Does gender-specific substance abuse treatment for women promote continuity of care?

Journal of Substance Abuse Treatment 32 (2007) 27 – 39

Regular article

Does gender-specific substance abuse treatment for women promote continuity of care? Ronald E. Claus, (Ph.D.)a,4, Robert G. Orwin, (Ph.D.)b, Wendy Kissin, (Ph.D.)b, Antoinette Krupski, (Ph.D.)c, Kevin Campbell, (Dr.P.H.)c, Ken Stark, (M.B.A, M.Ed)c a

Missouri Institute of Mental Health, University of Missouri School of Medicine, St. Louis, MO 63139, USA b Substance Abuse Research Group, Westat, Rockville, MD 20850, USA c Division of Alcohol and Substance Abuse, Washington State Department of Social and Health Services, Olympia, WA 98504, USA Received 28 February 2006; received in revised form 6 June 2006; accepted 15 June 2006

Abstract Research has stressed the value of providing specialized services to women and suggests the importance of treatment duration. This quasiexperimental retrospective study reports on the continuity of care for women with children who were admitted to long-term residential substance abuse treatment. Women were admitted to 7 agencies offering specialized, women’s only treatment (SP, n = 747) or to 9 agencies that provided standard mixed-gender treatment (ST, n = 823). Client and treatment data were gathered from administrative sources. We hypothesized that women in specialized treatment would demonstrate higher continuing care rates after controlling for treatment completion and length of stay. Women in SP programs (37%) were more likely than those in ST programs (14%) to continue care. Multivariate analyses revealed that SP clients who completed treatment with longer stays were most likely to continue care. The findings show that specialized treatment for women promotes continuing care and demonstrate the importance of treatment completion. D 2007 Elsevier Inc. All rights reserved. Keywords: Gender-specific treatment; Women; Substance abuse; Continuing care; Treatment completion

1. Introduction In the past two decades, gender-specific services emerged in response to the multidimensional profile of problems that women display upon admission to substance abuse treatment. The emergence of women-only treatment programs also reflects the recognition that traditional mixed-sex programs often fail to address women’s needs, and programs designed specifically for parenting women have been developed to address the additional needs faced by mothers and their children. Treatment models for women have become more prevalent within the field, but limited research

Portions of this research were presented at the annual meeting of the American Public Health Association. 4 Corresponding author. Tel.: +1 314 877 6470; fax: +1 314 877 6477. E-mail address: [email protected] (R.E. Claus). 0740-5472/07/$ – see front matter D 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jsat.2006.06.013

has been conducted to examine whether women receiving treatment from women-only programs differ in their characteristics and treatment outcomes from women receiving treatment from mixed-sex programs (Grella, 1999). There is mounting evidence that women admitted to women-only programs have better retention and better outcomes relative to traditional mixed-gender programs. However, most women in the United States are treated in nonspecialized mixed-gender settings, and little empirical research has measured the degree to which gender-specific programming is related to treatment outcomes. Although there are a handful of small-scale studies that looked at policy-relevant client outcomes—and some larger-scale studies that looked at program costs or retention—there have been no large-scale comparative studies on whether specialized programs that address parenting women’s needs lead to better outcomes compared to more traditional approaches; on program characteristics that are associated

28

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

with successful outcomes; and on whether these outcomes offset the costs of providing specialized services. 1.1. Gender-specific treatment Although much of the research on gender-sensitive treatment has examined services intended to meet women’s distinctive treatment needs, theoretical models that focus on addiction and recovery for women have also emerged in recent years (Velasquez & Stotts, 2003). Recent gendersensitive theoretical models view substance abuse in the context of women’s relationships, including broader relational and multigenerational systems. Women’s addiction has been described as more bsocially embeddedQ than men’s (Saunders, Baily, Phillips, & Allsop, 1993). Similarly, qualitative studies concerned with women’s recovery hinge on the repair of relationships with children and other family members and on the development of support systems to prevent relapse (Kearney, 1998; Lewis, 2004). Many women entering treatment have experienced violence, including child abuse, battering, or rape, which affects subsequent connections with others. These relationship bdisconnectionsQ may be associated with addiction, although, conversely, treatment services that build bconnectionsQ may be associated with women’s recovery (Comfort & Kaltenbach, 2000; Finkelstein, 1994). Because women entering treatment report social isolation, are more likely to have partners who are involved in drugs or alcohol, and have fewer friends than their male counterparts, family therapy approaches have been utilized (McComish, Greenberg, Ager, Chruscial, & Laken, 2000; McComish et al., 2003). Finally, the use of a relational model is consistent with a recent call to examine treatment outcomes as related to gender role and culture, rather than as related to gender per se (Hodgson & John, 2004). For parenting women and their children, several issues suggest a need for deviation from traditional models. Some traditional program models (e.g., therapeutic communities) tend to use a confrontational style that does not work with most women (Kauffman, Dore, & NelsonZlupko, 1995). Furthermore, women may benefit from a style of treatment that is less structured and less rigid (Hodgins, el-Guebaly, & Addington, 1997). Programs have historically been predominantly for men, and bmale cultural normsQ have dominated (Hodgins et al., 1997; Saunders et al., 1993). The recognition of interpersonal group dynamic communication is more important in the treatment of women. For example, women tend to be more expressive verbally and behaviorally in single-sex group sessions, and often yield to both women and men when interrupted in mixed-gender settings, whereas men tend to only yield to interruptions from other men (Hodgins et al., 1997). For women in such groups, issues left unaddressed can result in adverse psychological effects (Copeland & Hall, 1992). Women stress that a treatment environment that is safe for themselves and their children promotes

therapeutic effects (Lewis, 2004). Other environmental characteristics, such as comfort, size, privacy, location, and attractiveness, have a small but demonstrable influence on women’s engagement in treatment (Grosenick & Hatmaker, 2000). The ability of a substance-abusing mother to be accompanied by her child while on treatment is characteristic of specialized treatment. Although some have argued that children in a treatment facility distract a mother’s ability to bwork her programQ (thus delaying or adversely affecting her recovery), available evidence suggests that women who are allowed this provision demonstrate higher rates of retention (Chen et al., 2004; Coletti et al., 1992; Hughes et al., 1995). Treatment for women and their children necessitates an emphasis on a bfamily context,Q which provides for an enhanced interaction between mother and child and an enhanced quality of family/domestic environment (Washington State Department of Social and Health Services, 1999). Trauma histories are common among female substance abusers and must be addressed appropriately in treatment (Orwin, Maranda, & Brady, 2001). Other general recommendations for achieving successful outcomes among women, especially those with children, include a continuum of coordinated and family-focused services and interventions guided by female-specific substance abuse treatment models (McKay, Gutman, McLellan, Lynch, & Ketterlinus, 2003; Washington State Department of Social and Health Services, 1999). Importantly, the delivery of childcare and prenatal care, a focus on women’s topics, comprehensive programming, and the utilization of supplemental services have been found to differentiate specialized and traditional treatments and are positively associated with treatment completion, length of stay (LOS), and improved treatment outcomes (Ashley, Marsden, & Brady, 2003). The complex patterns and intricate interdependence of women’s substance abuse problems and outcomes support the need to identify gender-sensitive factors to address these issues (Green, Polen, Lynch, Dickinson, & Bennett, 2004). Relative to traditional mixed-gender programs, evidence exists that women in women-only programs have better retention (Anglin, Hser, & Grella, 1997; Washington State DSHS, 1999) and better treatment outcomes (Orwin, Francisco, & Bernichon, 2001; Orwin, Kissin, & Dugan, 2003). The vast majority of women, however, are served in mixed-gender programs (Grella & Greenwell, 2004). The growth of women-only programs in the early to mid 1990s—in part triggered by the bcrack babiesQ scare of the mid 1980s—peaked and was actually in decline by the end of the century (Grella, 1999; Grella & Greenwell, 2004). 1.2. Continuity of care In recent years, the need to provide a continuum of care to individuals with substance abuse problems has received increasing emphasis. By transferring clients to less restric-

