Cancer Institute or the National Institutes of Health. NCI had no role in the design, analysis or writing of this article.

Abbreviations

HFSSM

Household Food Security Survey Model

SNAP

Supplemental Nutrition Assistance Program

TANF

Temporary Assistance for Needy Families

US

United States

USDA

United States Department of Agriculture

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Ethical Standards Disclosure

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Institutional Review Board of the University of South Carolina. Verbal informed consent was obtained from all subjects and was witnessed and formally recorded.

Conflict of Interest

None

Declarations of Interest: None

Data deposition:

Not applicable.

Data availability statement:

The data will be available upon acceptance for publication of final results in accordance with the National Institutes of Health Grants Policy Statement, Section 8.2.3.1 Data Sharing Policy of Section 8.2 Availability of Research Results: Publications, Intellectual Property Rights, and Sharing Research Resources https://grants.nih.gov/grants/policy/nihgps/html5/section_8/8.2_availability_of_research_results_publications__intellectual_property_rights__and_sharing_research_resources.htm

References:

  1. Assari S, Lankarani MM, & Burgard S (2016). Black-white difference in long-term predictive power of self-rated health on all-cause mortality in United States. Annals of Epidemiology, 26(2), 106–114. 10.1016/j.annepidem.2015.11.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bauer L, & Schanzenbach DW (2018). Children’s exposure to food insecurity is still worse than it was before the Great Recession.
  3. Bickel G, Nord M, Price C, Hamilton W, & Cook J (2000). Guide to Measuring Household Food Security, Revised 2000. In United States Department of Agriculture, Food and Nutrition Service. [Google Scholar]
  4. Bureau, U. S. C. (2010). DP-1 Profile of General Population and Housing Characteristics: 2010, 2010 Demographic Profile Data.
  5. Chilton M, & Booth S (2007). Hunger of the Body and Hunger of the Mind: African American Women’s Perceptions of Food Insecurity, Health and Violence. Journal of Nutrition Education and Behavior, 39(3), 116–125. 10.1016/j.jneb.2006.11.005 [DOI] [PubMed] [Google Scholar]
  6. Coleman-Jensen A, Rabbitt MP, Gregory CA, & Singha A (2017a). Household Food Security in the United States in 2016. United States Department of Agriculture, Econmic Researh Service, ERR-237, 1–36. [Google Scholar]
  7. Coleman-Jensen A, Rabbitt MP, Gregory CA, & Singha A (2017b). Household Food Security in the United States in 2016 (pg 11). United States Department of Agriculture, Econmic Researh Service, ERR-237, 11. [Google Scholar]
  8. Concannon K (2013). Helping SNAP Recipients Prepare for November 1st Benefit Changes.
  9. Drucker ER, Liese AD, Sercy E, Bell BA, Draper C, Fleischer NL, … Jones SJ (2019). Food insecurity, childhood hunger and caregiver life experiences among households with children in South Carolina, USA. Public Health Nutrition, 22(14), 2581–2590. 10.1017/S1368980019000922 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dutko P, Ploeg M. Ver, & Farrigan T (2012). Characterics and Influential Factors of Food Deserts, ERR-140.
  11. Fleischer NL, Liese AD, Hammond R, Coleman-Jensen A, Gundersen C, Hirschman J, … Jones SJ (2018). Using systems science to gain insight into childhood food security in the United States: Report of an expert mapping workshop. Journal of Hunger and Environmental Nutrition, 13(3), 362–384. 10.1080/19320248.2017.1364194 [DOI] [Google Scholar]
  12. Fox MK, Hamilton W, & Lin B-H (2004). Effects of Food Assistance and Nutrition Programs on Nutrition and Health. In Economic Research Service/USDA (Vol. 4). [Google Scholar]
