Correlates of Six-Month Recidivism among Rural Incarcerated Women

Abstract

With growing rates of incarceration among women, it is crucial to understand factors associated with reincarceration. Rural incarcerated women have increased risks for strains and drug use. General Strain Theory posits that strains (drug use, psychosocial stressors) are associated with recidivism. Women experience as much, or more strain than men. The present secondary analysis examined intravenous drug use (IDU) and psychosocial factors (partner status, employment status, educational attainment, number and custody of children) as correlates of six-month recidivism among women incarcerated in rural jails. Baseline and six-month follow-up data for 361 women were examined. Two logistic regression models tested IDU and psychosocial characteristics as correlates of six-month recidivism, highlighting changes in correlates from baseline (Model 1) to follow-up (Model 2). Model 1 indicated number of children and partner status as significant correlates. Model 2 indicated number of children, employment, and IDU as significant correlates. Implications for desistence interventions were discussed.

Introduction

Incarceration rates of women in the United States (U.S.) have increased by more than 700% over the last four decades, representing a higher growth rate of incarceration for women compared to men (The Sentencing Project, 2020). Drug use is a common influencer of incarceration among women. For example, approximately 51% of incarcerated women meet criteria for a drug use disorder (Fazel et al., 2017) and 60% have a history of drug abuse (The Sentencing Project, 2007). Drug use and drug-related offenses have compounding effects on criminal re-offending. The rate of recidivism—or reoffending leading to re-entry into the criminal justice system—for drug offenders exceeds 75% (Durose et al., 2014). Notably, specific drugs, drug classes, and routes of administration are associated with higher risk of recidivism, including heroin, polydrug use, amphetamines, and injection drug use (IDU; Hakansson & Berglund, 2012). While criminal recidivism is a concern among incarcerated and formerly incarcerated women, with a five-year recidivism rate nearing 68% (Snyder et al., 2016), there is limited research examining factors associated with recidivism in jails among rural women.

Research among rural women suggests that factors associated with staying out of jail include age, employment, drug abstinence, housing, and prosocial peers (Staton et al., 2019). Rural women are an important subpopulation of incarcerated women to study because risky drug use (e.g., IDU; Hakansson & Berglund, 2012) is associated with recidivism and is prevalent in rural regions (Havens et al., 2011; Young et al., 2010). Reincarceration may have unique, gender- and region-specific implications among rural women. Given the growing incarceration rates among women, exploring factors associated with recidivism among this population is imperative.

While IDU is a predictor of recidivism, there are several other factors associated with recidivism that merit further study among rural populations of women. For example, psychosocial factors such as increases in educational attainment (Zgoba et al., 2008), having children (Kreager et al., 2010), marriage (Rocque et al., 2015), and employment (Skardhamar & Savolainen, 2014), are also shown to be associated with reduced recidivism. General Strain Theory (GST; Agnew, 1992) provides a framework to examine associations between drug use, psychosocial characteristics, and criminal recidivism. GST focuses on experiences of strain and the negative emotions and consequences caused by strains. Strains include loss of positive stimuli, imposition of negative stimuli, and inability to achieve ones’ goals (Ackerman & Sacks, 2012; Agnew, 1992). Loss of positive stimuli might include losing employment or strained/lost family relationships (e.g., with partners or children). Imposition of negative stimuli might include drug use, and inability to achieve ones’ goals might include not achieving educational, family/partnership, or employment goals. Presence of these three areas of strain is associated with criminal recidivism (Ackerman & Sacks, 2012), and research suggests that women experience as much, or more strain than men (Broidy & Agnew, 1997). However, literature examining psychosocial and injection drug-related correlates of recidivism has primarily focused on men and metropolitan contexts. Approximately a quarter of women will recidivate within the first six months of release (Snyder et al., 2016). As such, the current study examined psychosocial factors and IDU as correlates of six-month recidivism in jail among rural incarcerated women. Findings may inform targeted prevention interventions aimed to reduce reincarceration among this population.

Predictors of Recidivism

Drug Use and Injection Drug Use

Drug use may increase among individuals experiencing high strain (Sharp et al., 2012). Further, drug use contributes to recidivism (Dowden & Brown, 2002; Scott et al., 2014; Taylor et al., 2008), and women experience unique drug-related consequences, such that women are more likely than men to be incarcerated for drug-related offenses (Carson, 2020). Drug use is also directly related to criminality overall, not just drug offenses, among women (Bennett et al., 2008; Staton-Tindall et al., 2015). Drug use in rural regions may have unique patterns affecting recidivism, warranting further study among rural populations of women. Specifically, alternative routes of administration, such as IDU are prevalent among rural drug users (Havens et al., 2011; Young et al., 2010). IDU is a risky drug use behavior that is associated with recent criminal activity (Kinner et al., 2009), which may be an artifact of engaging in other risky behaviors that lead to criminal justice involvement. Importantly, despite levels of recent and lifetime drug use and IDU among rural users, rural residents often underutilize drug treatment (Oser et al., 2011; Warner & Leukefeld, 2001). As such, there is a need to understand the implications of IDU on reincarceration among rural women to better inform culturally appropriate interventions.

