Responding to the Rofecoxib Withdrawal Crisis: A New Model for Notifying Patients at Risk and Their Health Care Providers

Context The rofecoxib withdrawal on 30 September 2004 was one of the largest drug recalls in modern history. Contribution This article describes how a single institution used its electronic medical record system to immediately identify 11 699 patients with rofecoxib prescriptions. Within 1 hour of the announcement of the national withdrawal, the institution pulled rofecoxib from its medication stocks. Within 7 hours, it deactivated prescriptions in the electronic record and notified providers via e-mail. Within 22 hours, letters about the withdrawal were sent to patients. Cautions The authors did not measure the effects of the notification on patient and provider behavior and outcomes. The Editors On the morning of 30 September 2004, the U.S. Food and Drug Administration (FDA) issued a Public Health Advisory about the withdrawal of rofecoxib (1). This worldwide withdrawal by Merck & Co., Inc., was based on data from the Adenomatous Polyp Prevention on Vioxx (APPROVe) trial that suggested an increased risk for cardiovascular events in patients receiving rofecoxib, particularly those who had been taking the drug for longer than 18 months (1, 2). At the time of the withdrawal, approximately 2 million people were receiving rofecoxib nationwide, making it the largest prescription drug withdrawal in U.S. history (3). The withdrawal occurred at the pharmacy level, meaning that pharmacists were required to cease dispensing this agent and remove remaining stock from inventories (4). As in previous drug withdrawals from the market, both the manufacturer and the FDA undertook major initiatives to educate the general public as well as health care providers (1, 5). No federal laws govern the process of drug withdrawal (6, 7). Furthermore, the manner and timeliness of transmitting drug withdrawal information to pharmacists, providers, and patients vary (8-10). Reports have described persistent use of medications even after their withdrawal from the market (11, 12). Consequently, in addition to the drug manufacturer, the FDA, and the dispensing pharmacies, drug withdrawal notification may also become a responsibility of the institution providing the patient's care (6, 13, 14; Morrison CM, McMains MB. Legal necessity for early nationwide notification of pharmacies after drug recalls. Presented at the 143rd Annual Meeting of the American Pharmaceutical Association, 11 March 1996, Nashville, Tennessee). Fortunately, health information technology has now made it feasible for health care institutions to successfully identify and inform distinct patients affected by a drug withdrawal (9, 15). We decided to use the electronic medical record (EMR), variably referred to as a computerized patient record (CPR), electronic patient record (EPR), or electronic health record (EHR), to identify patients with an active rofecoxib prescription, notify them of the withdrawal, and notify their providers (16, 17). The Cleveland Clinic Foundation Institutional Review Board deemed the study protocol exempt from review. Methods Technology At the Cleveland Clinic Foundation, nearly all health care providers use a commercial EMR (Epic Systems Corp., Madison, Wisconsin) for clinical documentation, order entry, and prescriptions (18, 19). We queried our EMR clinical data warehouse (Oracle Database server, Oracle Corp., Redwood Shores, California) to identify patients with an active rofecoxib prescription. We extracted demographic characteristics, e-mail address, diagnosis history, primary and prescribing provider, and prescription dosage and duration for each patient into a data file. Communication to the Patients Our institution prepared a letter template that outlined the specifics of the withdrawal and instructed patients to stop rofecoxib and contact their health care providers for guidance. We also generated a mail-merge file of patient names and addresses consistent with our institutional privacy policy and the Health Insurance Portability and Accountability Act (HIPAA) (20). After the letters were generated, mailroom personnel used an automated system to stuff and stamp envelopes before delivery to the U.S. Postal Service for mailing via regular post. We also sent our Internet-based shared EMR users an e-mail similar to the letter and placed a warning and link of the FDA withdrawal on their Web site login page (21). Communication to the Health Care Providers We sent an e-mail (Microsoft Outlook, Microsoft Corp., Redmond, Washington) to our institution's health care providers that outlined the steps being taken by the Cleveland Clinic Foundation to notify patients about the rofecoxib withdrawal. We also informed the providers that they would shortly receive a list of their own patients who may be affected by the withdrawal. We also deactivated rofecoxib within our EMR medication formulary and created an automated computer alert triggered when a provider opened the electronic chart of