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

tive levels of care when sufficient progress has been made, providers seek to increase the duration of a treatment episode while limiting the amount of time clients spend in relatively intensive and expensive programs (McKay et al., 1998). The move from a residential program to an outpatient aftercare program is a key transition point in a treatment episode. Continuing care in an outpatient modality is intended to maintain the gains made in residential treatment, to provide support for participation in self-help groups and other activities that promote recovery, and to prevent relapse to substance use. Many clients fail to make this transition, and continuing care rates as low as 20% are common (Donovan, 1998). The reasons given for high dropout rate include the following: lack of motivation for further treatment, competing responsibilities with regard to family or employment, physical or psychiatric problems that may hinder participation, difficulty in travel to the continuing care site, and relapse to substance use (Donovan, 1998; McKay, Lynch, et al., 2004; Schmitt, Phibbs, & Piette, 2003). The challenges related to this key transition have sparked interest in the development of a variety of interventions to improve compliance with continuing care recommendations (Chutuape, Katz, & Stitzer, 2001; McKay, 2001). The role of continuing care in publicly funded community-based treatment programs has not been well studied. Correlational studies that examine the association between continuing care and substance use outcomes have consistently produced positive findings (Donovan, 1998; McCollister et al., 2003; McKay, 2001). These studies show that continuing care clients are less likely to be readmitted to treatment, and have better substance use and criminal justice outcomes. A review of controlled studies (McKay, 2001) provides weaker evidence that supports the effectiveness of some, but not all, continuing care interventions. The likelihood that clients enter continuing care appears linked to both individual and program characteristics, although, to date, most research studies on the transition from residential treatment to continuing care have focused on individual characteristics that put clients at risk for dropout. Individuals who are older or married are more likely to attend, whereas those with greater addiction or psychiatric severity may be less likely to attend (Donovan, 1998; Schmitt et al., 2003), but these client-level predictors are not always consistent (McKay et al., 1998). A recent study (McKay, Foltz, et al., 2004) found that a shorter index LOS in intensive outpatient treatment was related to higher continuing care participation, but it reported no such association for clients who began treatment as inpatients. We are aware of only one previous study that examined the influence of gender-specific residential treatment on continuing care achievement. Stevens and Patton (1998) reported that women who had their children with them in a modified therapeutic community were more likely to attend continuing care groups than women whose children

29

were not with them. In this small-sample study, women were randomly assigned to have or to not have their children with them; however, the authors reported univariate analyses that did not control for between-condition differences in LOS. 1.3. Length of stay Research indicates that increased LOS in substance abuse treatment generally leads to a broad range of improved outcomes, including increased abstinence from alcohol and drugs, gains in employment, and decreases in criminal behavior (Hubbard, Craddock, Flynn, Anderson, & Etheridge, 1997; Luchansky, Brown, Longhi, Stark, & Krupski, 2000). As well, clients with shorter LOS are at increased risk for poor outcomes, such as readmission to treatment (Luchansky, He, Krupski, & Stark, 2000; Moos, Brennan, & Mertens, 1994; Moos, Petit, & Gruber, 1995). A reduced likelihood of adverse birth outcomes for pregnant women receiving residential care has been observed for those who stay longer in treatment (Clark, 2001). For these reasons, identifying, understanding, and addressing the factors that affect treatment retention are critical to improving the delivery and outcomes of substance abuse treatment. Retention has been associated with demographic and predisposing factors, including severity of substance use and mental health problems, and presenting substance use problem (Ashley, Sverdlov, & Brady, 2004). Gender has a complex relation with treatment retention that is affected by personal, social, and treatment factors (Stark, 1992). Mertens and Weisner (2000) found that retention was positively associated with substance abuse severity for both men and women, but other predictors differed by gender. Likewise, Green, Polen, Dickinson, Lunch, and Bennett (2002) reported that the factors predicting LOS differed markedly for men and women. Although extant research identifies a variety of client characteristics related to treatment retention, many of these studies were conducted with small samples of women. That treatment duration is related to posttreatment outcomes seems clear, although researchers have debated the amount of treatment necessary in a particular modality to realize improvements. Although some studies have suggested that the association of LOS and drug-use outcomes is linear (Simpson, Savage, & Lloyd, 1979), other studies have argued that clients must stay in treatment past a particular retention bthresholdQ before favorable outcomes are observed (Simpson, Joe, & Brown, 1997). For treatment modalities with long expected stays, however, an arbitrary cutpoint may be misleading. Zhang, Friedman, and Gerstein (2003) observed an inverted U-shaped relationship between LOS and drug-use outcomes among long-term residential treatment clients. Their results showed that clients made appreciable improvement through usual residential treatment durations, but also

30

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

indicated that unusually long retention was increasingly predictive of less improvement. An increased LOS for women is associated with delivery of women-only treatment (Ashley et al., 2004; Grella, 1999; Roberts & Nishimoto, 1996), which may facilitate longer stays by creating an environment that is more focused on women’s issues. In particular, programs that allow women to bring their children into residential treatment have reported increased LOS (Hughes et al., 1995; Wobie, Eyler, Conlon, Clarke, & Behnke, 1997). Childcare, case management, and referral services are additional program features intended to increase LOS (Ashley et al., 2004). Hser, Maglione, Joshi, and Chao (1998) found that matching clients to treatment based on gender-specific needs resulted in longer stays in treatment. For women, these needs included childcare services, women-only treatment, and prenatal care services. 1.4. Treatment completion Keeping clients in treatment long enough to complete appropriate treatment goals is a major challenge for substance abuse treatment programs. Lengthening an individual’s time in treatment does not necessarily produce better outcomes, as clients who stay in treatment for a sufficient time but fail to complete treatment appear different from those who are successfully discharged (Knight, Logan, & Simpson, 2001). Treatment dropout rates vary widely across modalities and programs. Outpatient programs usually report higher dropout rates, as proportions of more than 70% may be observed, but reports on inpatient dropout rates have ranged from 19% to 67% (Wickizer et al., 1994). Treatment completion, often confounded with LOS, has been linked to improved substance use and social outcomes (Greenfield et al., 2004). Previous research in Washington State (Luchansky, He, et al., 2000) has found that treatment completers were less likely to be readmitted during the following year. This study also found that men and women who received a combination of inpatient and outpatient treatments had a lower readmission risk and that, compared to men, women had an increased risk of readmission. Furthermore, completion has been associated with higher posttreatment earnings. A longitudinal study (Luchansky, Brown, et al., 2000) found that, controlling for wages earned before treatment, individuals who completed treatment earned more than others, and those who also received vocational services earned the most. Finally, improved drug outcomes are demonstrated for those who are motivated enough to complete treatment and also continue with aftercare (Ghodse et al., 2002). Treatment dropout among women is influenced by environmental barriers such as inadequate childcare and lack of transportation (Howell, Heiser, & Harrington, 1999). Structural barriers, such as the inability of mothers to bring their children into treatment with them, provide other obstacles to participation (Hughes et al., 1995; Metsch

