Assessment of risk for food insecurity among African American urban households: utilizing cumulative risk indices and latent class analysis to examine accumulation of risk factors

Abstract

African American caregivers in low-income, urban communities have high rates of food insecurity. Unemployment, education, smoking, stress, and depressive symptoms are associated with household food insecurity. A cumulative risk model suggests that accumulation of risk may compound food insecurity risk, and certain risk factors are more likely to co-occur. This study utilizes two approaches to examine food insecurity risk among African American caregivers with an adolescent daughter—a cumulative risk index to examine accumulation of risk and food insecurity risk; latent class analysis (LCA) to determine if certain risk profiles exist and their relation to food insecurity risk. Caregivers completed surveys including demographic, psychosocial, and behavioral questions (to create a cumulative risk index) and a validated 2-item food insecurity screen. LCA was used to identify risk profiles. Logistic regression was used to examine relations between cumulative risk, risk profiles, and food insecurity risk. Each additional cumulative risk index factor was associated with a 54% increase in odds of risk of food insecurity. LCA identified three subgroups: high stress/depression (class #1), low education/low stress and depression (class #2), and low risk overall (class #3). Odds of food insecurity risk were 4.7 times higher for class #1, and 1.5 times higher for class #2 compared with class #3. This study contributes to understanding of how food insecurity risk relates to cumulative risk and risk profiles. Findings can be used to improve food insecurity risk screening in clinical settings, enhancing intervention/referral for food security risk and mental health among African American caregivers and their households.

Implications.

Practice: Practitioners should incorporate screening for both mental health and food insecurity in primary care settings.

Policy: Policymakers should consider incorporating mental health screening/treatment and educational supports for recipients of food assistance programs.

Research: Future research should examine cumulative risk among different populations—specifically rural communities, immigrant communities, and families with special needs.

Food insecurity occurs when households lack adequate access to sufficient food for maintaining a healthy, active life [1]. Based on 2017 estimates, approximately 12% of U.S. households are food insecure, with greatest risk among households with children (16%), households under 185% of the federal poverty guidelines (31%), and African American households (22%) [2]. Food insecurity is associated with adverse health outcomes among adults, including poor overall health [3], hypertension [4], diabetes [5], and metabolic syndrome [6]. Among adolescents, food insecurity is associated with difficulty accessing healthcare, education, and housing [7]; increased risk of mental illness [8], substance use, and income insecurity [7]; and greater risk of lower health-related quality of life [9]. Adolescents establish lifelong dietary practices as they transition into independence [10], and food insecure households are less likely than food secure households to have access to healthy foods [11].

Several caregiver-specific demographic, psychosocial, and behavioral risk factors are associated with overall household food insecurity: unemployment/underemployment [12], low education [13], depression [14], stress [15], and smoking [16]. Households with an adult who is unemployed or underemployed have a higher likelihood of being food insecure when compared with households with working or retired adults [17]. Having an education through high school is associated with lower rates of food insecurity, compared with not completing high school [18]. Having a smoker in the home increases likelihood of food insecurity risk [19], and food insecurity severity compared with nonsmoking households [20], with risk increasing over time [21]. Psychosocial well-being also plays a role in food security. Mothers who report food insecurity are also more likely to report symptoms of depression compared with those who are food secure [22].

Although each can be examined individually, the cumulative risk approach suggests that accumulation of factors, rather than individual characteristics, is associated with higher likelihood of negative health and well-being outcomes for children [23, 24] and adolescents [25]. Evans proposes a cumulative risk model to account for the complex interplay and natural covariation of risk factors [26]. Individual factors may have small effects when considered in isolation. However, when considered within the context of multiple factors, the impact is magnified [27–29]. For example, low education has been associated with food insecurity [18]. However, low education in the context of poverty, unemployment, and other risk factors may intensify the overall food insecurity risk. Caregivers in low-income families may be able to procure the food necessary to maintain household food security. However, in the context of stress and depression, the ability to attain the food necessary to preserve food security is seriously jeopardized [14, 15]. In addition, in most low-income communities, families are more likely to experience multiple risk factors than to experience single risk factors [26]. The cumulative risk model allows for the covariance of common risk factors that disproportionally occur in certain populations [26].

