Personal agency and alcohol abstinence self-efficacy among incarcerated women

Abstract

Incarcerated women with alcohol use disorders (AUDs) have unique treatment needs. Behavior change models emphasize self-efficacy in making changes to alcohol use, but have not been tested in samples of incarcerated women. Personal agency in several domains was examined as a correlate of alcohol abstinence self-efficacy in a sample of 173 incarcerated women with AUDs. Lower alcohol cravings (β = −0.19, p = .029), greater self-care (β = 0.17, p = .012), and less engagement in transactional sex (β = −0.48, p = .007) were associated with greater self-efficacy. Intrapersonal and interpersonal agency influence incarcerated women’s self-efficacy.

The population of incarcerated women has steadily increased in recent decades (Minton & Golinelli, 2014; Snyder, 2012). The growth of this population has prompted increased attention to the unique needs of and challenges faced by incarcerated women (Cortoni, 2017; Covington, 2007), with a mounting emphasis on the role of substance use problems. Substance problems are a pervasive and influential factor in a variety of important outcomes for incarcerated women, including poorer mental health, greater limitations in access to stable housing, employment, and less contact with treatment and social services compared to men (Alemagno, 2001; Langan & Pelissier, 2001; Messina, Burdon, Hagopian, & Prendergast, 2006; Pelissier, Camp, Gaes, Saylor, & Rhodes, 2003; Scott & Dennis, 2012).

While substance use is a general concern for incarcerated women, alcohol use is a specific and significant problem. For example, alcohol use among incarcerated women has been associated with physical health problems, mental health problems, risky sexual behavior, and criminal behavior (Jordan, Schlenger, Fairbank, & Caddell, 1996; Stein et al., 2009). In fact, evidence suggests that criminal behavior is more likely to be associated with alcohol use for women than for men (Bácskai, Czobor, & Gerevich, 2011; Martin & Bryant, 2001; Swan & Goodman-Delahunty, 2013) and alcohol use disorders are more highly associated with violent offense than drug use disorders (Kopak & Smith-Ruiz, 2014), making alcohol use an important behavioral target for criminal-justice based treatment.

Alcohol treatment development efforts often look to behavioral health models, such as theory of planned behavior (Berridge, Cheetham, McKay-Brown, & Lubman, 2015; Kelly, Leung, Deane, & Lyons, 2016; Norman et al., 2018) and the Theory of Reasoned Action (Brubaker, Prue, & Rychtarik, 1987; Clark, Ringwalt, Shamblen, & Hanley, 2011; Roberto, Shafer, & Marmo, 2014) as a guide for substance treatment development. However, the application of general models in specific populations can reveal unique elements or factors that operate differentially from the general population. For example, a growing body of research suggests that gender differences may contribute to unique pathways of entry into jails and prisons (Pasko & Mayeda, 2011; Spjeldnes & Goodkind, 2009; Weiser et al., 2009) and further suggests that women and men have different treatment needs following an incarceration period (Langan & Pelissier, 2001).

The examination of elements from theoretical models that are routinely used to guide intervention development in unique populations, such as that of incarcerated women, helps to elucidate whether associations found in the general population are also found in those specific populations. One well-known element included in many types of behavioral health models is personal agency, defined as the ability to originate and direct one’s own choices. Incarcerated women, compared to incarcerated men or compared to women in the general population, may have unique constraints on their personal agency (such as fewer job options, less education, fewer network members who do not drink or use drugs, abusive partners, childcare responsibilities, more health problems, and fewer life options in general (Messina, Burdon, & Prendergast, 2003), making this an important factor to explore as it relates to self-efficacy to quit alcohol.

If personal agency is more limited among incarcerated women, then the association between personal agency and belief in one’s ability to succeed in making behavioral changes (i.e., self-efficacy) may be unique, as well. As indicated in a recent meta-analysis across 40 studies, self-efficacy for behavioral change is influenced by one’s sense of personal agency in the general population, with a significant association emerging between perceived agency and alcohol-related intentions (r = .31; Cooke, Dahdah, Norman, & French, 2016). Indeed, individuals with alcohol use disorders (AUDs) frequently struggle with a sense of limited personal agency, in both intrapersonal (e.g., cravings) and interpersonal (e.g., social networks; Miller, Westerberg, Harris, & Tonigan, 1996) domains. Yet this association may have important nuances that are unique to incarcerated women with AUDs.

