Introduction

Food insecurity, defined as having limited access to food and ability to acquire food [1], and physical, sexual, and psychological violence are two major public health problems affecting women in the United States (U.S.). Nationally-representative data show that female-headed households experience significantly higher rates of food insecurity as compared to households in general (30% versus 13%) [2]. Women in the U.S. also experience high rates of violence with at least 25% of women having experienced physical or sexual violence over their lifetime[3]. Food insecurity can have significant impacts on women’s mental and physical health including depression and substance abuse, human immunodeficiency virus (HIV) infection, and HIV-related morbidity and mortality through nutritional, mental health (e.g., depression), and behavioral (e.g., non-adherence to HIV medication) pathways [4]. Similarly, violence can lead to physical injury, chronic disease, depression, post-traumatic stress disorder, substance abuse, and sexually-transmitted infections [5, 6]. Rates of violence and food insecurity among HIV-positive women are high. Between 14–20% of U.S. women reported physical violence in the past six months [7] and an estimated 50% of HIV-positive individuals, including women, are food insecure [4].

Decades of research has demonstrated a consistent link between markers of poverty such as income, employment, and education, and violence against women [8, 9]. Poverty may constrain women’s power in their relationships due to economic dependence on partners, making women less able to leave an abusive relationship. It is also plausible that financial difficulties serve as a trigger for violence in families or couples through the path of stress [9, 10]. Less research has examined this association specifically for food insecurity—which is related to, but conceptually different from, poverty. Economic abuses in the household have been linked to women’s food insecurity and these same conditions may precipitate violence [11].

There is evidence that food insecurity and violence are positively associated. According to the Centers for Disease Control, women who reported being food insecure in the past 12 months experienced higher levels of rape and physical violence than women who reported being food secure [3]. Of the small number of studies on the association between food insecurity and violence, most consist of cross-sectional studies with people receiving public assistance [12, 13], small community-based samples [14], or people outside of the US [11, 15, 16]. In a few cross-sectional studies with larger national samples, researchers have found a positive relationship between food insecurity and intimate partner violence.[17, 18] Yet, there is an urgent need for studies that utilize longitudinal samples of women to confirm these cross-sectional findings and additionally, to examine the association between persistent exposure to food insecurity and women’s risk for violence. Persistent life conditions, such as chronic food insecurity, can be even more detrimental to health than acute crises or events [19]. Moreover, chronic food insecurity can adversely affect psychological, physical, and social well-being—including interpersonal relationships [20]. To date, cross-sectional studies have been unable to disentangle persistent from shorter-term food insecurity and its association with violence.

Even less is known about the role of HIV status in the association between food insecurity and violence against women. In general, HIV-positive women experience high rates of violence, poverty, and food insecurity [4], and are at risk for stigma, discrimination, and poor mental and physical health because of their HIV status [21, 22]. Food insecurity is associated with higher CD4 cell counts, poorer adherence to ART, decreased viral suppression, and increased morbidity and mortality [4]. According to the theory of syndemics [23], HIV infection may magnify the association between food insecurity on violence and lead to synergistic relationships between mental and physical health. Yet only one study has considered the role of HIV infection in its evaluation of food security and violence. Montgomery and colleagues examined correlates of violence, including food insecurity, using cross-sectional data from the Women’s HIV Seroincidence Study in the U.S. (N = 2,099) [18]. There were only 30 HIV-positive women in the sample making it difficult to examine the role of HIV status.

In the current study, we used longitudinal data from a national cohort study of women at risk or living with HIV to: (1) investigate the association of current and prior food insecurity (i.e., persistent food insecurity) with sexual, physical, and psychological violence; and (2) examine whether the association between food insecurity and violence differs by HIV status. For aim 1, we hypothesized that women who are food insecure would be at higher risk for violence as compared to women who are food secure. We also hypothesized a dose-response relationship would exist between food insecurity and violence in terms of the duration (i.e. persistence) and severity of food insecurity. For aim 2, we hypothesized that the association between food insecurity and violence would be stronger among women who are HIV-positive.

