Latent class profile of psychiatric symptoms and treatment utilization in a sample of patients with co-occurring disorders

Objective: To identify symptom-based subgroups within a sample of patients with co-occurring disorders (CODs) and to analyze intersubgroup differences in mental health services utilization. Methods: Two hundred and fifteen patients with COD from an addiction clinic completed the Symptom Checklist 90-Revised. Subgroups were determined using latent class profile analysis. Services utilization data were collected from electronic records during a 3-year span. Results: The five-class model obtained the best fit (Bayesian information criteria [BIC] = 3,546.95; adjusted BIC = 3,363.14; bootstrapped likelihood ratio test p < 0.0001). Differences between classes were quantitative, and groups were labeled according to severity: mild (26%), mild-moderate (28.8%), moderate (18.6%), moderate-severe (17.2%), and severe (9.3%). A significant time by class interaction was obtained (chi-square [χ2[15]] = 30.05, p = 0.012); mild (χ2[1] = 243.90, p < 0.05), mild-moderate (χ2[1] = 198.03, p < 0.05), and moderate (χ2[1] = 526.77, p < 0.05) classes displayed significantly higher treatment utilization. Conclusion: The classes with more symptom severity (moderate-severe and severe) displayed lower utilization of services across time when compared to participants belonging to less severe groups. However, as pairwise differences in treatment utilization between classes were not significant between every subgroup, future studies should determine whether subgroup membership predicts other treatment outcomes.

Latent class profile of psychiatric symptoms and treatment utilization in a sample of patients with co-occurring disorders.

OBJECTIVE: To identify symptom-based subgroups within a sample of patients with co-occurring disorders (CODs) and to analyze intersubgroup differences in mental health services utilization. METHODS: Two hundred and fifteen patients with COD from an addiction clinic completed the Symptom Checklist 90-Revised. Subgroups were determined using latent class profile analysis. Services utilization data were collected from electronic records during a 3-year span. RESULTS: The five-class model obtained the best fit (Bayesian information criteria [BIC] = 3,546.95; adjusted BIC = 3,363.14; bootstrapped likelihood ratio test p < 0.0001). Differences between classes were quantitative, and groups were labeled according to severity: mild (26%), mild-moderate (28.8%), moderate (18.6%), moderate-severe (17.2%), and severe (9.3%). A significant time by class interaction was obtained (chi-square [χ2[15]] = 30.05, p = 0.012); mild (χ2[1] = 243.90, p < 0.05), mild-moderate (χ2[1] = 198.03, p < 0.05), and moderate (χ2[1] = 526.77, p < 0.05) classes displayed significantly higher treatment utilization. CONCLUSION: The classes with more symptom severity (moderate-severe and severe) displayed lower utilization of services across time when compared to participants belonging to less severe groups. However, as pairwise differences in treatment utilization between classes were not significant between every subgroup, future studies should determine whether subgroup membership predicts other treatment outcomes.