Psychosocial Characteristics

Several psychosocial factors, such as education level are associated with recidivism (Blomberg et al., 2011). Incarcerated men and women with a high school equivalency diploma have lower recidivism rates than incarcerated people without one (Zgoba et al., 2008). Among women specifically, there is a significant association between educational attainment and reduced arrest rates, which is thought to be related to partner and family formation as educational attainment increases, rather than influenced by job-related factors (Cano-Ubrina & Lochner, 2019). More research is needed to see if educational attainment is correlated with recidivism among rural populations of women, specifically given educational attainment in rural regions is lower than metropolitan regions (United States Department of Agriculture Economic Research Service, 2021) and incarcerated women have lower educational attainment than women in the general public (Couloute, 2018).

Similarly, employment status is associated with recidivism (Staton-Tindall et al., 2015), such that individuals with full- or part-time employment have decreased likelihood of recidivism. Among males, employment during an active phase of criminal behavior yields substantial reductions in criminal offending thereafter (Skardhamar & Savolainen, 2014). Literature representing the risk of unemployment on recidivism is particularly important since formerly incarcerated women have higher rates of unemployment compared to formerly incarcerated men (Couloute & Kopf, 2018). It is therefore important to examine the effect of employment status on recidivism among incarcerated women. Further, this research will inform re-entry program needs among female inmates re-entering rural communities, where there are economic disparities compared to metropolitan regions (United States Department of Agriculture Economic Research Service, 2021).

Children are also an important factor in predicting recidivism among women. For example, research suggests that the number of children reported by women is one of the strongest predictors of recidivism, such that increases in number of children is associated with decreased recidivism (Benda, 2005). Further, motherhood (i.e., having a child; Kreager et al., 2010) is associated with reductions in delinquency. These findings suggest that children may serve as a buffer against criminality among women, however, existing research is inconsistent. While some literature indicates that children may be a motivation for abstaining from drugs and crime (Harm & Phillips, 2001), other research shows that children living with their mother might also be a source of stress and home strain (Luecken et al., 1997).

Whether women have custody of children is also associated with recidivism. For example, previous literature among women indicates that not having custody of children is associated with nearly twice the likelihood of recidivism (Scott et al., 2014). Women with a history of drug use that have custody of their children are more likely to remain in mandated substance abuse treatment (Nishimoto & Roberts, 2001), which may indirectly affect engagement in crime. Though, children may also serve as a barrier for women to engage in treatment (Elms et al., 2018; Seay et al., 2017). Given inconsistent findings related to children, it is important to investigate the nuanced effect of children and child custody on recidivism among rural incarcerated women, particularly since women in rural contexts may have fewer environmental resources and may experience more strain related to children as a result.

Marriage is another aspect of family that is associated with recidivism among women. Desistance from crime is shown to be facilitated by the development of marital bonds (Rocque et al., 2015). Among a sample of majority male participants, in rural regions those who recidivated were more likely to have never been married (Staton-Tindall et al., 2015). Importantly, women in the criminal justice system are more likely to have never been married compared to women in the general population (Bloom et al., 2004). Therefore, it would be beneficial to expand marital status to partner status to see if similar associations with recidivism exist among rural women.

The Present Study

The aim of the present secondary analysis was to examine IDU, educational attainment, number of children, custody of children, partner status, and employment status as correlates of six-month recidivism among rural women incarcerated in jails. The present analysis also examined the effect of potential changes in partner status, custody of children, employment, and IDU in the six-months after release from baseline incarceration on six-month recidivism—particularly since these factors may not remain stable after release from incarceration. In both models, it was hypothesized that: (1) women who have injected drugs will have increased likelihood of six-month reincarceration; (2) as educational attainment increases the likelihood of six-month reincarceration will decrease; (3) partnered women will have less likelihood of six-month reincarceration, (4) women who were employed in the last six months will have less likelihood of six-month reincarceration, (5) as number of children increase, there will be a greater likelihood of six-month reincarceration, and (6) women who report not having custody of their children will have a greater likelihood of six-month reincarceration. Regarding number of children, despite inconsistent findings in the literature when examining the effect of children on reincarceration, in the context of GST, strain is associated with reincarceration and children may be a source of stress and strain (Luecken et al., 1997). Thus, it was hypothesized that more children may cause more strain, increasing reincarceration risk.