Interplay among risk factors may amplify effects on food insecurity. For example, stress increases the likelihood of smoking [30], and smoking is twice as likely among adults with depressive symptoms [31] and heightens financial strain among adults with depression [32]. Experience of any one risk factor may be related to food security status and each additional risk factor may further challenge caregivers’ ability to maintain household food security. Examining how risk accumulation relates to household food insecurity will increase the understanding of how individual risk variables work together to heighten the likelihood of food insecurity. Hernandez utilized a cumulative risk approach to examine the effect of four indices on household food insecurity: financial strain, poor maternal health/risky behaviors, family conflict, and parental disruption [33]. Higher scores on financial strain, poor maternal health/risky behavior, and parental disruption indices were associated with a higher risk of food insecurity. The current study expands on this research by examining how risk factors covary and relate to food security status. Cumulative risk models assume equal weighting of risk factors, thus additional analysis is needed to identify the relative strengths among factors associated with food insecurity risk.

Latent class analysis (LCA), an alternative to cumulative risk, may clarify relations among underlying risks [34]. LCA is a statistical approach that identifies patterns within data to detect homogenous subpopulations among a larger heterogeneous group [35]. With regard to food insecurity, identification of subgroups through LCA can pinpoint co-occurring risks that can be evaluated against food insecurity risk. In addition, classification of subgroups can assist practitioners in tailoring interventions to meet the needs of households at highest food insecurity risk. The current study examines food insecurity risk from both a cumulative risk and a LCA perspective. Research on cumulative risk with food insecurity as an outcome is limited [33] and the use of LCA to identify underlying risk profiles is novel. This study addresses food insecurity risk among low-income urban households with adolescent daughters. The study has three aims (a) to assess relations between a cumulative risk index and food insecurity risk; (b) to identify clustering among food insecurity risk factors, independent of food security risk status; and (c) to examine how latent profiles are associated with odds of food insecurity risk.

METHODS

The study is a cross-sectional analysis of baseline data collected from 2009 to 2012 as part of a randomized control trial testing an obesity prevention/health promotion intervention among girls in a large urban public school district [36]. Twenty-two schools that met inclusion criteria (public middle or K-8 school with >75% of students receiving free or reduced lunch and a population of >70% African American students) were recruited, including 789 girls. Eligibility included enrollment in sixth or seventh grade and no health or developmental conditions that prohibit physical activity or participation in a group intervention to promote dietary and physical activity behaviors. There were no exclusionary conditions for caregivers. The research was approved by institutional review boards at the university where the research was conducted and the public school district. Participating caregivers provided written informed consent for themselves and their daughters; daughters provided written assent. Caregivers completed self-administered, paper-based surveys reporting on household food insecurity risk, demographics, and psychosocial measures; 455 (58%) caregivers completed the survey.

Measures

Food insecurity risk

Household food insecurity risk was measured with a 2-item screen [36]: (a) “within the past 12 months, we worried whether our food would run out before we got money to buy more” and (b) “within the past 12 months, the food we bought just didn’t last and we didn’t have money to get more.” Answering affirmatively to either or both questions indicates household food insecurity risk, coded 0 = food secure; 1 = food insecure risk. The screen has excellent sensitivity (88.5%–98.7%), specificity (83.0%–92.7%), and convergent validity among households with young children [36], adolescents [7], and adults [37].

Cumulative risk

Caregivers were categorized as either low risk (coded 0) or high risk (coded 1) for five risk factors associated with food insecurity risk. Employment status was coded as employed full time (low risk) versus unemployed or underemployed (working less than 30 hr per week per U.S. Internal Revenue Service’s standards [38]; high risk). Education was separated by caregivers who had a high school diploma/GED or less (high risk) and caregivers who had some college or more (low risk). Having a smoker in the home was categorized as high risk and households with no smokers were categorized as low risk.

Stress was measured using the Perceived Stress Scale (PSS), with high scores representing high stress [39]. The PSS demonstrates predictive validity when compared with depressive symptoms and life-events [39]. In the current sample, the PSS demonstrated good internal consistency (Cronbach’s α = .77). Caregivers with sample-defined scores >75th percentile (scores > 28) on the 0–56 scale were categorized as high risk and caregivers ≤75th percentile (scores ≤ 28) were categorized as low risk, consistent with Sameroff’s [40] work with cumulative risk including measures without specific categories.

