Stimulants Concomitant with Other Psychotropic Classes: A Competing Risk Analysis in Medicaid-Insured Youth.

Objective: Pharmacoepidemiologic research shows increasing use of polypharmacy to manage behavioral treatment of youth. Methods to increase precision, for example, competing risk analysis, to capture psychotropic patterns of concomitant stimulant treatment changes over time have not been explored. Methods: A retrospective cohort study was derived from Medicaid enrollment data, prescription drug, and clinician-reported diagnosis claims data from 2007 to 2014. Youths aged 2-17 years with 1-7.5 years of continuous enrollment who were new users of stimulants were followed. Major outcomes include detailed changes of concomitant use according to the number of psychotropic classes (NPC); competing risk assessment of patient factors according to NPC; and time factors related to changes in NPC. Results: Among 30,294 new stimulant users, 75.5% remained on stimulant monotherapy and 24.5% had stimulant concomitant regimens. Among the latter, great flux was observed, revealing exposure to combinations changed substantially across time. As a proportion of all changes, retention of the maximum NPC was observed for 65.3% of 2 concomitant classes, 56.2% of 3 concomitant classes, and 57.1% and 56.2% of 4 and 5 concomitant classes, respectively. Median duration according to NPC showed a linear decrease in time from 223 days for 2 classes, 172 days for 3 classes, 141 days for 4 classes, and 113 days for 5 classes combinations. By contrast, the path to maximum NPC regimens took median times of 491-833 days as NPC increased from 2 to 4 concomitant classes. Competing risk analysis demonstrated significantly increased hazard ratios according to the number of concomitant classes for 12-17-year-olds, patients with foster care or disability coverage, and those with 3-4 years of continuous enrollment. Conclusions: Detailed NPC changes illustrate great flux in concomitant stimulant patterns among Medicaid-insured youth. Competing risk analysis brings more precise patient characteristics risk information to assess NPC changes compared with a binary model.