This cross-sectional study utilized baseline data from the Transitions Clinic Network (TCN, www.transitionsclinic.org), a multi-site prospective longitudinal cohort study of post-incarceration medical care.
Setting
The TCN is a national consortium of 24 primary care centers that serve the health needs of individuals returning from incarceration. The current study derives from 13 sites that participated in the longitudinal cohort study. Multi-disciplinary health care teams at each site include community health workers (CHWs) who have a history of incarceration and have been trained in health education, health system navigation, and motivational enhancement. CHWs link individuals released from prison or jail to primary care at TCN sites. Other features of TCN sites include: providers who have received training in best practices in caring for individuals with criminal justice involvement; ability to provide or refer for mental health and SUD services; and collaboration with social service providers, including housing, employment, and legal aid agencies. Individual TCN sites have been described in more detail [27, 28]. Many sites are part of integrated health systems with specialty SUD services, but most patients were referred to TCN sites to initiate primary care.
Participants
All new patients at TCN sites seen between May 2013 and February 2015 were screened for inclusion in the cohort study. Referrals of recently released individuals with chronic conditions came from three main sources: correctional agencies—specifically, prisons and parole and probation offices; community agencies, such as social service agencies and community-based organizations; and traditional sources such as other clinicians or self-referral from patients [29]. Inclusion criteria were: (1) recent release from prison (within 6 months); (2) presence of at least one chronic health condition warranting primary medical care, including SUD as a chronic health condition, or age equal to or greater than 50 years old; (3) ability to provide informed consent in English or Spanish; and (4) a plan to live in the area near the TCN program site for the duration of the study. Patients who planned to return to a previous primary care provider were excluded. All participants provided written informed consent, and data was protected by a certificate of confidentiality from the National Institutes of Health.
Data collection
We used data from baseline surveys for all participants of the TCN cohort study. Surveys were administered by trained research staff in person or via telephone. Data were stored in an online HIPAA-compliant portal and relevant clinical information was provided to primary care providers to facilitate medical care. Data included sociodemographic factors, self-reported incarceration history, past medical, mental health, and substance use history and treatment.
Measures
Substance use
Our primary outcome variable was any self-reported illicit substance use following release from incarceration. Use of each of the following substances post-release was assessed: heroin or other opioids, cocaine, cannabis, amphetamines, hallucinogens, MDMA, or illicit use of prescription medications. We assessed lifetime use, use since release, and frequency of use, but for this analysis, any use of any of these substances post-release was considered illicit substance use. A secondary outcome was any self-reported alcohol use assessed based on frequency and quantity of use post-release (i.e., the number of days per week and standard drinks per day when alcohol was consumed). Participants also self-reported whether they had ever been diagnosed with a SUD. For this analysis, we differentiated between presence of a drug use disorder (DUD) and alcohol use disorder (AUD).
Substance use disorder treatment
Participants who self-reported a DUD or AUD were also assessed for receipt of DUD and AUD treatment, respectively. Participants self-reported whether they received treatment while they were incarcerated and at the time of the survey. Participants indicated the type(s) of treatment they received by choosing from a list with the following options: Alcoholics Anonymous/Narcotics Anonymous (AA/NA) or self-help groups; pharmacotherapy; one-on-one counseling; or other, where participants could give free text responses. Participants were able to choose more than one type of treatment.
Psychiatric diagnoses
Participants were asked about prior psychiatric diagnoses, including SUDs. They also self-reported diagnoses of depression, bipolar disorder, post-traumatic stress disorder (PTSD) and schizophrenia. In addition to self-report, surveys included validated screening instruments for PTSD and Depression (Primary Care PTSD screen and the Patient Health Questionnaire) [30, 31].
Criminal justice history
Participants self-reported criminal justice involvement in several ways: time spent incarcerated during their most recent prison term, lifetime arrest and conviction counts, current parole/probation status, restricted incarceration status and the amount of time that had passed since release from incarceration.
Covariates
Other data collected included sociodemographic factors (age, binary gender, race/ethnicity, education, marital status), employment status and history (including employment, access to cash, benefits and other income sources), food security, housing security (concern for becoming homeless within 4 weeks), and housing type. The survey prompted participants to choose between eight different housing types, which we used to create four categories: unstable (street homeless; living in a shelter or single room occupancy hotel), institutional (drug treatment facility or other type of residential facility), “doubling-up” (staying with friends or family), and rent/own (renting or owning one’s own apartment or house).
Data analysis
First, we conducted descriptive statistics to assess the characteristics of the cohort. Next, we determined the proportion of participants reporting post-incarceration illicit substance use. Frequencies and proportions were assessed separately for each substance, and for the composite measure of any illicit substance use, which did not include alcohol use. Next, we built a multivariable logistic regression model with any illicit substance use as the outcome measure (dichotomous, yes/no). The entire sample (i.e., individuals with and without prior DUD or AUD) was included in the regression model. For model building, we explored factors associated in bivariate testing with post-release illicit substance use by using Chi square, student’s T test or Mann–Whitney tests. Covariates that were associated with post-release substance use (p < 0.10) were then included in the multivariable logistic regression model. After bivariate testing, the covariates that were included in the final regression model were: age, gender, race/ethnicity, housing type, time incarcerated at latest prison term, time to engagement with TCN site, parole status, depression, bipolar disorder, and DUD diagnosis. Finally, we performed sensitivity analyses to test the robustness of our multivariable regression model. In the first, we restricted the sample to only participants with a DUD diagnosis and repeated the modeling approach. Our goal was to determine whether factors associated with post-release illicit substance use differed between participants with and without a prior DUD diagnosis. In the second, we restricted the sample to only participants on parole and again repeated the modeling approach without parole status as an independent variable. Our goal was to determine whether overall substance use and associated covariates changed when excluding participants who were not monitored by parole.