Content uploaded by Vivian Brown
Author content
All content in this area was uploaded by Vivian Brown on Mar 05, 2019
Content may be subject to copyright.
Running head: TREATMENT RETENTION
Does Integrated Trauma-Informed Substance Abuse Treatment Increase Treatment Retention?
Hortensia Amaro, Ph.D.,1 Miriam Chernoff, Ph.D.,1 Vivian Brown, Ph.D.,2 Sandra Arévalo,
B.A.,1 Margaret Gatz, Ph.D.3
1 Institute on Urban Health Research, Bouvé College of Health Sciences, Northeastern
University,
2 PROTOTYPES Systems Change Center,
3 University of Southern California
Key words: Substance abuse treatment, Treatment retention, Trauma services, Women
Author Note
Corresponding Author: Hortensia Amaro, Distinguished Professor, Bouvé College of Health
Sciences, 360 Huntington Avenue, Stearns Suite 503, Boston, MA 02115. E-mail:
h.amaro@neu.edu; phone (617) 373-7601.
Treatment Retention
Abstract
This paper presents findings from a quasi experimental non-randomized group design
study with repeated measures that explored whether trauma-enhanced substance abuse treatment
results in longer treatment stays and improved outcomes compared with substance abuse services
as usual. Data represent 461 participants in two residential treatment sites of the Women, Co-
Occurring Disorders and Violence Study, which was sponsored by the Substance Abuse and
Mental Health Services Administration. Results indicate that after controlling for baseline
treatment-group differences, the intervention group had a 31% lower risk than the comparison
group for discontinuation of treatment within four months. Proportional hazards models show
that baseline mental health and trauma symptoms and alcohol and drug severity scores did not
predict either overall length of time in treatment or differences in retention between intervention
and comparison groups. Substance abuse and mental health symptoms improved with increased
duration of treatment in interaction with more severe baseline symptoms.
2
Treatment Retention
Does Integrated Trauma-Informed Substance Abuse Treatment Increase Treatment Retention?
High drop-out rates continueImproving retention continues to be a major challenge in
substance abuse treatment (Battjes, Onken, & Delaney, 1999; Dakof et al., 2003).; Festinger,
Lamb, Kountz, Kirby, & Marlowe, 1995). Crits-Christoph and Siqueland (1996) reported, “High
dropout rates are apparent even in our sample of relatively recent studies, in which researchers
have attempted to provide high quality therapy using treatment guides or manuals” (p. 753).
There is still debate whether these rates are much higher than the drop-out rates found among
outpatients in mental health centers. However, results of research with clients with addictive
disorders show a powerful association between dropping out and negative outcomes. Clients
who complete drug treatment are more likely to be alcohol and drug free, have lower
unemployment, and arrest, and relapse rates, cease intravenous drug use, and have lower relapse
rates (Simpson & Brown, 1999; Simpson, Joe, & Brown, 1997). Specifically, clients have more
favorable outcomes if they remain in treatment over 90 days (National Institute on Drug Abuse
[NIDA], 1997; Simpson et al., 1997).
Because of the implications for treatment effectiveness, methods for improving treatment
retention have been the focus of empirical study. The Drug Abuse Treatment Outcome Study
(DATOS) found that rates of retention of 90 days or more for long-term residential treatment
range from 21% to 75% for the poorest and the best programs, respectively. Numerous oOther
studies (Amaro, Nieves, Johannes, & Cabeza, 1999; DeLeon, Jainchill, & Melnick, 2000;
Donovan, Rosengren, Downey, Cox, & Sloan, 2001; Grella, Hser, Joshi, & Anglin, 1999; Haller,
Elswick, Dawson, & Schnoll, 1997; Knight, Logan, & Simpson, 2001; Roberts & Nishimoto,
1996; Rowan-Szal, Joe, & Simpson, 2000; Strantz & Welch, 1995; Hughes, et al., 1995; Veach,
3
Treatment Retention
Remley, Kippers, & Sorg, 2000) have corroborated the importance of retention and have also
shown the difficulty in retaining clients. Overall, retention rates reported in these studies
retention rates of residential programs 90 days or longer range from 41% to 56%.
Factors related to retention in substance abuse treatment are generally classified into
client or program in nature. According to Knight et al. (2001), client factors can be categorized
into five groups: 1) sociodemographics (age, gender, education, race/ethnicity), 2) substance use
(severity of the problem), 3) legal involvement (legal pressure or external motivation), 4)
psychological functioning, and 5) social relations. In general, studies have shown that risk of
In general, studies have shown that risk of drop out is greater for clients who are younger
(Scott-Lennox, Rose, Bohlig, & Lennox, 2000), who are female (Arfken , Klein, di Menza, &
Schuster, 2001; Bride, 2001; Kelly, Blacksin, & Mason, 2001), who have fewer years of
education (Knight et al., 2001), and whose race is not White (King & Canada, 2004; Knight et
al., 2001; Nelson-Zlupko, Dore, Kauffman, & Kaltenbach, 1996; Mertens & Weisner, 2000;
Milligan, Nich, & Carroll, 2004; Scott-Lennox, et al., 2000). In addition, clients with more
severe drug problems (Kelly, et al., 2001; Mertens & Weisner, 2000; Veach et al., 2000), with no
legal pressure (Joe, Simpson, Greener, & Rowan-Szal, 1999; Knight, et al., 2001), with more
psychological problems or higher psychiatric severity (Mertens & Weisner, 2000; Ross, Cutler,
& Sklar, 1997 ), with lower motivation for treatment (Joe et al., 1999; Simpson et al., 1997), and
clients who lack social support are also at higher risk for dropping out (Kelly, et al., 2001).
Other studies have found either no association between gender and risk for drop out
(Lundy, Gottheil, Serota, Weinstein, & Sterling, 1995) or lower drop-out rates among women
(Andersen & Berg, 2001). Environmental barriers including inability to bring young children
into treatment with them, inadequate childcare, lack of transportation, and lack of gender-
4
Treatment Retention
sensitive services are other factors contributing to treatment drop-out among women (Howard &
Beckwith, 1996; Hughes et al., 1995; Lewis, Haller, Branch, & Ingersol, 1996; Szuster, Rich,
Chung, & Bisconer, 1996).
Results are mixed on the relationship between retention and client psychological
attributes, experiences with trauma, and/or co-morbidity. Some studies have shown better
retention among persons with co-morbid disorders, especially depressive disorders (Martinez-
Raga, Keaney, Marshall ,Ball, & Strang, 2002; Saxon & Calsyn, 1995;) or social anxiety (Egelko
& Galanter, 1998). However, others have reported higher drop-out rates among women with Axis
I co-morbidity, especially those with severe mental illness (Brown, Huba, & Melchio, 1995;
Brown, Melchior, & Huba, 1999). Furthermore, clients in substance abuse treatment with co-
morbid externalizing disorders or personality disorders are more likely to drop out of treatment,
and psychiatric severity at admission to treatment is the best predictor of drug use at follow-up
(Greenberg, Otero, & Villanueva, 1994; Martinez-Raga, et al., 2002; Rowe, Liddle, Greenbaum,
& Henderson, 2004). Higher drop-out rates are also reported among persons with a history of
physical or sexual abuse or a history of child abuse/neglect (Claus & Kindleberger, 2002; Kang,
Deren, & Goldstein, 2002). Amaro, et al. (1999) also reported that Latina women from a
residential substance abuse program with a history of childhood abuse were more likely to drop
out in the early stages of treatment. In another study, Thompson and Kingree (1998) reported that
treatment completion in a residential substance abuse treatment program for low-income
pregnant women was related to Post Traumatic Stress Disorder (PTSD). Over one third (37%) of
the participants who manifested symptoms consistent with PTSD completed treatment compared
with 59% of those asymptomatic for PTSD; however, in this study exposure to a violent
traumatic event per se did not predict treatment completion. Similarly, Hien, Nunes, Levin, and
5
Treatment Retention
Fraser (2000) showed that occurrence of trauma or PTSD did not predict drop-out rates;
however, childhood physical and sexual abuse and adult exposure to violence were associated
with higher rates of PTSD, which in turn predicted poorer treatment adherence.
