Food Insecurity among Women Involved in the Criminal Legal System and Members of their Risk Network

Abstract

Background. People affected by the criminal legal system face significant challenges accessing food assistance. Methods. We estimated the prevalence of food insecurity among women involved in the criminal legal system (WICL) and members of their risk network using a secondary analysis. Results. Among 104 participants, over one-third were food insecure, 97% of whom lived in an economically disadvantaged ZIP Code. Compared with those who were not food-insecure (N=67), people who were food-insecure (N=37) were more often married, unemployed, and had a severe psychiatric disorder. In bivariate and multivariate models, food insecurity was associated with the same factors and not with race/ethnicity, education, number of dependents, monthly income, or type of CL involvement. Conclusion. Among WICL, we found widely prevalent food insecurity. In an economically disadvantaged ZIP Code, food insecurity compounds other social determinants to drive health disparities. Eliminating policies that restrict access to food assistance is important for ensuring that WICL can meet basic subsistence needs.

The U.S. has the highest incarceration rate in the world, including sixty-five million Americans who have been previously incarcerated and 2.3 million U.S. residents who are currently incarcerated in federal and state prisons.^1,2 Upon returning to their communities, individuals who have been previously incarcerated often face obstacles in securing housing, employment, and access to food assistance.³ Many states have imposed restrictions and limitations on or modified access to social benefits that include temporary bans, enrollment requirements in education programs, or benefit termination, which may exacerbate health outcomes due to the extended and time-consuming process of benefit re-enrollment.^1,4 These collateral consequences of incarceration mean that millions of returning citizens are denied access to the Supplemental Nutrition Assistance Program (SNAP) and Temporary Assistance for Needy Families (TANF) because of a prior drug felony conviction.^4,5 In several studies, prior incarceration is strongly associated with food insecurity, a condition that is defined as having limited access to adequate food that poses adverse health consequences to individuals (i.e., cancer, stroke, depression), and negatively affects overall well-being of family members.^1,6,7

Women involved in the criminal legal system often experience disrupted access to essential resources and stable employment, placing them at higher risk for food insecurity and making them more susceptible to negative health outcomes.^1,8 Women are the fastest-growing segment within the U.S. criminal legal system (CLS); women experiencing food insecurity are three times more likely to suffer from major depression and have a 66% higher need for mental health services compared with those who are food-secure.^1,6,8,9 Over 82% of women affected by the criminal legal system (WICL) have alcohol or substance use disorders which, when combined with a lack of basic subsistence needs (e.g., food insecurity), complicates engagement in care following prison-release for a variety of health conditions.^1,9

Despite the documented high prevalence of food insecurity among individuals affected by the CLS, the extent to which WICL experience other negative social determinants of health (SDOH) or other social vulnerabilities has not been fully analyzed. In addition, prior studies have not examined how specific environments affect the health and well-being of WICL. This is an important gap since the health consequences of a poor diet could exacerbate negative health outcomes, putting WICL at higher risk for chronic diseases.⁸ With the rapidly increasing numbers of WICL, there is a need for new research into how incarceration status affects food insecurity. This study aims to contribute to existing knowledge regarding health disparities experienced by WICL by examining the prevalence of food insecurity among WICL, assessing the SDOH that may be associated with food insecurity among WICL, and examining the potentially contributory environments in which formerly incarcerated women live.

Methods

Study setting and population

This is a secondary data analysis derived from a demonstration project, known as Project Empowering, which assessed the feasibility and acceptability of delivering HIV pre-exposure prophylaxis (PrEP) via electronic health care (eHealth) to WICL who were residing in the community (i.e., on probation, parole, or recently returned home from prison or jail.)¹⁰ The study was conducted in an urban setting in New England. Project Empowering was registered on Clinicaltrials.gov (NCT03293290) and details of the study have been previously published elsewhere.¹⁰

Study setting and participant recruitment

Project Empowering employed a patient-centered approach to recruitment, with outreach efforts to locations frequented by WICL, such as probation and parole offices, community outreach programs, and health centers. Enrollment occurred between December 2017 and May 2019, wherein potentially interested participants were screened by trained research assistants for the following inclusion criteria: (1) self-identification as female, (2) age 18 years or older, (3) self-reported HIV-negative status, and (4) involvement with the criminal legal system, defined as release from prison or jail within the past six months or under current community supervision such as probation, parole, or intensive pretrial supervision. Individuals currently taking, unwilling, or unable to take PrEP were excluded from the study, as the program aimed to assess PrEP initiation for HIV prevention.

