Gender-specific participation and outcomes among jail diversion clients with co-occurring substance use and mental health disorders

Abstract

Men and women with co-occurring substance use disorders and mental illness are at relatively high risk for becoming involved in the criminal justice system. Programs, such as post-booking jail diversion, aim to connect these individuals to community-based treatment services in lieu of pursuing criminal prosecution. Gender appears to have an important influence on risk factors and pathways through the criminal justice system, which in turn may influence how interventions like jail diversion work to engage men and women in treatment services and reduce recidivism. Different circumstances, levels of engagement, and outcomes by gender may be related to both person-level characteristics and external factors such as availability of gender-specific services and resources. This mixed-methods study identified specific ways in which men and women use services and reoffend after being diverted, and complemented those findings with in-depth insights from program clinicians about how program experiences and resources differ in important ways by gender. We matched and merged administrative records from 2007 to 2009 for 16,233 adults from several state agencies in Connecticut, and included data on demographic characteristics, clinical diagnoses, outpatient and inpatient behavioral health treatment utilization, arrest, and incarceration. Using propensity analysis, the 1,693 men and women who participated in the statewide jail diversion program were matched to respective comparison groups of nondiverted men and women. We used longitudinal multivariable regression analyses to estimate the effects of jail diversion participation on treatment utilization, arrest, and incarceration, separately for men and women. We conducted three focus groups with jail diversion clinicians from around the state (n=21) to gain in-depth insight from them about how circumstances, program experiences, and resources differ by gender in important ways; these subjective clinician insights complement the quantitative analyses of diversion outcomes for men and women. For both men and women, diversion was associated with reductions in risk for incarceration and increases in utilization of outpatient treatment services. For men only, diversion was associated with higher utilization of inpatient mental health care. No differences in treatment or criminal justice outcomes were observed in models that compared men and women directly. Major themes from the focus groups included: the existence of too few inpatient and residential resources for women with co-occurring disorders; different challenges to treatment engagement that men and women face; and a need for more effective, gender-specific services for all program participants. Results from this mixed-methods study offer information on gender-specific program outcomes and surrounding circumstances that can help programs to better understand and address unique risks and needs for men and women with co-occurring substance use and mental health disorders who are involved in the criminal justice system.

1. Introduction

Four to five million adults in the United States have co-occurring substance use and mental health disorders (CODs) (Epstein, Barker, Vorbuger et al., 2002; SAMHSA, 2006). Among them, women with substance use disorders are twice as likely to also have serious mental illness (SMI) as compared to their male counterparts—approximately 30 percent versus 16 percent, respectively (Epstein, Barker, Vorbuger et al., 2002). Evidence suggests that most people with CODs do not receive any treatment; and for those who do, it is often for substance use or for mental health, but rarely for both disorders (Epstein, Barker, Vorbuger et al., 2002; Kessler, Demler, Frank, et al., 2005). A related public health crisis among people with CODs is their striking overrepresentation in the criminal justice (CJ) system (Steadman, Osher, Robbins, et al., 2009; Abram, Teplin, & McClelland, 2003; Abram, 1990; Baillargeon, Penn, Knight et al., 2010; Wood, 2011). Recent estimates of prevalence of serious mental illness and rates of co-occurring substance use disorders in U.S. jails indicate that as many as 1.4 million people with CODs are booked into U.S. jails each year, accounting for nearly 11 percent of all new jail admissions (Steadman, Osher, Robbins, et al., 2009; Abram, Teplin, & McClelland, 2003; Proctor, 2012; Nowotny, Belknap, Lynch, 2014; Lynch, DeHart, Belknap, et al., 2014; Proctor, Hoffmann, & Raggio, 2018; Prince & Wald, 2018). So, while the prevalence of CODs is approximately three percent in the general population, it may be as high as 10 percent for men and 22 percent for women among jail inmates (Steadman, Osher, Robbins, et al., 2009).

Compared to men, women with CODs appear to have unique risks for entering the CJ system (Teplin, 1994; Teplin, Abram, McClelland, 1996; Zlotnick, 1997), which is consistent with the larger relative increase in the rate at which they are entering the CJ system (42 percent increase between 2003 and 2007 among women as compared to 24 percent increase among men), and nearly double the rate of CODs among incarcerated women as compared to men (Teplin, 1994; Teplin, Abram, McClelland, 1996; Zlotnick, 1997; Greenfeld & Snell, 1999; Muruschak, 2008; Bronson & Berzofsky, 2017; van Wormer & Bartollas, 2007; Nowotny, Belknap, Lynch, 2014). Women with serious mental illness also appear to have a higher relative likelihood of being arrested as compared to women without mental illness, potentially accounted for, in part, by discretionary policing practices that may affect men and women differently (Crocker, Hartford, & Heslop, 2009). Furthermore, evidence suggests that women in jail who have SMI are 50 percent more likely than their male counterparts to have substance dependence (Abram & Teplin, 1991), and commonly enter the CJ system on drug-related charges, at which time their mental health needs may often be identified for the first time (Hills, 2003; Peters & Osher, 2003). Many women with CODs face a complex combination of challenges, such as trauma and physical abuse in childhood that continues into adulthood, including intimate partner violence, underemployment and broad socioeconomic disadvantage, and caring for dependent children. Women under these circumstances often fear losing custody of their children, which can discourage them from seeking treatment (Teplin, Abram, McClelland, 1996; Zlotnick, 1997; van Wormer & Bartollas, 2007; Bloom, Owen, Covington, 2003; Jordan, Schlenger, Fairbank, et al., 1996; Lewis, 2006; Sacks, 2004; Knight, Logan, & Simpson, 2001; Glaze & Maruschak, 2010; Andrews, Bonta, & Wormith, 2006; Spjeldnes & Goodkind, 2009; DeHart, Lynch, Belknap, 2013; Lynch, DeHart, Belknap, et al., 2014; Stainbrook, Hartwell, James, 2015; Fedock, Fries, & Kubiak, 2013; Cusack, Herring, Steadman, 2013). The combination of these social-environmental challenges may exacerbate CODs, significantly increase women’s risk of entering the CJ system, and affect how well they engage in and respond to treatment interventions. By comparison, criminal offending among men appears to be driven less by addiction and major psychopathology, and more by antisocial personality and other criminogenic risk factors (Andrews, Bonta, & Wormith, 2006), which drive higher rates of offending among men than women in the general population as well (Fisher, Silver, & Wolff, 2006; Skeem, Manchak, & Peterson, 2011).