Method

Participants and Procedure

The present study was a secondary data analysis of female inmates from the rural mid-west and included data from the baseline and six-month follow-up samples of the parent study (see Staton et al., 2018). The parent study was approved by a university Institutional Review Board and obtained a federal certificate of confidentiality from the National Institutes of Health. Participants were women selected from three rural, Appalachia-area jails using a random selection design (described in Staton et al., 2018). Eligible participants included women housed at the jails who had at least two weeks to three months to serve on their sentence. Release dates were verified using on-line jail records. Eligible participants who were randomly selected and interested in participating in the study provided informed consent and were screened for substance use using the National Institute on Drug Abuse (NIDA)-modified Alcohol, Smoking, and Substance Involvement Test (ASSIST; NIDA, 2009), as well as answered questions related to risky sexual behavior. Private face-to-face baseline interviews were then conducted. A total of 440 eligible participants were screened for baseline interviews—of which 40 were released from jail before participating—resulting in a final sample of 400 women who completed the baseline interview (Staton et al., 2018). Baseline interviews included questions on demographics, drug use, psychosocial characteristics, arrest history, risky sexual behaviors, and HIV/HCV knowledge. Interviews were conducted using Computer Assisted Personal Interview (CAPI) software and participants were compensated $25.

The follow-up face-to-face interview was conducted at six months after the participant’s initial release from jail (i.e., release following baseline). Participants were reached for the follow-up using phone calls, mailings, and social media to gather updated information from the baseline interview (Staton et al., 2019). The follow-up interview was conducted in-person at agreed upon locations (Staton et al., 2019). A total of 361 (90.2%) participants completed the six-month follow-up interview, and the baseline and follow-up data for these participants were included in the current analyses. There were no significant differences on any of the variables or controls included in the present study between the 361 participants retained and the 39 participants not retained from baseline. Participants were compensated $25 after each interview, with an additional $25 bonus if they participated in both interviews.

Measures

Demographic and Psychosocial Characteristics

For Model 1, all sociodemographic information was collected at baseline with a series of items asking participants about educational attainment, parental status, partner status, and employment status. Education was measured with a single item at baseline asking participants, “What was the last grade or year you completed in school?” Responses were continuous with higher responses indicating higher educational attainment. Parental status was measured by two questions relating to children at baseline: “How many children have you ever had in your lifetime?” and “In the past six months, who has had legal custody of your child?” Responses to the first question were continuous with higher numbers indicating more children. On the second question, participants responded by indicating who had custody of their child(ren), which was coded as participant had custody (1) or participant did not have custody (0). Partner status was measured with a single item at baseline asking, “What is your current marital status?”, which was then coded to examine partner status. Specifically, responses were coded as partnered (1; e.g., married, remarried, living with someone as married [but not married], married but living apart) or not partnered (0). Employment status was measured at baseline by asking participants to identify what best described their work situation during the six months before the baseline incarceration. Employment status was coded as either employed (1; e.g., full-time job, part-time job, temporary/seasonal work) or unemployed (0).

Other demographic and descriptive information (e.g., age, race, income, history of incarceration, charges leading to current baseline incarceration, substance use history, and treatment history) were also measured at baseline to gather sample characteristics. For age, participants were asked their birthdate and their age was then calculated. Race was measured by asking participants to “Please select at least one race.” Participants then selected from, “White/Caucasian,” “African American/Black,” “Asian,” “Hispanic, Latino, or Chicano,” “Native American,” “Pacific Islander,” and “Other.” Individual and family household income were measured with two items asking participants to provide a dollar value for individual and family household income in the six months before baseline incarceration. Responses were continuous, with higher values indicating higher income. History of incarceration was measured with a single item asking participants, “Not counting juvenile offenses, how many times have you been incarcerated in your lifetime?” Responses were continuous, with higher values indicating more lifetime incarcerations. Regarding charge associated with the baseline incarceration, participants were asked, “What is the reason for your current incarceration?” and chose from several possible charges (e.g., shoplifting, probation/parole violations, driving while intoxicated, other). Finally, substance use and treatment history were measured with a series of items including, “Have you ever used [insert substance]?”, “Have you ever overdosed?”, and “Have you ever been to a substance abuse treatment program?” Participants responded “yes” (1) or “no” (0) to these items.

In the Model 2, changes in custody of children, partner status, and employment status in the six months after release from the baseline incarceration to the time of the six-month follow-up were examined to determine the effect on six-month reincarceration. In Model 2, sociodemographic data maintained from baseline were age, education, and number of children. To measure updated partner status, women were asked at the six-month follow-up, “What is your current marital status?”, which was coded to examine partner status. Updated custody of children was measured with a single item at the six-month follow-up asking, “In the past six months, who has had legal custody of your child?” Updated employment status at the six-month follow-up was measured by asking participants to identify what best described their work situation during the past six months. These variables were coded the same as the baseline coding.

