Sampling and recruiting
Survey data from PWID living in San Francisco and Los Angeles were collected for the Change the Cycle (CTC) study, a United States National Institute on Drug Abuse-funded randomized controlled trial performed between 2016 and 2017 to assess the efficacy of a behavioral intervention on injection initiation among PWID. Participants were recruited using targeted sampling of participants from previously identified street, community, and program sites, a method developed by Watters and Biernacki to sample populations who are difficult-to-reach with traditional sampling methods [22,23,24]. Targeted sampling involves the synthesis of secondary data, such as drug treatment and arrest data, and direct community observation (ethnography) to map out target neighborhood blocks for recruitment. Based on estimates of relevant population size of people who inject drugs, a representative number of individuals are recruited from each area by an outreach worker familiar with the neighborhoods and population. Inclusion criteria for this study required participants to be 18 years of age or older, to self-report injecting drugs in the past 30 days, and to have the evidence of recent drug use confirmed by staff inspection . The interviews were performed at private field sites leased by RTI and USC, which were easily accessible by foot or public transport. After providing written informed consent, survey participants completed a computer-assisted personal interview (Questionnaire Development System, Nova Research, Bethesda, MD) in which trained interviewers spent 40–60 min reading questions aloud and recording the answers on a laptop computer. Participants were reimbursed $15 for their time and participation. The six-month follow-up survey for CTC introduced questions concerning housing and relocation, thus it was the dataset for this cross-sectional analysis. The last six-month interview was collected on June 10th, 2018 and the analysis was performed from June 2018 to October 2019. The CTC intervention conducted at baseline included no content that would influence the exposures or outcomes of this analysis. The baseline sample for the CTC study included 979 individuals, of whom 601 participated in the 6-month follow-up survey (316 in San Francisco and 285 in Los Angeles) which included questions about housing status and residential relocation. All protocols were approved by the Institutional Review Board at the University of Southern California.
The main exposure variable was if study participants experienced residential relocation in the last 30 days. We asked each participant “In the last 30 days, how many times did you sleep in a different place or location (same type of place but different location)?” We created both dichotomized (those who moved in the past 30 days compared to those did not move in the past 30 days) and categorical distributions of our relocation variable. The categorical distribution was created by generating approximate quartiles based on frequency of residential relocation, yielding the following move categories: did not relocate, relocated< 3 times, relocated 3–9 times, and relocated >9 times in past 30 days.
The main outcome variables included exposure to violence, health outcomes, health behaviors, criminal legal system involvement, and access to services. Participants’ exposure to violence was assessed by asking whether they had their belongings stolen (“In the past 6 months, have any of your belongings been stolen?”), experienced physical assault (“In the last 6 months has anybody punched, slapped, kicked, or physically hurt you?”), experienced weaponized assault (“In the past 6 months, has anybody used a knife, gun, club, or other weapon against you?”), or had experienced sexual assault (“In the past 6 months, has somebody used physical force or threats to make you have vaginal sex, anal sex, or oral sex with them?”). The responses to these items were all coded as binary (yes vs. no).
We assessed participants’ health outcomes, health behaviors, and access to services by asking about overdose, injecting with syringes used by others, severe food insecurity, and access to substance use treatment. For overdose, we asked participants “In the last 6 months, have you overdosed?” and coded participant responses as binary (yes vs. no). For injecting with syringes used by others, we asked “In the last 6 months, how many times did you inject using syringes/needles that you know had been used by someone else (including a close friend or lover)?” Responses were coded as binary (whether participants had any injecting with used syringes vs. no injecting with used syringes in the past 6 months). To assess food insecurity during the prior 30 days we used a 10-question scale utilized by Schmitz et al. that consisted of questions about skipping meals, losing weight because of inability to access food, and concern about access to food [26, 27]. Participants were assigned one point for each food insecurity question they endorsed. Participants with 0 to 5 points were designated as not severely food insecure, while participants with 6 to 10 points were designated as severely food insecure. We assessed participants’ access to substance use treatment by asking the following question: “In the last 6 months, have you participated in any type of substance use treatment program (including methadone or alcohol treatment, but excluding NA, AA, or other self- help programs)?” Participant responses were coded as binary (yes vs. no).
Criminal-legal system involvement was assessed by asking participants if they had been arrested (“Have you been arrested in the last 6 months?”), or if they had been incarcerated (“In the last 6 months, have you been held overnight in jail?”). Both outcomes were operationalized as binary (yes vs. no).
We also included several demographic variables as potential confounding factors. These included gender (cismale, cisfemale, transgender, or other), sexual orientation (heterosexual, gay/lesbian, or bisexual), race/ethnicity (White, Latinx, Black, Asian/Pacific Islander, Native American, Mixed Race/Other) , homelessness (If respondents considered themselves to be homeless or unstably housed, or not), education (less than high school diploma or high school diploma or more), age (18–29, 30–39, 40–49, or 50 or older), relationship status (single, in a relationship but not living as married, or married/living as married), and monthly income (Less than $1,400, or $1,400 or more).
Descriptive statistics, including means and frequencies, were generated for all study variables. We analyzed the relationship between our relocation variable and our three aggregate outcome variables which were grouped a priori: exposure to violence, health (including health outcomes, health behaviors, and access to services), and involvement with the criminal legal system. Bivariate analyzes consisting of odds ratios and χ2 tests were conducted to assess associations between the relocation variable and outcome variables. Potential confounding by demographic variables (eg self-reported race/ethnicity, gender, income, age) was assessed with two-way χ2 tests (p<0.05) for each outcome variable. Variables found to be associated with both the explanatory variable (residential relocation) and outcome variables were included as potential covariates in multivariable logistic regression models generated for each outcome variable. Covariates not statistically significantly associated with the outcome in the models at p < 0.05 were removed in the final models. Data was analyzed in R, version 3.5.3 .