Over the last 20 years, drug courts—and more recently, mental health courts—have diverted people with substance use disorders who have criminal charges or convictions away from the CJ system and instead into treatment. Evidence for effectiveness has been mixed, but finds, overall, that participation in drug courts and mental health courts is associated with increased use of community behavioral health services, and that these courts are cost-effective in reducing recidivism and, to some extent, substance use among participants (US GAO, 2005; Brown, 2010; Steadman, Redlich, & Callahan, 2011; Cosden, Ellens, Schnell, et al., 2003; Lamberti, Weisman, Schwarzkopf, et al., 2001; Hoff, Baranosky, Buchanan, et al., 1999; Swartz, Wilder, Swanson, et al., 2010; Gilbert, Moser, Van Dorn, et al., 2010; Costopoulos, Swanson, Tyc, et al., 2019). The drug and mental health court models—where the primary focus is on either substance use or mental illness, respectively—often do not address the complex needs of justice-involved adults with CODs (Belenko, 2001; Cooper, 1997; Hagerdorn & Willenbring, 2003). Other types of diversion programs accommodate people with CODs, directing them to treatment for both disorders as well as recovery-oriented services such as supportive housing and supported employment (Becker, Drake, & Naughton, 2005; Bond, Drake, & Becker, 2008; Drake & Bond, 2008; Foster, LeFauve, Kresky-Wolff, et al., 2010; Rosenheck, Kasprow, Frisman, et al., 2003).

In Connecticut’s innovative jail diversion program, clinicians from community mental health agencies are based in the court and work with police, prosecuting attorneys, and judges to identify people (at arraignment) with substance use disorders, SMI, and often CODs who are appropriate for diversion into treatment services (Frisman, Sturges, Baranoski, et al., 2001). Evidence suggests that diversion participants with CODs are less likely to be re-incarcerated and spend fewer days incarcerated, though with inconsistent evidence of improvements in mental health and substance use outcomes (Frisman, Lin, Sturges, et al., 2006; Steadman & Naples, 2005; Case, Steadman, Dupuis, et al., 2009; Broner, Lattimore, Cowell, et al., 2004; Shafer, Arthur, & Franczak, 2004). Diversion participants’ clinical and CJ responses may vary independently, insofar as these outcomes may be driven by different psycho-pathologic, criminogenic, or other situational risk factors—alone or in combination (Swanson, Van Dorn, Swartz, et al., 2008; Swanson, Swartz, Essock, et al., 2002; Silver, 2006; Fisher, Silver, & Wolff, 2006).

While evidence for jail diversion is promising, we do not know whether jail diversion works differently for men and women with CODs, given their unique needs and pathways into the CJ system. That women are diverted at a higher rate than men (Naples, Morris, & Steadman, 2007) has long been the subject of speculation among diversion experts, and the continuing rapid increase of incarceration among women appears to be related mainly to drug use and convictions (Abram, Teplin, & McClelland, 2003; Abram & Teplin, 1991; Glaze & Kaeble, 2013). Race/ethnicity plays an important role in diversion selection, too, even when controlling for criminal offense type and severity—non-Hispanic whites, especially women, are most likely to be diverted (Naples, Morris, & Steadman, 2007), which is in contrast to the overrepresentation of men and people of color in the CJ population. (Spjeldnes & Goodkind, 2009).

These trends suggest that men and women have different risk factors and pathways through the system, which in turn may influence how interventions like jail diversion work to engage them in treatment services and reduce recidivism. Differential participation and outcomes by gender may be related to characteristics of the men and women themselves, driven by different histories, treatment needs and preferences, and life circumstances. Gender differences may also be driven by external factors, including tendencies of judges and other court personnel to treat men or women preferentially, or insufficient gender-specific services and resources.

This mixed-methods study aimed to identify gender differences in how men and women use services and reoffend differently after being diverted, and to gain insight from program clinicians about how program experiences and resources differ in important ways by gender.

2. Materials and methods

The Connecticut Department of Behavioral Health and Addiction Services (DMHAS) provided administrative records on demographic characteristics, clinical diagnoses, outpatient treatment utilization, and state-funded hospitalizations for mental health and substance use treatment. The Department of Social Services Medicaid program provided service claims for outpatient service utilization, emergency department (ED) and crisis center visits, and psychiatric and substance use hospitalizations in community hospitals. The Connecticut Department of Correction (DOC) provided data on periods of incarceration, periods of time under probation by the Judicial Branch, and arrest records by the Department of Public Safety. The Court Support Services Divisions, Judicial Branch provided diversion program participation data; referral data were not available, however. We matched, merged, and de-identified the data from these public agencies, originally for another study (NIH R01-MH086232) using Link King software V9, a public-domain software that integrates both probabilistic and deterministic matching algorithms to identify individual cases across all datasets that may include name misspelling, inversion of digits of the Social Security number or date of birth, or other differences between datasets. The Link King application has shown to have high accuracy for records linkage, with sensitivity at 96.6% and positive predictive at 96.1% (Campbell et al, 2008).

We identified adults 18 or older with recorded diagnoses of mental illness (schizophrenia spectrum, bipolar disorder, or major depressive disorder) and co-occurring substance use disorders, and with at least one night in DOC during the sample inclusion window. We then created two groups: the diversion group (those who were arrested and then participated in Connecticut’s jail diversion program) and the non-diversion group (those who were arrested but did not participate in the diversion program). We did not have data on individuals who were evaluated for the program and either declined or were determined to be ineligible. We had access to data from January 2005 to December 2009. To allow for at least 12 months of observation before the end of 2009, we included only individuals who were arrested and released to the community during January 1 2006–December 31 2008.

For the diversion group (n=1,693; n=553 women, n=1,140 men), the index arrest was defined as the temporally closest arrest preceding the diversion date by no more than 90 days. For the comparison group (n=14,540; n=4,480 women, 10,060 men), the index arrest was defined as the last observed arrest during 2007–2009. There were 16,233 adults who met all study criteria and were eligible to be included in the analytic sample.

We then created two propensity-score-matched subsamples—one each for men and women— to estimate gender-specific outcomes associated with diversion for individuals of similar propensity to participate in diversion.