Injection Drug Use (IDU)

In Model 1, IDU was measured at baseline by a single item. Participants were asked: “Have you ever injected any drugs in your lifetime?” Responses to this item were coded as: participant has injected drugs in their lifetime (1) or participant has not injected drugs in their lifetime (0). More recent IDU behaviors (i.e., in the six-months after release from baseline incarceration) was also examined. In Model 2, IDU was measured by a single item at the six-month follow-up indicating if participant injected any drugs other than those prescribed in the past six-months after release, using the same baseline coding.

Six-Month Recidivism

Six-month recidivism was measured by a single item during the six-month follow-up interview. The item indicated reincarceration at any time during the past six months after the release from the baseline incarceration. Participants were asked, “Our records indicate that you participated in our study a few months ago while you were in jail and then you were released. Have you been back in jail at any time during the past six-months since you were released?” Participants responded with “yes” (1) or “no” (0), with “yes” responses indicating past six-month reincarceration. This item was the outcome variable in both of the tested models.

Data Analysis

Data were analyzed using IBM SPSS version 27. The data were cleaned, coded, and assumptions for regression analyses were tested. There were no indicated outliers and skewness and kurtosis of the distributions were calculated and acceptable. Linear regression was then used to test for multicollinearity between predictor variables. All tolerance and variance inflation factor (VIF) statistics were acceptable, suggesting there was not multicollinearity between the independent variables. Univariate analyses (i.e., descriptive statistics and frequencies) were conducted to gather baseline sample characteristics. Main study analyses utilized multivariate binomial logistic regression to examine two models highlighting IDU, educational attainment, custody of children, number of children, partner status, and employment status as correlates of six-month recidivism. In Model 1, correlates were added from the baseline interview to examine the effect on six-month reincarceration. In Model 2, recent IDU, partner status, employment status, and custody of children in the six-months after release from the baseline incarceration to the time of the six-month follow-up interview were included in addition to baseline number of children and educational attainment to determine the effect on six-month reincarceration. In both models, age and number of adulthood incarcerations were added as covariates and variables were entered simultaneously. Participants with missing data were excluded from analyses.

Results

Sample Characteristics

Descriptive statistics for the main variables included in the models for both the baseline and follow-up samples are presented in Table 1. The women at baseline ranged from 18 – 61 years old (M = 32.81; SD = 8.24) and mostly identified as White (98%). The women were incarcerated for an average of 70.22 days (SD = 87.66) during the incarceration at baseline. Charges leading to the incarceration at baseline varied, but most common charges included drug crimes (25.3%), parole/probation violations (21.8%), theft-related charges (e.g., theft by deception, burglary, robbery, or theft by unlawful taking; [13.9%], theft specific to shoplifting [12.3%]), driving under the influence or other major driving violations (7.5%), child support (5.0%), disorderly conduct (4.8%), and other (9.4%). Most of the women (98.2%) had a previous history of incarceration, with an average of 16.22 (SD = 24.98) months of incarceration in their lifetime and an average of 5.04 (SD = 5.48) adulthood incarcerations in their lifetime. The average personal income was $8,467.15 (SD = $18,558.9), while the average family income was $16,471.79 (SD = $22,523.38) in the six months prior to the baseline incarceration. Additionally, 92.8% of the women reported polydrug use and 35.5% reported that they have previously overdosed. Approximately half of the women have been to a substance abuse treatment program. Finally, two-thirds of the women have injected drugs in their lifetime.

Table 1.

Sample Descriptive Statistics: Variables Included in Models

Variable	N (%) (Baseline)	M (SD) (Baseline)	N (%) (Follow-up)	M (SD) (Follow-up)
IDU	269 (75.6)	-	80 (22.2)	-
Last grade completed in school	-	11.1 (2.31)a	-	-
Number of children	-	2.2 (1.52)a	-	-
Custody of children	114 (32)	-	93 (30.4)	-
Partnered	126 (35.5)	-	108 (29.9)	-
Employed	83 (23.3)	-	76 (21.1)	-
6-month reincarceration	-	-	90 (26.2)b	-

Note:

^adenotes only measured at baseline;

^bdenotes only measured at follow-up

Logistic Regression Model 1

Results of the logistic regression model examining independent baseline correlates of six-month recidivism are presented in Table 2. Inclusion of the correlates and control variables collectively contributed to the prediction of six-month reincarceration, x²(8) = 18.26, p = .02. The Hosmer-Lemeshow (H-L) yielded, x² (8) of 8.38, p = .40, suggesting that the model was fit to the data well and 75.6% of the cases were correctly predicted given the model. Holding all other variables constant, number of children and partner status were significantly associated with six-month reincarceration. Number of children was positively associated with reincarceration (B = .23, p = .01), with an odds ratio of 1.26, indicating that the odds of reincarceration during the six-months after release from baseline increased by a multiplicative factor of 1.26 for each additional unit increase in number of children reported at time of baseline. Further, partner status was negatively associated with six-month reincarceration (B = −.57, p = .05). The log odds of reincarceration during the six-months after release from baseline was .57 for women reporting that they were partnered at the time of the baseline interview, indicating that partnered women were 43% less likely to be reincarcerated compared to unpartnered women. Baseline education, custody of children, employment status, lifetime IDU, and controls (age and total lifetime incarcerations) were not significantly associated with six-month reincarceration.