Upon enrollment, participants were invited to recruit their peers using an incentivized modified respondent-driven sampling approach. Recruited peers were adults over 18 years old who were self-reported HIV-negative and not on PrEP, regardless of their status of involvement with the CLS or gender, although most recruited peers were WICL. Everyone who was eligible and interested in participating completed written informed consent procedures, including authorization for the retrieval of health information from external sources. Yale University Institutional Review Board, which includes a prisoner representative, approved the protocol for this study, with additional approvals from the statewide Department of Corrections Research Advisory Committee, Judicial Branch Court Support Services Division of Adult Probation, and the area's largest drug treatment provider to facilitate recruitment. No participants were enrolled while they were incarcerated; all study activities took place in the community.

Following enrollment, participants underwent a baseline study interview conducted in English or Spanish by trained research staff in private settings, and data were entered directly into REDCap. Those meeting clinical eligibility criteria for PrEP were offered it and monitored for up to 12 months with quarterly study interviews. For the purposes of this analysis, only baseline and eligibility screening data were used. One participant was excluded from this analysis because a response to the food insecurity question was missing.

Measures

The primary outcome of interest was food security status, which was measured through a single question screener, "In the past 90 days, was there any time for two or more days when you didn't get anything, or barely anything, to eat?" Food insecurity was defined by a "yes" response. A fuller screening instrument for food insecurity was not used because it was not the primary focus of the parent study, and in the interest of minimizing participant burden for structured interviews, only a single screening question was selected.

The following sociodemographic and SDOH characteristics were considered as potential explanatory factors: age, gender, race/ethnicity, education, marital status, number of children in the household, current housing status, past three-year employment status, and monthly income. Categorical variables were created and variables stratified for analysis as follows:

Age was stratified into four categories by decade (21–30, 31–40, 41–50, 50+ years). Although participants aged 18 years and older were eligible for participation, there were no participants under aged 21 who enrolled.

Gender was stratified into two categories (female and male). One transwoman was excluded from the gender analysis because the study was underpowered to fully examine the experiences of transwomen and we wanted to avoid misclassification.

Race/ethnicity was stratified into four categories (non-Hispanic White, non-Hispanic Black, Hispanic/Latinx, and Other). "Other" included people who identified as American Indian, Alaskan Native, Asian/Pacific, or other non-Hispanic.

Education was stratified into two categories based on the number of years participants attended school (high school or less and some college/higher education).

Marital status was stratified into three categories (married, previously married, and never married). The variable "previously married" includes widowed, separated, and divorced.

The number of children was stratified into three categories (no children, 1–3, and 3+).

Housing status was stratified into five categories (independent living, dependent living, homeless, rehabilitation center, and transitional housing). "Independent living" included residing in a permanent hotel, rented apartment, or owned home, and "dependent living" was defined as relying on someone or something else for aid and support and included staying with family and friends. "Homeless" included being completely homeless (i.e., residing in a place not meant for human habitation) or residing in a shelter.¹¹ "Residential drug treatment" was defined as residing in a drug treatment facility, residential drug treatment program, or sober house. "Transitional housing" included residing in a transitional or other residential program (housing program, halfway house, or long-term shelter), or a time-limited, single-roomoccupancy hotel. Participants were instructed to select a single response from each category that most accurately represented their usual pattern.

Employment patterns in the past three years were stratified into four categories (full/part-time, unemployed, retired/disability, and other). The variable "other" included not being in the workforce because of residing in a carceral setting or other institution or because of being retired or disabled. Participants were instructed to select the single response from each category that most accurately represented their usual pattern.