2.1. Measures

2.1.1. Outcomes

We measured the length of time (survival) to outpatient treatment, hospital-based healthcare utilization, and criminal offending by the number of days from the first community day after the index arrest to the first instance of the outcome within a one-year observation period. We measured outpatient treatment in three ways: crisis service utilization; intensive outpatient therapy, residential, case management and social rehabilitation for mental health treatment services; and similar services for substance use treatment. We measured hospital-based healthcare in three ways: any inpatient mental health treatment; any inpatient substance use treatment; and any visits to the ED. We measured criminal offending in two ways: any criminal arrest conviction (excluding technical violations and moving violations), and any days in jail or prison.

2.1.2. Explanatory variables: Covariates

We adjusted estimates for demographic and clinical characteristics, index arrest type and severity, and observed criminal justice involvement and service utilization over the 12 months preceding the arrest. Demographic and clinical characteristics included: age at time of index arrest; race/ethnicity (African American, Hispanic/Latino, and non-Hispanic white or other); educational need, measured by DOC on a scale of 1 (lowest) to 5 (highest); primary SMI diagnosis (schizophrenia spectrum disorder, bipolar disorder, or major depressive disorder); and substance dependence type (opioid vs. other). We characterized the most serious charge of the index arrest using a severity scale of 1 (highest) to 7 (lowest), based on offense type (felony or misdemeanor), class (A, B, C, D), and sentencing guidelines (for unclassified felonies). We also described index arrest using four dichotomous indicators corresponding to drug, prostitution, violent, and felony charges. To adjust for pre-index criminal justice involvement, we included count variables measuring jail days, arrests, and probation days; dichotomous variables indicating jail, arrest, and probation; an indicator of any jail, arrest, or probation; and indicators of 14 crime classifications (e.g., drug, property, DWI, various classifications of violence). To adjust for pre-index service utilization, we included dichotomous and count versions of inpatient (IP) treatment for mental health (MH), SA (substance abuse), and medical treatment; outpatient (OP) for MH, SA, and medical; use of ED or crisis services; and dichotomous variables for Medicaid enrollment and SSI/SSDI receipt. We also adjusted for index year (year of arrest), which was measured as a categorical variable (2007, 2008, 2009).

2.2. Analysis

To create propensity-matched groups, we used all the covariates listed above in a logistic regression model to obtain propensity scores, then carried out a 1:1 case-control match using a local optimal, or “greedy”, algorithm to identify matching pairs for each study group, in which we matched each diversion participant in sequential order to the non-diverted comparison group participant with the closest propensity score (Coca-Perraillon, 2007). The pool of potential matches included 16,233 cases, after we performed a a 1:1 propensity matching. We removed an additional 283 participants because they were not in the community during the observation period and thus not eligible for the current study. The 283 with no community days for observation were more likely to be men and less likely to have been diverted (logistic regression, not shown). As shown in Table 1, the differences between this overall group and the group of 505 female and 1006 male clients was significantly different on most of the variables used for matching. However, after we completed the matching, we found no significant differences. Thus, it appeared that we had identified a suitable matching group of 508 female and 995 male non-diverted cases.

Table 1.

Characteristics of female sample: Pre- and post-propensity score matching.

	1a. Female pre-matching					1b. Female post-matching
	Non-diverted (n=4,480)		Diverted (n=553)			Non-diverted (n=508)		Diverted (n=505)
	n	%	n	%	p	n	%	n	%	p
Demographic and clinical characteristics
Age (mean, SD)	(34.32)	(10.06)	(36.37)	(9.77)	***	(35.41)	(10.06)	(36.15)	(9.73)
Educational need (mean, SD)	(2.24)	(0.75)	(2.17)	(0.78)	*	(2.18)	(0.79)	(2.16)	(0.78)
Race/ethnicity
White	2491	55.60	331	59.86		276	54.33	307	60.79
Black	1250	27.90	153	27.67		148	29.13	133	26.34
Hispanic	712	15.89	66	11.93		81	15.94	62	12.28
Other	27	0.60	3	0.54		3	0.59	3	0.59
SMI diagnosis					***
Schizophrenia	1916	42.77	128	23.15		132	25.98	128	25.35
Bipolar	1629	36.36	173	31.28		152	29.92	171	33.86
Depression	935	20.87	252	45.57		224	44.09	206	40.79
Substance use disorder					*
Non-opioid	2472	55.18	278	50.27		233	45.87	258	51.09
Opioid	2008	44.82	275	49.73		275	54.13	247	48.91
Index arrest characteristics
Severity of charge (1–7, 1=most severe) (mean, SD)	(4.68)	(1.63)	(4.80)	(1.47)		(4.78)	(1.52)	(4.80)	(1.50)
Drug charge	856	19.11	93	16.82		80	15.75	89	17.62
Prostitution charge	129	2.88	53	9.58	***	41	8.07	40	7.92
Violent charge	601	13.42	85	15.37		83	16.34	77	15.25
Felony charge	1442	32.19	171	30.92		138	27.17	159	31.49
Pre-index criminal justice characteristics
Any arrest	840	18.75	326	58.95	***	271	53.35	280	55.45
Number of arrests (mean, SD)	(0.32)	(0.82)	(1.13)	(1.35)	***	(1.05)	(1.46)	(1.06)	(1.30)
Any jail	751	16.76	166	30.02	***	163	32.09	144	28.51
Jail days (mean, SD)	(14.83)	(51.07)	(25.35)	(60.42)	***	(23.82)	(58.22)	(24.99)	(61.59)
Any probation	624	13.93	113	20.43	***	92	18.11	100	19.80
Probation days (mean, SD)	(41.11)	(109.21)	(54.35)	(118.79)	*	(47.69)	(111.96)	(53.83)	(118.79)
Any jail, arrest, or probation	1627	36.32	397	71.79	***	329	64.76	350	69.31
Pre-index service utilization and insurance
Any Medicaid enrollment	3738	83.44	458	82.82		418	82.28	414	81.98
Any SSI	727	16.23	193	34.90	***	170	33.46	157	31.09
Any IP for SA	550	12.28	104	18.81	***	83	16.34	89	17.62
Any IP for MH	410	9.15	149	26.94	***	122	24.02	117	23.17
Any IP for medical	773	17.25	92	16.64		99	19.49	80	15.84
Any OP for MH	2859	63.82	418	75.59	***	361	71.06	370	73.27
Any OP for SA	2081	46.45	311	56.24	***	266	52.36	276	54.65
Any OP for medical	3204	71.52	393	71.07		366	72.05	350	69.31
Any ED/crisis	901	20.11	221	39.96	***	177	34.84	189	37.43

An additional 23 variables were included in the propensity model but not shown here (types of pre-index arrest charges, number of days utilizing various services, number of days with insurance).

