Quality Management Using Data Analytics: An Application to Pharmaceutical Regulation

The U.S. government regulates consumer products through its various federal agencies. One such agency is the Food and Drug Administration (FDA) that governs the approval and safe public use of pharmaceutical products. If a drug is found unsafe, the FDA can issue a recall or a black box warning (BBW). This regulatory decision directly affects an operational decision: providers' production technology, affecting their treatment choices. Existing methods for monitoring drug safety are geared towards identifying unknown adverse drug reactions (ADRs) and suffer from several shortcomings such as reliance on limited data. There is a lack of data-driven approaches to evaluate a drug's association with a specific ADR. We propose a data-driven approach that fills this gap. We demonstrate the workings of our approach using a controversial BBW on a diabetes drug that warned prescribers of an increased risk of heart attack and cardiovascular mortality with the drug. Our findings, based on a large and comprehensive dataset, suggest that the drug was not harmful. On the contrary, we find that individuals who used the drug were less likely to die from cardiovascular complications or experience a heart attack. Our approach is robust to multiple specifications, avoids selection bias, and is complementary to existing drug surveillance systems. Further, our approach offers policymakers a decision support system to carefully assess drug safety in a real-world setting. Our approach can be extended to other consumer products that are subject to recalls and/or warnings such as toys, food, and automobiles.