Monthly income was stratified into two categories (below federal poverty level and above federal poverty level). A monthly income below $1,215 per individual is defined as being below the federal poverty level.¹²

Psychiatric disorders were assessed using the Addiction Severity Index, Psychiatric (ASI-P) sub-scale which is commonly used to evaluate need for and response to treatment for psychiatric disorders among people with substance use disorders. People with a score greater than or equal to 0.22 were considered to have severe psychiatric disorder, while those with scores less than 0.22 were considered not to have severe psychiatric disorder.¹³

Current CLS involvement was stratified into two categories (on probation/parole and/or released from prison/jail in the last six months vs. no probation/parole/prison/jail). The six-month period was determined based on the parameters of the original study, so participants reporting current probation or parole status or release from prison/jail within the last six months were categorized as CLS-involved. For the purposes of this analysis, we only considered current or recent CLS involvement, rather than remote (i.e., lifetime) CLS involvement.

As for neighborhood or community-level characteristics, we used the Social Deprivation Index (SDI) to quantify the severity of deprivation experienced by WICL based on the ZIP Code provided during the baseline interview.¹⁴ The SDI is an area-level deprivation measure, using seven demographic characteristics to assess health outcomes, including the percentage of: individuals living in poverty, with less than 12 years of education, in single-parent households, in rented housing units, in overcrowded housing units, and in households without a car, and unemployed adults younger than 65 years of age.¹⁴ The SDI centile scale ranges from 1–100, where a score of 1 is considered the least disadvantaged or least economically disadvantaged ZIP Code, while 100 represents the most economically disadvantaged ZIP Code. In this analysis, participants were identified as residing in an economically disadvantaged ZIP Code if their SDI score was higher than 75. ZIP Codes may be centered around cities or reflect certain communities, but we did not collect specific data on participants' communities, neighborhoods, or cities. Instead, we used ZIP Codes as a proxy for the built environment.

Statistical analysis

Data analysis was performed using IBM SPSS Statistics 28.0 (IBM Corp. Released 2021. IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY: IBM Corp). A descriptive analysis was conducted to characterize the study sample at baseline. To compare participants who were food-insecure with those who were not, chi-squared tests were used for categorical variables. A p-value of <.05 was considered statistically significant. ZIP Code-level economic disadvantage was defined by the SDI scale as a score higher than 75.

We used bivariate logistic regression to examine the association between each variable and the primary outcome of food insecurity. We developed an adjusted regression model, including variables significant at p<.05 in the bivariate analysis. We also built an adjusted regression model that incorporated all variables significant at p<.05 in the bivariate analysis, along with CLS-involvement, a priori, but the initial model was a better fit based on the goodness of fit test, Aikikes criteria, and Bayesian criteria.

Results

Table 1 shows the baseline characteristics of the 104 enrolled participants that included 37 index participants and 67 recruited peers. There were no significant differences by recruitment source in terms of demographic characteristics, as we have described previously.¹⁰ Many (43%) participants were in their 30s, female (70.2%), and non-Hispanic White (56.7%). Most participants had a high school education or less (72.1%), were never married (60.2%), and had one to three children in their household (52.9%). The sample population overall had a high rate of unmet basic subsistence needs, as evidenced by 23.1% being homeless, 41.3% being unemployed, and nearly all (93.3%) subsisting below the federal poverty level. There was a high prevalence of severe psychiatric disorders (53.8%). As would be expected based on the inclusion criteria of index participants and peer-recruited participants within their risk network, more than half of the study sample, 61%, was on probation, parole, or recently released from prison or jail.

Overall, 35.5% of the study population were food-insecure (Table 1) and there were no significant differences in food insecurity by whether participants were index participants or recruited risk network members (data not shown). Comparing those who were not food-insecure (n=67) with those who were food-insecure (n=37), no statistically significant differences were detected in terms of: age, gender, race/ethnicity, education, number of children in the household, housing status, monthly income, or

Table 1.