et al., 2001; Szuster, Rich, Chung, & Bisconer, 1996). Conversely, the provision of gender-sensitive services such as parenting classes and vocational training can facilitate retention in residential settings (Chen et al., 2004; Howell et al., 1999). Program-level factors that decrease dropout include the provision of case management services, availability of vocational services, and engagement of clients in the development of treatment plans (Hser et al., 1998); a short waiting list for admission (Claus & Kindleberger, 2002); and high clinical staff:client ratios and small caseloads (Stark, 1992). 1.5. Study overview and hypotheses This article reports one of several studies based on an evaluation of the effectiveness of specialized treatment programs for parenting women in Washington State (see also Kissin, Orwin, Grella, & Garfield, 2003; Orwin, Kissin, Claus, Grella, & Williams, 2004; Orwin et al., 2003). The overall design is a retrospective quasi-experiment comparing outcomes for women in two long-term residential treatment modalities: specialized gender-specific (SP) and standard mixed-gender (ST). The present study addresses whether specialized treatment is more effective than standard treatment in linking women to outpatient aftercare upon exiting residential care (i.e., in achieving continuity of care). The hypotheses are as follows: Hypothesis 1. Women in the specialized modality will achieve significantly higher rates of continuity of care than women in the standard modality. Hypothesis 2. The effect of modality on continuity of care will persist even after controlling for differences between conditions in terms of: (a) client characteristics, (b) prior treatment histories, (c) program completion rates, and (d) LOS. Hypothesis 3. Higher completion rates and longer LOS will significantly increase the modality effect. The hypotheses, if confirmed, will support the generalized inference that specialized substance abuse treatment leads to higher rates of continuity of care for parenting women, with the highest rates achieved by women in specialized modality who complete and stay longer in treatment.

2. Materials and methods To be eligible for the study, a woman must: (1) have entered a publicly funded long-term residential substance abuse treatment in Washington State between January 1994 and mid-2000; (2) have a child less than 18 years old at the time of admission; and (3) not be pregnant (or immediately postpartum) on admission. Pregnant and postpartum women were excluded because so few entered ST treatment (2 pregnant and 16 postpartum women during the study

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

period); consequently, their inclusion would compromise a comparison of the two modalities. It should be noted that nearly two thirds (62%) of women referred to SP programs were also classified as parenting; thus, the study covers the majority of women admitted to specialized programs. To qualify for public funding, a woman had to be at or below 185% of the federal poverty level at the time of entry into treatment and had to be referred by an assessment center before placement (or assessed by the center within 72 hours of placement in treatment). These criteria yielded a sample of 747 women with children in the SP condition and 823 in the ST condition.1 2.1. Program descriptions In the early 1990s, Washington State legislatively mandated a specialized long-term residential treatment modality designed specifically for pregnant, postpartum, and/or parenting women (PPW) who require intensive longterm substance abuse treatment. It allowed children up to the age of 6 years to live with their mothers during the course of treatment and to be offered child placement, care and supervision, and required specialized services for women and children. Upon successful completion, clients were eligible to receive up to 18 months of transitional housing. Standard long-term residential programs included all other long-term residential programs serving publicly funded clients in the state; they admitted men and women without children, as well as PPW, and were not specifically designed to meet PPW needs. Prospective clients were assessed at a Division of Alcohol and Substance Abuse (DASA) Assessment Center. An extensive set of client characteristics was assessed upon entry to DASA, and women determined to meet eligibility requirements were referred to a program site. Client characteristics, including problem severity, as well as more idiosyncratic factors (e.g., a woman might insist on entering the program site nearest her family), influenced the assessment center’s decision. Primary data collection efforts included site visits, telephone interviews, or site visits and telephone interviews with key staff members from the largest programs (five SP and four ST) to collect qualitative program data. Key staff members included program directors, clinical directors, counselors, and, as appropriate, childcare staff. Differences were also reflected in the programs’ own admissions policies. Compared with ST program directors, SP program directors reported that they were less likely to exclude women who were minors, who received methadone maintenance, or who had an acute

1

During the study period, 13% of SP clients and 20% of ST clients reenrolled in the same treatment modality. However, bcrossoversQ (women treated in both modalities) were relatively rare (3% total across both groups).

31

psychiatric condition. SP program directors also described a treatment philosophy that was more women-oriented, less confrontational, and more lenient regarding relapse. The cost of SP treatment to the state is approximately twice that of ST treatment. During the enrollment period for the study (1994–2000), the SP cost was US$100.56 per day compared to US$49.42 per day for ST. Study clients were distributed across seven SP and nine ST programs, and included small and large providers in both conditions (from 1.0 to 51.7 mean annual admissions). The median number of women who were served by SP programs was 71 (range, 5–315), whereas that for ST programs was 44 (range, 2–362). Mixed-gender ST agencies served primarily men (67.8%; range, 51.5–81.9%), although one smaller ST agency exclusively served women. SP programs generally had higher staff:client ratios (1:7–9) than did ST programs (1:8–15). 2.1.1. Core services All long-term residential treatment programs (SP and ST) offer a combination of chemical dependency education, treatment, and other services in a residential setting. The duration of treatment varies on an individual basis and differs by agency. By state regulation, each residential service provider is expected to ensure that each client receives: (1) education regarding alcohol, other drugs, and other addictions for at least 2 hours/week; (2) individual and group counseling by a chemical dependency professional for a minimum of 2 hours/week; (3) education on social and coping skills; (4) social and recreational activities; (5) assistance in seeking employment, when appropriate; (6) patient record review and update, at least monthly; (7) assistance with reentry living skills; and (8) a living arrangement plan. 2.1.2. Specialized services for women SP programs are also required to address chemical dependency issues specific to women and their children. This includes the provision of (1) same-gender clinical (individual and group) services; (2) chemical dependency education tailored to women (e.g., codependency issues); (3) case management and social services, including interfacing with the criminal and juvenile justice system and child welfare services, as well as specialized counseling, as needed, for eating disorders, sexual assault, domestic violence, childhood abuse, and family dysfunction (these may be referred to outside providers); (4) medical/health services, including at least 10 hours of education regarding child and adult nutrition, pregnancy, labor and delivery, lactation, HIV/AIDS, birth control, exercise, smoking, sexually transmitted disease, and fetal alcohol syndrome/ effect; (5) mental health services, as needed; and (6) reentry preparation (general equivalency diploma preparation and job readiness); as well as other women-centered components (e.g., food service that meets or exceeds the nutritional requirements for pregnant and lactating women and children, as defined by the Department of Health).

32

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

2.1.3. Specialized services for children SP programs also provide therapeutic childcare services for the health and welfare of the children accompanying parents who participate in residential substance abuse programs. Providers are expected to deliver services for the care, protection, and treatment of children at risk for abuse, neglect, and eventual substance abuse. Services include the following elements: (1) developmental assessment using recognized standardized instruments; (2) play therapy; (3) behavioral modification; (4) individual counseling; (5) self-esteem building; and (6) family intervention to modify parenting behavior and/or the child’s environment to eliminate or prevent dysfunctional behavior in children. Finally, SP program graduates are eligible to receive up to 18 months of transitional housing services for themselves and their children in drug-free and alcohol-free residences. 2.2. Data source DASA’s management information system, the Treatment and Assessment Report Generation Tool (TARGET), provided records of each admission, discharge, and referral at each agency and modality for all clients who received publicly funded substance abuse services in Washington State. TARGET also yielded a broad variety of demographic and descriptive data collected at treatment entry for each client. Treatment completion was determined from discharge codes, and LOS was calculated from admission and discharge dates indicated in TARGET. Project procedures, including the use of TARGET data, were approved by the Institutional Review Board of the University of Missouri and of Westat, and the Human Subjects Review Board at the Washington State Department of Social and Health Services. 2.3. Definition of continuity of care Clients were considered to have achieved continuity of care if they enrolled in a less intense treatment modality within 30 days of discharge from the index episode of residential treatment. The 30-day cutoff was selected based on an examination of admission and discharge records and has been previously used by DASA (Luchansky, He, et al., 2000) and others (McLellan, Weinstein, Shen, Kendig, & Levine, 2005; TOPPS-II Interstate Cooperative Study Group, 2003).2 This requirement was used to ensure that a continuing care episode reflected a planned and intended transfer rather than the onset of a new episode of care. Overall, about one fourth of the sample (25.4%; n = 398) achieved continuity of care. Most continuing care clients

2 Other studies have used a 14-day cutoff (e.g., McKay, Foltz, et al., 2004); however, sensitivity analyses have yielded only small differences in classification rates for cutoff intervals between 7 and 30 days (Orwin, personal communication). This is because most clients who transition do so within 1 week of residential discharge, often on the same day.