Despite findings of low retention rates among women with co-occurring disorders, there
has been a dearth of research on how interventions might be modified to improve treatment
retention for this client population. Haller, Miles, and Dawson (2002) reported that only 36% of
women with severe addiction and psychiatric disorders completed treatment, whereas 57% of
women with a less severe clinical profile completed treatment. However, women with Cluster B
personality disorders (externalizing psychopathology) were retained longer (76% completed
treatment) when “environmental barriers” (i.e., lack of childcare, transportation, and gender-
sensitive services such as parenting classes and vocational training) were removed. In addition,
Kelly et al. (2001) found that higher numbers of stressful and chaotic factors including
psychiatric illness, current domestic violence, and the involvement of child protective services
were present in the lives of women who did not complete treatment, suggesting that attention to
these factors might improve retention. Cosden and Cortez-Ison (1999) also found that women
who reported sexual abuse stayed in their programs for shorter periods of time than those who
did not report it.
For persons with co-occurring disorders, there is growing agreement that an integrated treatment
approach is the most appropriate model (Barrowclough et al., 2001; Hellerstein, Rosenthal, &
Miner, 2001; RachBeisel, Scott, & Dixon, 1999; Ziedonis & Stern, 2001) in contrast to
treatments that address each disorder separately. Some of the components for successful
integrated treatment models include stages of change components, motivational interviewing,
and cognitive-behavioral interventions (Ouimette, Brown, & Najavits, 1998; Ziedonis & Stern,
6
Treatment Retention
2001), plus screening for trauma and health problems and an approach that builds on safety and
empowerment in the community., as well as education about sexuality and pregnancy prevention
(RachBeisel, et al., 1999). For persons with substance use disorder and PTSD, especially, it has
been suggested that comprehensive Substance Use Disorder (SUD)-PTSD treatment should
address substance use and directly intervene with PTSD symptoms and quality of life concerns
such as vocational issues and social support (Najavits, Weiss, & Shaw, 1997; Ouimette, et al.,
1998).
The question that remains unaddressed is whether an integrated treatment program
improves retention. This study seeks to answer the following questions related to retention in
treatment: 1) Do integrated trauma services lead to longer retention in residential treatment
among women with co-occurring disorders? 2) Does the effect of the intervention on retention
differ by severity in symptoms of addiction, mental health or trauma? 3) Do longer treatment
stays lead to better 6-month addiction, mental health, and trauma outcomes?
Data presented in this paper reflect the experiences of 461 residential treatment
participants in two research sites of the Women, Co-Occurring Disorders, and Violence Study
(WCDVS), sponsored by the Substance Abuse and Mental Health Services Administration
(SAMHSA). The WCDVS is the first large-scale federal initiative that addresses the significant
lack of appropriate services for women with co-occurring substance use and mental health
disorders who have experienced trauma and that also includes their children. This federal
initiative entailed a nine-site multi-model intervention study with a quasi-experimental design,
including baseline, six 6-month follow-up, and 12-month follow-up interviews with the women
in the study. Each site contributed data from both an intervention and comparison group.
7
Treatment Retention
Intervention sites provided integrated, consumer-involved, trauma-informed and
comprehensive services (see Huntington, Moses, & Veysey, in press, for more detail). The
comprehensive services included outreach, assessment, crisis intervention, trauma-specific
counseling, ongoing treatment, parent skills training, resource coordination and advocacy, and
peer-run services. The present analyses included two of the nine Women and Violence research
sites with residential treatment programs that collected information on actual admission and
discharge dates. Women participated in one of two manualized, trauma-specific group
interventions:---Seeking Safety, a 31 sessions/4-months long intervention (Najavits, Weiss,
Shaw, & Muenz, 1998), at the Los Angeles site (LA) and Trauma Recovery and Empowerment, a
25 sessions/6-months long intervention (Harris, 1998), at the Boston site (B). Both of these
multi-session trauma group interventions focused on maintaining personal safety, empowerment
and coping skills, and understanding the links among substance abuse, mental health problems,
and trauma. The intervention’s comprehensive services included residential services where
women could live with their children, childcare, transportation, and gender-sensitive services that
have been shown to contribute to lower treatment drop-out rates among women (Howard &
Beckwith, 1996; Hughes et al, 1995; Lewis et al, 1996; Szuster et al, 1996). Intervention
participants received treatment from five agencies: two 12-months residential treatment agencies
from the LA site, and two 12-months and one 9-months residential treatment agencies from the B
site. The comparison condition represented residential substance abuse care as usual in that
site’s region. It also varied from site to site and sometimes included some of the same program
elements offered by the integrated-services intervention. However, none of these usual care
agencies provided trauma-specific treatment (McHugo, Kammerer, Jackson, Markoff, Gatz,
Larson, Mazelis, & Hennigan, in press). Comparison participants received treatment from seven
8
Treatment Retention
agencies: one 9-months, and four 6-months residential treatment agencies from the LA site, and
one 6-months and one 9-months residential treatment agencies from the B site.
Methods
Sample
This paper examines baseline and 6-month interview data from 461(N=198 in
intervention group, N=263 in comparison group) women in residential treatment and enrolled at
the Boston (N=92) and Los Angeles (N=369) sites of the WCDVS. (For a full description of the
study design, see McHugo, Kammerer, et al., [in press] and Giard, et al., [in press]. For a full
description of women’s characteristics at baseline, see Becker, et al., [in submission].)
Study participants in this sample were recruited into the study and received baseline
interviews within 60 days of admission to residential treatment: 35% of participants received
baseline interview within two weeks, 52% within 30 days, and 12% between 31 and 60 days.
About 29% of participants were in 6-month programs, whereas 30% were in 9-month treatment
programs and 41% were in 12-month programs. Average age was 33 years in both the
intervention and comparison groups. Average education (in years) was 11 for both groups.
Across groups, the most common race/ethnicity identifications were Hispanic, black non-
Hispanic, and white non-Hispanic (33%, 24%, and 34%, respectively), with no differences
between intervention and comparison groups.
Measures
Background characteristics. Respondents were asked closed-ended questions about
demographic information, employment status prior to treatment entry, history of homelessness,
court-ordered treatment, jail experience , relationship status (married or partnered), number of
children, custody status of children, and loss of parental rights.
9
Treatment Retention
Dependent measures of mental health, trauma, and addiction symptoms. Mental health
symptoms were assessed using the global severity index (GSI) of the Brief Symptom Inventory,
a 53-item symptom checklist (Derogatis, 1993). The GSI is a mean measure for perceived
severity of 53 mental health symptoms in the past 7 days (each symptom ranging from 0-4, not at
all to extremely, with higher scores indicating more severe symptoms). For GSI (n=441),
Cronbach alpha= 0.961. In addition, the following measures of psychiatric history were obtained:
number of psychiatric hospitalizations and use of prescription psychotropic medications.
Trauma symptoms were measured using the Post Traumatic Stress Disorder Scale (PSS:
Foa, Cashman, Jaycox, & Perry, 1997). Respondents were asked how often in the past month (on
a scale of 0 [not at all or only once] to 3 [five or more times or almost always]) they have
experienced a list of problems sometimes experienced after a traumatic event. For PSS (n=443),
the Cronbach alpha was 0.896. Traumatic events were measured using an adaptation of the Life
Stressor Checklist – Revised (LSC-R) (Wolfe, Kimerling, Brown, Chrestman, & Levin, 1996).