  13. Grineski SE, Morales DX, Collins TW, & Rubio R (2018). Transitional dynamics of household food insecurity impact children’s developmental outcomes. Journal of Developmental and Behavioral Pediatrics, 39(9), 715–725. 10.1097/DBP.0000000000000598 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Heflin CH, & Butler JS (2013). Why Do Women Enter and Exit From Material Hardship? Journal of Family Issues, 34(5), 631–660. 10.1177/0192513X12442822 [DOI] [Google Scholar]
  15. Hernandez DC, & Jacknowitz A (2009). Transient, but not persistent, adult food insecurity influences toddler development. Journal of Nutrition, 139(8), 1517–1524. 10.3945/jn.109.105593 [DOI] [PubMed] [Google Scholar]
  16. Hernandez DC, Marshall A, & Mineo C (2014). Maternal depression mediates the association between intimate partner violence and food insecurity. Journal of Women’s Health, 23(1), 29–37. 10.1089/jwh.2012.4224 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Ip EH, Saldana S, Arcury TA, Grzywacz JG, Trejo G, & Quandt SA (2015). Profiles of food security for US farmworker households and factors related to dynamic of change. American Journal of Public Health, 105(10), e42–e47. 10.2105/AJPH.2015.302752 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Jackson DB, & Vaughn MG (2017). Parental History of Disruptive Life Events and Household Food Insecurity. Journal of Nutrition Education and Behavior, 49(7), 554–560.e1. 10.1016/j.jneb.2017.04.010 [DOI] [PubMed] [Google Scholar]
  19. Jansen EC, Kasper N, Lumeng JC, Brophy Herb HE, Horodynski MA, Miller AL, … Peterson KE (2017). Changes in household food insecurity are related to changes in BMI and diet quality among Michigan Head Start preschoolers in a sex-specific manner. Social Science and Medicine, 181, 168–176. 10.1016/j.socscimed.2017.04.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Jones SJ, & Frongillo EA (2006). The Modifying Effects of Food Stamp Program Participation on the Relation between Food Insecurity and Weight Change in Women. The Journal of Nutrition, 136(4), 1091–1094. 10.1093/jn/136.4.1091 [DOI] [PubMed] [Google Scholar]
  21. Jyoti DF, Frongillo EA, & Jones SJ (2005). Food Insecurity Affects School Children’s Academic Performance, Weight Gain, and Social Skills. The Journal of Nutrition, 135(12), 2831–2839. 10.1093/jn/135.12.2831 [DOI] [PubMed] [Google Scholar]
  22. Kennedy S, Fitch CA, Warren JR, & Rivera Drew JA (2013). Food Insecurity During Childhood: Understanding Persistence and Change Using Linked Current Population Survey Data. University of Kentucky Center for Poverty Research Discussion Paper Series, DP2013–03, 2013–2016. [Google Scholar]
  23. Laraia BA (2013). Food Insecurity and Chronic Disease. Advances in Nutrition, 4(2), 203–212. 10.3945/an.112.003277 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Layes A, Asada Y, & Kepart G (2012). Whiners and deniers - What does self-rated health measure? Social Science and Medicine, 75(1), 1–9. 10.1016/j.socscimed.2011.10.030 [DOI] [PubMed] [Google Scholar]
  25. Liese A, Ma X, Hutto B, & Et A (2017). Food Shopping and Acquisition Behaviors in Relation to BMI among Residents of Low-Income Communities in South Carolina. International Journal of Environmental Research and Public Health, 14(9), 1075. 10.3390/ijerph14091075 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Lown EA, Schmidt LA, & Wiley J (2006). Interpersonal violence among women seeking welfare: Unraveling lives. American Journal of Public Health, 96(8), 1409–1415. 10.2105/AJPH.2004.057786 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Ma X, Liese AD, Hibbert J, & Et A (2017). The Association between Food Security and Store-Specific and Overall Food Shopping Behaviors. Journal of the Academy of Nutrition and Dietetics, 117(12), 1931–1940. 10.1016/j.jand.2017.02.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Mossey JM, & Shapiro E (1982). Self-rated health: a predictor of mortality among the elderly. American Journal of Public Health, 72(8), 800–808. 10.2105/AJPH.72.8.800 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Nord M, Andrews M, & Winicki J (2002). Frequency and duration of food insecurity and hunger in US households. Journal of Nutrition Education and Behavior, 34(4), 194–200. 10.1016/S1499-4046(06)60093-6 [DOI] [PubMed] [Google Scholar]
