Methods: Data were collected from a sample of homeless adults (Baseline, N=421) across three timepoints in their initial year in PSH (3-months, 6-months, 12-months). At each wave, a structured interview gathered participants’ demographics (i.e. age, race, gender, education), inpatient/outpatient psychiatric service utilization, physical health conditions, and positive past-month screening for a psychiatric disability (measured by the Modified-Colorado Symptom Index [MCSI]) and probable post-traumatic stress disorder (measured by the PC-PTSD). Social network interviews asked participants to nominate past 3-month individuals they interacted with (network size), as well as collected information about nominated individuals including roles they served (i.e. romantic partner, relative, doctor, emotional health counselor, case manager, neighbor), their proximity to participants, and support they provided (i.e. instrumental, tangible, emotional, conflict). Generalized linear mixed models assessed: 1) Changes in MCSI and PTSD symptoms over time (controlling for demographics, services, and health conditions), and 2) Social network characteristics associated with changes. To differentiate cross-sectional and longitudinal effects of time-varying social network correlates, models included within-person means and mean deviations at each time point.
Results: At baseline, 50% of the sample met criteria for probable PTSD and 65% for a psychiatric disability. Compared to baseline, residents were less likely to screen positive on the MCSI and PC-PTSD across all follow-ups. Positive screens on the MCSI continuously decreased over time (56%, 54%, 50%) while probable PTSD declined in the initial 3-months (40%) with marginal decreases at remaining follow-ups (39%, 38%). Differences across time remained significant in controlled models. Gaining a romantic partner was associated with a longitudinal increase in a positive MCSI screening (within-subject effect; OR=1.59; 95% CI: 1.08-2.33). Cross-sectionally (between-subject effect), emotional health counselors were associated with an increased likelihood in positive MCSI (OR=16.07; 95% CI: 5.86-44.05) and PC-PTSD (OR=4.46; 95% CI: 2.35-8.46) screenings, while doctors were protective against positive MCSI (OR=.29; 95% CI: 0.11-0.77) and PC-PTSD (OR=0.41; 95% CI: 0.21-0.78) screenings, and case managers were protective against positive MCSI screenings (OR=0.49; 95% CI: 0.26-0.94). Conflicting social network members were positively associated with MCSI screenings (OR=3.11; 95% CI: 2.00-4.84) and PC-PTSD (OR=1.87; 95% CI: 1.43-2.45), cross-sectionally.
Conclusions/Implications: These unique findings lend evidence to the impact of PSH on homeless persons’ health; specifically, housing may facilitate positive changes in residents’ mental health symptomatology. Despite changes, there remains a persistent number of positive screenings. Social network interventions that target residents’ interpersonal exchanges and build social support may facilitate further mental health improvements.