Table 2.

Six-month Recidivism Logistic Regression Model with Baseline Correlates

Variable	B (SE)	Wald	Exp(B)	95% CI Exp(B)
Age	−.03 (.19)	2.77	.97	.93 – 1.00
Lifetime incarcerations	.02 (.02)	1.10	1.02	.98 – 1.07
Education	.05 (.06)	.82	1.05	.94 – 1.19
Number of children	.23 (.09)*	6.52	1.26*	1.06 – 1.51
Custody of children	−.13 (.28)	.20	.88	.51 – 1.54
Partner status	−.57 (.29)*	3.93	.57*	.33 – .99
Employment status	−.45 (.34)	1.76	.64	.34 – 1.24
IDU	.43 (.35)	1.55	1.54	.78 – 3.03

Note:

^*p ≤ .05

Logistic Regression Model 2

Results of the logistic regression model examining how more recent partner status, custody of children, employment, and IDU since release from baseline incarceration as correlates of six-month recidivism are presented in Table 3. Inclusion of correlates and control variables collectively contributed to the prediction of six-month reincarceration, x²(8) = 43.39, p < .001. The Hosmer-Lemeshow (H-L) yielded, x² (8) of 10.10, p = .26, suggesting that the model was fit to the data well and 75.3% of the cases were correctly predicted given the model. Holding all other variables constant, number of children, employment, and IDU were significantly associated with six-month reincarceration. Number of children was positively associated with reincarceration (B = .28, p = .01), with an odds ratio of 1.32, indicating that the odds of reincarceration during the six-months after release from baseline increased by a multiplicative factor of 1.32 for each additional unit increase in number of children that the women reported. Employment in the six months after release from the baseline incarceration until the time of the six-month follow-up was negatively associated with six-month reincarceration (B = −1.22, p = .01), with an odds ratio of .30, indicating that women who became employed were 70% less likely to be reincarcerated than women who were not employed. Finally, IDU in the six months after release from the baseline interview was positively associated with six-month reincarceration (B = 1.27, p < .001), with an odds ratio of 3.55, indicating that women with IDU in the six months after release from the baseline incarceration were over three times more likely to be reincarcerated at the six-month follow-up than women who had not injected drugs. Age, total adulthood incarcerations, education, and custody of children were not significantly associated with six-month reincarceration, though custody of children (B = −.62, p = .08) and education (B = .12, p = .08) approached significance. Partner status was no longer a significant correlate of reincarceration (B = −.17, p = .60), indicating that potential changes in partner status after release from the baseline incarceration impacted whether the women were reincarcerated.

Table 3.

Six-month Recidivism Logistic Regression Model with Follow-up Correlates

Variable	B (SE)	Wald	Exp(B)	95% CI Exp(B)
Age	−.03 (.02)	1.52	.98	.94 – 1.02
Lifetime incarcerations	.01 (.02)	.16	1.01	.96 – 1.06
Education	.12 (.07)	3.06	1.12	.99 – 1.28
Number of children	.28 (.11)*	6.67	1.32*	1.07 – 1.62
Custody of children	−.63 (.35)	3.16	.54	.27 – 1.07
Partner status	−.17 (.33)	.27	.84	.45 – 1.60
Employment status	−1.21 (.48)*	6.41	.30*	.12 – .76
IDU	1.27 (.33)**	14.78	3.55**	1.86 – 6.78

Note:

^*p ≤ .05;

^**p ≤ 001

Discussion

The present study aimed to examine correlates of six-month recidivism, as well as the potential changes in correlates over time on six-month recidivism among a sample of rural women incarcerated in jails. Results from Model 1 indicated that family factors (i.e., having more children and being unpartnered) increased the likelihood of reincarceration in the six-months after release. After examining how more recent custody of children, employment, partner status, and IDU after being released effect reincarceration in Model 2, number of children remained significant, and employment status and IDU became significant indicators of six-month reincarceration. Findings support GST, which posits that experiencing strain can increase the likelihood of criminal activity (Agnew, 1992; Eitle, 2010). Specifically, results of the current study suggest that the level of strain experienced (e.g., related to children, partner relationships, employment, and risky drug use) influences the likelihood of criminal recidivism among women in rural regions. These findings have important implications for desistence and prevention interventions.