(n = 376; 95%) entered outpatient care, whereas a few (n = 22; 5%) entered another less intense treatment modality (mainly a recovery house). 2.4. Analytic approach 2.4.1. Addressing group nonequivalence The core of research design is a quasi-experimental comparison of specialized treatment and mixed-gender long-term residential treatment on outcomes for women. The referral of women to the two conditions was nonrandom, as is typical of large health service studies that are based on the observation of systems as they operate in normal practice. Consequently, at the analysis stage, steps were necessary to minimize the degree that the pretreatment nonequivalence could bias effect estimates. Propensity scoring (Rosenbaum & Rubin, 1983) was chosen for this purpose. Although not addressing every concern, propensity scoring methods bring tangible improvements on earlier methods for drawing valid effect inferences from nonexperimental comparisons of alternate treatments. Propensity scores are often estimated by using a logistic model to predict treatment group membership from confounding covariates, which outputs the conditional probability of being in a particular group (in our case, SP vs. ST) given an individual’s values on a set of observed covariates. When subsequently included in a regression model, the propensity score carries all the information from the complex covariate model in a single variable, consuming only 1 df. Most important, propensity scoring facilitates a direct examination of the extent to which confounds have been removed, through tests of balance. Simulations, studies of actual data, and formal proofs have shown that subclassification of the propensity score into about five strata or bquintilesQ is generally sufficient to assess the quality of adjustment for all covariates that went into its estimation, no matter how many there are (Rubin, 1997). A model is considered balanced when the distribution of predictor variables across groups is similar within each quintile and when the total number of statistically significant differences observed is roughly comparable to what would be expected by chance under random assignment. In this study, potential predictors of group membership representing demographic, addiction, mental health, and other descriptive variables were taken from TARGET baseline interviews. A logistic regression model was fitted to estimate the probability of being in the specialized treatment condition as a function of 29 baseline variables. Within each propensity score quintile, differences between conditions on the 29 main effect variables were examined using one-way analysis of variance or chi-square tests to assess balance. The distribution of cases in outer quintiles was highly skewed, substantially reducing the beffectiveQ sample size (the harmonic mean) per condition. Because this limited the available statistical power to assess balance, a small effect size difference, rather than a significance level

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

difference, was used as a standard. Results from the initial propensity model revealed that 9.7% of contrasts (14 of 145) had greater than small effect size differences between treatment conditions, mostly concentrated in outer quintiles. Interaction terms were added to the model, selected by testing all possible interactions where any main effect was out of balance in any quintile. In the revised model, only 6.2% of contrasts (9 of 145) had greater than small effect size differences between conditions (i.e., the model was essentially in balance). 2.4.2. Effect estimation Hierarchical logistic regression was used to analyze the association between treatment condition and receipt of continuing care while controlling for group nonequivalence with propensity scores. First, we entered control variables, which included the propensity score and variables that represented substance abuse treatment received prior to the index admission. Next, treatment group (SP vs. ST) was entered as a main effect. To control for confounding factors related to treatment group, we also entered LOS and treatment completion as main effects. Because the relationship between treatment duration and posttreatment improvements after residential treatment may be nonlinear (Zhang et al., 2003), a quadratic LOS term was used. Finally, twoway and three-way interactions between treatment modality, treatment completion, and LOS were entered into the model.

3. Results Bivariate differences between treatment modalities in continuing care rates and other variables are described first, followed by multivariate analyses that predict receipt of continuing care. 3.1. Bivariate associations 3.1.1. Client characteristics Similarities and differences between women at SP and ST programs are presented in Table 1. Women admitted to SP programs were younger than those admitted to ST programs (29 vs. 34 years old). They were less likely to be white (68% vs. 74%), but they somewhat more likely identified themselves as Native American (11.1% vs. 7.7%) or another racial/ethnic group (6.3% vs. 3.0%). African Americans comprised 15% of both groups. Women in SP programs were both more likely to have never married (52% vs. 34%) and less likely to be divorced (20% vs. 34%). Women who were married, widowed, or separated comprised about 30% of both groups. In addition, women in SP programs were somewhat less likely to have obtained a high school diploma (45% vs. 53%) and were less likely to have been involved in the criminal justice system (50% vs. 64%). Furthermore, more SP clients had children at home (53% vs. 21%), although fewer had children else-

33

Table 1 Client characteristics for women in specialized and standard treatment modalities Characteristics Age (years) [M (SD)] Race (%) White Black Native American Other Marital status (%) Married Divorced Never married Separated Widowed High school diploma (%) Criminal justice involvement (%) Children at home (%) Children elsewhere (%) Living alone (%) Public assistance (%) Monthly household income (US$) [M (SD)] Drug choice (%) Heroin Alcohol Cocaine Marijuana Methamphetamine Previous mental health treatment (%) Taking psychiatric medications (%) Current domestic violence (%) Past domestic violence (%)

Specialized

Standard

t or m2

29.4 (6.4)

34.2 (7.7)

13.5444

67.9 14.6 11.1 6.3

74.4 15.0 7.7 3.0

13.144

8.4 19.8 51.7 17.9 2.1 44.7 49.9

11.6 34.4 34.5 15.5 4.0 52.9 63.7

70.7444

53.3 71.1 16.7 72.3 358 (494)

20.5 86.2 49.6 58.8 234 (413)

10.744 25.9444 182.8444 53.9444 189.7444 31.5444 5.5444

20.5 68.4 56.2 53.4 42.7 26.1

25.6 76.1 59.3 36.7 29.7 45.6

5.74 11.7444 1.5 44.2444 31.7444 64.5444

14.9

28.1

40.4444

5.8 75.4

10.2 76.4

9.344 0.165

4 p b .05. 44 p b .01. 444 p b .001.

where (71% vs. 86%). Consistent with their family status, women in SP programs were more likely to have received public assistance (72% vs. 59%), and they reported higher household incomes (on average, US$358 vs. US$234 per month). Relatively small differences in alcohol and drug use were reported. Overall, two thirds (66.6%) used both alcohol and drugs, about one quarter (27.6%) used only drugs, and few (5.8%) used only alcohol. Among primary, secondary, or tertiary drugs, women admitted to SP programs were more likely than those in ST treatment to identify methamphetamine or marijuana use; women admitted to ST programs were somewhat more likely than those in SP programs to report alcohol or heroin use. Although the rates of past domestic violence were high among both groups, utilization of mental health treatment and current use of psychiatric medication were higher for women entering ST programs. 3.1.2. Preindex treatment utilization Treatment admission, utilization, and discharge data from the 2-year period before the index residential episode were

34

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

examined. Preliminary inspection showed that treatment utilization during this time was generally low. During any particular month of the 2-year pretreatment window, 10 –15% of the sample was enrolled in treatment, and the groups averaged fewer than 5 days of care per month. During this window, 66.1% of SP clients and 57.1% of ST clients received some preindex treatment, v 2(1) = 13.52, p b .05, / = .09. Among those with an admission, ST clients averaged slightly more total months with a treatment day than SP clients (6.1 vs. 5.2; t = 2.72, p b .05, d = 0.17). During the 6-month period immediately preceding their treatment entry, ST clients were more likely than SP clients to be admitted for detoxification (42.9% vs. 20.3%), v 2(1) = 91.42, p b .05, / = .24, but SP clients were more likely to receive outpatient care (50.9% vs. 40.6%), v 2(1) = 16.66, p b .05, / = .10. About one third of each group (overall, 33.0%) received residential treatment during this time.