Depressive symptoms were measured by the 21-item Beck Depression Inventory II (BDI-II) [41]. The BDI-II displays strong criterion validity regarding measurement of depressive symptoms among African Americans [42]. The BDI-II displayed good internal consistency in the current sample (Cronbach’s α = .90). For cumulative risk, caregivers with moderate/severe depressive symptoms were categorized as high risk (scores ≥ 20) and caregivers with minimal/mild depressive symptoms, based on the standardization sample, were categorized as low risk (scores < 20).

Cumulative risk index was calculated based on caregivers’ endorsement of demographic (employment and education), behavioral (smoking), and psychosocial (stress and depression) risk factors. Caregivers received a score of 1 for each factor endorsed; cumulative risk was calculated by summing the items, yielding total scores of 0–5 (0 = no risk; 5 = maximum risk). Prorated scores were computed for 57 (13%) cases with 1 missing item, resulting in a sample of 438 cases (97%) with a cumulative risk score [43].

Covariates

Income status was comprised of multiple indications of socioeconomic status or poverty: household income at or below 135% of the federal poverty guidelines, receipt of cash assistance, receipt of free lunch or breakfast for the adolescent, or receipt of medical assistance for the caregiver. Household income data included income from work, TANF, or Social Security/Disability. Caregivers who met one or more of these conditions was categorized as having low-income (coded 1) versus not having low-income (coded 0). Race was also included, categorized as White or Asian (coded 0) and African American, Latino/Hispanic, or Other (coded 1).

Data analysis

LCA was conducted using Mplus (version 7.4) to determine underlying patterns of risk. LCA models with classes ranging from two to four were conducted, comparing each model with the model with one less class to determine if an additional class resulted in improved fit using the Lo–Mendell–Rueben (LMR) and Bootstrap Likelihood Ratio Test (BLRT). Goodness of model fit was demonstrated by p >.05 [44, 45]. Fit indices were compared across models to determine the most appropriate model fit using the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and adjusted Bayesian Information Criterion (aBIC). Smaller values of AIC, BIC, and aBIC indicate better model fit [44]. Class probability was calculated for each participant and each participant was assigned a class based on the class with the highest probability. Two multivariate logistic regression analyses were conducted using SPSS 22, each with risk of food insecurity as the dependent variable and race and income status as covariates. In the first model, the cumulative risk index ranging from 0 to 5 was the predictor variable. In the second model, the LCA results were used to create a variable that classified each participant into a risk category.

RESULTS

Sample characteristics

Most caregivers were mothers (85%) and adolescents lived fulltime in nearly all households (94%); the remaining girls split time with another caregiver. The majority of caregivers and adolescent daughters were African American (91%) and the majority were living at or below 135% of the federal poverty guidelines or receiving services that indicated low-income status (88%). Approximately half (52%) of the households experienced food insecurity risk. Households at risk of food insecurity did not differ significantly from food secure households on race or caregiver employment. However, education level (p = .004) was significantly lower, and low-income status, caregiver stress, depressive symptoms, and smoking were significantly higher among food insecurity risk households compared with food secure households (p < .001; see Table 1).

Table 1.

Comparison of sample characteristics and food insecurity risk

	Overall % (n = 450)	Food secure % (n = 203)	At risk of food insecurity % (n = 219)	χ  2	p
Overall sample		48.1	51.9
Cumulative risk indicators
Employment				0.09	.422
Employed	52.1	48.4	51.6
Unemployed/Underemployed	47.9	47.0	53.0
Education				7.46	.004
Some college or greater	55.2	49.6	50.4
High school or less	44.8	40.3	59.7
Smoking				12.10	<.001
No smoker in the home	62.4	55.6	44.4
Smoker in the home	37.6	37.1	62.9
Stress				34.56	<.001
PSS ≤75th percentile	70.0	57.4	42.6
PSS >75th percentile	30.0	26.2	73.8
Depressive symptoms				15.65	<.001
Minimal or mild	84.4	52.2	47.8
Moderate or severe	15.6	25.8	74.2
LCA class profiles				21.02	<.001
Class #3—high stress/depression	12.4	22.6	77.4
Class #2—low education/low stress/depression	40.2	45.3	54.7
Class #1—low risk overall	47.3	57.3	42.7
	Overall μ	Food secure μ	At risk of food insecurity Μ	p
Cumulative risk index	1.76	1.38	2.09	<.001