Incarceration, by its very nature, diminishes personal agency. And while the period of incarceration is an obviously restricted time of behavioral freedom, incarcerated individuals regularly encounter behavioral restrictions even after jail or prison discharge (Eley, 2007; Pope, Smith, Wisdom, Easter, & Pollock, 2013; van Dooren, Claudio, Kinner, & Williams, 2011). For example, individuals with a criminal record experience impediments to gaining employment and securing housing, limiting their options for ways to meet basic needs. In such cases, individuals may feel compelled to engage in illegal activities or sex trading to get basic needs met, even if they would prefer not to do so. Beyond the general limitations to personal agency conferred by an incarceration history, women may be particularly vulnerable to various affronts to their power in ways that are influenced by their gender (Wingood & DiClemente, 2000), such as discrimination, sexual assault, and economic disempowerment.

The evidenced association among the general population between personal agency and self-efficacy for quitting alcohol may operate differently in a sample of incarcerated women. In order to more effectively inform health services and health policy development, the current study sought to explore the association between personal agency and self-efficacy among a sample of incarcerated women with AUDs.

Factors influencing personal agency

Although personal agency is broadly associated with self-efficacy in the general population, this construct may also have unique factors that are differentially associated with incarcerated women’s self-efficacy for quitting alcohol use. For example, personal agency may be experienced differently in intrapersonal (i.e., agency to make choices for oneself) versus interpersonal (i.e., agency to make choices influencing relationships) domains. Therefore, exploring intra- and interpersonal factors influencing personal agency and exploring whether those unique factors are associated with self-efficacy may be valuable in understanding the impact of personal agency on self-efficacy to quit alcohol use.

Various individual factors may influence women’s capacity to make independent and free choices. For example, minority ethnic and racial status may magnify the challenges experienced during the re-entry period because individuals of a minority background are more likely to encounter institutionalized racism and other impediments to personal agency (Mann, 2002). AUD and addictions to additional substances may also reduce personal agency as symptoms of compulsivity and impulsivity can influence choices related to substance use (Gullo et al., 2014). Urges to drink (e.g., DSM-V criteria for urges directly queries whether individuals felt the “need” to drink), for example, have an important influence over one’s perceived ability to control drinking (Larimer, Palmer, & Marlatt, 1999; Miller et al., 1996). Finally, engagement in self-care related behaviors (e.g., sleeping well, eating a healthy diet, taking care of mental and physical health) has been found to be associated with better substance-related outcomes (Neale et al., 2016) suggesting that individuals with higher levels of personal agency in engaging in self-care behavior may feel more confident in efforts to quit alcohol.

With respect to interpersonal (i.e., relationally oriented) factors influencing personal agency, previous research underscores the importance of social networks in alcohol-related outcomes (Miller et al., 1996). When it comes to personal agency, the ability to select participation in or exit from a social network reflects an important area of personal agency that influences alcohol-related outcomes (Hearn, Whitehead, Khan, & Latimer, 2015; Staton-Tindall et al., 2011). Incarcerated women, for example, note the importance of personal agency in choosing sober networks (Jennifer E. Johnson et al., 2013). Examining choice in sexual partners may also capture one element within the broader construct of interpersonal agency. Engagement in transactional sex, for example, represents one of the viable means that disadvantaged women have of providing for economic needs (El-Bassel, Simoni, Cooper, Gilbert, & Schilling, 2001). Furthermore, transactional sex may be a means of gaining access to substances to which women are already addicted, thereby limiting the perceived freedom in selecting sex partners for women with AUDs.

