Changes in Prescribing Practices Following the 2004 Pediatric Antidepressant Black Box Warning: Estimating the Influence of Non-Clinical Child Characteristics

The effect of the FDA’s 2004 pediatric antidepressant black box warning remains a source of debate. Some scholars argue that the black box warning scared providers into withdrawing beneficial treatment while others argue that the warning was the right response to a documented risk. Nonetheless, results from multiple studies show that the warning changed prescribing practices:  after 2004, antidepressant treatment among children fell. However, exactly how prescribing practices changed is less understood. Since antidepressant treatment among children in the early 21st century required that prescribers make decisions in situations of uncertainty, decision theory suggests that prescriber choices after the black box warning may have been shaped by non-clinical information, specifically child characteristics.

 This project uses seven years of data from the Medical Expenditure Panel Survey (MEPS), a nationally representative survey that documents U.S. health care expenditures, to analyze how antidepressant treatment changed in accordance with children’s clinical and non-clinical characteristics. Past-year antidepressant treatment was defined according to receipt of an antidepressant drug, children’s mental health was measured using the parent-reported Columbia Impairment Scale and clinical diagnosis, and non-clinical characteristics included child age, sex, race, ethnicity, household income quintile, insurance coverage, and geographic region. Descriptive univariate and bivariate statistics and pooled regression models (2001-2003 compared to 2005-2007) were generated, accounting for MEPS survey weights.

 Results show the prevalence of antidepressant treatment was similar across periods at approximately 1.3 percent of the population, and that the average age of recipients was 14 years. After 2004, the proportion of treated children rose among males and dropped among white non-Hispanic, middle income, and privately insured children. For example, treatment was approximately proportionally distributed by insurance coverage between 2001-2003, but publicly insured children were overrepresented after 2004, increasing their share from 31 to 45 percent. Perhaps most striking was the change in geographic representation. While the Northeast accounted for 30 percent of treated children before 2003, this dropped to less than 10 percent by 2005. Conversely, while the South accounted for 25 percent of treated children before 2003, it accounted for 43 percent after 2004. Further differences were measured within region.

 Multiple variable models replicated this pattern:  although white children were close to four times as likely to experience antidepressant treatment before 2003, controlling for mental health and all other covariates, this disproportionate likelihood decreased by half by 2005-2007. The likelihood of treatment did not differ by insurance status between 2001-2003, but publicly insured children were more likely to be treated after 2005. Finally, while Midwestern, Southern, and Western regions were associated with significantly lower likelihoods of antidepressant treatment than the Northeast before 2003, children in the Midwest and South were more than twice as likely to be treated after 2004.

 These results suggest that although the black box warning was not differentially targeted, the effects of warning may not have operated similarly across patient populations. The unexplained differences in prescription patterns, based on non-clinical characteristics, suggests the need for more comprehensive oversight of prescribing choices among child mental health practitioners.