  30. Ploeg M. Ver, Breneman V, Farrigan T, Hamrick K, Hopkins D, Kaufman P, … Kim S (2009). Access to Affordable and Nutritious Food: Measuring and Understanding Food Deserts and Their Consequences, Report to Congress.
  31. Ribar DC, & Hamrick KS (2003). Dynamics of Poverty and Food Sufficiency. In Economic Research Service/USDA. [Google Scholar]
  32. Ricks JL, Cochran SD, Arah OA, Williams JK, & Seeman TE (2016). Food insecurity and intimate partner violence against women: Results from the California Women’s Health Survey. Public Health Nutrition, 19(5), 914–923. 10.1017/S1368980015001986 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ryu JH, & Bartfeld JS (2012). Household food insecurity during childhood and subsequent health status: The Early Childhood Longitudinal Study - Kindergarten Cohort. American Journal of Public Health, 102(11). 10.2105/AJPH.2012.300971 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. SAS. Statistical Analysis System. (2013). Cary, NC, USA: SAS Institute Inc. [Google Scholar]
  35. Schwartz N, Buliung R, & Wilson K (2019). Disability and food access and insecurity: A scoping review of the literature. Health & Place, 57, 107–121. [DOI] [PubMed] [Google Scholar]
  36. Seligman HK, Laraia BA, & Kushel MB (2010). Food insecurity is associated with chronic disease among low-income NHANES participants. The Journal of Nutrition, 140(2), 304–310. 10.3945/jn.109.112573 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Sharpe PA, Bell BA, Liese AD, Wilcox S, Stucker J, & Hutto BE (2020). Effects of a food hub initiative in a disadvantaged community: A quasi-experimental evaluation. Health and Place, 63, 102341. 10.1016/j.healthplace.2020.102341 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Sharpe PA, Liese AD, Bell BA, Wilcox S, Hutto BE, & Stucker J (2018). Household food security and use of community food sources and food assistance programs among food shoppers in neighborhoods of low income and low food access. Journal of Hunger & Environmental Nutrition, 13(4), 482–496. 10.1080/19320248.2017.1364188 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. United States Census Bureau. (2010). DP-1 Profile of General Population and Housing Characteristics: 2010, 2010 Demographic Profile Data.
  40. USDA. (2017). Food Access Research Atlas.
  41. Whitaker RC, & Sarin A (2007). Change in food security status and change in weight are not associated in urban women with preschool children. The Journal of Nutrition, 137(9), 2134–2139. 10.1093/jn/137.9.2134 [DOI] [PubMed] [Google Scholar]
  42. Wilde PE, Nord M, & Zager RE (2010). In Longitudinal Data From the Survey of Program Dynamics, 16.9% of the U.S. Population Was Exposed to Household Food Insecurity in a 5-Year Period. Journal of Hunger & Environmental Nutrition, 5(3), 380–398. 10.1080/19320248.2010.504115 [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1666860-supplement-1.xml (250B, xml)

Data Availability Statement

The data will be available upon acceptance for publication of final results in accordance with the National Institutes of Health Grants Policy Statement, Section 8.2.3.1 Data Sharing Policy of Section 8.2 Availability of Research Results: Publications, Intellectual Property Rights, and Sharing Research Resources https://grants.nih.gov/grants/policy/nihgps/html5/section_8/8.2_availability_of_research_results_publications__intellectual_property_rights__and_sharing_research_resources.htm

RESOURCES