Results showed that the more children women reported having at baseline, the more likely they were to report reincarceration after the initial release from incarceration in both models. While the current study did not examine motherhood as it pertains to the relationship with children, rather it examined the number of children women reported having, this finding is in juxtaposition to research indicating being a mother as a predictor of desistance from crime (Kreager et al., 2010). Notably, Kreager et al. (2010) examined criminal desistance over a period of 10 years, suggesting that motherhood may be protective against longer-term, rather than short-term recidivism. Another notable distinction is that Kreager et al. (2010) focused on criminal desistance, whereas the present analyses focused on recidivism. Whether or not women are reincarcerated does not necessarily imply criminal desistance, which may explain why children were not protective against recidivism in the current study. Further, the average number of children in the present study was 2.2. It is possible that the protective effects of motherhood are for women with one child, and the more children women have may increase the level of strain (Luecken et al., 1997), predicting criminality. Nonetheless, the present findings suggest that there should be targeted efforts for reducing recidivism among women with children, particularly among mothers with a substantial drug use history. In the current sample, 92.8% of the women reported polydrug use. As such, residential drug treatment programs for mothers who are nearing release from incarceration may be beneficial (Johnston, 1995). Moreover, programs targeting parenting skills and prison nursery programs may also help reduce strain experienced by incarcerated women with children. These programs are also shown to be associated with low recidivism (Carlson, 2018; Goshin et al., 2014; Hoffman et al., 2010), and help women positively develop in their role as a mother, which extends post-release (Tuxhorn, 2021). Utilization of skill-based programs are particularly important for incarcerated women, since women are more likely to be the primary caretaker for dependent children (Bloom et al., 2004). The benefits of these programs may extend to women in jails; however, more research is needed.

While number of children was a significant correlate of recidivism, whether women had custody of their children was not significant in either model. At baseline, 114 women reported having custody of children, and during the follow-up only 93 women reported having custody in the six-months after release, indicating some women reported losing custody since baseline. Nonetheless, custody of children was not associated with reincarceration. This finding may be impacted by the interpretation of “custody,” and who had alternative child custody. Of the women with children who reported not having custody of their child(ren) at both time-points, the majority of the children were placed with the child’s other parent/participant’s spouse or another relative, while less than 4% were placed with the Department of Child and Family Services (DCFS) or a court. Women with children placed in the custody of a relative might experience less strain and difficulty reunifying than women with children placed with DCFS. Further, there may be race- and region-specific nuances for strain related to child custody because loss in custody did not appear to be related to recidivism in either model, despite fewer women reporting having custody at the follow-up. Longitudinal research is needed to examine the impact of child custody and reunification on criminal recidivism among rural women.

Baseline partner status was significantly associated with six-month recidivism, such that partnered women were less likely to recidivate, supporting our hypothesis. This finding is consistent with previous research indicating that marriage protects against criminal behavior (Rocque et al., 2015). Present findings suggest that decreased likelihood of recidivism extends beyond married couples to partnered couples among rural women. This is an important finding because compared to the general population, women in the criminal justice system are more likely to have never been married (Bloom et al., 2004). Notably, fewer women reported being partnered at the follow-up interview (see Table 1), and in Model 2, partner status was no longer associated with recidivism. This finding suggests that changes in partner status may have impacted whether women recidivated. The differences in the association between partner status and recidivism in Model 1 compared to Model 2 may be related to relationship quality. Specifically, research examining desistance from crime indicates that the quality of relationships is associated with decreased drug use and criminality (Fleming et al., 2010; Forrest, 2014; Visher et al., 2009), though the current analysis did not examine relationship quality. Future research should examine the impact of partnership quality on recidivism among women. Overall, programs for rural incarcerated women aimed to reduce criminal recidivism may benefit from re-entry programs that address building healthy, positive relationships in efforts to reduce recidivism among rural women.

Interestingly, lifetime IDU measured at baseline was not a significant correlate of six-month recidivism, though, more recent IDU measured at the follow-up was associated with an over three times greater likelihood of reincarceration. The follow-up finding is consistent with previous research positing that IDU predicts recidivism (Hakansson & Berglund, 2012). Recent IDU, rather than lifetime IDU, likely better captures whether women are actively engaging in risky behavior, including criminal behavior, that can lead to incarceration. Similarly, the insignificant effect of lifetime IDU on recidivism appears to be related to cessation versus relapse of IDU in the six-months post-release from incarceration. Drug relapse is a common concern among individuals who inject drugs (Xia et al., 2015) and lower rates of cessation successes are observed among individuals with a history of incarceration (Evans et al., 2009). Other factors associated with IDU relapse include homelessness (Linton et al., 2013) and risky-sexual behaviors (e.g., sex trade; West, 2016). As such, criminal desistance interventions for women with a recent history of IDU may benefit by minimizing risks by increasing engagement in substance abuse and relapse prevention programs in conjunction with helping women maintain stable housing, which may have residual effects on criminal recidivism.