Materials and methods

Analysis

Analysis models included all data collected over six visits. We used two-level logistic regression to model the association between food insecurity and experiences of violence, with women as a random effect. We used random rather than fixed effects because we expected that time-invariant differences between individuals could influence whether women experience violence, and we wanted to directly estimate the effects of these variables on women’s risk for violence. Random effects assume that the error term across individuals is not correlated with the predictors, thus allowing for time-invariant variables to play a role as explanatory variables.[27] We included a lagged variable for prior food insecurity in addition to a variable for current food insecurity. Since the models were additive, this allowed us to examine the association between persistent food insecurity (prior and current) and violence. The term “current” refers to the same visit at which violence is also assessed, whereas “prior” refers to a lagged visit which occurred six months before the current visit.

The models adjusted for potential confounders based on existing literature and theory: marital status (binary variable consisting of married/cohabitating or unmarried), currently in a relationship with a partner (yes/no), number of dependent children under care (continuous variable), employment status (employed/unemployed), annual household income (variable consisting of eight income categories), race/ethnicity (categorical variable with four types), stable housing status (yes/no), age (continuous variable), depressive symptoms (continuous variable measured using the Center for Epidemiological Studies Depression scale[28]), alcohol use (categorical variable with four levels: non-drinker, 0–7 drinks/week, 7–12 drinks/week, >12 drinks/week), any illicit drug use (such as cocaine, crack, heroin, amphetamines, club drugs, or methadone, but excluding marijuana), transactional sex (exchanging sex for food, money, or shelter; categorical variable with three levels: not having any sex; had sex, but no transactional sex; had transactional sex), and HIV status (positive/negative). Finally, to test for statistical interaction between HIV status and food security, we included an interaction term in the multivariable models.

Food insecurity information was available on 2,553 women at the baseline visit (Table 1). The final multivariable analysis used longitudinal data from 2,343 women who had complete data on all variables. Those excluded due to missing data (less than 10%) did not differ on key variables. All analyses were performed using Stata 15.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of participants in the women's interagency HIV study (Food insecurity sub-study): United States, 2013–16.

https://doi.org/10.1371/journal.pone.0213365.t001

Results

At baseline (N = 2,553), the mean age was 47 years old, 64% had a high school education or less, 72% were non-Hispanic Black, and 62% reported having a current partner (Table 1). Consistent with the WIHS study design, 71% were HIV-positive. Almost 45% of women experienced food insecurity (defined as having marginal, low, or very low food security) in the past six months. In the past six months, 4.5% of women experienced sexual or physical violence and 4.4% of women experienced psychological violence.

The bivariate logistic regression models showed that women who were HIV-positive were less likely to experience sexual or physical violence (odds ratio (OR) = 0.60; 95% confidence interval (CI): 0.41, 0.88) and were less likely to experience psychological violence (OR = 0.55; 95% CI: 0.36, 0.83), compared to HIV-negative women. There was a significant dose-response relationship between current food security and both measures of violence (Table 2). The odds of experiencing sexual or physical violence were 2.18 times greater for women with marginal food security (95% CI: 1.47, 3.23), 2.87 times greater for women with low food security (95% CI: 1.96, 4.20), and 8.64 times greater for women with very low food security (95% CI: 6.04, 12.36), as compared to women with high food security. Similarly, a dose-response relationship was observed between current food security and psychological violence; corresponding unadjusted ORs for marginal, low, and very low food security were 2.84 (95% CI: 1.89, 4.26), 3.14 (95% CI: 2.08, 4.72), and 6.70 (95% CI: 4.48–10.02). There was also a significant dose-response relationship between prior food security and both measures of violence (see Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Bivariate associations between food insecurity and violence: Women's interagency HIV study, United States, 2013–16.

