Flexibly Modeling Age Trends in the Prevalence of Co-occurring Patterns of Substance Use and Mental Health Disorders Using Time-varying Effects and Latent Class Analysis

Substance use disorders (SUDs) and mental health disorders may change and co-occur in complex patterns across adult ages, but these processes can be difficult to capture with traditional statistical approaches. To elucidate disorder prevalence and comorbidities across adult ages by using time-varying effect models (TVEMs), latent class analysis (LCA), and modeling latent class prevalences as complex functions of age. Data were drawn from participants who are 18-65 years old in the National Epidemiologic Survey on Alcohol and Related Conditions III (n = 30,999; 51% women) and a subsample who reported a past-year post-traumatic stress disorder (PTSD), mood, anxiety, or SUD based on DSM-5 diagnoses (n = 11,279). TVEM and LCA were used to examine age trends and comorbidity patterns across ages. SUD prevalence peaked at age 23 (31%) and decreased thereafter, while mental health disorder prevalence was stable (20%-26% across all ages). The prevalence of five classes of individuals based on specific combinations of mental health and SUDs varied by age: the Alcohol Use Disorder class had the highest prevalence at age 26, whereas the Mood and Anxiety Disorder classes peaked around age 63. Interestingly, the Poly-Disorder class prevalence was greatest at age 18 but decreased sharply across young adulthood; however, the prevalence of the other high comorbidity class, PTSD with Mood or Anxiety Disorder, remained fairly constant across age, peaking at age 44. Multimorbid mental health disorders (excluding SUDs) persist in prevalence across adult ages. LCA, TVEM, and their integration together hold substantial potential to advance addiction research.