Assessment Center, or other services. Due to the practical circumstances surrounding unplanned discharges, referrals were not routinely provided to women who did not complete treatment. Among treatment completers, women in SP agencies (n = 336) received, on average, 2.6 (SD = 1.5) referrals, whereas women in ST agencies (n = 440) received 1.7 (SD = 0.8) referrals—a large effect size difference (t = 11.16, p b .05, d = 0.78). That SP agencies provided more referrals than ST agencies represents a programmatic difference between treatment conditions consistent with the provision of enhanced SP services described earlier (e.g., case management and social services, including many involving referrals to outside providers). Furthermore, this increased linkage with resources is consistent with the increased supplemental services often associated with specialized treatment programs elsewhere (Ashley et al., 2003).

3.1.3. LOS in index treatment On average, women who attended specialized programs stayed in treatment longer than those who attended traditional programs (104.5 vs. 74.4 days; t = 8.36, p b .05, d = 0.42). The amount of treatment women received varied broadly for both specialized (range, 1–367; SD = 67) and traditional (range, 1– 400; SD = 80) groups. Site differences were also observed. Mean LOS varied widely across both traditional (range, 34 –95 days) and specialized (range, 77–118 days) programs. This variance was, in part, related to the different expectations that programs had for clients. At the program level, the expected LOS ranged from 5.5 to 12 months in specialized programs (n = 7, Mdn = 6 months) and from 2 to 12 months in traditional programs (n = 9, Mdn = 6 months).

3.1.6. Continuity of care As noted previously, women who enrolled in outpatient treatment (or in another step-down treatment modality) within 30 days of discharge from long-term residential treatment were considered to have achieved continuity of care. Overall, about one fourth of the sample (25.4%; n = 398) achieved continuity of care after their residential stay. Notably, SP clients (37.2%; n = 278) were more likely than ST clients (14.2%; n = 120) to continue care, v 2(1) = 106.6, p b .05, / = .26. Because treatment completion rates differed across groups, we also tested its association with continuity of care. As expected, women who completed residential treatment were more likely than noncompleters (40.6% vs. 10.5%), v 2(1) = 189.1, p b .05, / = .35, to go on to outpatient aftercare. However, SP treatment completers were far more likely to continue with treatment (65.5%) than either women who completed ST programs (23.9%) or women who did not complete either SP (14.6%) or ST care (6.3%).

3.1.4. Treatment completion Overall, about half (49.4%; n = 776) of the sample completed residential treatment. Women in specialized programs were somewhat less likely to complete treatment (45% vs. 53.3%), v 2(1) = 10.94, p b .05, / = .08. Substantial within-group differences were again noted, as mean program completion rates ranged widely for both specialized (22.2–80.0%) and traditional (13.9–78.7%) agencies. Particular program practices and expectations appeared to influence these findings. For instance, one program, which had the shortest expected LOS (60 days), admitted predominantly women who were mandated to treatment; among traditional programs, this agency had the highest completion rate (78.7%). Excluding this outlier program, the median programs in both groups had similar completion rates (both 45%). 3.1.5. Posttreatment referrals At discharge, agencies made referrals to another health provider and to self-help groups, and referrals for medical or dental services, mental health services, vocational rehabilitation or job placement, further assessment at the ADATSA

3.2. Multivariate prediction of continuity of care To identify the unique effects of specialized treatment on the achievement of continuing care, we conducted a hierarchical logistic regression. In the first step (details not shown), we controlled for sample differences using propensity scores and also controlled for the modality and amount of substance abuse treatment received in the 2 years prior to the index treatment. In the second step, the main effects for treatment condition and treatment completion were entered, along with linear and quadratic terms for LOS. In the final two steps, two-way and three-way interaction terms were entered. Over and above the main effects, a significant three-way interaction between condition, completion, and LOS was observed (see Table 2). To help interpret this interaction, LOS was characterized via a median split as high or low, and the proportion of women who achieved continuity of

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

35

Table 2 Logistic regression predicting continued care achievement by women in SP and ST residential modalities

Table 3 Summary of hierarchical logistic regression predicting continued care achievement

Variables

Step

Variables

R2

DR 2

Dm2

1 2 3 4

Preindex covariates Main effects Two-way interactions Three-way interaction

.074 .333 .349 .354

.074 .259 .016 .005

80.27444 321.94444 21.86444 5.914

Main effects Treatment condition (SP vs. ST) Treatment completion LOS (days) LOS squared Two-way interactions Condition  Completion LOS  Completion LOS  Condition Three-way interaction Condition  LOS  Completion Constant

B

SE

OR

Wald

0.682

0.310

1.98

0.997 0.008 0.0001

0.253 0.004 0.00002

2.71 1.008 0.9999

0.565 0.008 0.0005

0.396 0.005 0.005

1.76 0.99 1.00

2.036 2.838 0.01

0.015

0.006

1.015

5.7424

2.231

0.261

4.8514 15.48444 4.3024 28.743444

73.03444

Note. Preindex covariates are not shown. 4 p b .05. 444 p b .001.

care was calculated for possible variable combinations. With this approach (see Fig. 1), we observed that treatment completers with long LOS in specialized treatment (67%) were most likely from among all other groups (6–29%) to receive continuing care. In addition, we found significant main effects for specialized treatment (odds ratio [OR] = 2.0) and treatment completion (OR = 2.7). We also found that, controlling for other effects, the association between LOS and continuing care was nonlinear. Significant linear and quadratic LOS terms indicate an inverted U-shaped

Fig. 1. Proportion of women achieving continuing care as a function of treatment condition, LOS, and completion status.

4 p b .05. 444 p b .001.

relationship, whereby women with unusually long LOS showed a decreasing likelihood of receiving continuing care. Overall, regression accounted for 35.4% of the variance in continuing care (Table 3). Together, treatment condition, completion status, and LOS (the main effects) accounted for approximately 25.9% of outcome variance. The propensity score and previous treatment variables that were used to control for pretreatment differences in the sample accounted for 7% of continuing care variance.

