Predicting persistence to antidepressant treatment in administrative claims data: Considering the influence of refill delays and prior persistence on other medications.

backgroundMany patients with major depressive disorder (MDD) who begin antidepressant treatment discontinue use before for six months, the recommended minimum treatment length. This study sought to identify predictors of six-month antidepressant persistence including predictors utilizing patients' electronic prescription records.

methodsCommercially insured children (3-17 years) and adults (18-64 years) with MDD who initiated antidepressant treatment, 1/1/2003-2/28/2010, were assessed for six-month persistence (based on prescriptions' days supply, allowing a 30-day grace period). Antidepressant persistence prediction models were developed separately for children and adults. Two additional measures, days without medication between the first and second antidepressant fill (children and adults) and prior persistence on other medications (adults only), were added to the models, concordance (c) statistics were compared and risk reclassification evaluated.

resultsAmong children (n=8837 children) and adults (n=47,495) with MDD, six-month antidepressant persistence was low and varied by age (37%, 18-24 years to 52%, 3-12 and 50-64 years). Independent baseline predictors of persistence were identified, with model c-statistics: children=0.582, adults=0.584. Patients with more days without medication between fills were less likely to be persistent (10-30 vs. 0 days, children: RR=0.72, adults: RR=0.74), as were adults not previously persistent to other medications (RR=0.73). LIMITATIONS: The definition of six-month persistence is dependent on correct days supply values and the grace period utilized; potential predictors were limited to measures available in claims data.

conclusionsSix-month antidepressant persistence was low and overall prediction of persistence was poor; however, days without medication between fills and prior persistence on other medications marginally improved the ability to predict antidepressant persistence.