After creating the comparison group for female and male cases, we carried out Cox proportional hazards regression analysis to estimate survival time to study outcomes, beginning with the first community day after the index arrest. The first community day marks the end of incarceration related to the arrest for people in the comparison group, and, for the diversion group, completion of any inpatient treatment that the program arranged as part of the diversion treatment plan. Beginning observation on this day increases comparability of the diverted and non-diverted groups, as outcomes will not be obscured by institutionalization associated with the arrest or diversion. In propensity-matched subsamples, survival models estimated effects associated with diversion within each gender group, controlling for index year and any variables that remained unbalanced following propensity-matching.

As a sensitivity analysis, we re-estimated models with a reduced sample that excluded individuals who were known to have been out of the community prior to the modeled outcome. For example, if an individual went to jail after 30 days of observation, and then entered inpatient treatment for mental health after 90 days, it is possible that the experience of going to jail delayed the inpatient admission by reducing the number of days at risk of this outcome. Similarly, entry into jail after 30 days and no record of inpatient treatment for mental health could indicate that an individual was in jail for the entire observation period with no risk of inpatient admission. For every modeled outcome, if the number of days until earliest entry into jail or inpatient treatment was lower than the number of days until the modeled outcome event (or end of study window if no event), we removed that individual from our analysis. For example, in the analysis of inpatient admission for mental health treatment, we excluded individuals who we observed going to jail or being admitted for substance use or medical treatment prior to the outcome (or end of observation period if no outcome).

We also conducted a series of Cox proportional hazards regression analyses to be able to directly compare men and women in our outcomes of interest, pooling the two propensity samples for men and women, respectively, and measuring differences in outcomes by gender with an interaction term for diversion status × gender in each model, and controlling for index year and any covariates that remained unbalanced after propensity matching. We used SAS 9.4 PROC LOGISTIC for propensity analysis and PROC PHREG for survival analysis.

2.3. Focus groups: Sample, methods, and analysis

To enrich interpretation of quantitative results, we conducted focus groups with diversion clinicians in Connecticut in 2015, after having obtained administrative data for data construction but before conducting quantitative analysis. We recruited clinicians via email from a roster of Connecticut jail diversion clinicians covering three geographic regions of the state and 10 treatment agencies (N=30) to participate in a 90-minute session; they were offered free lunch. Two facilitators asked open-ended questions to learn more about their perceptions of the key gender-specific factors that influence selection of participants into the jail diversion program, their engagement once initiated, and their likelihood of success. Seventy percent of those recruited (21/30) participated in three focus groups (4–11 per group), one each in Hartford, New Haven, and Middletown, CT. Focus groups lasted an average of 75 minutes each, were audio-recorded and transcribed (with identifiers removed and names replaced with pseudonyms), and entered into NVivo. We coded transcripts first to tag distinct chronological stages of the diversion pathway (e.g., initial referral to diversion, outcomes after diversion) and then coded to capture themes related to gender as discussed for each diversion stage. Two coders developed and applied thematic codes using a consensus process. First, text associated with each diversion stage was open-coded by one coder to identify gender-related themes, and these themes were consolidated to form a draft codebook, following the principles of grounded theory (Lofland, Snow, Anderson, & Lofland, 2005; Strauss & Corbin, 1990). Then, the second coder applied draft codes, and the two coders met to reconcile differences, revise the codebook, and code as needed to achieve consensus. A single text segment could be coded as relevant to multiple diversion stages and thematic codes. The most prominent themes were identified, with prominence defined as occurrence in two or more focus groups, or extensive discussion by multiple members in a single focus group. For the current analysis, we selected quotations from the most prominent themes from any diversion stage according to their relevance to the outcomes modeled in survival analysis and grouped into broader themes. The Duke University School of Medicine and DMHAS Institutional Review Boards approved this study.

3. Results

3.1. Quantitative results

Demographic and clinical characteristics, index arrest characteristics, and observed criminal justice involvement and service utilization over the 12 months prior to arrest are presented in Table 1 (women) and Table 2 (men).

Table 2.

Characteristics of male sample: Pre-and post-propensity score matching.

	2a. Male pre-matching					2b. Male post-matching
	Non-diverted (n=10,060)		Diverted (n=1,140)			Non-diverted (n=995)		Diverted (n=1,006)
	n	%	n	%	p	n	%	n	%	p
Demographic and clinical characteristics
Age (mean, SD)	(36.06)	(10.86)	(36.20)	(11.22)		(36.52)	(11.04)	(36.01)	(11.31)
Educational need (mean, SD)	(2.39)	(0.77)	(2.40)	(0.77)		(2.35)	(0.76)	(2.41)	(0.78)
Race/ethnicity					***
White	5296	52.64	574	50.35		550	55.28	528	52.49
Black	2199	21.86	355	31.14		252	25.33	283	28.13
Hispanic	2513	24.98	205	17.98		189	18.99	189	18.79
Other	52	0.52	6	0.53		4	0.40	6	0.60
SMI diagnosis					***
Schizophrenia	4868	48.39	197	17.28		189	18.99	197	19.58
Bipolar	2669	26.53	286	25.09		289	29.05	280	27.83
Depression	2523	25.08	657	57.63		517	51.96	529	52.58
Substance use disorder
Non-opioid	5785	57.50	683	59.91		556	55.88	602	59.84
Opioid	4275	42.50	457	40.09		439	44.12	404	40.16
Index arrest characteristics
Severity of charge (1–7, 1=most severe) (mean, SD)	(4.61)	(1.61)	(4.94)	(1.36)	***	(4.93)	(1.50)	(4.90)	(1.37)
Drug charge	1818	18.07	153	13.42	***	138	13.87	137	13.62
Prostitution charge	3	0.03	1	0.09		0	0	1	0.1
Violent charge	1577	15.68	165	14.47		161	16.18	146	14.51
Felony charge	3551	35.3	300	26.32	***	266	26.73	277	27.53
Pre-index criminal justice characteristics
Any arrest	2138	21.25	681	59.74	***	552	55.48	558	55.47
Number of arrests (mean, SD)	(0.35)	(0.84)	(1.21)	(1.47)	***	(1.05)	(1.42)	(1.05)	(1.31)
Any jail	2687	26.71	448	39.3	***	372	37.39	370	36.78
Jail days (mean, SD)	(32.75)	(77.98)	(43.50)	(83.65)	***	(40.24)	(80.69)	(42.44)	(84.99)
Any probation	1785	17.74	267	23.42	***	203	20.4	219	21.77
Probation days (mean, SD)	(50.92)	(118.97)	(57.54)	(120.02)		(51.64)	(116.67)	(56.27)	(120.51)
Any jail, arrest, or probation	4712	46.84	842	73.86	***	699	70.25	712	70.78
Pre-index service utilization and insurance
Any Medicaid enrollment	7333	72.89	885	77.63	***	754	75.78	770	76.54
Any SSI	1690	16.8	441	38.68	***	332	33.37	344	34.19
Any IP for SA	1789	17.78	201	17.63		198	19.9	179	17.79
Any IP for MH	943	9.37	310	27.19	***	240	24.12	231	22.96
Any IP for medical	1016	10.1	125	10.96		126	12.66	103	10.24
Any OP for MH	4491	44.64	792	69.47	***	651	65.43	660	65.61
Any OP for SA	4602	45.75	565	49.56	*	505	50.75	478	47.51
Any OP for medical	5311	52.79	650	57.02	**	545	54.77	559	55.57
Any ED/crisis	2086	20.74	454	39.82	***	351	35.28	362	35.98