a patient with an active rofecoxib prescription. Finally, we placed a warning and a link of the FDA withdrawal on our institutional Web site. Clinical Data and Analysis To identify comorbid conditions, we used International Classification of Diseases, 9th Revision, diagnosis codes entered by physicians during patient encounters: peptic ulcer disease (531534.9), bleeding peptic ulcer (531.0/2/4/6, 532.0/2/4/6, 533.0/2/4/6, 534.0/2/4/6), ischemic heart disease (410414.99), stroke (433.xx437.xx, excluding 435.9), transient ischemic attack (435.9), hypertension (401.xx), and congestive heart failure (428.xx). To identify the subset of patients at a potentially higher relative risk for developing cardiovascular events from rofecoxib, we categorized patients as either aspirin indicated or aspirin not indicated on the basis of the criteria for secondary cardiovascular prophylaxis used in the Vioxx Gastrointestinal Outcomes Research (VIGOR) trial (that is, history of myocardial infarction, angina, cerebrovascular accident, transient ischemic attack, angioplasty, or coronary bypass grafting) (22). We also queried the EMR data warehouse to determine the number of triggered clinical alerts and the number of rofecoxib prescriptions discontinued since the withdrawal. Results Patient Characteristics Table 1 shows characteristics of the identified patients. Of the 11699 patients with an active prescription for rofecoxib, most were receiving 25 mg (n= 10318), the strength used in the APPROVe trial. Duration data were available for 11613 patients; of these, 5203 (45%) had had prescriptions for more than 18 months and 1051 (9%) had a prescription dating back more than 3 years. Of note, 297 (2.5%) patients had a history of peptic ulcer disease and 867 (7.4%) patients had a documented history of ischemic heart disease. In addition, 1207 (10.3%) patients were deemed aspirin indicated according to criteria for secondary cardiovascular prophylaxis. This finding suggests that these patients were at higher risk for a cardiovascular event due to rofecoxib. Table 1. Patient and Health Care Provider Characteristics Provider Characteristics We identified 842 providers at our institution who had prescribed rofecoxib to the identified patient population. The most frequent prescribing providers were primary care providers (internists, family practitioners, and general pediatricians) (Table 1). The number of patients identified per prescribing provider varied from 1 to 122 (average, 14). The most common diagnoses that providers linked to their prescription were osteoarthritis, joint pain, and rheumatoid arthritis. Notification Timings Table 2 chronicles our response to the rofecoxib withdrawal. Approximately 1 hour after the withdrawal was announced, the Cleveland Clinic Foundation's Drug Information Center informed the various department heads, including the Chief Pharmacy Officer and the Chief Information Officer, via e-mail. A multidisciplinary ad hoc team of clinicians, pharmacists, administrators, and informaticians devised an action plan for identifying and notifying patients and providers affected by the withdrawal. Within 7 hours, we deactivated rofecoxib from our EMR medication formulary and created the automated computer alert. By 7 a.m. the next morning, we sent letters to our patients via regular post. Finally, hardcopy letters identifying a provider's list of patients with an active rofecoxib prescription were sent on 1 October via interoffice mail. Table 2. Timeline of Cleveland Clinic Foundation's Response to Rofecoxib Withdrawal Impact As of 18 November 2004, clinical alerts triggered warnings for 4511 (38.6%) distinct patients during office, telephone, and refill encounters. Of these patients, 3717 (82.4%) had their prescription discontinued by the provider during that encounter. Eight letters were returned unopened and 2 were returned with replies indicating that the patient had died. Discussion With the health care environment becoming more complex and specialized, the utility of EMR technology to facilitate communication about important health care issues is critical. This report describes an immediate response to a major drug withdrawal using EMR patient data in combination with readily available, automated communication tools. We identified and notified 11699 patients receiving rofecoxib who were actively managed in our EMR. We also identified 842 prescribing providers within a short time after the FDA Public Health Advisory and provided them a list of their patients. Our response should not be viewed as a limited answer to a rare problem but rather as a generalizable model. The rofecoxib withdrawal is not an isolated crisis; it reflects an ever-growing problem in modern health care (23). For example, between 1997 and 1998, drug withdrawals affected nearly 20 million people (24, 25). With more and more drugs entering the market, there is a concern that this trend may increase (23). Historically, it appears that the FDA and the drug manufacturers shoulder the burden of managing