The current investigation included data from 173 incarcerated women with AUDs. The focus of this study was on exploring the association between two variables: personal agency and self-efficacy to make changes to alcohol use. Because personal agency may operate differently in an incarcerated sample of women, this study sought to take a more granular approach to examining the association between personal agency and alcohol quitting self-efficacy. Factors hypothesized to influence one’s ability to direct choices related to drinking behaviors included (a) demographic characteristics, (b) severity of substance use, (c) self-care, (d) social network choice, and (e) agency in selecting sexual partners. Each of these factors influencing personal agency was examined in association with our primary outcome of interest: self-efficacy in quitting alcohol use.

Method

Study design

Data for the current study were taken from the baseline assessment of a larger trial examining an intervention that connects incarcerated women with a twelve-step volunteer during pretrial jail detention (Johnson, Schonbrun, & Stein, 2014; Schonbrun, Johnson, Anderson, Caviness, & Stein, 2017). Recruitment of participants from pretrial jail detention began in 2013 and is currently ongoing. The larger study had Institutional Review Board approval, participation was voluntary, and confidentiality was assured through a Federal Certificate of Confidentiality. Baseline assessments were conducted at the jail. Before participation in the study, participants were told that: (a) not participating/participating would have no impact on their time in jail or on their sentencing process; (b) the study had a Certificate of Confidentiality thereby ensuring that all responses would be kept confidential within the research study; and (c) except when danger to self or other was identified, no information provided during the study would be shared with jail staff, officers of the court, parole officers or any other individual involved in the correctional setting. The larger trial is registered at Clinicaltrials.gov (NCT01970293).

Study eligibility criteria included: (a) an age of 18 years or older; (b) unsentenced or sentenced to 60 days or fewer, (c) residence in close proximity to recruitment site and participant plan to remain in the area for the next 6 months, (d) endorsing DSM-5 criteria for an AUD (i.e., two or more DSM-5 criteria met) in the last six months, (e) participant not expecting to attend residential alcohol or drug treatment upon release, and (f) English proficiency. Participants were also required to provide the names of at least two locators (i.e., individuals who would be able to help study staff locate the participant after the participant’s jail discharge).

Participants

A total of 500 women were screened for the current study. Three hundred and eight women were ineligible for participation for reasons including expectations of being sentenced, not living in close proximity to the recruitment site, or not meeting criteria for an alcohol use disorder. Of the 192 eligible women, 19 individuals were excluded from the study for additional reasons including declining to participate after consenting procedures had been completed, or incompetency. A total of 173 individuals were included in this analysis.

Measures

Alcohol self-efficacy was measured using a single item on a continuous 10-point scale. Participants were asked: “How successful do you expect to be in quitting using alcohol at this time? Be realistic about this, based on your past experiences and present strength of motivation.” Possible responses on alcohol-self-efficacy ranged from 1, lowest expectation of success in quitting alcohol, to 10, highest expectation of success in quitting alcohol. While previous research has employed multiple-item scales for measuring alcohol self-efficacy (DiClemente, Carbonari, Montgomery, & Hughes, 1994), there is evidence for the convergent, discriminant, and predictive validity of a single item measure (Hoeppner, Kelly, Urbanoski, & Slaymaker, 2011).

Demographic variables:

Demographic characteristics assessed included age, and race/ethnicity.

Alcohol use disorder criteria.

The Structured Clinical Interview (SCID) for alcohol symptoms (SCID; Spitzer, Williams, Gibbon, & First, 1990; Williams, Gibbon, First, & Spitzer, 1992) was used to assess AUD criteria. Entry eligibility for the larger study required that women met criteria for an AUD in accordance with the DSM-5. A score of total number of criteria (out of 11 possible) endorsed was tabulated.

Other drug use.

Data on cocaine and opiate use, the most common drugs of abuse in this population (Johnson & Zlotnick, 2012) was collected using the drug module of the Addiction Severity Index (ASI; McLellan et al., 1992). Any reported use of cocaine and opiates in the past 90 days was coded as positive (i.e., dichotomized into a yes/no variable for past 90 day use).