An additional 23 variables were included in the propensity model but not shown here (types of pre-index arrest charges, number of days utilizing various services, number of days with insurance).

3.1.1. Characteristics of sample

Propensity-score-matching using 49 independent variables yielded balanced samples of women (n=505 diverted, n=508 non-diverted counterparts; Table 1) and men (n=995 diverted and n=1,006 non-diverted counterparts; Table 2). Among matched women (Table 1b), the diverted and non-diverted groups were similar on all but three variables: two pre-index service utilization variables (medical inpatient days and ED/crisis events) and one pre-index variable (any serious violent charge). Among matched men (Table 2b), the diverted and non-diverted groups were similar on all but one variable: pre-index days hospitalized for medical treatment.

In both female and male propensity-matched samples, the mean age was about 36 and more than half were white. The modal diagnosis for both male and female propensity-matched samples was major depression, followed by bipolar disorder and schizophrenia spectrum disorders, with approximately half having a diagnosis of opioid use disorder. The female propensity-matched sample included 8% with prostitution charges at index, whereas the male matched sample had virtually none. Index arrest severity was higher among men, as was the percentage with drug charges. However, the percentage with violent charges and felony charges appeared similar across genders. We did not test statistically differences between male and female propensity-matched samples.

3.1.2. Diversion outcomes

Table 3 presents outcomes for propensity-matched women (Table 3a) and propensity-matched men (Table 3b). Among women, diversion was associated with a 21% lower risk of incarceration at any time during the survival window (HR=0.79, p<.01), and with spending fewer days in jail, on average (86 days vs. 99 days). There was no statistical difference in the risk of subsequent arrest for diverted women compared to non-diverted women. We found a similar pattern of risk for jail and arrest for men: diverted men had a 27% lower risk of incarceration (HR=0.73, p<.0001) than non-diverted men in the follow-up period, and also spent fewer days in jail (120 days vs. 138 days). Similar to the pattern for women, there was no statistical difference between diverted and non-diverted men in arrest risk in the main models. Among diverted women, more individuals were arrested than went to jail (56% vs. 53%), but among non-diverted women, more individuals went to jail than were arrested (60% vs. 57%). We also observed this pattern for diverted and non-diverted men.

Table 3.

Criminal justice and treatment outcomes in propensity-matched samples arrested in Connecticut and released to the community 2007–2009.

Table 3a. Propensity-matched women (n=1,013)
	Survival models estimating effect of diversion
Outcome	Hazard Ratio	95% C.I.	p
Jail	0.79	(0.66, 0.94)	0.008	**
Arrest	1.03	(0.86, 1.23)	0.751
IP-SA	0.96	(0.72, 1.29)	0.781
IP-MH	1.01	(0.76, 1.34)	0.968
ED	1.06	(0.84, 1.35)	0.606
Crisis	1.24	(0.83, 1.87)	0.297
OP-MH	1.70	(1.46, 1.97)	<.0001	***
OP-SA	1.01	(0.85, 1.19)	0.944

Table 3b. Propensity-matched men (n=2,001)
	Survival models estimating effect of diversion
Outcome	Hazard Ratio	95% C.I.	p
Jail	0.73	(0.65, 0.82)	<.0001	***
Arrest	1.10	(0.97, 1.24)	0.126
IP-SA	0.94	(0.76, 1.17)	0.591
IP-MH	1.36	(1.10, 1.67)	0.004	**
ED	1.04	(0.87, 1.24)	0.695
Crisis	1.77	(1.34, 2.33)	<.0001	***
OP-MH	1.71	(1.52, 1.91)	<.0001	***
OP-SA	1.11	(0.98, 1.26)	0.109

Survival models adjust for index year and any covariates that remained unbalanced after propensity-matching.

IP=inpatient hospitalization; MH=mental health treatment; SA=substance use disorder treatment; ED= Emergency Department visit; OP=Outpatient treatment.

The effect of diversion on inpatient hospitalization for mental health and/or substance use differed by gender, but emergency room use did not. Diversion was associated with a greater risk of inpatient hospitalization for mental health (IP-MH) treatment (HR=1.36, p<.01) among men only. Approximately 24% of diverted men were admitted for IP-MH compared to 19% of non-diverted men, and they waited fewer days before admission (114 days vs. 136 days). There was no statistically significant association of diversion with risk of IP-MH in the female sample, nor with risk of inpatient hospitalization for substance use disorder (IP-SA) treatment for either gender. Among both diverted men and women, IP-MH was more common than IP-SA. By contrast, among the non-diverted, IP-SA was more common than IP-MH. In survival models, diversion for neither the male nor female propensity-matched samples did not affected the risk of emergency room visits.

