Warfarin is an oral anticoagulant used to prevent and treat thromboembolism, and its use has increased substantially over time, largely because of its indication for atrial fibrillation in an aging population [14,15]. Due to the variable dose-response relationship of warfarin among patients, a narrow therapeutic index, and the potential for numerous drug and dietary interactions, warfarin requires ongoing monitoring during therapy [14]. Monitoring the International Normalized Ratio (INR), a measure of warfarin's effect on clotting factors and the blood's propensity to clot, is essential for maintaining the drug within its narrow therapeutic window. Bleeding events are a significant cause of morbidity and mortality with warfarin therapy [7,15], and the risk of bleeding is largely influenced by the INR value [14]. Safety issues related to anticoagulant use have recently been reexamined by many healthcare systems in light of the Joint Commission's new anticoagulation National Patient Safety Goal 3E requiring action to "reduce the likelihood of patient harm associated with the use of anticoagulation therapy [16]."

Although the rate of ADEs can vary depending on data collection methodology, several studies have reported that anticoagulants, including warfarin, are among the medication classes most frequently associated with ambulatory ADEs and ADEs requiring hospital admission [3,9-13]. Using a computer-based event detection application, Jha and colleagues identified hospital admissions due to ADEs. Anticoagulants were implicated in 9% (7/76) of these ADEs; however, warfarin was the most common individual agent identified in all events [11]. In a review of elderly patients by Gurwitz et al, anticoagulants accounted for 8% of ambulatory ADEs, the fifth most common medication class involved in events [10]. As part of the National Electronic Injury Surveillance System-All Injury Program, trained coders from nine participating hospitals identified ADEs noted in emergency department records. Anticoagulants were associated with 4.7% of ED visits; more strikingly, they were the third leading cause (15.4%) of ADEs resulting in hospitalization. Together, warfarin and insulin accounted for 16% of all outpatient ADEs and 33% of outpatient ADEs in patients aged ≥ 50 years [9].

Not only is warfarin involved in a large proportion of ambulatory ADEs, but its effect on patient outcomes is also substantial. In our study, the median highest INR at the time of the event was 6.1, and more than half (52.1%) of these cases resulted in a bleeding event. The majority (64.8%) were defined as major bleeds, with intracranial hemorrhages accounting for 17.5% of these. The intensity of warfarin therapy is known to be strongly associated with bleeding risk, and an INR ≥ 4.5 is the single greatest risk factor for bleeding [14].

Patients and health systems are also affected by hospital admissions stemming from these warfarin-related events. Due to the nature of our surveillance system, all events were identified in patients already admitted to the hospital for reasons that could include coagulopathy, other diagnoses, or both. Interestingly, more than half (59.2%) of admissions were due at least in part to correction of coagulopathy. Budnitz et al reported on ambulatory ADEs identified via surveillance of ED records in geriatric patients. Among warfarin-related ADEs, not only did 73% result in bleeds, but 44.2% of cases required hospitalization [13]. Other studies have noted the propensity for warfarin-related ADEs to lead to hospitalization [9,11].

Such potentially avoidable hospitalizations can be very costly to the healthcare system. In our study, for each admission due to coagulopathy regardless of bleeding, the median cost of care was $10,419, with a median LOS of 4.8 d. Of course, other diagnoses and complicating factors may have influenced both cost of care and LOS, but a broad generalization suggests that if the ADEs and pADEs had not occurred, more than half of the hospital admissions could have been avoided, saving the hospital and payers nearly $1.7 million over 19 months. Moreover, 73.9% of bleeding events required fresh frozen plasma to rapidly reverse the INR, and 51.1% required blood transfusions. Similarly, Jha and colleagues noted 76 ambulatory ADEs led to hospital admissions totalling 738 patient days, with a median 5-day LOS and an average cost of $16,177 per visit [11]. Interestingly, 24 patients in our study experienced a total of 28 additional warfarin-related ADEs identified by ADE-S either before or subsequent to the study period. Ten of these events occurred within 6 months following the study. These potentially avoidable warfarin-related events result in significant healthcare costs to the DUHS, and such problems are likely to be found in other organizations as well.