History of childhood physical and sexual abuse and history of adult physical and sexual abuse
were derived from answers to the LSC-R, and a scale of lifetime exposure to stressful events
(LESE) was constructed by summing across events (McHugo, Caspi, et al., in press). For LESE
(n=430), the Cronbach alpha was 0.710.
Addiction severity was measured using the Addiction Severity Index (ASI) alcohol and
drug composite scores (adapted; McLellan, Luborsky, Woody, & O'Brien, 1980). ASI composite
scores, ranging from 0 to 1, are based on the amount of use, related problems (e.g. cravings,
withdrawal, wanting but unable to stop, being bothered by these problems), and perceived
importance of treatment. For ASI Alcohol Composite (n=461), the Cronbach alpha was 0.895
based on weighted items used in computing the ASI alcohol composite measure. For ASI Drug
10
Treatment Retention
composite (n=458), the Cronbach alpha was 0.695 based on weighted items used in computing
the ASI drug measure. Women were also asked number of previous substance abuse treatment
starts.
Dependent measure of days in treatment. Length of stay (LOS) was computed in days from
program specific admission and discharge data. Women who were discharged and readmitted
within 30 days were counted as having one treatment episode. Comparison and intervention
programs were not matched for the length of stay required for program completion. More
participants in the intervention group were in longer programs; specifically, there were five
programs of 6-months duration and two programs of 9-months duration in the control group,
compared to the intervention group where there was one program of 9-months duration and four
programs of 12-month duration. Therefore, in the analyses, the length of stay (LOS) maximum
number of days was capped at 120 days because a substantial proportion of women in the
shortest (6-month) treatment programs left after 5 months but had completed the programs
successfully.
Statistical Methods
All statistical tests were 5% sized, two-sided tests of the specified null hypotheses against
the two-sided alternative. Events were defined as discharge from the substance abuse treatment
program, censoring at 120 days (4 months)
To test for baseline differences, we used t-tests for normally distributed variables, non-
parametric Wilcoxon rank sum tests for continuous non-normally distributed variables (such as
the ASI composite scores) and ordered, categorical variables (such as the number of substance
abuse [SA] treatment starts), and chi square tests for categorical variables.
11
Treatment Retention
To test treatment effect on length of treatment, we first estimated conditional probabilities
of remaining in treatment using Kaplan Meier product-limit estimates with censoring at 120
days. Second, we used Cox proportional hazards models to test for treatment group differences in
hazard rates for the risk of program discharge, controlling for client characteristics that differed
at baseline (with p < 0.05) and for the interaction effect between two of these, which were highly
correlated (e.g., jail experience and mandated to treatment; Hosmer and Lemeshow, 1999). To
test the effect of addiction, mental health, or trauma symptom severity on treatment retention we
used the final proportional hazards regression models, adding each symptom or severity measure
and its interaction with treatment group and tested the null hypotheses that these parameters were
equal to zero, in order to determine if there was a statistical association between symptom
severity with time in treatment and whether the intervention worked better in certain groups of
women, for example, those with higher or lower baseline mental health or trauma symptoms or
baseline substance use.
To assess the effect of longer treatment stay on 6-month addiction, mental health, and
trauma outcomes, we tested the relation of time in treatment with 6-month outcomes using
general linear models, first including all pairs of two-way interactions among length of stay,
treatment group, and baseline symptom value as well as the three-way interaction among these
variables. Retaining the interaction between treatment group and length of stay, we excluded
interaction effects sequentially, first choosing the model effect with the least significant p-value,
until we included all main effects and only interaction effects with p-values < 0.15 (Scheffe,
1959). We performed sensitivity analyses to better understand regression results, re-analyzing
these data after excluding outlying observations that were more than three standard deviations
from the mean improvement. Since treatment engagement might affect outcome, we also
12
Treatment Retention
included a measure of engagement in the final regression models, specifically, the percentage of
trauma group sessions attended (SS in LA and TREM in Boston). Because treatment group and
engagement parameters were confounded (participants in the comparison group did not attend
trauma sessions), we re-analyzed two datasets, first including women in both treatment groups
and secondly, only the women in the intervention groups, with both models substituting the
engagement measure for the treatment group parameter.
Results
Baseline Differences Between Groups
Table 1 shows comparison of intervention and comparison groups at baseline. Treatment groups
differed in history of homelessness (p=0.036), mandated treatment (p<0.0001), recent jail
experience (p<0.0001), number of substance abuse treatment starts (p=.004), and number of days
from admission to the baseline interview (p< 0.0001). There were no differences at baseline
between control and intervention groups regarding the use of mental health or SA medications, at
baseline 34% of all participants were taking psychotropic medications, and 4% were taking SA
medications. However, intervention group participants who did not take medications within 3
months prior to the baseline interview were more than twice as likely as the control group to start
taking mental health medications 3 months prior to the 6-month interview (OR [95% CI], 2.29,
[1.16, 4.54]) .No differences in other variables were found between the two groups: intervention
and services as usual.
Table 1 also reports mental health (BSI-GSI), trauma (PSS and LESE), and addiction
(ASI) severity: BSI (GSI) scores are interpreted by comparison to normative data available for
both clinical and non-clinical sample of adults (mean (SD) for Psychiatric out-patient 1.32
(0.72); for Psychiatric in-patients 1.36 (0.86); and for Non-patients 0.30 (0.31) (Derogatis,
13
Treatment Retention
1980)). BSI (GSI) scores for both intervention, 1.13 (0.69); and control, 1.20 (0.73), fell within
the ranges for psychiatric out-patient and psychiatric in-patient normative scores. The LESE
(trauma history interview) has no established summary variables, so the ones we created for the
WCDVS are without normative data (McHugo, Kammerer, et al., In press). For the PSS, both the
intervention (20.8) and comparison (20.9) groups fell in the moderate to severe category (PTSD
symptom severity: <=10 mild, 11-20 moderate, 21-35 moderate to severe, >=36 severe)
(McHugo, Kammerer, et al., In press). Composite scores of ASI that range from 0 (no problem)
to 1.0 (extreme problem) were 0.02 (0, 0.30) and 0.04 (0, 0.51) respectively for intervention and
control groups.
Table 1 here
Does the intervention increase treatment retention after we control for baseline differences
between treatment groups?
By 4 months, 45% of the intervention group and 50% of the comparison group had left
treatment. Crude difference in length of time in treatment was not statistically significant (Log
rank p=0.203, Wilcoxon p = 0.163 (see Figure 1). In the intervention group, the 75th percentile
(the time point at which the probability of still being in treatment is 0.75) is 63 days after
admission (95% CI: 49, 80 days; n=198 with 55% censored) for the integrated-services treatment
group (e.g., intervention group) and 49 days (95% CI: 41, 60 days; n=263, with 50% censored)
for the comparison group.
Figure 1 here
The first level model controlled for treatment group and for the variables that were
significantly different between the two groups at baseline: homelessness in past 6 months, jail
experience within past 6 months, history of mandated treatment, number of substance abuse
14
Treatment Retention
treatment starts, and days from admission to baseline interview. Results are shown in Table 2.
Proportional hazards model assumptions were met, with p=0.584 for test of non-zero, time-
dependent covariate. After controlling for baseline differences, the women in the intervention
group were more likely to have longer treatment stays, with a 31% lower risk of dropping from
treatment at any time than that of the comparison group women (N=448, p=0.021, HR=0.691;
53% censored observations). Table 2 shows other significant predictors for treatment program
drop out: Women who were court-mandated tended to stay in treatment longer (HR = 0.65; the
drop-out risk was 35% lower for these women; p<.007). Women with no previous substance
abuse treatment had a higher risk of drop out (HR=2.5) than women with 5 or more previous
substance treatment starts (p<.0001 for overall test), while women with a moderate number of
treatment starts (1-2 or 3-4) tended to stay in treatment longer than did women with 5 or more
starts (HR=0.6 to 0.7, or risk was 30% to 40% less).