Social Network Choice:

A seven-item scale assessing the degree to which women interact with others who they believe may put their sobriety at risk in order to obtain housing, transportation, and money was developed for this study. Our previous qualitative work (Johnson et al., 2013; Johnson et al., 2015) with incarcerated women guided the item development of this scale that includes items like “How much do you have to interact with people who are risky to your sobriety in order to get housing, transportation, or money?”

Self-care:

Guided by our previous work with this population, a nine-item assessment tool measuring self-worth and self-care was developed (Johnson et al., 2013; Johnson et al., 2014). Items included “I take good care of myself” and “I put myself in dangerous situations” and “I am worth protecting.” Items were answered on a 5-point scale ranging from strongly agree to strongly disagree.

Urges to drink:

The Penn Alcohol Craving Scale (PACS) is a five-item self-report measure of alcohol craving (Flannery, Poole, Gallop, & Volpicelli, 2003; Flannery, Volpicelli, & Pettinati, 1999). Items reflect the frequency, intensity, and duration of craving. In the current study, participants were asked about their experiences of craving for alcohol since their arrival at the jail. Items are scaled from 0 to 6 with higher mean scores indicating higher levels of craving.

Involvement in Transactional Sex.

Two items evaluating engagement in transactional sex were taken from the Risk Assessment Battery (Metzger, Woody, & Navaline, 1993). Women were asked how often they had traded sex for money, or sex for drugs, in the 90 days prior to their current incarceration period. Response options ranged from (0) never to (7) more than once a day. The two items were collapsed and dichotomized to create a variable reflecting whether or not (i.e., yes/no) participants had traded sex for either drugs or money in the prior 90 days.

Analytic approach

Descriptive statistics summarizing the characteristics of the sample are presented in Table 1 and product-moment correlation coefficients describing the unadjusted associations of alcohol self-efficacy with demographic characteristics, severity of AUD, past 90-day cocaine and opioid use, and measures of personal agency are presented in Table 2. Multiple linear regression was employed to estimate the adjusted associations of these same variables with alcohol self-efficacy. Due to heteroskedasticity in the data, the Huber-White robust variance estimator was applied for all tests of significance when estimating 95% confidence intervals. We report fully standardized coefficients for continuous covariates, and y-standardized coefficients for categorical covariates.

Table 1.

Sample characteristics (n = 173).

Variable	n (%)	M (±SD)	Median	Range
Age		35.6 (±9.9)	34	18–68
Race (White)	120 (69.4)
Latina (Yes)	18 (10.4)
No. AUD criteria met		7.4 (±2.4)	8	3–11
Past 90 day cocaine use (Yes)	80 (42.6)
Past 90 day opiate/opioid use (Yes)	60 (31.9)
Transactional sex (Yes)	74 (42.8)
Social network choice		2.3 (±2.3)	1.6	0–7
Self-care		31.2 (±5.5)	31	13–44
Urges		2.9 (±1.7)	3	0–6
Alcohol self-efficacy		8.0 (±1.9)	8	1–10

Note. AUD = alcohol use disorder.

Table 2.

Product–moment correlations (n = 173).

	1	2	3	4	5	6	7	8	9	10	11
1	1.00
2	0.00	1.00
3	−0.02	−0.10	1.00
4	−0.11	−0.03	−0.23**	1.00
5	−0.08	0.06	−0.01	0.13	1.00
6	−0.06	−0.00	−0.02	−0.02	0.20**	1.00
7	−0.11	−0.17*	0.05	0.02	0.13	0.41**	1.00
8	0.01	−0.12	0.08	−0.09	0.16*	0.20**	0.32**	1.00
9	0.25**	−0.05	−0.17*	−0.01	−0.19*	−0.07	−0.12	−0.07	1.00
10	−0.21**	−0.09	−0.07	0.07	0.33**	0.326*	0.29**	0.35**	−0.22**	1.00
11	−0.26**	0.03	−0.08	0.05	0.26**	0.37**	0.35**	0.18*	−0.20**	0.19*	1.00

Note. Variables: 1 = alcohol self-efficacy; 2 = years age; 3 = race (White); 4 = ethnicity (Latino/a); 5 = number of AUD criteria met; 6 = past 90-day cocaine use (yes); 7 = past 90 day opiate/opioid use (yes); 8 = social network choice; 9 = self-care; 10 = PACS; 11 = transactional sex.