Quality improvements in ambulatory care can be accomplished by strengthening the information technology (IT) infrastructure [20]. The main process areas that should be targeted include prescribing and monitoring, since errors in these processes respectively account for 41.0%-64.7% and 26.0%-72.7% of preventable ambulatory ADEs [3,10,21]. For warfarin-related events in this study where contributing factors could be identified retrospectively, we found drug interactions, particularly between antibiotics and warfarin, to be most common. Antibiotics are known to contribute to deviations in INR; thus, initiation of such medications in outpatients without close monitoring may result in supratherapeutic INRs and possible harmful outcomes. Ambulatory e-prescribing with drug interaction checking can help reduce medication errors [22,23]. In addition, ambulatory EHRs can have further additive benefits by providing electronically-available provider notes and medication reconciliation. In this study, 73.2% of the events for which a prior record existed in our EHR occurred in patients seen at an inpatient or outpatient DUHS healthcare visit within the previous 4 weeks; however, 40.0% did not have a documented follow-up plan within that same period. EHRs can help bridge the communication gap during transitions in care and assist providers with decision-making and lab test follow-up [20]. This is extremely important not only for warfarin use, where monitoring is critical, but also applies broadly to clinical care given the segmented nature of health care.

However, the impact of quality improvement strategies for the safe management of warfarin therapy may be most felt in terms of fostering patient empowerment and engaging patients in the ownership of their care. Patient nonadherence has been reported as being among the most common drug therapy problems associated with hospital admissions [3]. In our study, evidence of non-compliance was limited to specific documentation by providers and thus the true rate of nonadherence is likely under-reported. Furthermore, the infrastructure of the ambulatory care environment is typically segmented and logistically complex, with patients often forced to seek their care from various providers or specialists, pharmacies, and laboratories spanning multiple health systems.

For these reasons, it is clearly essential that patients be equipped with the knowledge and understanding of their role in their own care. Recognizing this, the Joint Commission instituted National Patient Safety Goal 13, which encourages "...patients' active involvement in their own care as a patient safety strategy" [16]. Patient education is also encouraged in the CHEST guidelines, and at least one study reported lower warfarin-related hospitalizations for bleeding in patients who received education regarding warfarin therapy [14,24].

Health portals or personal health records can also be leveraged to encourage active patient involvement in improving drug safety. Health portals not only provide education via access to consumer medication and health knowledge resources, but also allow access to patients' medical information (e.g., displaying current medication lists and laboratory values) [25]. Moreover, interactive portals could permit patients to enter information, such as INR values or medication changes, obtained from various facilities, to maintain an accurate, up-to-date health record [25,26]. The full potential of portals has yet to be explored, but thus far, patients report satisfaction with using many of their features [26-28].

Though not suitable for all patients, home monitoring of INR using point-of-care testing (POCT) also remains a viable option for patient involvement [29-32]. Although 73.2% of our events were reported in patients who had a healthcare visit within the previous month, INR elevations still occurred. Deviations in warfarin control may be due to events occurring between healthcare visits that the patient may be first to recognize and mitigate if given proper knowledge and tools. Furthermore, use of POCT has shown reductions in bleeding and thromboembolic complications [33].

Additional improvements in warfarin dosing may also be made possible by the clinical application of pharmacogenomic guided therapy [34-36]. Single nucleotide polymorphisms (SNPs) in the CYP2C9 and VKORC1 genes have been associated with a wide variability in patient response to warfarin dosing and a potentially greater risk for adverse events [34,36]. A randomized controlled trial of the clinical safety and efficacy of genotype-guided warfarin dosing (the European Genetics of Anticoagulant Therapy Trial [EU-PACT]) is currently under way [37].