Table 2 here
Does the intervention effect on retention differ by severity in addiction, mental health, or trauma
symptoms?
Using proportional hazards regression, we tested each symptom scale individually to see
if there was a differential intervention effect on retention. The risk of drop out did not differ for
women with different severity levels of baseline symptoms (main effects p-values > 0.19 for
ASI, GSI, and PSS). Also, there was no differential effect of treatment on retention by baseline
symptom severity itself (interaction effect p-values between treatment and each of ASI, GSI, PSS
measures all >0.23).
Do longer treatment stays lead to better 6-month mental health, addiction severity, and trauma
outcomes?
15
Treatment Retention
Longer treatment stay was associated with greater symptom improvements for GSI
(Table 3a) and ASI measures (Tables 3c and 3d), but not for PSS measures (Table 3b). For
example, baseline symptom severity moderated the effect of length of treatment, generally with
more improvement seen in women who started with more severe symptoms. As Table 3a shows,
for GSI, the interaction between length of treatment and baseline symptom severity differed by
treatment group; specifically, the positive effect of longer stay in treatment on symptom outcome
was greater for women with higher GSI values at baseline (p=0.002). However, this effect
differed by treatment group, with greater strength in the intervention group (three-way
interaction p-value=0.022).
Tables 3a, 3b, 3c and 3d here
In addition, length of stay affected symptom improvement through interaction with
baseline ASI measures. Women with longer lengths of stay and higher baseline ASI drug and ASI
alcohol measures showed greater improvement at 6 months (p<0.001 and p<0.002, respectively;
Tables 3c and 3d).
Sensitivity analyses after excluding outliers (for GSI and PSS; there were no outliers for
ASI measures) showed similar results as noted above. For GSI, only the interaction between
length of stay and baseline value remained significant at p<0.05; however the trend in the p-
values and parameter estimates matched the analysis of the full dataset.
Seventy-two percent of intervention participants attended at least one trauma group
session, with a mean number of sessions of 13 (SD=7), representing, on average, 42% of the
targeted number of sessions. Analyses of PSS that incorporated the percent of treatment
engagement led to similar results and did not explain the lack of significance of the effect of
length of stay on the PSS outcome.
16
Treatment Retention
Discussion
After controlling for baseline differences in the two groups, integrated trauma
intervention services led to better retention than that found for treatment as usual (e.g., the
comparison group). In the integrated treatment group (e.g., the intervention group), 55% of
women remained in treatment for more than 120 days compared to 50% in the comparison
group. Although the crude rates were not statistically significant, after we adjusted for baseline
group differences, the higher retention in the intervention group was statistically significant.
Although perhaps clinically modest, these findings compare favorably with 90-day retention
rates reported in other studies of long-term residential substance abuse treatment (Amaro et al.,
1999; DeLeon, et al., 2000; Donovan, et al., 2001; Grella, et al. , 1999; Haller, et al., 1997;
Hughes, et al., 1995; Knight, et al., 2001; Roberts & Nishimoto, 1996; Rowan-Szal et al., 2000;
Strantz & Welch, 1995; Veach et al., 2000). Even though there are a number of recent studies on
integrated treatment efficacy, we could not find any controlled studies that reported the effect of
integrated services on substance abuse treatment retention.
In contrast to studies (Broome, Flynn, & Simpson, 1999; Haller, et al., 2002; Kelly, et al.,
2001) that reported individuals with more complex psychiatric and addiction presentations have
shorter lengths of stay in treatment, we found no association between baseline symptom severity
and treatment retention. It is possible that our findings differ from those previously reported
because women eligible for this study had to have diagnosed mental health and substance abuse
disorders and a history of trauma, making this sample different from those previously studied.
We found that retention was higher among women with mandated treatment and those
with a moderate number of prior treatment episodes. No studies with co-occurring samples were
found that reported on the effect of legal pressure or number of prior treatment episodes on
17
Treatment Retention
treatment retention. Previous studies on retention among women have reported that women who
are mandated to treatment via the criminal justice system (especially by child protection services
and the welfare system) remain in treatment longer (Nishimoto & Roberts, 2001). However, as
Nishimoto & Roberts (2001) state, the literature on the effect of legal pressure on retention is
mixed, with some studies showing better retention for those who have been coerced into
treatment and others showing negative outcomes related to coercion. Regarding previous
substance abuse treatment, women in this study with no previous substance abuse treatment
episodes had shorter treatment stays. To be eligible for the current study, women had to have at
least two previous treatment episodes in either the mental health or substance abuse system.
Women with no previous substance abuse treatment may have been those with a more substantial
psychiatric treatment history; this suggests that mental health burden may be a risk factor for
earlier drop out, a finding consistent with those of previous studies. On the other hand, we also
found that women with many previous substance abuse treatment episodes were at greater risk of
drop out. They may reflect a group of women with many failed treatment efforts, a finding that is
also consistent with those of previous studies.
The importance of treatment retention is shown by the finding that longer retention was
associated with better 6-month outcomes on three of the four key symptom measures for women
with more severe baseline symptoms. This finding is consistent with those reported from
previous studies (NIDA, 1997; Simpson et al, 1997, Simpson et al., 1999; Simpson, 1979;
Simpson, 1981). Somewhat puzzling to us is why the effect of length of stay on 6-month
outcome was not stronger for the intervention group than the comparison, which we expected.
In the case of mental health symptoms, we saw some evidence of more improvement among
intervention-group women than comparison, for those women with worse baseline status.
18
Treatment Retention
However our results also suggested further questions .For example, given an average length of
stay, it seemed that improvement for the women in the intervention group was not as dependent
on baseline status as it was for women in the comparison group. At the same time, for mental
health and trauma symptoms, our results also suggested a possibly non-linear relationship
between length of stay and improvement, with intervention group women who stayed in
treatment for moderate lengths of time showing less improvement than expected (graphical
analysis, not shown). We need to explore the mechanisms underlying how length of stay affects
outcome and whether there may be critical time points during treatment that impact on treatment
outcomes.
The finding that participation in the trauma specific group did not explain the lack of
significance of the effect of length of stay on the trauma symptomatology might be explained by
the fact that the trauma groups were only one part of the intervention, which included a ‘package’
of trauma-relevant services that all women were exposed to and that permeated the broader
treatment milieu. Future studies are needed to assess the relative contribution of various
treatment components in this intervention.
An interesting question that arises from the positive effect on length of stay on mental
health symptoms and the concomitant lack of effect on trauma symptomatology is whether
mental health and trauma specific symptoms follow a different or similar course of improvement.
For example, it is possible that general mental health symptoms may be more immediately
responsive to the intervention while changes in trauma symptoms may take longer to manifest or
be mediated through initial improvements in mental health. These questions need to be
addressed in future studies.
19
Treatment Retention
Although this is one of the largest intervention studies to date on women with co-
occurring mental health and substance abuse disorders and trauma, there are a number of study
limitations that warrant consideration. The sample may not be completely representative of
women with co-occurring disorders and a history of exposure to trauma because it was drawn
from only two cities and featured publicly funded treatment programs. Women in publicly
funded treatment programs may represent those with the most severe co-occurring disorders,
conditions that interfere with social function and the ability to have adequate private insurance
coverage. There was no random assignment, and there were baseline differences in the sample.