^*p < .05,

^**p < .01.

Results

Participants averaged 35.6 (±9.9) years of age; 69.4% were White, and 10.4% were Latina (Table 1). On average participants met 7.4 (± 2.4, Median = 8) DSM-V AUD criteria. About 42.6% reported use of cocaine in the past 90-days, and 31.9% reported use of any illicit opiate or opioid during that same period. Seventy-four (42.8%) reported they had engaged in transactional sex. Descriptive statistics for the cravings, self-care, and social network choice are also reported in Table 1. The mean alcohol self-efficacy score was 8.0 (± 1.9, Median = 8; range: 1–10).

Product-moment correlations among variables used in the multivariate regression model are reported in Table 2. As predicted, self-care (r = 0.25, p < .01) was significantly associated with higher alcohol self-efficacy. Alcohol cravings were significantly (r = −0.21, p < .01) associated with lower alcohol self-efficacy and engaging in transactional sex (coded 1 on the dichotomous indicator) was also significantly correlated (r = −0.26, p < .01) with lower alcohol self-efficacy. Additional significant bivariate associations include that between cocaine (r = 0.37, p < .01) and opiate use (r = 0.35, p < .01), and transactional sex. Cravings were associated with number of AUD criteria (r = 0.33, p < .01) endorsed, and cocaine (r = 0.20, p < .01) and opiate use (r = 0.32) were associated with higher likelihood of engaging with people who put their sobriety at risk in order to be able to meet their own basic needs (e.g., for housing). Contrary to prediction, minority status, AUD severity (number of criteria met), cocaine use, opiate/opioid use, and social network choice were not associated with self-efficacy to quit alcohol use.

Finally, controlling for the statistical effect of other variables of interest (see Table 3), alcohol self-efficacy was positively and significantly associated with self-care (β = 0.17, p = .012) and inversely and significantly associated with alcohol craving (β = −0.19, p = .029). Controlling for other variables in the regression model, women who reported engaging in transactional sex had lower adjusted mean alcohol-self efficacy (β = −0.48, p = .007) than those who had not engaged in transactional sex. Alcohol self-efficacy was not associated significantly with demographic characteristics, AUD severity, cocaine or opioid use, or social network choice in the multivariate model.

Table 3.

Regression of alcohol self-efficacy on demographic characteristics, substance use, and measures of empowerment (n = 173).

Variable	βa (95% CI)	t (p = )b
Age	−0.00 (−0.15; 0.14)	−0.01 (.993)
Race (White)	−0.10 (−0.43; 0.22)	−0.62 (.533)
Latino/a (Yes)	−0.28 (−0.85; 0.28)	−0.98 (.328)
No. AUD criteria met	0.06 (−0.10; 0.22)	0.71 (.479)
Used cocaine (yes)	0.12 (−0.21; 0.45)	0.72 (.474)
Used opiates/opioids (yes)	−0.03 (−0.38; 0.33)	−0.15 (.883)
Choice index	0.12 (−0.04; 0.27)	1.44 (.152)
Self-care	0.17 (0.04; 0.31)	2.53 (.012)
PACS	−0.19 (−0.37; −0.02)	− 2.21 (.029)
Transactional sex (Yes)	−0.48 (−0.84; −0.13)	−2.72 (.007)
Constant	0.36 F10,140 = 2.63 (p = .005) R2 = 0.16

^aReported coefficients were fully standardized for continuous covariates and y-standardized for categorical covariates.

^bt-statistics, p-values, and 95% confidence interval estimates were based on the Huber-White robust variance estimator.

Bold indicates p < .05.