Cumulative risk

The mean cumulative risk index score was 1.76 (SD = 1.34). Households at risk of food insecurity (M = 2.09, SD = 1.37) had higher cumulative risk scores than food secure households (M = 1.38 SD = 1.18; t_412.46.= −5.13, p < .001). A logistic regression model was tested to examine relations of the cumulative risk index with food insecurity risk (see Table 2). Income status was not significantly associated with food security status in the multivariate model. Cumulative risk was significantly related to household food security status; each additional risk factor in the cumulative risk index was associated with a 54% increase in odds of food insecure risk (OR = 1.54; 95% CI: 1.30, 1.83; p < .001). Caregivers who identified as African American, Latino, or Other were 3.4 times more likely to be at risk of being food insecure than caregivers who identified as White or Asian (95% CI: 1.48, 7.83; p = .004).

Table 2.

Logistic regression analyses of cumulative risk and latent classes as predictors of food insecurity risk

	OR (p)	95% CI
Model 1 (n = 411)
Cumulative risk index	1.54 (<.001)	1.29, 1.83
Race	3.41 (.004)	1.48, 7.83
Low-income	1.88 (.066)	0.96, 3.70
Model 2 (n=412)
Class #3—low risk (reference)
Class #2—low education/low stress and depression	1.52 (.056)	0.99, 2.35
Class #1—high stress/depression	4.65 (<.001)	2.24,9.65
Race	2.96 (.009)	1.31, 6.69
Low-income	2.59 (.004)	1.34, 4.98

LCA results

A comparison of LMR, BLRT, and fit indices indicated that a three-class model provides best overall fit in these data (see Table 3). When comparing class probabilities across the two, three, and four-class model, both three- and four-class models demonstrated interpretable solutions. The three-class model was retained for further analysis and interpretation based on non-significant LMR and BLRT statistics for the four-class model (see Table 3). In the three-class model, caregivers in class #1 (n = 56, 12%) experienced high probability for stress (probability = 1.0) and moderate probability of depression (probability = 0.71). Caregivers in class #2 (n = 181, 40%) reported high probability for low education (probability = 0.99) and low probability for stress (probability = 0.18) and depression (probability = 0.10). Caregivers in class #3 (n = 213, 47%) had low probability for unemployment (probability = 0.30), low education (probability = 0.00), smoking (probability = 0.26), stress (probability = 0.16), and depression (probability = 0.01: see Fig. 1).

Table 3.

LCA model fit statistics

	Models
	Two-class	Three-class	Four-class
Parameters	11	17	23
Ho Likelihood	−1,280.99	−1,258.7	−1,254.22
AIC	2,583.985	2,551.38	2,554.437
BIC	2,629.187	2,621.24	2,648.95
aBIC	2,594.277	2,567.29	2,575.956
Pearson χ 2	67.051	19.708	10.944
Pearson χ 2df	20	14	8
Pearson χ 2p value	<.001	.1396	.2049
LR χ 2	64.63	20.026	11.082
LR χ 2 df	20	14	8
LR χ 2p value	<.001	.1293	.1971
Entropy	0.661	0.861	0.917
BLRT	−1,332.21	−1,282	−1,258.69
BLRT p value	<.001	<.001	.28
VLMR	−1,332.21	−1,281	−1,258.69
VLMR p value	<.001	.0001	.0559

Fig 1.

