Mental disorders pattern in staff of a military unit in Iran: the role of metabolic syndrome on latent class membership

Abstract

Introduction Mental disorders are among the most prevalent health problems of the adult population in the world. This study aimed to identify the subgroups of staff based on mental disorders and assess the independent role of metabolic syndrome (MetS) on the membership of participants in each latent class. Methods This cross-sectional study was conducted among 694 staff of a military unit in Tehran in 2017. All staff of this military unit was invited to participate in this study. The collected data included demographic characteristics, anthropometric measures, blood pressure, biochemical parameters, and mental disorders. We performed latent class analysis using a procedure for latent class analysis (PROC LCA) in SAS to identify class membership of mental disorders using Symptom Checklist-90. Results Three latent classes were identified as healthy (92.7%), mild (4.9%), and severe (2.4%) mental disorders. Having higher age significantly decreased the odds of belonging to the mild class (adjusted OR (aOR = 0.21; 95% confidence interval (CI): 0.05–0.83) compared to the healthy class. Also, obesity decreased the odds of membership in mild class (aOR = 0.10, 95% CI: 0.01–0.92) compared to healthy class. On the other hand, being female increased the odds of being in severe class (aOR = 9.76; 95% CI: 1.35–70.65) class in comparison to healthy class. Conclusion This study revealed that 7.3% of staff fell under mild and severe classes. Considering educational workshops in the workplace about mental disorders could be effective in enhancing staff’s knowledge of these disorders. Also, treatment of comorbid mental disorders may help reduce their prevalence and comorbidity.

Patterns of Substance Use During Early Pregnancy and Associations With Behavioral Health Characteristics

“…Using the fitted LCA model, each individual in the study sample is then probabilistically assigned to a class based on their observed data, resulting in subgroups of individuals who are most similar to each other and different from individuals in the other subgroups. [24][25][26] We fit the LCA model 3 times varying the number of classes from 2 to 4 and included only the substance use variables in the models. To select the best model among the three models, we compared the likelihood-ratio statistic G 2 , the Akaike information criteria (AIC), the Bayesian information criterion (BIC), and Entropy.…”

“…In addition to considering model fit statistics, we also considered the interpretability of the models. 23,25,26 After identifying the substance use clusters using our final LCA model, we calculated the prevalence of sociodemographic and behavioral health characteristics in each substance use cluster. We then used modified Poisson regression to calculated adjusted prevalence ratios (aPRs) and corresponding 95% confidence intervals to statistically test for differences in prevalence of behavioral health characteristics (modeled as the outcome) between the clusters (modeled as the exposure) while adjusting for maternal age, neighborhood deprivation index (reference = 1st quartile), race/ethnicity (reference = non-Hispanic White), and parity (reference = 1 or more).…”

“…The LCA model parameters include the latent class membership probabilities and the parameters that relate the latent classes to the observed variables; the parameters are estimated by maximum likelihood. Using the fitted LCA model, each individual in the study sample is then probabilistically assigned to a class based on their observed data, resulting in subgroups of individuals who are most similar to each other and different from individuals in the other subgroups 24–26 . We fit the LCA model 3 times varying the number of classes from 2 to 4 and included only the substance use variables in the models.…”

“…Lower G 2 , AIC, BIC, and log-likelihood values and higher Entropy values indicate better model fit. In addition to considering model fit statistics, we also considered the interpretability of the models 23,25,26 …”