BASELINE CHARACTERISTICS OF THE STUDY SAMPLE, BY FOOD SECURITY STATUS, (N=104)

CLS involvement. The food-insecure group included a higher proportion of married individuals than the food-secure group (24.3% vs. 4.5%, p=.01). There were also differences in employment patterns over the past three years, with higher unemployment among people who were food-insecure (56.8% vs. 32.8%, p=.02). There was a significantly higher prevalence of severe psychiatric disorders observed among people who were food-insecure (70.3% vs. 44.8%, p=.01).

We evaluated neighborhood and community-level characteristics potentially affecting food insecurity with the SDI scale. Among participants, 88% (n=91) resided in economically disadvantaged ZIP Codes. Moreover, 97% of women who reported being food-insecure lived in economically disadvantaged ZIP Codes, accounting for 35 out of 37 participants.

Table 2 shows the results of logistic regression modeling. In bivariate models of food insecurity, none of the following were significantly correlated with food insecurity status: age, gender, race/ethnicity, education, number of children in the household, housing status, monthly income, and CLS involvement. In bivariate models, however, food insecurity was inversely associated with being previously or never married compared with being married (p=.03); food insecurity was directly correlated with being unemployed as opposed to employed (p=.03) and directly correlated with having a severe psychiatric disorder (p=.01). In multivariate models, compared with people who were married, those who were previously married had a lower odds of being food-insecure (aOR=0.16; 95% CI=1.26, 29.22; p=.02). Compared with people without a severe psychiatric disorder, those with a severe psychiatric disorder had a 2.87 higher odds of being food-insecure (aOR=2.87; 95% CI=1.14, 7.27; p=0.02). After adjusting for these other characteristics, employment status was no longer significantly associated with food insecurity in the multivariate model.

Table 2.

BIVARIATE AND MULTIVARIATE LOGISTIC REGRESSION MODELS OF FOOD INSECURITY (N=104)

Discussion

Though previously published studies had described associations between prior incarceration and food insecurity, there were no prior studies to our knowledge that focused on the experience of food insecurity among WICL and members of their risk network, nor considered the individual SDOH and built environment-level characteristics that are associated with food insecurity in this population. To address this gap, we assessed individual- and ZIP code-level correlates of food security status among WICL and members of their risk network (most of whom were also WICL). We found a relatively high prevalence of food insecurity, with over one-third of study participants reporting being food-insecure. In contrast to our original hypothesis, we did not find a statistically significant association between current CLS involvement and food insecurity, either because there is no true association, the sample size was too small to detect an association, or because of potential confounders. We did find that marital status, employment patterns in the past three years, and having a severe psychiatric disorder were each significantly associated with food insecurity, painting a fuller picture of people's experiences and barriers to accessing resources after returning to communities from prison or jail. There may have been additional confounders that were unmeasured.

Social determinants of health affect health access and resource availability.⁷ Despite a high prevalence of poverty in the study sample (with nearly all participants identified as living below the federal poverty level), it was interesting that we did not find significant differences in poverty level by food-insecurity status. We found that individuals who were previously married had lower odds of food insecurity than those currently married, which potentially reflects the demands of supporting a partner who may also be unemployed, impoverished, or affected by the CLS. In contrast with this finding, prior research suggests that having a social support system can enhance food security by providing another source of household income.¹

We also found that employment patterns were associated with food insecurity. Criminal legal system involvement may prevent individuals from finding employment and attaining socioeconomic security.¹⁵ Participants who were unemployed were more likely to experience food insecurity than those employed full or part-time, perhaps because of restrictions in the Supplemental Nutrition Assistance Program (SNAP) or other social welfare benefits, which pose barriers to eligibility for individuals with prior drug-related criminal convictions, requiring additional steps to qualify for benefits.^4,16 The Re-Entry Support through Opportunities for Resources and Essentials Act of 2023 (RESTORE Act), otherwise known as the Farm Bill, includes important provisions that would repeal the federal SNAP felony ban in all states and allow people to apply for SNAP benefits within 30 days of release from prison or jail.¹⁶ If passed into law, these stipulations would fill a much-needed resource gap for returning citizens to address food insecurity.