Although we adjusted for these factors in statistical analyses, there may have been other
unmeasured differences. There may also be bias introduced with reported drug and alcohol use
because underreporting is common. Also, a high proportion of cases were censored because of
differences in length of treatment programs. We used drop out as a proxy for unsuccessful
treatment rather than a direct measure of unsuccessful treatment. Still, it is virtually always the
case that individuals who left treatment before the 4-month mark left with unsatisfactory
progress. Furthermore, due to sample size limitations, we were not able to control for site
although the specific interventions used differed. Statistical models exploring how length of stay
influenced outcomes assumed linear effects and might not detect other kinds of influences.
Finally, secondary analysis of data reduces power.
In summary, women with co-occurring disorders and trauma histories respond to trauma-
enhanced residential substance abuse treatment, regardless of problem severity, by remaining in
treatment longer. Substance abuse and mental health symptoms improve with increased length
of stay in treatment in interaction with more severe baseline symptoms. These findings lend
support to a growing consensus that for persons with co-occurring disorders, an integrated
20
Treatment Retention
treatment approach is the most appropriate (Barrowclough et al., 2001; Hellerstein et al., 2001;
Ouimette et al., 1998; RachBeisel et al., 1999; Ziedonis & Stern, 2001). The findings have
important implications for substance abuse treatment because treatment retention and outcomes
among women with co-occurring disorders have generally been poor, even in long-term
residential treatment programs.
21
Treatment Retention
References
Amaro H., Nieves R., Johannes S.W., & Cabeza N.M.L. (1999). Substance abuse treatment:
Critical issues and challenges in the treatment of Latina women. Hispanic Journal of
Behavioral Sciences, 21(3), 266-282.
Andersen S., & Berg J.E. (2001). The use of a sense of coherence test to predict drop-out and
mortality after residential treatment of substance abuse. Addiction Research and Theory,
9(3), 239-251.
Arfken, C.L., Klein, C., di Menza, S., & Schuster, S.R. (2001). Gender differences in problem
severity at assessment and treatment retention. Journal of Substance Abuse Treatment,
20(1), 53-57.
Barrowclough, C., Haddock, G., Tarrier, N., Lewis, S.W., Moring, J., O'Brien, R., Schofield, N.,
& McGovern, J. (2001). Randomized controlled trial of motivational interviewing,
cognitive behavior therapy, and family intervention for patients with comorbid
schizophrenia and substance use disorders. American Journal of Psychiatry, 158(10), 1706-
1713.
Battjes, R.J., Onken, L.S., & Delany, P.J. (1999). Drug abuse treatment entry and engagement:
Report of a meeting on treatment readiness. Journal of Clinical Psychology, 55(5), 643.
Becker, M. A., Noether, C., Larson, M. J., Gatz, M., Brown, V., Heckman, J., & Giard, J. (in
submission). Characteristics of women engaged in treatment for trauma and co-occurring
disorders: Findings from a national multi-site study. Journal of Health Services Research
and Policy.
Bride, B.E. (2001). Single-gender treatment of substance abuse: Effect on treatment retention
and completion. Social Work Research, 25(4), 223-232.
22
Treatment Retention
Broome, K. M., Flynn, P. M., & Simpson, D. D. (1999). Psychiatric comorbidity measures as
predictors of retention in drug abuse treatment programs. Health Services Research, 34(3),
791-806.
Brown, V.B., Huba, G.J., & Melchio, L.A. (1995). Level of burden: Women with more than one
co-occurring disorder. Journal of Psychoactive Drugs, 27(4), 339-346.
Brown, V.B., Melchior, L.A., & Huba, G.J. (1999). Level of burden among women diagnosed
with severe mental illness and substance abuse. Journal of Psychoactive Drugs, 31(1), 30-
40.
Claus, R.E., & Kindleberger L.R. (2002). Engaging substance abusers after centralized
assessment: Predictors of treatment entry and dropout. Journal of Psychoactive Drugs,
34(1), 25-31.
Cosden, M., & Cortez-Ison, E. (1999). Sexual abuse, parental bonding, social support, and
program retention for women in substance abuse treatment. Journal of Substance Abuse
Treatment, 16(2), 149-155.
Crits-Christoph, P., & Siqueland, L. (1996). Psychosocial treatment for drug abuse: Selected
review and recommendations for national health care (review). Archives of General
Psychiatry, 53(8), 749-756.
Dakof, G.A., Quille, T.J., Tejeda, M.J., Alberga, L.R., Bandstra, E., & Szapocznik, J. (2003).
Enrolling and retaining mothers of substance-exposed infants in drug abuse treatment.
Journal of Consulting and Clinical Psychology, 71(4), 764-772.
DeLeon, G. H. J., Jainchill, N., & Melnick, G. (2000). Therapeutic communities: Enhancing
retention in treatment using "Senior Professor" staff. Journal of Substance Abuse
Treatment, 19, 375-382.
23
Treatment Retention
Derogatis, L.R. and Melisaratos, N. (1983). The Brief Symptom Inventory: an introductory
report. Psychological Medicine, 13, 595-605.
Derogatis, L.R. (1993). Brief symptom inventory: Administration, scoring, and procedures
manual. (4th ed.). Minneapolis MN: NCS Pearson, Inc.
Donovan, D. M., Rosengren, D.B., Downey, L., Cox, G.C., & Sloan, K.L. (2001). Attrition
prevention with individuals awaiting publicly funded drug treatment. Addiction, 96, 1149-
1160.
Egelko, S., & Galanter, M. (1998). Impact of social anxiety in a "therapeutic community."
Oriented cocaine treatment clinic. American Journal on Addictions, 7(2), 136-141.
Festinger, D.S., Lamb, R.J., Kountz, M.R., Kirby, K.C., & Marlowe, D. (1995). Pretreatment
dropout as a function of treatment delay and client variables. Addictive Behaviors, 20(1),
111-115.
Foa, E.B., Cashman, L., Jaycox, L., & Perry, K. (1997). The validation of a self-report measure
of posttraumatic stress disorder: The Posttraumatic Diagnostic Scale. Psychological
Assessment, 9, 445-451.
Giard, J., Nadlicki, T., Hennigan, K., Vogel, W., Rinehart, D., Huntington, N., Veysey, B., &
Mazelis, R. (in press). Development and implementation of the cross-site evaluation of the
Women, Co-occurring Disorders and Violence Study. Journal of Community Psychology.
Greenberg, W.M., Otero, J., & Villanueva, L. (1994). Irregular discharges from a dual diagnosis
unit. American Journal of Drug and Alcohol Abuse, 20(3), 355-371.
Grella, C. E., Hser, Y., Joshi, V., & Anglin, M. D. (1999). Patient histories, retention, and
outcome models for younger and older adults in DATOS. Drug and Alcohol Dependence,
57, 151-166.
24
Treatment Retention
Haller, D. K. J., Elswick R.K., Dawson K.S., & Schnoll, S.H. (1997). Perinatal substance
abusers: Factors influencing treatment retention. Journal of Substance Abuse Treatment
14(6), 513-519.
Haller, D.L., Miles, D.R., & Dawson, K.S. (2002). Psychopathology influences treatment
retention among drug-dependent women. Journal of Substance Abuse Treatment, 23, 431-
436.
Harris, M. (1998). T.R.E.M. Trauma Recovery and Empowerment: A Clinician's Guide
for Working with Women in Groups. New York: The Free Press.
Hellerstein, D.J., Rosenthal, R.N., & Miner, C.R. (2001). Integrating services for schizophrenia
and substance abuse. Psychiatric Quarterly, 72(4), 291-306.
Hien, D.A., Nunes, E., Levin, F.R., & Fraser, D. (2000). Posttraumatic stress disorder and short-
term outcome in early methadone treatment. Journal of Substance Abuse Treatment, 19(1),
31-37.
Hosmer, D.W.Jr., & Lemeshow, S. (1999) Applied survival analysis: Regression modeling of
time to event data. John Wiley and Sons, Inc. NY.
