Antidepressant Adherence Patterns over Time Among Adults with Depression — an Application of Group Based Trajectory Modeling Using Commercial Claims Data from 2007 — 2020

Background: Antidepressant adherence patterns are variable and may be affected by substance use disorders (SUDs). Few studies use time-varying methods to address the association between SUD and variable patterns of antidepressant treatment adherence.

Objectives: To identify the association between SUD and antidepressant adherence trajectory subgroups.

Methods: A new user cohort of US adults 18-64 years-old was derived from a 10% sample of IQVIA PharMetrics Plus for Academics commercial claims database. New users had at least 1 antidepressant anytime 01/2007 through 06/2020 with no evidence of use in the prior 6 months (i.e., baseline). The date of first prescription claim was the index date. Inclusion criteria were: 1) evidence of a depression diagnosis at the index date or in the prior 6 months and 2) continuous medical and prescription benefits in the 6 months before and after the index date. Group-based trajectory modeling (GBTM) identified antidepressant adherence subgroups. The model incorporated a biweekly continuous outcome measure —the proportion of days covered (PDC)— over a 6-month period after the index date (i.e., follow-up) and a censored normal distribution for the outcome. Bayesian Information Criterion (BIC) and log Bayes Factor were used to select the number of trajectory subgroups. We used multinomial logistic regression to compare baseline demographics (e.g., age, sex) and clinical covariates (e.g., SUD, mental health) across trajectory subgroups.

Results: Among the 82,408 new antidepressant users, 5 adherence trajectories emerged: 1) early decline (n=17,471; 21%), 2) delayed decline (n=11,124; 14%), 3) gradual decline (n=8,819; 11%), 4) decline-to-increase (n=9,029; 11%), and 5) persistent high (n=35,965; 44%). Mean PDC during follow-up varied from 16%-89% across trajectory subgroups. Compared to persistent high, those in the early decline group were more likely to be younger and less likely to have a major depressive disorder diagnosis or use psychotherapy. Early decline, delayed decline, and decline-to-increase had a higher proportion of adults with SUD relative to the other groups. SUD was significantly associated with all groups (early decline [OR:1.6, 95%CI: 1.5,1.7], delayed decline [OR:1.5, 95%CI: 1.4,1.6], gradual decline [OR:1.4, 95%CI: 1.3,1.5]) and decline-to-increase [OR:1.5, 95%CI: 1.4, 1.6] relative to the persistent high group.

Conclusions: Groups that demonstrated a decline in antidepressant adherence had a 40-60% greater odds of including individuals with SUD. Efforts to improve antidepressant adherence should address co-existing SUD.