Discussion

Behavioral health models have identified personal agency an important correlate of self-efficacy for quitting alcohol (Cooke et al., 2016). However, such models have not yet been tested in samples of incarcerated women with AUDs, limiting their utility in guiding health intervention and health policy development for this growing and underserved population. Accounting for other variables, we found that three measures of personal agency were associated with self-efficacy for quitting alcohol: Lower alcohol cravings, greater self-care, and non-engagement in transactional sex. Personal agency at both intrapersonal and interpersonal levels might then be implicated as a target for change among incarcerated women with AUDs. Making changes to habitual behaviors is challenging (Bandura, 2006), so individuals must have sufficient ability to direct choice in order to be able persist in the face of sometimes difficult or painful challenges. Incarcerated women, by virtue of their imprisonment, have personal agency removed or limited in many important ways, but particular behaviors and alcohol needs exacerbate this loss of agency.

Almost half of the study sample reported having engaged in transactional sex in the three months prior to incarceration in order to gain access to money or substances. This rate is consistent with that reported in other studies focusing on incarcerated women (29%–48%; Farel et al., 2013; Schonbrun, Johnson, Anderson, & Stein, 2015). Transactional sex may be an important marker of the kinds of restrictions to personal agency experienced by women who are involved in the criminal justice system, and may have limited opportunities to secure gainful employment. In other words, transactional sex may reflect an important area where personal agency is impeded for women who have been incarcerated and who struggle with AUDs. In line with this hypothesis, we found that engagement in transactional sex remained significantly associated with alcohol self-efficacy, even after controlling for the effect of other variables of interest. The effect of decreased personal agency in selecting sexual partners may influence self-efficacy for quitting alcohol use because alcohol may be used both to facilitate sex with customers (Li, Li, & Stanton, 2010) and to cope with the stressors of the occupation (Gossop, Powis, Griffiths, & Strang, 1994), and further exploration of this important domain of personal agency is warranted.

In our sample, the bivariate association between transactional sex and cocaine/opiate use reached a medium effect size (Cohen, 1988). The effect is reflected in previous research documenting that sex work and substance use are associated both with each other and with involved in the criminal justice system (El-Bassel et al., 2001; Schonbrun, Johnson, Anderson, & Stein, 2016). This predictable association highlights the importance of delving into further explorations of the relationship between substance use and/or addiction with engagement in transactional sex. It is also unsurprising that an association between higher craving for alcohol and greater number of endorsed AUD criteria emerged; the relationship between these two constructs highlights craving as a core marker of disordered alcohol use. Clinical and neurological research demonstrates the important role that craving plays in AUDs (e.g., Schacht, Anton, & Myrick, 2013; Yoon, Kim, Thuras, Grant, & Westermeyer, 2006). Finally, both opiate and cocaine use were associated with a greater likelihood of interaction with peers who put sobriety at risk; participants interacted with these peers in order to meet basic needs, such as for housing, transportation, and employment. This finding underscores the importance of how “people, places, and things” can put sobriety at risk. Indeed, initial drug use, treatment entry, and relapse can each be influenced by social triggers or cues (Davey, Latkin, Hua, Tobin, & Strathdee, 2007; Gorsuch & Butler, 1976; Latkin, Knowlton, Hoover, & Mandell, 1999; Lloyd et al., 2005; Moos, 2007).

Indices of personal agency in self-care and craving management also emerged as having significant associations with alcohol abstinence self-efficacy, even after statistically controlling for the effect of other variables. One’s ability and skillset in engaging in self-care sets the stage for both skills and confidence to make changes to drinking behavior. Previous research in areas as wide-ranging as diabetes, HIV, and caregiver fatigue management, suggest that self-care is a critical part of managing stress (Chou, Holzemer, Portillo, & Slaughter, 2004; Corless et al., 2012; Gharaibeh, Gajewski, Alsmadi, & Boyle, 2016; Soto et al., 2015; Tomasulo, 2002), and that heightened levels of stress may be associated with unhealthy coping behaviors (Rao, 2009; Rodriquez, Gregorich, Livaudais-Toman, & Pérez-Stable, 2017; Srivastava, 2008). Indeed, alcohol use may serve as a specific method for stress management (Condit, Kitaji, Drabble, & Trocki, 2011; Magrys & Olmstead, 2015; Scott-Sheldon et al., 2013), which for women who are experiencing a chaotic and stressful period as they re-enter their communities after an incarceration period, may be an important area of intervention. Conversely, reports of higher levels of urges to drink were negatively associated with self-efficacy to quit alcohol. Symptoms of alcohol cravings have been reported as an important influence on one’s ability to control drinking (Larimer et al., 1999; Miller et al., 1996). A recent functional magnetic resonance imaging study demonstrated that the neural circuitry involved in behavioral control is differentially impacted by the severity of an AUD (Claus, Ewing, Filbey, & Hutchison, 2013), further limiting behavioral agency in a population such as ours with high levels of severity.