The diversion program encouraged use of outpatient treatment and crisis services. Eighty-eight percent of diverted women attended outpatient treatment for a mental health disorder (OP-MH) during the follow-up period (mean number of visits=36.20, SD=50.54). By contrast, 74% of non-diverted women used OP-MH in the follow-up period (mean visits=27.72, SD=57.58). Among men, 80% of those who were diverted attended OP-MH during the follow-up period (mean number of visits=45.97, SD=65.38); where 63% of non-diverted men did so (mean number of visits=31.11, SD=60.17). Fewer days elapsed before OP-MH for diverted individuals than for their non-diverted counterparts, with diverted women starting OP-MH after 44 days and diverted men after 37 days, on average. In survival models, diversion was associated with utilization of OP-MH for both propensity-matched women (HR=1.70, p<.0001) and men (HR=1.71, p<.0001).

Sensitivity analyses to exclude individuals who were out of the community showed consistent results, with a few exceptions. In the sensitivity analysis, diverted women had a 20% lower risk of arrest (H.R. 0.80, p<.05) compared to non-diverted women. Aside from this finding, results of the sensitivity analysis corroborated the primary analysis in statistical significance and direction of effect.

The pooled propensity models that directly tested for differences in program outcomes between men and women indicated that there were no statistically significant differences evident (results not shown).

3.2. Qualitative results

Twenty-one jail diversion clinicians participated in three focus groups, including 20 women and one man. Participants included 15 non-Hispanic white women, two Hispanic white women, two African American women, one Asian woman, and one non-Hispanic white male. We identified prominent thematic codes in bold text and grouped them into four broad themes: lack of inpatient/residential treatment options for women with CODs; the need for gender-specific services; women’s engagement and compliance challenges; and men’s engagement and compliance challenges.

3.2.1. Lack of inpatient and residential treatment options for women with CODs

The most widely endorsed theme across all three focus groups was the lack of inpatient and residential beds available to women with co-occurring mental health and substance use disorders. Fewer beds of all kinds were available for women (e.g., inpatient, detox, step-down, sober housing), causing difficulty for clinicians trying to secure treatment slots for a treatment plan that the court would approve, and thereby resulting in longer jail stays for these women waiting for a spot, assuming one could be found. “For a client with co-occurring issues that may need more intensive services, possibly even inpatient, there’s options for men, and there’s no options for women, in the state” (FG3). Participants in all three focus groups contrasted the lack of options for women to the availability of Sierra Center in New Haven, a supervised, residential option for men with CODs involved in the criminal justice system; “I don’t know any true female co-occurring programs” (FG1).

Even if the court did not require an intensive treatment program, it might require housing services to release certain defendants. For individuals who were “homeless or stepping down from a crisis … we have 10 beds and only two are dedicated to females, and we’re competing with so many others in the agencies for those two female slots. … And right, there the female has a disadvantage” (FG1). A lack of housing and childcare could threaten the success of outpatient treatment while diverted: “If they don’t have stable housing, the outpatient is lost. They’ll come, but I don’t know how much good it’s doing all the time if they have nowhere to go” (FG 2). There was consensus among respondents across all three focus groups that “there seems to be a lot more out there, to place men” (FG2). In one clinician’s words, finding a bed for a female client could be a “crap shoot” (FG3). Participants in all three focus groups observed that women with children were at a particular disadvantage in securing housing. There were fewer “women and family shelters” (FG1), few residential programs would allow women to bring their children (FG2), and only “one halfway house … [where] women can have kids re-united with them” (FG3).

3.2.2. Gender-specific services are needed to address everyone’s needs

Respondents from all three groups cited a need for more gender-specific services. Clinicians observed that “lumping men and women into the same types of groups” (FG3) meant that neither gender group’s needs were met; “they’re not very targeted to the unique issues that women face and the unique issues that men face” (FG3). Others expressed similar sentiments: “[T]here’s not enough services that are gender-focused, and even if they [women] are in a women’s program it’s usually just a house for women. And there might be, maybe, one trauma group? And … I don’t know how much training they’ve had …” (FG2). Clinicians cited problems with having men and women together, particularly in groups focused on trauma or anger management: “If you’re in a group, and you’re a woman who has experienced trauma … and you’re there with a bunch of men, nine times out of ten you’re not going to feel comfortable sharing. You can’t share, you just can’t” (FG2). Similarly, clinicians noted that the court frequently ordered anger management groups, but “the majority of people in it are men. And then you throw a couple of women in there with trauma issues, even though they’re the aggressor because they’re the one with the charges, they just aren’t successful at all. They last about two groups and then they’re out. And then they get in trouble because they’re not abiding by the court-ordered anger management” (FG2).

3.2.3. Engagement and compliance challenges: Women

In general, participants in all three groups described women’s needs as more difficult for clinicians to manage: “It’s definitely more challenging dealing with women than men” (FG3). Having dependent children made diversion cases far more complicated: “There’s just so many pieces to the puzzle that the court is not aware of” (FG3); “They accepted the diversion, O.K.; now there’s 800 million things you now have to think about and take care of” (FG3). Women with children had a harder time complying with court conditions due to a lack of childcare. For women with children, “it’s harder for them to keep up with compliance with appointments and stuff, more than males” (FG1), often because “they’re not able to bring their kids with, so they can’t come, either” (FG1). A lack of childcare also detracted from treatment when children were brought to appointments.

[W]e have one woman who has three kids that she, WHEN she can come, she brings the kids and it’s really hard to do … especially the trauma-related work that she is in desperate need of with three kiddos in the room. And we’re not allowed, at this facility, to do anything that looks like childcare, so you can’t have staff take the kids at least not you know, you could get caught doing that, so it’s a big challenge from an access standpoint. (FG2)

Male caregivers, while less common, were also affected: “We’ve had clients, both male and female, who have not been able to attend their treatment appointments due to a lack of daycare” (FG1). However, according to one clinician, male caregivers had an advantage because they were more likely to be assigned to evening programs when potential caregivers are more likely to be available.

