Overview. This study utilized baseline data from a hybrid type 1 trial examining the implementation and effectiveness of a peer-led healthy lifestyle program for overweight/obese (BMI ≥25) clients with SMI in three supportive housing agencies . Supportive housing is an important service sector for people with SMI because it combines community-based housing with a range of services addressing clients’ physical and mental health, substance use treatment, and community integration needs .
Sample. Trained research assistants (RAs) conducted in-person screenings to assess client eligibility: supportive housing resident, aged 18 years or older, English or Spanish speaking, diagnosed with an SMI (e.g., schizophrenia, schizoaffective disorder, mood disorders), overweight/obese (i.e., BMI ≥ 25), and willing to obtain medical clearance if randomized to the intervention. Exclusionary criteria included failing a capacity to consent questionnaire , posing a danger to self or others, receiving detoxification services for alcohol/drug abuse, reporting conditions contraindicated with participating in a weight loss intervention (e.g., recent cardiac event)  (See  for a complete list of medical contraindications), and for those 65 years or older, screening positive on the Mini-Cog Cognitive Impairment screener [35, 36]. Baseline interviews were conducted by RAs, lasted 1.5 hours on average, and participants were reimbursed $25 for their time. All measures used in the interviews had previously been published and all participants provided written informed consent. All study procedures were approved by the [University] Institutional Review Board and the [City] Department of Public Health IRB.
Participant Characteristics. Participants were asked to self-report demographic (e.g., sex, race/ethnicity, age, history of homelessness) and clinical characteristics, such as lifetime physician diagnosis of SMI and drug/alcohol abuse or dependence, as well as names of medications they were currently prescribed. Antipsychotic medication was categorized as first or second-generation antipsychotic (SGA).
CVD Risk Factors. Obesity was defined as a BMI of >30.0 kg/m2 which was calculated using participants’ height and weight ((weight [lbs] * 703) / height(in2)), as measured by the RAs using wall tape and digital scales. Having high blood pressure, diabetes, and high cholesterol were defined as participants’ self-report of a lifetime physician diagnosis of these conditions, as indicated by “Yes” answers to a series of questions asking “Has a doctor ever told you that you have…” for each condition [37, 38]. Smoking was defined by self-report of current smoking, whether daily or not. Dichotomous variables were created to indicate the presence (yes or no) of each of these five risk factors. A sum score reflecting the total number of co-occurring risk factors for each participant (ranging from 0: overweight only to 5: all risk factors present) was calculated [25, 39].
Health-Related Quality of Life. The Physical and Mental Health Composite Scores (PCS & MCS) from the 12-item Short Form Health Survey (SF-12; ) were used to assess participants’ physical and mental HRQoL. Scores range from 0 to 100, with higher scores indicating better health. The SF-12 has demonstrated reliability among populations with mental illness and co-occurring medical conditions , as well as validity among those who have experienced both homelessness and mental illness .
Behavioral and Psychiatric Functioning. The Behavior and Symptom Identification Scale (BASIS-24) consists of 24 items that ask participants to rate the frequency or amount of difficulty they experienced with a range of psychiatric symptoms and substance use issues in the past week . It consists of six sub-scales comprised of 3 to 6 items assessing depression/functioning, relationships/interpersonal competence, psychotic symptoms, alcohol/drug use, emotional lability, and self-harm. Responses are rated on a 5-point likert scale ranging from 0 (no difficulty/none of the time/never) to 4 (extreme difficulty/all of the time/always). An algorithm using a weighted average generates an overall scale score ranging from 0 to 2.83 as well as subscale scores. The BASIS-24 has demonstrated excellent external reliability and fair to excellent internal consistency and validity , including among racial/ethnic minorities .
Health Locus of Control. The Multidimensional Health Locus of Control (MHLC) scale consists of 24 questions that yield scores along four sub-scales which assess the degree to which individuals believe that internal and/or external factors (Chance, Powerful Others, God) contribute to their health. Sample questions include “If I take care of myself, I can avoid illness” and “Most things that affect my health happen to me by accident.” Participants rated their degree of agreement with each item from 1 (strongly disagree) to 6 (strongly agree) . Responses are summed for each sub-scale to produce scores ranging from 6 to 24, with higher scores indicating greater endorsement of that dimension. The MHLC has demonstrated adequate reliability and validity in measuring locus of control beliefs, though statistical associations with related constructs tend to vary .
Statistical Analysis. Analyses utilized baseline data from participants enrolled in the clinical trial irrespective of subsequent random assignment to intervention or usual care. Univariate analyses were used to describe the distribution of individual risk factors and the cumulative number of CVD risk factors. Bivariate analyses (e.g., Student’s t-tests, chi-squares) explored relationships between participant characteristics and each risk factor. Considering the number of CVD risk factors as a count variable, Poisson regression with robust standard errors was used to explore its correlates. No problem of overdispersion was found using a likelihood ratio test of over-dispersion parameter alpha by running the same regression model using a negative binomial distribution (Long and Freese, 2004). Finally, ordinary least squares regression was used to explore the correlates of PCS and MCS. Selection of potential correlates for regression analyses was informed by findings from existing research examining CVD risk factors and/or HRQoL among people with SMI (e.g. Newcomer and Hennekens, 2007; Lim and Lee, 2018; Neil et al., 2018). Variables were included in regression models if they were associated with the respective dependent variable (CVD risk factors, PCS, MCS) in bivariate analyses with a p-value of <0.1 (results available upon request). Model diagnostics including tests for normality, homoscedasticity, multicollinearity, outliers and influential cases indicated none of the model assumptions were violated (Kutner et al., 2004). Participants with missing data on any variables in the multivariate models were excluded. Eventually all models had less than 10% missing data, with 9.9%, 4.8%, and 3.8% for each multivariate model. A two-sided p-value of 0.05 or less was used to indicate statistical significance. All analyses were performed in Stata 15.0 (StataCorp LP, College Station, TX).