Despite the high rates of alcohol and illicit substance use, cravings, and engagement in sex for money or substances, participants in this sample, on average, reported high self-efficacy in quitting alcohol use. The scaling of the efficacy measure presented a lowest possible alcohol efficacy of 1 and highest possible efficacy of 10, and participants scored a mean 8.0 (SD = 1.9). Self-efficacy may be higher during the time of incarceration in part because incarceration is a forced time of sobriety although self-efficacy ranges across samples (Saxena, Grella, & Messina, 2015; Swopes, Davis, & Scholl, 2017). Incarceration may nevertheless be a period during which individuals may be prompted to consider behavioral change, including substance-related behavioral change. Indeed, evidence suggests that diminished use of substances may be evident during the postrelease period (Tangney et al., 2016). Along these lines, the time of incarceration may be leveraged as a powerful time of intervention, both to take advantage of high levels of self-efficacy and to prompt women to consider movement in self-efficacy.

Taken together, it appears that personal agency in terms of selecting sexual partners, managing AUD symptoms, and engaging in self-care is associated with confidence in making alcohol consumption changes in a sample that experiences serious restrictions to agency. Perceived agency, even among a sample of incarcerated women has variability, and is therefore a worthwhile target for alcohol interventions. Helping women to identify the areas where they do have control or agency, for example in managing their symptoms using mindfulness techniques, may prove valuable. Moreover, working with women to help increase personal agency at a behavioral level, for example by providing both psychoeducation and skills training in selecting social networks, securing legal employment, managing alcohol cravings, and finding accessible ways to engage in self-care may support better alcohol outcomes.

Our study had a number of limitations. First, our findings were cross-sectional rather than longitudinal. Future research should continue to examine personal agency among marginalized populations, including incarcerated women. Future investigations would also benefit from longitudinal investigations of agency and self-efficacy, and should include a direct analysis of the influence of personal agency on actual alcohol use during the re-entry period. Second, most women had severe AUD and our findings might not apply to women with less severe alcohol problems. Future research should broaden out from this investigation to include additional important variables, such as perceived agency in other kinds of relationships (e.g., friendships) and perceived levels of stress. Finally, our sample was incarcerated and thus our findings may not apply to women who are not in jail; we recommend that future research explore the association between agency and self-efficacy in similar populations including women in prison (versus jail) settings and incarcerated women with other forms of substance use disorders (vs. AUDs).

Conclusions

Because personal agency among criminally involved populations is limited relative to the general population, and because these limitations may be heightened among women (vs. men) who are incarcerated, it is important to develop treatment interventions that target these variables. Indicators of personal agency including lower alcohol cravings, greater self-care, and less engagement in transactional sex were associated with greater self-efficacy to quit alcohol use postincarceration. Interventions that improve women’s personal agency directly (i.e., education and job training) and reduce potential need for transactional sex (including addiction treatment, safe housing, and childcare support) may be especially helpful. In addition, interventions fostering self-care and providing strategies for managing cravings may be useful in enhancing confidence in making changes to alcohol use post-jail discharge. Dialectical behavior therapy, as one prominent example, offers a module that emphasizes self-care as it relates to emotion dysregulation and impulsivity. Developing skills in self-soothing and in “improving the moment” may offer an opportunity to improve self-care, whereas applying mindful attention to emotions and distress tolerance techniques may offer avenues to managing cravings more effectively. Approaches that provide women with more life choices and/or skills for managing cravings are worthy of direct evaluation.