People who are unemployed face especially high risks of food insecurity when they experience restricted access to government assistance and financial support. Our results indicate that individuals residing in drug treatment facilities are less prone to food insecurity than those living independently. In 2006, the U.S. Department of Agriculture (USDA) released guidance authorizing drug and alcohol treatment and rehabilitation centers to participate in Food and Nutrition Services, which may have contributed to the higher food-security rates observed among participants in our study.¹⁷ This policy may have improved access to nutritional support for residents, mitigating food insecurity.

Beyond individual-level SDOH, we used the SDI to gauge social deprivation at the community level and assess the connection between deprived ZIP Codes and food insecurity among participants. Figure 1 shows the prevalence of food insecurity across various towns in Connecticut, which maps closely onto the SDI scale, highlighting specific areas as one of the top 10 towns with the highest incidence rate of food insecurity in Connecticut, with 22.0% of households self-reporting food-insecurity.¹⁸ A striking 97% of participants reporting food insecurity lived in socially deprived ZIP Codes, comprising 35 out of 37 participants. This highlights how SDOH characteristics affect food-security outcomes for this population. Various literature suggests that

Figure 1.

Percentage of Connecticut households reporting food insecurity by town or town-cluster (Map from Boehm et al. 2019).

residents of impoverished communities disproportionately experience food insecurity due to social and environmental factors that prevent adequate access to nutrition, with detrimental health outcomes.^19,20 The prevalence of food insecurity among individuals living in economically disadvantaged areas leads to health disparities, especially among those with involvement in the CLS.⁸ Following incarceration, many people affected by the CLS return to the same communities from which they came because that is where their support network resides, and others return to similarly disadvantaged areas,^21,22 emphasizing the need for targeted interventions to address interconnected challenges.

Despite deploying a novel focus on the impact of women's CLS involvement in terms of food insecurity, this study had limitations. In this secondary data analysis, we relied on available data that included a single-item question screen to assess food security. More detailed and standardized food insecurity assessments do exist, such as that recommended by the U.S. Department of Agriculture, and should be considered for future studies.^23,24 We did not have access to data on food insecurity prior to incarceration nor any details of the duration of their incarceration. A small sample size limited our ability to fashion more complex modeling, and the study setting in resource-rich New England, where there is Medicaid expansion, may reduce generalizability to areas where there are fewer community-based resources. The study did not focus on previous CLS involvement beyond the six-month period, nor involvement of a partner and/or family member in the CLS. In this secondary analysis, we used ZIP Code as a proxy for built environment, but did not have any available participant-level data on neighborhood or community. This may have underestimated the impact of the built environment, as a ZIP Code can encompass widely varying SDOH. Despite these limitations, we found high rates of food insecurity that correlated with other social needs and would likely only be higher in other areas with fewer available community resources. To further understanding of food insecurity among WICL, we should consider structural violence as a root cause of health disparities. Although we did not find that race and ethnicity were associated with food insecurity in our sample, we acknowledge that structural racism can indirectly affect individual access to basic resources. There is a need to develop tailored interventions to address people's social, health, and subsistence needs in a holistic way. Future research should address factors that are linked to food insecurity or present barriers to obtaining employment, receiving assistance, or receiving health care.

Conclusion

In conclusion, among a recruited sample of WICL and members of their peer-recruited risk network who were enrolled in a PrEP demonstration project, we found a high prevalence of food insecurity. Study findings suggest that women returning home from prison or jail, or under community supervision, reside in ZIP codes with high levels of economic disadvantage. Beyond poverty, food insecurity compounds other negative social and structural determinants of health and potentially drives health disparities, demanding focused intervention. The RESTORE Act provision to repeal a federal ban on SNAP benefits for people with drug-related felony convictions is one important way to address food insecurity among people affected by the CLS.

Funding

Support for the parent study provided by a Gilead Investigator Sponsored Award to JPM. Funding source had no role in the decision to undertake this analysis, interpretation of findings, drafting of the manuscript, or the decision to submit the manuscript for publication.

This analysis was developed as part of Ms. Robles's thesis project for a Master of Public Health in Social and Behavioral Sciences at Yale School of Public Health. The thesis was adapted to prepare the manuscript for publication.