2.1. Study design
Data utilized in this paper were from baseline interviews conducted as part of a randomized controlled trial conducted between 2016 and 2018 (for parent study details, see [40]). In the current study, we did not focus on intervention effects; therefore, we controlled for group assignment in the analysis. The parent study was a parallel-group trial (NCT02977988) designed to compare retention days between women randomly assigned to one of two study conditions during residential SUD treatment: (a) a mindfulness-based intervention and (b) education regarding the neurobiology of addiction, which served as the control group. Baseline interviews occurred prior to randomization and intervention delivery. Randomization into the control and intervention groups occurred via a simple randomization procedure. All participants received SUD treatment as usual in a therapeutic community setting without affecting standard level of care typically provided to patients [39, 40]. Treatment retention data for the study were abstracted from treatment site records upon patients’ discharge from the program and were based on treatment entry date and the date of discharge. The three types of referral status (CJ, CPS, and nonmandated) were likewise obtained from clinical records.
2.2. Study site
The study site was a publicly-funded residential SUD treatment program for women diagnosed with a SUD located in Southern California. The program provides interagency collaborative support services for women referred through the CJ and CPS systems, along with women-specific SUD treatment for those that are self-referred. Alongside SUD treatment, integrated services include mental health counseling, trauma-informed care, and services to those with health problems such as HIV/AIDS. While clients were able to remain in treatment for up to 12 months, the average length of stay was 4 months. The residential SUD treatment program provides dedicated women’s groups, family education and therapy, trauma-informed care, case management, nutritional education and support, and health and wellness programs. The program further provides integrated childcare, resources for children, medical and dental treatment for both women and children, GED courses, an Early Head Start program funded by the federal government, and a preschool program funded by the school district.
2.3. Participants and procedures
All participants (N = 245) were adult women (aged 18–65) diagnosed with a SUD, alcohol use disorder (AUD), or both and who spoke English. Women were mandated by the CJ and CPS systems or nonmandated (i.e., self-referred) into residential treatment. Women could leave the program at any time, although those mandated by the CJ or CPS systems could face legal or child custody consequences if the full range of treatment mandated by the CJ or CPS systems was not completed. The facility intake counselor identified clients who were eligible and informed the women about the study. The women who assented were contacted by a study interviewer who made appointments with prospective participants, conducted the informed consent and HIPAA process, and performed the baseline assessment interview. Trained research interviewers using computer-assisted interview procedures collected data from all study participants [39, 40].
2.4. Inclusion and exclusion criteria
Eligibility criteria were as follows: participant housed at SUD program treatment site, female, adult (18–65 years old), able to speak fluent English, diagnosed with a SUD, and signed an informed consent form agreeing to participate in research. Exclusion criteria were as follows: inability to comprehend or sign informed consent to participate due to language barrier, severe cognitive impairment, untreated psychotic disorder, severe mental health condition, past-30-day suicidality based on clinical assessment, currently incarcerated, participating in other research, > 6 months pregnant, > 65 years of age, and not willing to sign HIPAA form or be audio recorded during interviews and intervention sessions [39, 40].
2.5. Data sources and measures
2.5.1. Demographic, clinical, referral, and retention data
Baseline demographic information was obtained via in-person interview and included age, race and ethnicity, parental status, housing status, education, and employment status. Admitted patients met one-on-one with a clinician coordinator who assessed for mental health diagnoses, substance or alcohol use disorders, and suicidality using the DSM-5 [41]. Nonmandated patients self-referred at their own initiative or as a result of a recommendation from a nonmandating provider, family member, or friend. For women who were mandated, legal and child custody status was monitored by CJ and CPS systems throughout treatment with updates in addition to coordinated and integrated interagency support and case management services.
2.5.2. Self-report measures: Depression, anxiety, stress, and post traumatic stress symptomatologies
Psychological symptoms were measured via the Depression, Anxiety and Stress Scale (DASS-21) and the Posttraumatic Symptom Scale-Interview (PSS-I) [42, 43]. The DASS-21 features three self-report subscales designed to measure the emotional states of depression (7 items), anxiety (7 items), and stress (7 items) on a 4-point Likert scale. The PSS-I was utilized to assess frequency of posttraumatic symptom severity in the last 30 days and includes re-experiencing (5 items), avoidance (7 items), and arousal (5 items) on a 4-point Likert scale. Participants replied regarding any traumatic event experienced, not an explicit traumatic event as the traditional instructions recommend.
2.6. Data analysis
Descriptive characteristics of pretreatment factors were examined for the three referral groups (CJ mandated, CPS mandated, and nonmandated). Utilizing t-tests and analysis of variance for comparison of means for continuous variables and chi-square tests for comparison of percentages with categorical variables, we compared sociodemographic and pretreatment characteristics among the referral groups. If significant differences between groups were found, pairwise comparisons (i.e., CPS versus CJ, CPS versus nonmandated, CJ versus nonmandated, and mandated versus nonmandated) were conducted. For continuous variables, Tukey’s honestly significant difference method was used to test all possible pairwise differences of means at the same time among the three referral sources (CJ, CPS, and nonmandated). For categorical variables, chi-square tests were conducted for each mandating agency and nonmandated group. If the chi-square p-value was less than .017 (i.e., alpha .05 divided by 3) based on number of group comparisons, the difference was considered statistically significant.
All study variables were tested for confirmation in meeting core model assumptions of normality, linearity, and homoscedasticity. A log10 transformation was conducted for the dependent variable, number of days in treatment, to reduce skewness. Additionally, a square-root transformation was conducted for independent variables of age, posttraumatic stress scores, and stress scores to reduce skewness and improve normality.
To address Hypotheses 1, 2, and 3, a multiple linear regression model-building method was used to determine if mandated (CJ or CPS) or nonmandated referral status predicted increased days in treatment. The primary independent variable of interest was referral status, and number of days in treatment was the outcome variable. Univariate analyses were conducted to determine potential associations between being CJ mandated, CPS mandated, or nonmandated and the number of days in treatment.
Next, potential covariates and control variables were evaluated to determine if there was a greater than 10% change in the effect estimate of the mandating variable (CJ or CPS) or nonmandating variable and the number of days in treatment. Based on prior work, potential covariates and control variables tested for possible inclusion in the final models were (a) age, race and ethnicity, being a mother, housing, education, and employment status and (b) number of mental health diagnoses, posttraumatic stress symptomology, depression, anxiety and stress; and type of SUD diagnoses. Control variables determined to be retained in final models were age, stress (symptomolgy), and number of mental health diagnoses.
Addressing Hypothesis 4, all models were examined for stress, having one or more mental health diagnoses, and their association with increased days in treatment among the three treatment referral groups. For all final multiple linear regression models, parameter estimates and corresponding p-values are presented. Percentages representing log10 back-transformed parameter estimates are further presented for clarity of interpretation. All statistical analyses were conducted using SAS version 9.4.