4. Discussion We hypothesized that women in specialized residential treatment programs would be more likely than women in ST residential programs to achieve continuity of care through outpatient aftercare treatment after exiting from their residential stay. This prediction was confirmed. Overall, one fourth of the sample (25.4%) received continuing care, with more than twice as many women in specialized treatment (versus women in standard programs) attending outpatient aftercare (37% vs. 14%). This finding is consistent with a previous study (Stevens & Patton, 1998), which found higher aftercare attendance for women who attended residential treatment with their children. That specialized treatment was associated with this enhanced proximal outcome is an important finding, given growing evidence that continuity of care leads to greater distal improvements in substance abuse, employment, and criminal justice domains. The effect of specialized treatment persisted when it was tested with multivariate models that simultaneously control for group nonequivalence and factors that influence continuity of care, namely, treatment completion and LOS. In particular, the three-way interaction we observed indicated that the effect of specialized care was augmented by treatment completion and a longer LOS. Women-only programs have been found to promote longer LOS relative to mixed-gender programs (Anglin et al., 1997; Ashley et al., 2003). This research is consistent with this study, where women in specialized programs had longer mean stays (104 vs. 74 days). Despite this disparity in average retention, treatment completion rates were similar for the two modalities. This suggests that specialized programs were effective in meeting women’s needs to successfully

36

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

keep them engaged in treatment. Even when completion and LOS were controlled, however, we observed that specialized treatment had a unique effect on continuing care. 4.1. Examination of alternative explanations We took additional steps to confirm that the relationship between specialized treatment and continuity of care was not a sample artifact. Because clients were not randomly assigned to treatment, we need to consider that causes other than treatment (most notably self-selection) may have been responsible for the observed group differences on outcomes. Conditional on a propensity score constructed from a suitably rich set of covariates and pretreatment variables, however, it may be reasonable to rule out most other potential causes of outcomes (Hirano & Imbens, 2001). By controlling for overt biases in assignment, a propensity score allows for an analytical adjustment that attempts to undo the problem created by an unequal opportunity to receive treatment. Although other unmeasured factors could, in principle, explain some of the associations observed between specialized treatment and continuing care, the magnitude of such a hidden bias would have to be large. As noted previously, the observed and planned LOS varied widely both within and between treatment conditions. In particular, one agency offered a relatively brief regimen (2 months) in a locked unit, whereas several other agencies planned to treat women for up to 1 year. To account for these differences, we calculated the ratio between the observed LOS for each woman and the planned LOS at the agency where she was treated. This LOS ratio was then used to determine whether the apparent effects of specialized treatment might have actually been caused by different program expectations for LOS. In fact, when the observedto-expected LOS variable was substituted into the original multivariate model (results not shown), we observed the same three-way interaction along with comparable ORs for specialized treatment and treatment completion. That is, there was no evidence that differences in planned LOS were responsible for group differences on outcomes. We also examined the effects of specialized treatment using multilevel models. Measurements of client behavior within sites are typically correlated (i.e., all else being equal, clients within the same site are more likely to respond similarly than clients at different sites). Failure to account for this correlation—called intraclass correlation or design effect—and the additional variance attributable to random effects can result in underestimated standard errors and spurious statistical significance. We first examined two-level models (clients nested within sites) that are analogous to our final model, which did not readily converge. We next examined models that were simplified by (1) removing predictors that were highly correlated, and by (2) restricting the sample to agencies with sufficient sample size at the client level. These simplified multilevel models produced results that were similar to those of our final logistic model.

As expected, standard errors of key effect parameters were larger, but effects remained significant at p b .05. Consequently, there is no indication that failing to account for intraclass correlation affected the substance of the findings. Finally, we considered the possible confounding effect of transitional housing on continuing care achievement. Transitional housing, available for up to 18 months to women who completed specialized treatment, was not widely utilized (n = 40) but did display a small positive association with continuing care (/ = .18). When transitional housing was included in our multivariate model, however, the effects of specialized treatment on continuing care were essentially unchanged (results not shown). In sum, the study design and analytic approach provided many strengths, as well as certain limitations. The study used administrative data from a large population of women who received SP or ST residential treatment for substance abuse problems. Although the study used a quasi-experimental design, group nonequivalence was addressed by constructing propensity scores, which successfully removed group confounds. This approach, drawing upon the rich set of variables collected upon entry, adjusted for group differences and also controlled for individual factors related to continuity of care. Our multivariate models, however, did not control for some factors potentially related to continued care such as motivation for treatment and motivation for recovery, and we did not attempt to model between-program differences to explain the achievement of continuing care. We did consider that arbitrary program differences in policies regarding retention both within and between treatment modalities could influence the findings. Important, we found that, when program heterogeneity in expected LOS was controlled, specialized treatment remained a significant predictor. Multilevel models that accounted for withinagency clustering also provided similar and consistent results. Taken together, the findings suggest that neither selection bias nor arbitrary structural differences between modalities can explain the observed superiority of specialized treatment in achieving continuity of care. They further suggest that specialized treatment delivered in other settings would provide similar results. 4.2. Treatment implications The present study highlights the value of specialized treatment, although the mechanism by which specialized treatment acts cannot be precisely determined. The expanded and women-focused services that mark specialized treatment (Ashley et al., 2003) may contribute to the results we obtained. As implemented in Washington, specialized treatment agencies provide substance abuse education tailored to women, case management and social services, mental health services, and job training services in an environment that supports women and their children. Furthermore, we found that specialized agencies provide women with significantly more service referrals on treat-

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

ment completion than do traditional programs. Hence, it is possible that the women-specific services and service linkages put forward by specialized agencies facilitate the transition of their clients to continuing care (Marsh, D’Aunno, & Smith, 2000). It is also possible that other treatment features are involved. For instance, childcare and women-only treatment may be bundled with other particular services, which are the key factors that result in higher continuity. The administrative database used in this study did not allow us to examine this possibility. In addition, the slightly more favorable client:staff ratio in specialized agencies could also play a role (Stark, 1992). A quadratic association was observed between LOS and continuing care, such that women with very short and very long LOS were the least likely to continue. An inverted U-shaped relationship with LOS has also been observed for the drug-use outcomes of long-term residential clients (Zhang et al., 2003). The inverted LOS curve seen in this study argues against a retention threshold that promotes continuing care achievement. It also illustrates that, even when providers use very long treatment residential stays, many clients choose to forgo continuing care. Finally, women who completed residential treatment were the ones most likely to continue with outpatient aftercare, but even some clients who did not complete their index episode enrolled in continuing care. Because the duration of a treatment episode may be more closely related to treatment outcomes than the service intensity delivered to a client (Moos, 2003), interventions that promote continuing care are especially important. Clients who attend continuing care are more likely to abstain from substance use, are less likely to be readmitted to treatment, and are more likely to have more favorable criminal justice outcomes. In recent years, more women have entered substance abuse treatment, but women-specific treatment programs comprise a minority of programs within the treatment system, and some pieces of evidence suggest that the priority given to specialized treatment for women may be diminishing (Grella & Greenwell, 2004). In this light, this study has implications for the development and utilization of residential substance abuse treatment programs for women. We found that women in specialized programs who were engaged in treatment for longer periods and who completed residential treatment were more likely to achieve continuing care. Thus, specialized programs may better address the unique substance abuse treatment needs of many women, especially those with children. Treatment completion is a critical clinical goal, given the positive association with continuing care and abstinence (Green et al., 2004) and the negative association with readmission (Luchansky, He, et al., 2000) that it affords. Because women with short treatment stays are particularly at risk for adverse outcomes, successful efforts to keep women in treatment until completion could benefit them and the treatment system as a whole. The effect of specialized treatment on continuing care should be interpreted in the context of its impact on other

37

treatment outcomes. Our initial descriptive examinations of employment, income support, arrests, and child welfare patterns confirm that women in both specialized and standard treatment modalities reached positive outcomes that were largely sustained for the 4-year follow-up period (Orwin et al., 2004). For example, in both groups, employment decreased immediately after treatment entry, but quickly rebounded to dramatically surpass pretreatment levels. A pattern similar to our continuing care findings emerged when the interactions between treatment condition and completion status were considered: Women who completed specialized treatment were twice as likely as noncompleters to be employed after 2 years, whereas women who completed standard treatment were no more likely than noncompleters to be employed. The broader importance of successfully engaging women in specialized care is suggested by these initial findings. The mechanism by which gender-sensitive treatment benefits women, however, awaits further study. Acknowledgments The authors gratefully acknowledge the assistance of Tracy Williams and Sue Green. This study was conducted with support from the National Institute on Drug Abuse (grant 1 R01 DA15094-01 to Orwin).