LCA—three-class model probabilities by risk variable

In further examining the three-class model, participants had the highest probability of being in class #3, the lowest risk group (M = 0.45, SD = 0.48), followed by class #2, the group with low education/low stress and depression (M = 0.39, SD = 0.45). Participants had the lowest probability of being in class #1, the group with high stress and depression (M = 0.16, SD = 0.30). Chi-square analyses indicate a difference in rate of food insecurity risk across class membership (χ² = 21.02, p < .001), with 77% of participants in class 1 (high stress/depression) reporting food insecurity risk compared with 55% among participants in class #2 (low education/low stress and depression), and 43% of participants in class #3 (low presence of risk factors; see Table 1).

A logistic regression model examined relations between latent class profiles and food insecurity risk (see Table 2). Class was significantly associated with household food insecurity risk: caregivers in class #1 (high stress/depression) had odds over four and a half times higher for food insecurity risk in comparison to caregivers in class #3 (low overall risk; OR = 4.65; CI 95%: 2.24, 9.65; p < .001). Caregivers in class # 2 (low education/low stress/depression) had a 52% increase in odds of food insecurity risk compared with class # 3 (OR = 1.52; CI 95% 0.99, 2.35; p = .056). Caregivers who identified as African American, Latino, or Other were three times more likely to experience food insecurity risk compared with caregivers who identified as White or Asian (CI 95%: 1.31, 6.69; p = .009) and caregivers who were identified as low-income were 2.6 times more likely to experience food insecurity risk than caregivers who were not low-income (CI 95%: 1.34, 4.98; p = .004).

Discussion

This study examined how the accumulation of known risk factors for food insecurity relate to household food insecurity risk among caregivers of adolescents. Two separate modeling approaches yielded unique and complimentary findings. In the cumulative risk model, as expected, higher cumulative risk scores were associated with higher odds of food insecurity risk, suggesting that cumulative risk is a useful tool in understanding food insecurity risk. Both Hernandez and the current study identified the cumulative role of psychosocial factors, including stress, depression, and anxiety in food insecurity [33]. Second, the LCA identified a model with the following components (three-class model): (a) low presence of risk factors, (b) low education/low stress and depression, and (c) high depression and stress. Comparing the three classes identified in the LCA, the class of low education/low stress and depression, and the class of high depression and stress both had elevated odds of food insecurity risk. Evidence suggests that depression and stress have a bidirectional relationship with food insecurity [22, 46]. Martin and colleagues (2016) report that food insecure adults, especially woman, have higher rates of mental illness (including depression), particularly in the context of stress. In qualitative interviews, low-income mothers reported that the constant stress of navigating “trade-offs” related to poverty and food insecurity contributed to depression, stress, and anxiety [47].

A cumulative risk approach provides an opportunity to assess food insecurity risk among families, allowing practitioners to screen for and monitor food insecurity risk among households who experience multiple risks. Health promotion efforts to address food insecurity risk can use a cumulative risk model to identify factors amenable to intervention [48]; addressing one risk factor may have multiplicative effects on overall risk. Greater risk matters, but which factors are experienced matter as well. Class #2 (low education [i.e., HS/GED or less] and low stress/depression) also had an increased likelihood of food insecurity risk in comparison to class #3 (overall low presence of risk factors). Although this finding could be interpreted as a function of low-income status, the overall sample was predominantly high risk for low-income (88%). Thus, education plays a role in food insecurity risk beyond socioeconomic status, consistent with prior research [18]. One possibility is that education beyond high school relates to an increase in coping strategies. Education serves to protect individuals from food insecurity, regardless of type of work [18].

Limitations and strengths

There are some important limitations to consider. The screening tool for food insecurity identifies families who are at risk of food insecurity, and does not specify the severity of food insecurity. Additionally, food insecurity is typically a transient experience [49]. In the current study, underemployment/unemployment of the survey respondent was used as a proxy for household underemployment/unemployment. It is possible that additional income earners in the household were not accounted for in calculating this risk variable. However, research among low-income families indicates that changes in the number of employed household members affects food security severity [50], with multiple household earners decreasing the odds of food insecurity, independent of household income[18]. In addition, the findings may not be generalizable beyond urban households with early adolescent girls and the cross-sectional design does not allow for inference of causality. There are also several strengths to consider. This study addresses an understudied area of research, contributing to awareness of how accumulation of risk and underlying risk profiles relate to food insecurity risk. The study increases our understanding of food insecurity risk among urban families with adolescent daughters. In past research, food insecurity has been a component of cumulative risk showing associations with negative health consequences for children [23, 24]. However, only one article was found that examined how cumulative risk relates to food insecurity as an outcome [33].