Employment status at baseline was not significantly associated with six-month recidivism, but employment status reported after release from the baseline incarceration was associated with recidivism. The association between employment and recidivism at the follow-up is consistent with previous literature (Skardhamar & Savolainen, 2014; Staton-Tindall et al., 2015), and shows that regaining employment after release from incarceration resulted in a 70% less chance of reincarceration among rural women. In the context of GST, women who became employed may have felt less strain, while women who were unemployed after release may have felt pressure to find gainful employment, increasing strain and the likelihood of criminal behavior and/or drug use leading to reincarceration. However, this study did not examine whether women felt pressure to become employed or if they were choosing to be unemployed. Future research should seek to determine if wanting to obtain stable employment but not being able to influences recidivism among rural incarcerated women. Nonetheless, programs aimed to reduce recidivism among rural women would likely benefit from integrating employment services to help rural incarcerated women with community re-entry.

Finally, education was not a significant correlate of six-month recidivism in either model. This finding is not consistent with previous literature (e.g., Blomberg et al., 2011; Zgoba et al., 2008). The benefits of education against recidivism may be specific to educational and vocational services provided in jail or prison (Zgoba, 2008), rather than level of attainment before incarceration, which was used for the current analysis. As such, participating in meaningful activities and developing prosocial networks in jail may be the catalyst for reduction of recidivism. Within-facility education opportunities may be particularly important in rural regions where there are fewer economic resources and growth. Further, it is important to distinguish the effectiveness of education programs among women in jails, such as the current sample, compared to women in prisons.

Strengths, Limitations, and Future Research Directions

The present study is not without limitations. One potential limitation of this secondary analysis is that it did not examine relationship quality and the impact of relationship quality, or ending of “poor quality” relationships, on reincarceration. Previous research suggests that prosocial peers is associated with staying out of jail among rural women (Staton et al., 2019). As such, examining the quality of partnerships among rural women in future research would be important to explore to determine the effect on reincarceration. This analysis also examined potential drug-related and psychosocial correlates of six-month reincarceration, but it did not examine the level of distress women felt related to correlates. It would be important for future research to examine distress levels related to recent IDU, unemployment, being unpartnered, and having children, as well as women’s relationship with their children and the importance of their role as a mother. Highlighting these aspects will help determine if the distress levels are the catalyst for criminal behavior leading to reincarceration. Moreover, this study focused on six-month recidivism as it is a period with high recidivism rates among women, however, future research should also examine correlates of long-term recidivism. Current research examining longer-term recidivism (i.e., 12-months), highlights several key predictors of staying out of jail, including drug use abstinence, health care utilization, and prosocial peers (Staton et al., 2019). More research focusing on recidivism over time can help inform the most relevant interventions to help rural women stay out of the criminal justice system. Further, the present analysis focused on correlates of recidivism among women retained from baseline interview to the six-month follow-up. Even though there were no significant differences between these women on any of the variables or controls and we had access to statewide tracking systems, there is a small possibility that someone who was not retained was reincarcerated at the six-month follow-up, which is a potential limitation. Similarly, the current analysis was not exhaustive in examining variables that could be associated with reincarceration or potentially change over time to contribute to the likelihood of reincarceration. Future research should examine other factors that may contribute to reincarceration among rural women to help inform criminal desistance interventions. Potential variables to further consider may include housing security, other substance use, substance using peer-groups, seriousness of offenses, and treatment engagement.

Despite limitations, there are also several strengths of the current research. This is a secondary analysis of data collected among a relatively large sample of underrepresented women that examined not only baseline correlates, but also correlates after release from the baseline incarceration. Findings highlighted factors associated with reincarceration among rural incarcerated women, which helps focus development of targeted interventions to encourage desistance from crime. Future research should seek to examine the effectiveness of introduced desistance strategies in minimizing recidivism among women in rural regions. Moreover, this study focused on six-month recidivism, which is a period where recently released women are vulnerable for reincarceration (Snyder et al., 2016). Identifying risk factors for reincarceration during this period may help highlight specific needs for successful community reintegration and reduce the risk of repeated criminal behavior after six-months post-release, where reincarceration for women continues to increase (Snyder et al., 2016).

Implications and Conclusions

The present secondary analysis highlighted significant psychosocial and drug-related correlates—as well as correlates after release from incarceration—of six-month recidivism among a sample of rural, incarcerated women. Findings suggest that strains experienced by women, such as unemployment, family factors, and IDU, influence reincarceration. Desistance from crime interventions among this population may benefit from focusing on women with children and women who recently have injected drugs to improve supports and relapse prevention strategies to minimize recidivism. This study also highlights that women who are unpartnered at baseline and unemployed at the follow-up have a greater likelihood of reincarceration. Helping women build skills for building healthy positive relationships and connecting women with employment services as part of re-entry programs may reduce the likelihood of six-month reincarceration, which may better increase their likelihood for long-term desistance. Integrating these strategies in detention centers, community programs, and/or supervision regulations can foster successful re-entry and minimize recidivism among women through managing strains by improving positive stimuli and helping women achieve their goals.