Howard, J., & Beckwith, L. (1996). Issues in subject recruitment and retention with pregnant and
parenting substance-abusing women. In E.R. Rahdert (Ed.), Treatment for Drug-Exposed
Women and Their Children: Advances in Research Methodology (pp. 68-86). NIDA
Research Monograph 166. Rockville MD: National Institute on Drug Abuse.
Hughes, P. C. S., Neri, R.L., Urmann, C.F., Stahl, S., Sicilian, D.M., & Anthony, J.C. (1995).
Retaining cocaine-abusing women in therapeutic community: The effect of a child live-in
program. American Journal of Public Health, 85(8), 1149-1152.
25
Treatment Retention
Huntington, N., Moses, D.J., & Veysey, B. (in press). Developing and implementing a
comprehensive approach to serving women with co-occurring disorders and histories of
trauma. Journal of Community Psychology.
Joe, G.W., Simpson, D.D., Greener, J.M., & Rowan-Szal G.A. (1999). Integrative modeling of
client engagement and outcomes during the first 6 months of methadone treatment.
Addictive Behaviors, 24(5), 649-659.
Kang S.Y., Deren S., & Goldstein M.F. (2002). Relationships between childhood abuse and
neglect experience and HIV risk behaviors among methadone treatment drop-outs. Child
Abuse and Neglect, 26(12), 1275-1289.
Kelly, P.J., Blacksin, B., & Mason, E. (2001). Factors affecting substance abuse treatment
completion for women. Issues in Mental Health Nursing, 22,287-304.
King, A.C., & Canada, S.A. (2004). Client-related predictors of early treatment drop-out in a
substance abuse clinic exclusively employing individual therapy. Journal of Substance
Abuse Treatment, 26(3), 189-195.
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(l), 1-18
Lewis R.A., Haller D.L., Branch D., & Ingersol K.S. (1996). Retention issues involving drug-
abusing women in treatment research. In E.R. Rahdert (Ed.) Treatment for Drug-Exposed
Women and Their Children: Advances in Research Methodology (pp. 110-122). NIDA
Research Monograph 166. Rockville MD: National Institute on Drug Abuse.
26
Treatment Retention
Lundy, A., Gottheil, E., Serota, R.D., Weinstein, S.P., & Sterling R.C. (1995). Gender differences
and similarities in African-American crack cocaine abusers. Journal of Nervous and
Mental Disease, 183(4), 260-266.
Martinez-Raga, J., Keaney, F., Marshall, E.J., Ball, D., Best, D., & Strang, J. (2002). Unplanned
versus planned discharges from in-patient alcohol detoxification: retrospective analyses of
470 first-episode admissions. Alcohol and Alcoholism, 37(3), 277-281.
McHugo, G., Caspi, Y., Kammerer, N., Mazelis, R., Jackson, E., Russell, L., et al. (In press).
Trauma Measures Qualitative Work Group: The assessment of trauma history in women with
co-occurring mental health and substance use disorders and a history of interpersonal
violence. Journal of Behavioral Health Services & Research.
McHugo, G.J., Kammerer, N., Jackson, E., Markoff, L., Gatz, M., Larson, M.J., Mazelis, R., &
Hennigan, K. (in press). Women, Co-Occurring Disorders, and Violence Study: Evaluation
design and study population. Journal of Substance Abuse Treatment.
McLellan, A.T., Luborsky, L., Woody, G.E., & O'Brien, C.P. (1980). An improved diagnostic
instrument for substance abuse patients: The Addiction Severity Index. Journal of Nervous
and Mental Disorders, 168, 26-33.
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(10),
1525-1533.
Milligan, C.O., Nich, C., & Carroll, K.M. (2004). Ethnic Differences in Substance Abuse
Treatment Retention, Compliance, and Outcome from Two Clinical Trials. Psychiatric
Services, 55,167-173.
27
Treatment Retention
Najavits, L.M., & Weiss, R.D., & Shaw SR. (1997). The link between substance abuse and
posttraumatic stress disorder in women: A research review. American Journal on
Addictions, 6(4), 273-283.
Najavits, L.M., Weiss, R.D., Shaw, S.R., & Muenz, L.R. (1998). "Seeking safety": Outcome of a
new cognitive-behavioral psychotherapy for women with posttraumatic stress disorder and
substance dependence. Journal of Traumatic Stress. 11(3), 437-456.
National Institute on Drug Abuse. (1997). NIDA NOTES. Study sheds new light on the state of
drug abuse treatment nationwide. Vol 12(5), September/October. Retrieved May 15, 2004
from http://www.drugabuse.gov/NIDA_Notes/NNVol12N5/Study.html
Nelson-Zlupko, L., Dore, M.M., Kauffman, E., & Kaltenbach, K. (1996). Women in recovery:
Their perceptions of treatment effectiveness. Journal of Substance Abuse Treatment, 13(1),
51-59.
Nishimoto, R.H., & Roberts, A.C. (2001). Coercion and drug treatment for postpartum women.
American Journal of Drug and Alcohol Use, 21, 161-181.
Ouimette, P.C., Brown, P.J., & Najavits, LM. (1998). Course and treatment of patients with both
substance use and posttraumatic stress disorders. Addictive Behaviors, 23(6), 785-795.
RachBeisel, J., Scott, J., & Dixon, L. (1999). Co-occurring severe mental illness and substance
use disorders: A review of recent research. Psychiatric Services, 50(11), 1427-1434.
Roberts, A.C. & Nishimoto, R.H. (1996). Predicting treatment retention of women dependent of
cocaine. American Journal of Drug and Alcohol Abuse, 22(3), 313-333.
Ross, H.E., Cutler, M., & Sklar, S.M. (1997). Retention in substance abuse treatment: Role of
psychiatric symptom severity. American Journal on Addictions, 6(4), 293-303.
28
Treatment Retention
Rowan-Szal, G. A., Joe, G. W., & Simpson, D. D. (2000). Treatment retention of crack and
cocaine users in a national sample of long term residential clients. Addiction Research,
8(1), 51-64.
Rowe, C.L., Liddle, H.A., Greenbaum, P.E., & Henderson, C.E. (2004). Impact of psychiatric
comorbidity on treatment of adolescent drug abusers. Journal of Substance Abuse
Treatment, 26, 129-140.
Saxon, D.A, & Calsyn, A.J. (1995). Effects of psychiatric care for dual diagnosis patients treated
in a drug dependence clinic. American Journal of Drug and Alcohol Abuse, 21(3), 301-11.
Scott-Lennox, J., Rose, R., Bohlig, A., & Lennox R. (2000). The impact of women's family
status on completion of substance abuse treatment. Journal of Behavioral Health Services
& Research, 27(4), 366-379.
Scheffe, H. (1959). The analysis of variance. New York, John Wiley and Sons, Inc.
Simpson, D. D. (1979). The relation of time spent in drug abuse treatment to posttreatment
outcomes. American Journal of Psychiatry, 136, 1449-1453.
Simpson, D. D. (1981). Treatment for drug abuse: Follow-up outcomes and length of time spent.
Archives of General Psychiatry, 38(8), 875-880.
Simpson, D. D., Joe, G. W., & Brown, B. S. (1997). Treatment retention and follow-up outcomes
in the Drug Abuse Treatment Outcome Study. Psychology of Addictive Behaviors, 11(4),
294-307.
Simpson, D. D. & Brown, B. (Guest Eds.). (1999). Special issue: Treatment process and
outcome studies from DATOS. Drug and Alcohol Dependence, 57(2).
Strantz, I., & Welch, S (1995). Postpartum women in outpatient drug abuse treatment: Correlates
of retention/completion. Journal of Psychoactive Drugs, 27(4), 357-373.