Study implications

When contemplating the role of household risk and food insecurity in the context of behavioral health, the bidirectional relationship between food insecurity and risk factors, such as depression, stress, and smoking, may interfere with interventions. Future food assistance policies should consider how to incorporate mental health screening and treatment as one facet of addressing food insecurity risk. Lessening the effects of food insecurity may lead to lower levels of depressive and stress, and given the bi-directional relationship between the variables, potentially further lower risk of food insecurity.

Given the significance of education in the latent class approach, additional research is necessary to understand how differing levels of education relate to food security risk. Connecting caregivers with a high school diploma/GED with additional vocational training or college education is an intervention that families and practitioners should consider, and policy efforts should focus on improving access to education for low-income caregivers. For participants in food insecurity risk households, who identify smoking cessation as a potential intervention, smoking cessation programs [51] or antidepressants for smoking cessation may be beneficial [52].

The study sample experienced high levels of food insecurity risk, thus recommendations include screening for food insecurity risk in families with multiple risk factors. Through identification of food insecurity and risk, practitioners and caregivers can work together to identify points of intervention, build financial capital, advance jobs and education, and alleviate financial burdens related to stress and food insecurity. Cumulative risk will provide multiple pathways for intervention, and allow service providers and caregivers to determine possible interventions tailored to their experiences. The screening tool used in this study provides an accurate and quick means to identify food insecurity risk among caregivers with young children [36], adolescents [7], and adults [37].

Translational implications

Prior research indicates that treating depressive symptoms among low-income women, such as those participating in the current study, can be complicated by barriers encountered in everyday life, including procurement of food [53, 54]. Although psychotherapeutic options can be effective, ignoring barriers can decrease the efficacy of interventions. Glasgow and colleagues [55] suggest that better understanding how and why interventions are implemented in practice settings, and necessary adaptations for success, are a key component of improving interventions. With regard to food insecurity and behavioral health, this means understanding the bidirectional nature of food insecurity and mental health, particularly depression and stress, as food insecurity may interfere with successful depression interventions. Adaptation of interventions that allow for the complex lives of families with children can improve the efficacy of interventions [56]. Healthcare settings are an ideal location to screen for food insecurity risk, as food insecurity is associated with higher healthcare utilization [57] and primary care providers indicate that standardized, brief screening for food insecurity risk fits easily into otherwise full schedules [58]. In addition, qualitative research among caregivers indicates that discussing food insecurity with healthcare providers lessens feelings of frustration and despair among caregivers [59]. The current study contributes additional contextual knowledge regarding food insecurity risk and its relationship with behavioral health that may help inform the implementation of interventions.

This study took place in an urban setting, with a high risk, low-income sample of African American families with adolescent daughters. Future research is needed to determine if findings hold across geography and demographics. It is also possible that different racial groups have different experiences with food insecurity risk and cumulative risk. Future research should explore potential differences across community context, urbanization, and immigration status. Future research should also expand on the cumulative risk model, including additional risk factors with established relations with food insecurity, such as housing insecurity and costs [60, 61], social capital [3], immigration status [62], and having a child with special needs [63].

Risk of food insecurity is a complex construct, with multiple interrelated factors that may amplify likelihood of food insecurity risk when combined, and multiple ways of examining risk. Significant findings related to cumulative risk and food insecurity risk provide public health practitioners, researchers, and policy makers with useful information about how accumulation of risk contributes to greater odds of food insecurity risk and points of intervention that may have a cumulative effect in lessening food insecurity. In addition, the current study suggests underlying risk profiles for food insecurity. Caregivers who experience high risk of mental health issues, particularly stress and depression, had the highest risk of food insecurity in the current study, indicating a need to attend to the roles of stress and depression in gaining and maintaining food security among low-income caregivers

Compliance with Ethical Standards

Conflict of Interest: None declared.

Ethical Approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent: Informed consent was obtained from all individual participants included in the study. This article does not contain any studies with animals performed by any of the authors.