References Anglin, M. D., Hser, Y. I., & Grella, C. E. (1997). Drug addiction and treatment careers among clients in the Drug Abuse Treatment Outcomes Study (DATOS). Psychology of Addictive Behaviors, 11, 308 – 323. Ashley, O. S., Marsden, M. E., & Brady, T. M. (2003). Effectiveness of substance abuse treatment programming for women: A review. American Journal of Drug & Alcohol Abuse, 29, 19 – 53. Ashley, O. S., Sverdlov, L., & Brady, T. M. (2004). Length of stay among female clients in substance abuse treatment. In C. L. Council (Ed.), Health service utilization by individuals with substance abuse and mental disorders (pp. 107 – 132). Rockville, MD7 Substance Abuse and Mental Health Services, Administration Office of Applied Studies (DHHS Publication No. SMA 04-3949, Analytic Series A-25). Chen, X., Burgdorf, K., Dowell, K., Roberts, T., Porowski, A., & Herrell, J. M. (2004). Factors associated with retention of drug-abusing women in long-term residential treatment. Evaluation and Program Planning, 27, 205 – 212. Chutuape, M. A., Katz, E. C., & Stitzer, M. L. (2001). Methods for enhancing transition of substance dependent patients from inpatient to outpatient treatment. Drug and Alcohol Dependence, 61, 137 – 143. Clark, H. W. (2001). Residential substance abuse treatment for pregnant and postpartum women and their children: Treatment and policy implications. Child Welfare, 80, 179 – 198. Claus, R. E., & Kindleberger, L. R. (2002). Engaging substance abusers after centralized assessment: Predictors of treatment entry and dropout. Journal of Psychoactive Drugs, 34, 25 – 32. Coletti, S. D., Hughes, P. H., Landress, H. J., Neri, R. L., Sicilian, D. M., Williams, K. M., et al. (1992). PAR village. Specialized intervention for cocaine abusing women and their children. Journal of Florida Medical Association, 79, 701 – 705.

38

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39

Comfort, M., & Kaltenbach, K. A. (2000). Predictors of treatment outcomes for substance-abusing women: A retrospective study. Substance Abuse, 21, 33 – 45. Copeland, J., & Hall, W. (1992). A comparison of predictors of treatment drop-out of women seeking drug and alcohol treatment in a specialist women’s and two traditional mixed sex services. British Journal of Addiction, 87, 883 – 890. Donovan, D. (1998). Continuing care: Promoting the maintenance of change. In W. R. Miller, & N. Heather (Eds.), Treating addictive behaviors, (2nd ed. pp. 317–336). New York7 Plenum. Finkelstein, N. (1994). Treatment issues for alcohol- and drug-dependent pregnant and parenting women. Health & Social Work, 19, 7 – 15. Ghodse, A. H., Reynolds, M., Baldacchino, A. M., Dunmore, E., Byrne, S., Oyefeso, A., et al. (2002). Treating an opiate-dependent inpatient population: A one-year follow-up study of treatment ompleters and noncompleters. Addictive Behaviors, 27, 765 – 778. Green, C. A., Polen, M. R., Dickinson, D. M., Lunch, F. L., & Bennett, M. D. (2002). Gender differences in predictors of initiation, retention, and completion in an HOM-based substance abuse treatment program. Journal of Substance Abuse Treatment, 23, 285 – 295. Green, C. A., Polen, M. R., Lynch, F. L., Dickinson, D. M., & Bennett, J. D. (2004). Gender differences in outcomes in an HOM-based substance abuse treatment program. Journal of Addictive Diseases, 23, 47 – 70. Greenfield, L., Burgdorf, K., Chen, X., Porowski, A., Roberts, T., & Herrell, J. (2004). Effectiveness of long-term residential substance abuse treatment for women: Findings from three national studies. American Journal of Drug and Alcohol Abuse, 30, 537 – 550. Grella, C. E. (1999). Women in residential drug treatment: Differences by program type and pregnancy. Journal of Health Care for the Poor and Underserved, 10, 216 – 229. Grella, C. E., & Greenwell, L. (2004). Substance abuse treatment for women: Changes in settings where women received treatment and types of services provided, 1987–1998. Journal of Behavioral Health Services & Research, 31. Grosenick, J. K., & Hatmaker, C. M. (2000). Perceptions of the importance of physical setting in substance abuse treatment. Journal of Substance Abuse Treatment, 18, 29 – 39. Hirano, K., & Imbens, G. W. (2001). Estimation of causal effects using propensity score weighting: An application to data on right heart catheterization. Health Services and Outcomes Research Methodology, 2, 259 – 278. Hodgins, D. C., el-Guebaly, N., & Addington, J. (1997). Treatment of substance abusers: Single or mixed gender programs? Addiction, 92, 805 – 812. Hodgson, R., & John, B. (2004). Gender, gender role and brief alcohol interventions. Addiction, 99, 3 – 4. Howell, E. M., Heiser, N., & Harrington, M. (1999). A review of recent findings on substance abuse treatment for pregnant women. Journal of Substance Abuse Treatment, 16, 195 – 219. Hubbard, R. L., Craddock, S. G., Flynn, P. M., Anderson, J., & Etheridge, R. M. (1997). Overview of 1-year follow-up outcomes in the Drug Abuse Treatment Outcome Study (DATOS). Psychology of Addictive Behaviors, 11, 261 – 278. Hughes, P. H., Coletti, S. D., Neri, R. L., Urmann, C. F., Stahl, S., Sicilian, D. M., et al. (1995). Retaining cocaine-abusing women in a therapeutic community: The effect of a child live-in program. American Journal of Public Health, 85, 1149 – 1152. Hser, Y. I., Maglione, M., Joshi, V., & Chao, B. (1998). Effects of treatment program and client characteristics on client retention. Rockville, MD7 Center for Substance Abuse Treatment. Kauffman, E., Dore, M. M., & Nelson-Zlupko, L. (1995). The role of women’s therapy groups in the treatment of chemical dependence. American Journal of Orthopsychiatry, 65, 355 – 363. Kearney, M. H. (1998). Truthful self-nurturing: A grounded formal theory of women’s addiction recovery. Qualitative Health Research, 8, 495 – 512.

