We thank Dr. Golden and colleagues for their response to our recent article. They brought up several interesting issues, but we believe the critical issue for patient safety is identified by Dr. Krishnamurthy in his online letter (1). The estimated 59 000 emergency department visits each year caused by adverse events from the use of insulin, warfarin, and digoxin are harms that should be addressed. We stand by our conclusions that performance measures and interventions targeting the use of warfarin, insulin, and digoxin could prevent more emergency department visits for adverse events than measures targeting the use of "potentially inappropriate" medications.

Dr. Golden and colleagues suggest that by relying on the diagnoses of emergency department physicians, we probably did not identify some adverse events—such as falls for which a sedating antihistamine might have been a contributing factor—that may have been identified if we had performed complete chart reviews. We indeed acknowledged the limitations that adverse events "diagnosed and treated in other settings (for example, in primary care offices, in urgent care centers, or during hospitalizations) or not treated in any health care facility were not included" and that our surveillance methods were "probably less sensitive than [those of] research studies involving chart review" (2). The data we presented were not based on epidemiologic associations, but rather on counts of physician diagnoses and treatments rendered. Exhaustively reviewing charts to identify antihistamine use in a patient who has fallen and then designating the antihistamine as a contributing factor post hoc is problematic. In fact, a recent review of prescribing for the elderly (3) concluded that the evidence that potentially inappropriate prescriptions are associated with adverse patient outcomes is "mixed and contradictory."

Performance measures and interventions to improve medication safety for older patients are important. These measures should be based on scientific soundness, feasibility, and relevance, particularly the ability to improve measurable patient outcomes (4). The estimated 5.3% of emergency department visits attributed to medications that are potentially inappropriate in certain circumstances and the fraction of emergency department visits from antiplatelet agents among patients with gastric ulcers or concomitant anticoagulation represent important subsets of patients. The Beers criteria provide useful guidance for optimizing medication selection and may identify safety issues for further study. But with limited resources available for medication safety, national public health surveillance data on numbers of and risks for emergency department visits provide useful information to help focus and prioritize safety efforts and measure impact in the future.