29
Treatment Retention
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 & Misuse, 31(8), 1001-1013.
Thompson, M.P.., & Kingree, J.B. (1998). The frequency and impact of violent trauma among
pregnant substance abusers. Addictive Behaviors, 23(2), 257-262.
Veach, L. J., Remley, T.P.Jr., Kippers, S.M., & Sorg, J.D. (2000). Retention predictors related to
intensive outpatient programs for substance use disorders. American Journal of Drug and
Alcohol Abuse, 26(3).
Wolfe, J., Kimerling, R., Brown, P. J., Chrestman, K. R. & Levin, K. (1996). The Life-Stressor
Checklist–Revised. (In B. H. Stamm (Ed.), Instrumentation in stress, trauma, and
adaptation (pp. 198-200). Northbrook, IL: Sidran Press.
Ziedonis, D.M., & Stern, R. (2001). Dual recovery therapy for schizophrenia and substance
abuse. Psychiatric Annals, 31(4), 255-264.
30
Treatment Retention
Acknowledgments
This study was funded under Guidance for Applicants (GFA) No. TI 00-003 entitled Cooperative
Agreement to Study Women with Alcohol, Drug Abuse and Mental Health (ADM) Disorders Who
Have Histories of Violence: Phase II from the Department of Health and Human Services, Public
Health Service, Substance Abuse and Mental Health Services Administration’s three centers:
Center for Substance Abuse Treatment, Center for Mental Health Services, and Center for
Substance Abuse Prevention (U.S. Health and Human Services, March 2000). We would like to
acknowledge the contributions of members of the clinical and research teams and the study
participants, without whom this study would not have been possible.
31
Treatment Retention
Table 1
Baseline characteristics of women by treatment group (N=461)
Characteristic Intervention Comparison P valuea
N=198 N=263
Demographics
Employed prior to
treatment entry 0.0611
No 173 87.81% 214 81.36%
Yes 24 12.18% 49 18.63%
Individual income below
poverty line 0.1907
No 26 13.40% 46 17.96%
Yes 168 86.59% 210 82.03%
History of homelessness 0.0361*
Never homeless 46 23.46% 77 29.27%
Homeless in past but not
in last 6 months. 56 28.57% 49 18.63%
Homeless in last 6
months 94 47.95% 137 52.09%
Criminal Justice
Mandated treatment
<.0001 *
Voluntary treatment 81 40.90% 162 61.83%
Mandated, criminal
court 89 44.94% 60 22.90%
Mandated, civil 28 14.14% 40 15.26%
Jail experience <.0001 *
Never in jail 24 12.30% 64 24.33%
Past jail experience 44 22.56% 97 36.88%
In jail in last 6 months 127 65.12% 102 38.78%
Family Characteristics
Has partner/spouse 0.1618
No 133 67.17% 160 60.83%
Yes 65 32.82% 103 39.16%
Has children by age 0.6803
32
Treatment Retention
Characteristic Intervention Comparison P valuea
N=198 N=263
Never had children or
none living 29 14.64% 37 14.06%
Child/children < 18 yrs 153 77.27% 210 79.84%
Child/children >= 18 yrs 16 8.08% 16 6.08%
Has custody of children 0.4603
No 110 55.55% 137 52.09%
Yes 88 44.44% 126 47.90%
Ever lost parental rights 0.2620
No 154 78.57% 194 74.04%
Yes 42 21.42% 68 25.95%
Mental Health
Lifetime number of
psychiatric
hospitalizations
0.8170
None 133 67.17% 176 67.17%
125 12.62% 39 14.88%
2-5 29 14.64% 38 14.50%
More than 5 11 5.55% 9 3.43%
Global Severity Index
GSI M(SD) 198 1.13 (0.69) 263 1.20 (0.73) 0.2941
Prescription medication
in last 3 months for
mental health
problems 0.9078
No 128 65.64% 174 66.15%
Yes 67 34.35% 89 33.84%
Prescription medication
in last 3 months for
substance abuse (SA)
0.879
33
Treatment Retention
Characteristic Intervention Comparison P valuea
N=198 N=263
problems
No 188 96% 253 96%
Yes 8 4% 10 4%
Trauma
Lifetime Exposure to
Stressful Events
(LESE) M(SD) 198 16.0 (4.7) 263 15.8 (4.3) 0.6021
Post Traumatic Stress
Disorder Symptom
Scale (PSS), Total
Score, M(SD) 189 20.8 (10.8) 254 20.9 (12.1) 0.9631
History of physical or
sexual child abuse 0.3887
No 56 28.28% 65 24.71%
Yes 142 71.71% 198 75.28%
History of physical or
sexual abuse as adult 0.3526
No 30 15.15% 32 12.16%
Yes 168 84.84% 231 87.83%
Alcohol and Drug
Addiction
No. of substance abuse
treatment starts .0043 *
None 40 20.51% 24 9.30%
1-2 52 26.66% 63 24.41%
3-4 33 16.92% 60 23.25%
5 or more 70 35.89% 111 43.02%
Addiction Severity
Index (ASI), Alcohol
198 0.02 (0,0.30) 263 0.04 (0,0.51) 0.080
34
Treatment Retention
Characteristic Intervention Comparison P valuea
N=198 N=263
composite score,
Median (Interquartile
range)
Addiction Severity
Index (ASI), Drug
composite score,
Median (Interquartile
range) 196 0.21(0.04,0.33) 262 0.25(0.15,0.34) 0.056
Study Enrollment
No. days from
admission to baseline
interview M(SD) 198 17.8 (10.9) 263 22 (11.3) <.0001 *
% Completed SA TX
Program **
Yes 46 23.23% 87 33.08% <0.05*
No 152 76.77% 176 66.92%
35
Treatment Retention
Characteristic Intervention Comparison P valuea
N=198 N=263
Note. a P-values are for two-sample, two sided t-test except for the following variables:
Chi-square tests: Race/Ethnicity, Employed, Income below poverty line, Living situation,
Mandated to treatment, Jail experience, Has partner, Has children, Has custody of children, Ever
lost parental rights, Prescription medication, History of child abuse, History of adult abuse.
Wilcoxon nonparametric rank sum tests, two sided: Lifetime history of psychiatric
hospitalizations, ASI, substance abuse treatment history.
* p < 0.05. ** Programs were not matched for length of stay in the control and intervention
groups, different lengths of stay were required for treatment completion in each group. More
participants in the intervention group were in longer programs. Women in the intervention group
were less likely to complete (OR=0.6122; 95% CI=0.403, 0.930).
36
Treatment Retention
Table 2
Cox Proportional Hazards Model of Time in Treatment, controlling for client characteristics that
differ at baseline between treatment groups (N=448, 53% censored)
Variable DF
Parameter
Estimate
Standard
Error
Chi-
Square
Pr >
ChiSq
Hazard
Ratio
Treatment group
(Intervention)
1 -0.37009 0.16056 5.3127 0.0212 * 0.691
Mandated treatment 1 -0.43366 0.16127 7.2305 0.0072 * 0.648
Jail experience (not in last
6 months)
1 -0.04216 0.18907 0.0497 0.8235 0.959
Recent jail experience
(within last 6 months)
1 -0.30076 0.19968 2.2686 0.1320 0.740
Homeless (not in last 6
months)
1 -0.03479 0.19631 0.0314 0.8593 0.966
Homeless (within last 6
months)
1 -0.16954 0.16511 1.0544 0.3045 0.844
No previous substance
abuse treatment starts
1 0.89817 0.18615 23.2807 <.0001 * 2.455
One to two treatment
starts
1 -0.38063 0.19111 3.9668 0.0464 * 0.683
3-4 treatment starts 1 -0.43659 0.20410 4.5755 0.0324 * 0.646
Days, admission to
baseline interview
1 -0.04219 0.00733 33.1421 <.0001 * 0.959
Per 7 days + -0.29533 0.05131 * 0.7443
Note. In the proportional hazards regression model, referents are Treatment, the comparison group; Living
situation, Never homeless; Jail experience, None; Substance abuse treatment starts, 5 or more.