Apart from a lack of access related to childcare, clinicians mentioned a variety of ways in which women were less engaged and compliant than men. Women with trauma (most women in diversion, according to participants) also found it difficult to trust providers (FG2), to engage with common substance use treatment approaches emphasizing powerlessness (FG2), and to feel comfortable opening up in mixed gender groups (FG3, as noted earlier). Similarly, another clinician said he had “problems engaging [women] into services”, and after describing female clients’ multiple needs—“children under DCF [Department of Children and Families] custody… homeless[ness], and no transportation, no telephone”—he noted that they may not be ready to engage in treatment (FG3). In addition, as noted above, some women with children might withhold information from treatment providers to avoid potential DCF involvement: “… some women might seem as if they’re holding stuff back because you do have to be careful about what you say, even in treatment, because as much as that is a therapeutic relationship, that therapist does have to report when they start to hear that there are children in the home and this is going on” (FG1). Clinicians offered another reason for female noncompliance was their lower motivation to succeed than men, given that women’s charges were typically less serious and so the consequences of returning to the prosecution pathway was less burdensome (FG1).

3.2.4. Engagement and compliance challenges: Men

Clinicians also discussed ways in which men were less engaged and compliant than women. One clinician noted that while women with children had trouble attending due to a lack of childcare, “sometimes we have males that just aren’t compliant at all … they’re just like ‘nope, I don’t want to do it’ and that’s it … even in terms of what I do in terms of benefits and stuff like that. Females are a little more on top of calling … they want to get connected to services” (FG1). Several respondents described men as less motivated than women; “they may not follow through as diligently with whatever the jail diversion recommendation is because they may feel that it doesn’t matter that much anyways” (FG1). Other participants described men as having a “greater sense of hopelessness” (FG1), being more “laid back” (FG1), being more apt to “throw in the towel” (FG1) and “let the dice fall where they may” (FG1). By contrast, women seemed to take diversion more seriously. Several clinicians attributed this greater seriousness to women’s sense of responsibility to succeed for their children and other family members, to become a better parent and, in some cases, to regain custody of their children. One clinician also said women applied themselves more diligently because they correctly understood that their treatment options were rare and therefore valuable; whereas men with more treatment options might “leave against medical advice because they feel like they can check themselves back in prior to another court date, and generally they’re right” (FG1).

Clinicians noted a variety of other ways in which men had more challenges with engagement and compliance. Although women had fewer housing options, they were also less likely to need them. Men were more likely to be homeless than women, and men in homeless shelters were “roaming around” (FG1) and therefore difficult for the diversion team to connect with. Precisely because women with children tended to have more complex needs, in some localities there were additional forms of support to help them succeed, such as specialized case management services (FG3) or shelters that allowed clients to remain during the day (FG1). Clinicians also noted that men were at higher risk of violating the court’s terms simply because they had more court conditions with which to comply—“urine screens, AIC [Alternative Incarceration Center], not to return to this address … need to call into the bail commissioner once a week … have electronic monitoring” (FG1).

Table 4 summarizes gender-related themes from the three focus groups, organized along the steps into and through the diversion program.

Table 4.

Most prominent^* gender-related themes at each stage of diversion pathway.

Stage of diversion pathway	Gender-related themes	N (focus groups)
Charges, arrest, and detention	Men are more likely to be arrested or in lock-up than women	3
	Men have more serious charges	3
Referral to diversion clinician for consideration	Women are more likely to self-refer and be referred to diversion	3
Client interaction with diversion clinician	Women and men discuss different topics (e.g., men avoid discussing trauma)	3
	Female clients may hide information	2
Client clinician’s decision to seek diversion	Lack of treatment resources for women may stop a clinician from recommending diversion	1
Features of the diversion plan itself (e.g., court requirements, treatment, and other services)	Court imposes more requirements on women than men	3
	Court imposes more requirements on men than women	1
	Fewer beds for women (IP, detox, step-down, sober housing)	3
	Fewer beds for women with children	3
	Men more likely to be homeless or lack basic needs	2
	Not enough gender-specific groups/treatment	2
	Women and men may receive different services/treatment	3
Client acceptance of diversion plan and diversion itself	Male pessimism about diversion and court	2
	Women more likely to want diversion	2
	Harder for women to accept diversion plan	2
	Women’s decision to accept diversion is affected by children and family	2
Client interaction with judge/court	Court more sympathetic to women, especially young women	3
	Children may impact court decision to divert women vs. hold them	3
	Physical characteristics of men and women affect court interaction with them	2
	Court interaction with women is no different than with men	3
	Judge’s gender may affect interaction with defendant	2
Securing treatment or other services after court has approved diversion plan	Women’s post-diversion needs are more challenging to meet than men’s	3
Client compliance, participation, and outcomes	Children and other relationships are a barrier to compliance, esp. for women	3
	Children/family may or may not be a motivator for women	3
	Women less likely to comply, participate, be open	3
	Men not as diligent, compliant, more helpless, demoralized, shut down	2
	Women and men are similar, can’t generalize about differences	2

^*Prominence: discussed in more than one focus group, or widely discussed in a single group and conceptually important. Participants were not unanimous; less prominent themes are not listed here and may be at odds with those in the Table.

4. Discussion

Jail diversion programs offer important opportunities to redirect individuals with behavioral health disorders from standard criminal prosecution and often incarceration, and instead into community-based substance use and mental health treatment, with individualized treatment plans designed to meet participants’ needed level of care. In Connecticut’s statewide diversion program, clinical staff from DMHAS are embedded in the community courts, working alongside the court team to identify and provide program services to participants. Diversion participants often have very complex treatment needs and histories, which courts must understand and attended to for programming to be successful. Gender is one dimension around which needs, histories, and resources vary fundamentally, potentially affecting program participation and chances for success. While we were unable to formally integrate the quantitative and qualitative data in this study, they do complement each other—with qualitative themes emerging that were consistent with and informed the quantitative outcomes—and offer insights into gender considerations in diversion programming.

In the administrative data analyses, among both men and women, diversion program participants had increased and earlier use of outpatient treatment than nonparticipants. While there was no difference between diverted and non-diverted groups in utilization of substance use services, it is important to note that treatment for CODs (e.g., integrated dual disorder treatment) in Connecticut is provided in mental health agencies, which is consistent with the increase in outpatient mental health treatment that we observed. As a complement to the utilization data, the qualitative data offer information about how the circumstances and motivators for treatment engagement differ between men and women. While outpatient service use increased for both diverted men and women, clinicians in the focus groups provided deeper insight into gender-specific barriers to treatment engagement that they have observed among their diversion clients, such as men being reluctant to discuss trauma histories, and for some women, responsibilities for dependent loved ones negatively affecting treatment adherence. Where the increase in outpatient service utilization associated with the program is beneficial, these more nuanced understandings of how those services are used, and what prevents optimal engagement, could help to inform programs’ gender-specific outreach to further improve program outcomes.

