Long-run Effects of Medication Risk Warnings: Antidepressant Risk Warnings for Youth

Background: In October 2004, the Food and Drug Administration (FDA) directed pharmaceutical companies to issue black-box warning labels on antidepressants about the potential link between the use of antidepressants and suicidal ideation among youth.

Objectives:  The objective of this paper is to analyze the long-run trends in antidepressant use among youth after the black-box warning. There has been a decline in antidepressant use among youth right after the warning in 2004, however, the antidepressant use started to increase in 2007 after the initial response to the warning faded away. We also explore whether the long run usage trends are different between the severe and non-severe populations.

Data: For our analysis, we use youth aged 5-17 from the Medical Expenditure Panel Survey (MEPS), covering years 2002-2009. MEPS is a nationally representative survey of the non-institutionalized US population. Final sample includes 52,442 observations.  All rates and model estimates are weighted to account for sample design and non-response. The analysis is performed using three different samples: all kids aged 5-17, severe kids and non-severe kids. Mental health severity is determined by Columbia Impairment Scale (CIS), where a score of CIS>16 is categorized as severe. The dependent variable of interest is any antidepressant use at a given year. 
Methods: We compare the pre-warning years (2002-03) to the warning years (2004-07) and  post warning years (2008-09) using multivariate logit models, and survey weights to account for the complex sampling design. We use recycled predictions methods to estimate the regression adjusted yearly predicted probability of use, and 500 bootstraps to estimate the standard errors of the predicted probabilities.

Preliminary Findings: There is a significant decline in the probability of using any antidepressants during the warning years compared to pre-warning years. This decline in the probability of use is driven by the decreased rates of use in the non-severe population. In the post warning years, however, the probability of use is not significantly different than the pre-warning years for all three samples of youth, showing that the initial impact of the warning faded away with time. For the non-severe youth population, the rates of use increase up to the pre warning levels four years after the warning (1.12% in 2003, 1.09% in 2008).