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

ABSTRACT

Background: 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.

Objective: 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.

Methods: 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.

Results: 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.

Conclusions: 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.

KEYWORDS:

• Substance use disorder
• psychiatric comorbidity
• latent class analysis
• time-varying effect models

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

Samuel W. Stull was supported by the National Science Foundation [grant DGE1255832], Stephanie T. Lanza was supported by the National Institute on Drug Abuse [grant P50 No. DA039838], and Ashley N. Linden-Carmichael was supported by the National Institute on Alcohol Abuse and Alcoholism [grant K01 No. AA026854]. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation, National Institute on Drug Abuse, or National Institute on Alcohol Abuse and Alcoholism.