Medication Use and Safety in the Pediatric Population: Real-World Evidence From Healthcare Databases

Abstract

Background: There is limited data to guide treatment decisions in children and adolescents. Real-world evidence from healthcare databases has the potential to fill this knowledge gap and inform medication use and safety in pediatric patients. Methods: We explored three ways that healthcare databases can be used to generate real-world evidence in the pediatric population. Healthcare utilization data were used to (1) describe the patterns of medication use in routine care-- in particular, antidiabetic medication use; (2) make causal inferences on medication safety-- we evaluated the safety of selective serotonin reuptake inhibitors (SSRIs) with respect to type 2 diabetes; and (3) build prediction models that can be developed into a tool for minimizing the biases that may occur in nonrandomized studies. Three nationwide US claims databases were used: the Medicaid Analytic eXtract (2000-2014, publicly insured), the IBM MarketScan database (2003-2017, privately insured), and the Optum© Clinformatics® Data Mart (2004-2019, privately insured). Results: In routine care, the use of non-insulin antidiabetic agents doubled from 2004 to 2019; however, treatment episodes tended to be short (79.6% of initiators experienced an early treatment interruption). Next, our nonrandomized study of medication safety found that children and adolescents initiating SSRIs may be at a small increased risk of developing type 2 diabetes (as-treated adjusted hazard ratios: 1.33; 95% CI: 1.21-1.47 among publicly insured; 1.10; 95% CI: 0.88-1.36 among privately insured, after careful control for potential biases). Finally, predictive models were used to develop a pediatric comorbidity index that provides a summary measure of disease burden. This index performed well in predicting the 1-year risk of hospitalization (c-statistic: 0.72, 95% CI: 0.71-0.72) as well as other outcome measures and can be used for risk adjustment in nonrandomized studies of children and adolescents. Conclusion: We demonstrate that healthcare databases can be leveraged to improve the understanding of medication use and safety in the pediatric population, and we propose a risk score that can be used to improve the validity of real-world evidence in this population.