Biographies

Brittany Miller-Roenigk, PhD is a Lyman T. Johnson Postdoctoral Fellow at the University of Kentucky in the Department of Educational, School, and Counseling Psychology. Dr. Miller-Roenigk earned her MA and PhD in Psychology (Clinical) from the University of Cincinnati. She received additional clinical and research training at Yale University, where she completed the Clinical and Community Psychology Internship and researched topics such as substance use intervention and consultation needs among racial and ethnic minority adult education students. Overall, her research focuses on substance use and mental health disparities and outcomes among underserved populations, such as racial and ethnic minorities, towards the aim of informing culturally adapted and tailored interventions. Dr. Miller-Roenigk has published in the areas of psychosocial, mental health, and cultural factors of substance use/misuse, treatment needs among Black adults who misuse prescription opioids, and HIV/HCV knowledge among rural incarcerated women.

Bridgette Peteet, Ph.D. is a Professor in the Department of Psychology at Loma Linda University and a California Licensed Clinical Psychologist. Prior to her appointment in 2019, she was a faculty member at the University of Cincinnati (UC) for 11 years. She earned her bachelor’s degree in psychology and criminal justice from Kent State University and her Master’s and doctorate in clinical psychology from UC. Dr. Peteet runs the Resilience and Disparities (RAD) Lab, which investigates health inequities in substance use disorders (SUD) using a community-inclusive and culturally-responsive framework. She is the Principal Investigator (PI) for a $3 million HRSA Scholarships for Disadvantaged Students grant and the Co-PI on a Graduate Psychology Education grant to provide specialized training in SUDs and to build pathways to diversify the scientific workforce. Dr. Peteet is an active member and former Secretary for the American Psychological Association Society for the Psychological Study of Culture, Ethnicity, and Race (APA Division 45). She is a past honoree of the Division 45 Charles and Shirley Thomas Award and as a UC faculty Diversity Ambassador for her institutional and national contributions to diversity mentorship, programming, and research. Dr. Peteet is also a facilitator of culturally conscious health practices and has long-term experience consulting with community organizations, all towards improving the health and well-being of historically marginalized people.

Danelle Stevens-Watkins, PhD is the Acting Dean of the College of Education, the Assistant Vice President for Research in Diversity and Inclusion, the Director of Graduate Studies, and a Professor in the Department of Educational, School, and Counseling Psychology at the University of Kentucky. She is also a licensed psychologist in Kentucky. She earned her doctorate degree in counseling psychology from the University of Kentucky and her MA in Clinical Psychology from Spalding University. Dr. Stevens-Watkins has numerous publications and has secured $4.2 million dollars in grant and contract funding as principal investigator (PI). She has been funded continuously through NIH and National Institute on Drug Abuse (NIDA) for the past decade. Dr. Stevens-Watkins is currently the principal investigator on a National Institute of Health (NIH), National Institute on Drug Abuse (NIDA) research grant examining structural, social, and cultural factors affecting opioid and other prescription drug misuse among Black Americans. As PI, she recently completed a NIDA funded project examining mental health, drug abuse, and HIV risk among criminal justice-involved Black Americans.

Michele Staton, PhD is a Professor in the Department of Behavioral Science at the University of Kentucky and a Faculty Associate of the UK Center on Drug and Alcohol Research. Dr. Staton’s research focuses primarily on justice-involved individuals with substance use disorders in rural communities, with a specific focus on high-risk related health consequences including HIV and HCV. Dr. Staton has published in the areas of women offenders, substance use disorder treatment, and availability and access to rural treatment. She currently serves as the PI on a number of NIH funded projects including the NIH/NIDA funded Kentucky hub of the Justice Community Opioid Innovation Network (JCOIN). As the only JCOIN project focused on women, the study aims to increase utilization of medication to treat opioid use disorder among women during the transition from jail to the community. Other projects focus on adapting and feasibility testing an evidence-based HIV prevention intervention for a Facebook platform for rural women re-entering the community from jail, as well as a study focused on increasing utilization and retention of treatment using extended-release naltrexone (Vivitrol^®) with rural offenders transitioning from jail to community supervision. As a social worker, Dr. Staton also works closely with a number of state treatment agencies to conduct evaluations for federally funded projects through agencies including SAMHSA and ACF. She also serves as the PI for a state contract with the Kentucky Department of Corrections to evaluate corrections-based substance abuse programs across the state.