Kissin, W. B., Orwin, R. G., Grella, C. E., & Garfield, T. J. (2003). Similarities and differences between mothers in women-only and mixedgender programs in Washington State. Presentation at the annual meeting of the College on Problems of Drug Dependence, Miami, FL. Knight, D. K., Logan, S. M., & Simpson, D. D. (2001). Predictors of program completion for women in residential substance abuse treatment. American Journal of Drug and Alcohol Abuse, 27, 1 – 18. Lewis, L. M. (2004). Culturally appropriate substance abuse treatment for parenting African American women. Issues in Mental Health Nursing, 25, 451 – 472. Luchansky, B., Brown, M., Longhi, D., Stark, K., & Krupski, A. (2000). Chemical dependency treatment and employment outcomes: Results from the dADATSAT program in Washington State. Drug and Alcohol Dependence, 60, 151 – 159. Luchansky, B., He, L., Krupski, A., & Stark, K. D. (2000). Predicting readmission to substance abuse treatment using state information systems: The impact of client and treatment characteristics. Journal of Substance Abuse, 12, 255 – 270. Marsh, J., D’Aunno, T. A., & Smith, B. D. (2000). Increasing access and providing social services to improve drug abuse treatment for women with children. Addiction, 95, 1237 – 1247. McCollister, K. E., French, M. T., Inciardi, J. A., Butzin, C. A., Martin, S. S., & Hooper, R. M. (2003). Post-release substance abuse treatment for criminal offenders: A cost-effectiveness study. Journal of Quantitative Criminology, 19, 389 – 407. McComish, J. F., Greenberg, R., Ager, J., Chruscial, H., & Laken, M. A. (2000). Survival analysis of three treatment modalities in a residential substance abuse program for women and children. Outcomes Management for Nursing Practice, 4, 71 – 77. McComish, J. F., Greenberg, R., Ager, J., Essenmacher, L., Orgain, L. S., & Bacik, W. J., Jr. (2003). Family-focused substance abuse treatment: A program evaluation. Journal of Psychoactive Drugs, 35, 321 – 331. McKay, J. R. (2001). Effectiveness of continuing care interventions for substance abusers: Implications for the study of long-term treatment effects. Evaluation Review, 25, 211 – 232. McKay, J. R., Foltz, C., Leahy, P., Stephens, R., Orwin, R., & Crowley, E. (2004). Step down continuing care in the treatment of substance abuse: Correlates of participation and outcomes effects. Evaluation and Program Planning, 27, 321 – 331. McKay, J. R., Gutman, M., McLellan, A. T., Lynch, K. G., & Ketterlinus, R. (2003). Treatment services received in the CASAWORKS for Families program. Evaluation Review, 27, 629 – 655. McKay, J. R., Lynch, K. G., Shepard, D. S., Ratichek, S., Morrison, R., Koppenhaver, J., et al. (2004). The effectiveness of telephone-based continuing care in the clinical management of alcohol and cocaine use disorders: 12-Month outcomes. Journal of Consulting and Clinical Psychology, 72, 967 – 979. McKay, J. R., McLellan, A. T., Alterman, A. I., Cacciola, J. S., Rutherford, M. J., & O’Brien, C. P. (1998). Predictors of participation in aftercare sessions and self-help groups following completion of intensive outpatient treatment for substance abuse. Journal of Studies on Alcohol, 59, 152 – 162. McLellan, A. T., Weinstein, R. L., Shen, Q., Kendig, C., & Levine, M. (2005). Improving continuity of care in a public addiction treatment system with clinical case management. American Journal on Addictions, 14, 426 – 440. Mertens, J. R., & Weisner, C. M. (2000). Predictors of substance abuse treatment retention among women and men in an HMO. Alcoholism: Clinical and Experimental Research, 24, 1525 – 1533. Metsch, L., Wolfe, H., Fewell, R., McCoy, C., Elwood, W., Wohler-Torres, B., et al. (2001). Treating substance-using women and their children in public housing: Preliminary evaluation findings. Child Welfare, 80, 199 – 220. Moos, R. H. (2003). Addictive disorders in context: Principles and puzzles of effective treatment and recovery. Psychology of Addictive Behaviors, 17, 3 – 12.

R.E. Claus et al. / Journal of Substance Abuse Treatment 32 (2007) 27 – 39 Moos, R. H., Brennan, P. L., & Mertens, J. R. (1994). Diagnostic subgroups and predictors of one-year readmission among late-middle-aged and older substance abuse patients. Journal of Studies on Alcohol, 55, 173 – 183. Moos, R. H., Petit, B., & Gruber, V. (1995). Longer episodes of community residential care reduce substance abuse patients’ readmission rates. Journal of Studies on Alcohol, 56, 433 – 443. Orwin, R. G., Francisco, L., & Bernichon, T. (2001). Effectiveness of women’s substance abuse treatment programs: A meta-analysis. Fairfax, VA7 Center for Substance Abuse Treatment. Orwin, R. G., Kissin, W. B., Claus, R. E., Grella, C. E., & Williams, T. (2004, June). Specialized versus standard chemical dependency treatment for women with children: Attending to heterogeneity in a retrospective multisite study. Presentation at the annual meeting of the College on Problems of Drug Dependence, San Juan, Puerto Rico. Orwin, R. G., Kissin, W. B., & Dugan, M. K. (2003, November). Specialized versus standard chemical dependency treatment for women with children: Preliminary findings from a retrospective multisite quasiexperiment. Presentation at the annual meeting of the American Evaluation Association, Reno, NV. Orwin, R. G., Maranda, M., & Brady, T. (2001). The impact of prior physical and sexual victimization on substance abuse treatment outcomes: A reanalysis of the National Treatment Improvement Evaluation Study (prepared for the Center for Substance Abuse Treatment, Substance Abuse and Mental Health Services Administration). Arlington, VA7 Author. Available at http://www.icpsr.umich.edu/ SAMHDA/NTIES/NTIES-PDF/REPORTS/Violence_2.pdf. Roberts, A. C., & Nishimoto, R. H. (1996). Predicting treatment retention of women dependent on cocaine. American Journal of Drug and Alcohol Abuse, 22, 313 – 333. Rosenbaum, P., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41 – 55. Rubin, D. B. (1997). Estimating causal effects from large data sets using propensity scores. Annals of Internal Medicine, 127 (8 Pt. 2), 757 – 763. Saunders, B., Baily, S., Phillips, M., & Allsop, S. (1993). Women with alcohol problems: Do they relapse for reasons different to their male counterparts? Addiction, 88, 1413 – 1422.

39

Schmitt, S. K., Phibbs, C. S., & Piette, J. D. (2003). The influence of distance on utilization of outpatient mental health aftercare following inpatient substance abuse treatment. Addictive Behaviors, 28, 1183 – 1192. Simpson, D. D., Joe, G., & Brown, B. (1997). Treatment retention and follow-up outcomes in the Drug Abuse Treatment Outcome Study (DATOS). Psychology of Addictive Behaviors, 11, 294 – 307. Simpson, D. D., Savage, L. J., & Lloyd, M. R. (1979). Follow-up evaluation of treatment of drug abuse during, 1969–1972. Archives of General Psychiatry, 36, 772 – 780. Stark, M. J. (1992). Dropping out of substance abuse treatment: A clinically oriented view. Clinical Psychology Review, 12, 93 – 116. Stevens, S. J., & Patton, T. (1998). Residential treatment for drug addicted women and their children: Effective treatment strategies. Drugs & Society, 13, 235 – 249. Szuster, R. R., Rich, L.L, Chung, A., & Bisconer, S. W. (1996). Treatment retention in women’s residential chemical dependency treatment: The effect of admission with children. Substance Use and Misuse, 31, 1001 – 1013. TOPPS-II Interstate Cooperative Study Group. (2003). Drug treatment completion and post-discharge employment in the TOPPS-II Interstate Cooperative Study. Journal of Substance Abuse Treatment, 25, 9 – 18. Velasquez, M. M., & Stotts, A. L. A. (2003). Substance abuse and dependence disorders in women. In M. Kopala, & M. Keitel (Eds.), Handbook of counseling women. Thousand Oaks, CA: Sage Publications, Inc. Washington State Department of Social and Health Services and Department of Health. (1999). Washington State Mom’s Project Perinatal Research and Demonstration Project—The Mom’s Project final report. Olympia, WA: Washington State Department of Social and Health Services. Wickizer, T., Maynard, C., Atherly, A., Frederick, M., Koepsell, T., Krupski, A., et al. (1994). Completion rates of clients discharged from drug and alcohol treatment programs in Washington State. American Journal of Public Health, 84, 215 – 221. Wobie, K., Eyler, F. D., Conlon, M., Clarke, L., & Behnke, M. (1997). Women and children in residential treatment: Outcomes for mothers and their infants. Journal of Drug Issues, 27, 585 – 606. Zhang, Z., Friedmann, P. D., & Gerstein, D. R. (2003). Does retention matter? Treatment duration and improvement in drug use. Addiction, 98, 673 – 684.