https://doi.org/10.1371/journal.pone.0213365.t002

After adjusting for potential confounders, the odds of experiencing sexual or physical violence were 3.12 times greater for women with very low food security at the current visit as compared to women with high food security at the current visit (95% CI: 1.88, 5.19; see Table 3). The odds of experiencing sexual or physical violence were 7.05 times greater for women with very low food security at both the current and prior visits compared to women with high food security at both visits (the natural logarithms of 3.12 and 2.26 for current and prior very low food security were summed and then exponentiated to obtain 7.05). Similarly, the odds of experiencing psychological violence were 3.03 times greater for women with very low food security at the current visit compared to women with high food security at the current visit (95% CI: 1.67, 5.50). The odds of experiencing psychological violence were 5.72 times greater for women with very low food security at both visits compared to women with high food security at both visits. Other covariates were significantly associated with both forms of violence: age, employment, and stable housing were negatively associated with violence, while having a current partner, depression, alcohol use, engaging in transactional sex, and any illicit drug use were positively associated with violence (see Table 3). HIV seropositive status was no longer significantly associated with violence in the adjusted models (the association became non-significant when adding prior food insecurity to the models containing all other variables including income). There was no evidence of statistical interaction between HIV status and food insecurity on violence.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted Odds Ratios (AOR) and 95% Confidence Intervals (CI) for associations between food insecurity and violence: Women's interagency HIV Study, United States, 2013–16.

https://doi.org/10.1371/journal.pone.0213365.t003

Discussion

In this longitudinal cohort of women, many of whom were living with HIV, we found that food insecurity—either at the prior or current visit—was positively associated with current violence even after adjusting for socio-economic status. This finding held for sexual or physical violence, and psychological violence. We also provide the first longitudinal evidence that being food insecure at a prior study visit is associated with subsequent experiences of violence. Additionally, women who experienced persistent, severe food insecurity were more likely to experience violence than women who experienced shorter periods of severe food insecurity.

There are several explanations for our findings. First, although food insecurity differs from poverty and we controlled for markers of poverty in the analysis, parallels can be drawn from existing literature on violence and poverty of which more is known. Poverty places women at increased risk for sexual and physical violence through the path of stress [8]. Food insecurity could also invoke stress by causing hunger and worry about having sufficient access to food, and feelings of deprivation and alienation [1], these in turn may serve as a trigger for violence in families or couples [9, 10]. Second, food insecurity may make it difficult for women to leave abusive relationships due to dependence on their partners for food [29]. Finally, food insecurity may be related to violence through sexual behaviors such as transactional sex. Transactional sex has been described as a survival means to obtain food [29] and women who are food insecure are more likely to engage in risky sexual behaviors including transactional sex [30, 31]. Other research has found a positive association between transactional sex and multiple forms of intimate partner violence [32, 33]. Not only can food insecurity, transactional sex, and other psychosocial factors such as substance abuse independently contribute to women’s risk for violence, but can co-occur together as a syndemic and have a compounding effect [34]. Finally, we found that persistent food insecurity is a stronger predictor of violence than shorter-term food insecurity. Persistent food insecurity that is difficult to overcome may cause significant stress, feelings of powerlessness, and social isolation—conditions that may ultimately foster violence [20].

This is also the first study to examine the role of HIV status in the association between food insecurity and violence. Contrary to our hypothesis, HIV status did not amplify the association between food insecurity and violence. HIV-positive women reported significantly less alcohol use, drug use, and risky sex, and may have better access to social services (including counseling for victims of violence) due to their HIV status and engagement in HIV care. This may be especially relevant to HIV-positive women in WIHS, many of whom have been receiving services for HIV infection for years. Use of services for substance abuse and mental health among HIV-positive individuals is high and may be higher in comparison to the general population [35]. It is possible that for some HIV-positive women, these services may provide skills and resources needed to avoid abusive relationships.

Acknowledgments

The authors would like to thank Lila Sheira for her assistance with the data used in this manuscript.