Methods: The analytic sample included respondents reporting past-year homelessness (N = 704) from the third wave of the National Epidemiological Survey on Alcohol and Related Conditions (NESARC-III). Twenty-four personality traits were preselected as indicator variables for the LCA, corresponding with DSM-5 borderline, schizotypal, and antisocial PD. Bayesian Information Criterion, Akaike’s Information Criterion, log likelihood, and bootstrapped likelihood ratio test (BLRT) were examined to aid in determining the best fitting model. One- to five-latent class models were estimated. Model selection was informed by prior theory and the interpretability of class solutions. Multinomial regression was conducted to test differences in class membership by behavioral health characteristics, specifically lifetime mood and anxiety disorders, PTSD, AUD, SUD, and suicide attempt. Analyses were weighted to account for the NESARC-III’s complex sampling design.
Results: The four-class solution was optimal. The largest class (35.44%) is characterized by high levels of all personality traits relative to other classes, especially paranoia and suspiciousness, unusual behaviors or appearance, intense and unstable relationships, impulsivity, norm violations and anger, and irritability. The second-largest class (26.51%) was also characterized by high levels of impulsivity, risk taking behavior, norm violations, but had substantially lower traits that typically characterize BPD and SPD. The third-largest class is characterized by minimal personality impairment (24.40%). The final class is characterized by relational instability and identity diffusion (13.65%). The uniformly severe class had statistically significantly higher risk of all psychiatric conditions relative to the minimal impairment class. The class characterized by impulsivity and norm violations had significantly higher risk of AUD (RRR = 2.38, 95%CI: 1.40, 4.05) and SUD (RRR = 6.55, 95%CI: 2.71, 15.82). The group characterized by identity diffusion and relational instability had higher risk of mood (RRR = 3.85, 95%CI: 1.76, 8.46) and anxiety (RRR = 4.90, 95%CI: 1.69, 14.23) disorders as well as lifetime suicide attempt (RRR = 3.10, 95%CI: 1.15, 8.35).
Conclusions and Implications: The present study explored subtypes of personality traits among a representative sample of adults reporting past-year homelessness, finding four distinct classes with differential relationships to several behavioral health conditions. PEH may benefit from intervention to promote healthy personality functioning. Clinicians and homeless service workers should consider assessment of personality traits and functioning to effectively engage potential clients into treatment services.