Overall test of simultaneous equality to 0 for model effects: Jail , p=0.224; Homelessness, p=0.538;
Substance abuse treatment starts, p<0.0001 *
* p < 0.05. + 95% CI for Hazard ratio for each additional week between admission and baseline interview:
37
Treatment Retention
(0.673,0.823) indicates the effect is statistically significant, p < 0.05.
38
Table 3a
Effect of time in treatment on improvement in mental health outcomes [as measure by GSI] from
baseline to six months
Parameter Estimate Standard
Error
p-value
for t-test
p-value for F-test
(if different) b
GSI, RSQ= 0.2677, N=322
Length of Stay (days) a 0.0001 0.0022 0.9672 0.0519
Treatment group (Intervention) 0.8751 0.4825 0.0707 na
Interaction treatment group X
Length of Stay (Intervention)
-0.0090 0.0044 0.0433 * na
Baseline GSI 0.2779 0.1756 0.1146 0.6479
Interaction treatment group X
baseline GSI (Intervention)
-0.7178 0.3479 0.0399 * na
Interaction Length of Stay X
baseline GSI
0.0013 0.0017 0.4533 0.0024*
Interaction treatment group X
Length of Stay X baseline GSI
(Intervention)
0.0075 0.0032 0.0216 * na
39
Parameter Estimate Standard
Error
p-value
for t-test
p-value for F-test
(if different) b
Note. Tests of covariates: Living status (p =0.050) was marginally significant with F tests and significant with
t-tests (p =0.040 for parameter representing Homeless but not in last 6 months). These results suggest that
women who had been homeless but not in last 6 months had slightly less improvement (parameter estimate=
-0.193, SE=0.093). All other tests of covariates had p > 0.15 for t-tests, p > 0.21 for F tests.
A sensitivity analysis of these data after excluding values outside of the mean +/- 3 SD (n=319) showed
similar results with decreased significance for model effects. Only the length of stay*GSI interaction
parameter had p-value < 0.05 (F test). However the trend in the p-values and parameter estimates matched the
analysis of the full dataset.
a Length of stay is censored at 120 days (see text for explanation). bF-tests may differ from t-tests when a
continuous parameter is tested since the F-test tests the hypothesis that the average slope across treatment
groups is zero whereas the t-test tests the hypothesis that the slope of the parameter for the comparison
treatment group is zero.
40
Table 3b
Effect of time in treatment on improvement in trauma symptom outcomes from baseline to six
months
Parameter Estimate Standard
Error
p-value
for t-test
p-value for F-test
(if different) b
PSS, RSQ=0.2737, N=310
Length of Stay (days) a 0.0147 0.0374 0.6942 0.7517
Treatment group (Intervention) -1.4900 3.7658 0.6926 na
Interaction treatment group X
Length of Stay (Intervention)
-0.0053 0.0363 0.8840 na
Baseline PSS 0.2964 0.1536 0.0546 na
Interaction Length of Stay X
baseline PSS
0.0022 0.0015 0.1487 na
Note. Tests of covariates: All other tests of covariates had p > 0.13 for t-tests, p > 0.20 for F tests.
A sensitivity analysis of these data after excluding values outside of the mean +/- 3 SD (n=308) showed
similar results.
a Length of stay is censored at 120 days (see text for explanation). bF-tests may differ from t-tests when a
continuous parameter is tested since the F-test tests the hypothesis that the average slope across treatment
groups is zero whereas the t-test tests the hypothesis that the slope of the parameter for the comparison
treatment group is zero.
41
Table 3c
Effect of time in treatment on improvement in substance use outcomes from baseline to six
months
Parameter Estimate Standard
Error
p-value
for t-test
p-value for
F-test (if
different) b
ASI Drug, RSQ= 0.7340, N=319
Length of Stay (days) a-0.0001 0.0003 0.7223 0.3008
Treatment group (Intervention) 0.0525 0.0297 0.0787 na
Interaction Length of Stay X
Treatment group
(Intervention)
-0.0003 0.0003 0.2643 na
Baseline ASI Drug 0.5558 0.0998 <0.0001 * na
Interaction Length of Stay X
ASI Drug
0.0035 0.0010 0.0003 * na
Note. Tests of covariates: Being mandated to treatment was marginally significant (p=0.049), with women
who were mandated having reduced improvement (parameter estimate= -0.022, SE=0.011). All other tests
of covariates had p > 0.35 t-tests and p > 0.57 for F tests. There were no outlying observations more than 3
SD from the mean improvement.
a Length of stay is censored at 120 days (see text for explanation). bF-tests may differ from t-tests when a
continuous parameter is tested since the F-test tests the hypothesis that the average slope across treatment
groups is zero whereas the t-test tests the hypothesis that the slope of the parameter for the comparison
treatment group is zero.
42
Table 3d
Effect of time in treatment on improvement in alcohol use outcomes from baseline to six months
Parameter Estimate Standard
Error
p-value
for t-test
p-value for
F-test (if
different) b
ASI Alcohol, RSQ=0.7601, N=322
Days in treatment (Length
of Stay)a
-0.0001 0.0004 0.8090 0.7386
Treatment group
(Intervention)
-0.0077 0.0518 0.8821 na
Interaction Length of Stay
X Treatment group
(Intervention)
0.0004 0.0005 0.4490 na
Baseline ASI Alcohol 0.5881 .0732 <0.0001 * na
Interaction Length of Stay
X ASI Alcohol
0.0023 0.0007 0.0016 * na
Note for Tables 3a, 3b, 3c and 3d. Regression models controlled for variables with baseline differences
with p < 0.05: 1) Living situation (never homeless, homeless but not in last 6 months, homeless in last 6
months); 2) voluntary or mandated treatment program; 3) jail experience (never in jail, in jail but not in
last 6 months, in jail during last 6 months); 4) Lifetime substance abuse treatment history (none, 1-2
starts, 3-4 starts, 5 or more starts); 5) days between admission to substance abuse program and baseline
interview. Models included dummy dichotomized variables for each possible level of a response, with
none as the reference for living situation, jail experience and with 5 or more starts as the reference for
substance abuse treatment history. Voluntary treatment is the reference. The comparison treatment group
is the reference.
T-tests were marginally significant for the category of one-two SA treatment starts (p=0.063), with these
43
Parameter Estimate Standard
Error
p-value
for t-test
p-value for
F-test (if
different) b
women having slightly more improvement compared to the other categories (parameter estimate=0.037;
SE=0.020). There were no other statistically significant relationships between covariates and symptom
improvement for the ASI alcohol measure (p>0.108 for t-tests and p > 0.135 for F tests). There were no
outlying observations more than 3 SD from the mean improvement.
The basic model included these variables plus the interaction between length of stay and treatment group,
which tested a relationship of interest, and all other two- and three-way interactions among treatment
group, symptom level at baseline, and length of stay. Except for the interaction between treatment group
and length of stay, which was retained in the final models, higher level interactions were excluded one by
one until only interactions with p < 0.15 were included in the final model.
a Length of stay is censored at 120 days (see text for explanation). bF-tests may differ from t-tests when a
continuous parameter is tested since the F-test tests the hypothesis that the average slope across treatment
groups is zero whereas the t-test tests the hypothesis that the slope of the parameter for the comparison
treatment group is zero.
* p < 0.05; General linear model. na; Where p-value for F test result is the same as for the t-test.
44
Figure 1. Probability of Remaining in Treatment
45