For men, diversion was associated with an increase in inpatient mental health care, which could indicate that the program provided them better access to the level of care they needed compared to men not in the program, who could continue to decompensate in the community or become incarcerated if they did not have sufficient access to psychiatric hospitalization when needed. That there was no difference in rate of psychiatric hospitalizations between women in the diversion and comparison groups could be attributed to there not being enough inpatient treatment resources for women in particular, especially those with CODs, and a lack of residential and inpatient resources that accommodate caregiving for dependent children. The scarcity of inpatient and residential resources for women was a prominent theme among clinicians in the focus groups; a diversion program cannot divert people to resources that are unavailable.

For both men and women, diversion was associated with a reduced risk for incarceration, but no difference in risk for new arrests. While reductions in both CJ outcomes would be desirable, reducing incarceration for this population is an important benefit in itself, especially considering this population is at high risk for recurring justice involvement. It may be that diversion program staff helped to divert new arrest charges to dismissal or alternatives to incarceration (e.g., community service) by working with prosecutors and judges and advocating for their program clients, thereby avoiding jail time associated with new, minor criminal offenses. In focus groups, diversion clinicians noted that it was hard to comply and succeed in diversion, with some saying it was harder for men and others saying it was harder for women. The similar chance of arrest among both genders may reflect a lack of resources to carry out the diversion program as intended.

While we observed variable effects of diversion on some outcomes for men and women, respectively, in the pooled models that allowed a direct comparison, there were no significant differences in outcomes by gender. It is possible that differential referral patterns that we were not able to observe in these data could influence who is selected into the program and affect outcomes. The qualitative data are helpful here, too, by revealing gender-specific circumstances and characteristics associated with clients’ experiences in the program that would otherwise be obscured in average-effect quantitative models and that could be used to improve programming.

The qualitative data from the three focus groups with diversion program clinicians provided additional insights into gender-specific contexts, experiences, and resources that helped to illuminate differences in programming for men and women, in a way that was complementary to the understanding of outcomes from the administrative data. A clear theme from diversion clinicians was that men and women face very different gender-related challenges along the diversion pathway, including who is referred to the program, access to services through diversion, intensity and circumstances around engagement in services, and interaction with court personnel.

The strong theme about inadequate services for women, especially inpatient and residential treatment, points to the lack of availability of these services (not enough beds) as well as inadequate accommodations in services that facilitate treatment engagement, such as residential programs where women can live with their children, childcare in outpatient treatment, and trauma-informed treatment. Furthermore, the lack of inpatient and residential treatment services for women creates an inequity, where one potential consequence is women being less likely to be diverted, according to one focus group, though our administrative data did not reflect this; 12% of women (553/4480) and 11% of men (1,140/10,060) were diverted. Focus group results suggest that at different points along the diversion pathway, there might be bias toward or away from men or women. This finding, taken together with the finding of a similar proportion diverted for each gender, suggests that the differences may cancel each other out, numerically speaking. Another potential inequity stemming from a lack of beds and also cited in focus groups is a higher likelihood of women spending time in jail while awaiting a treatment slot or not being offered diversion at all. While we were not able to confirm these potential inequities stemming from inadequate inpatient services for women in our administrative data, these focus group themes were consistent with the quantitative outcome that indicated diversion was not associated with increased use of inpatient treatment among women.

Focus group data also suggest that to optimize client engagement in the program and services, there are unique barriers for men and women that treatment needs to address. Men more often face practical barriers, like homelessness, that make consistent treatment engagement very challenging; and also attitudinal barriers that undermine engagement but that might be improved with interventions like motivational interviewing. For women, child custody loss can lead to hopelessness and despondency in some cases, while in others, hopes of reunification can be a strong motivator to fully engage in treatment. Trauma is a near universal experience among diverted women, and often complicates treatment engagement and success.

These analyses had some limitations that are important to consider in interpreting the results. Given our focus on community-based outcomes, the sample excluded people with arrests within the study window who were not released to the community, potentially due to more serious offenses. Assuming some of those people would have been eligible for diversion, that would make the comparison group systematically lower-risk by excluding people with more serious crimes, thereby suppressing the estimate of the diversion program’s true effect. It is also important to note that although we carried out multiple comparisons, we decided not to adjust the alpha value as we believe equal attention should be paid to protecting against Type II error. Given that the current study is exploratory, our main concern is to uncover promising evidence to study further. Also, the quantitative data do not allow for full triangulation with the qualitative data, given that the administrative data do not contain relevant contextual information about such concerns as homelessness, the presence of children, and whether the participant received gender-specific services. We did not have information about people who may have moved out of state or died during the study period. That could potentially bias effect estimates; however, we suspect that there were not enough of those circumstances to affect our results. Additionally, although the comparison group was well-matched through propensity scoring, it is possible that there were some unmeasured, but important, differences between the study groups that influenced both their chances of being referred for diversion, or willingness to accept diversion among those who are referred, and our outcomes of interest. Furthermore, we were unable to include a measure for co-occurring PTSD diagnosis in our analyses because it was inconsistently recorded in these data, which precluded us from making direct associations with PTSD and study outcomes. Although these data are several years old, many circumstances related to services and programming at the mental health–criminal justice interface, including in CT’s diversion program have remained steady, and many of the key challenges persist. Focus groups represented three state regions and 70% of the diversion clinicians we recruited, and where one-on-one interviews may have been less prone to group influence, participants shared candidly about their different experiences and client base. Last, direct clinical measures or self-reported data would have provided an informative complement to the administrative records, but were unavailable and beyond the scope of this study.

This mixed-methods study demonstrated that jail diversion programs can successfully engage clients in treatment and reduce their risk for incarceration. It also revealed that there are important gender-specific needs and differences in circumstances during program participation—including, insufficient programming and treatment resources for women—that must be addressed to optimize program outcomes.

Highlights.

• For both men and women, jail diversion was associated with reductions in risk for incarceration and increases in utilization of outpatient treatment services

• Jail diversion was associated with higher utilization of inpatient mental health care for men only

• Jail diversion clinicians highlighted the existence of too few inpatient and residential resources for women with co-occurring disorders, and a need for more gender-specific services