Call me maybe? The right time and place for BPAs in EHRs

This editorial refers to ‘Alert-based computerized decision support for high-risk hospitalized patients with atrial fibrillation not prescribed anticoagulation: a randomized, controlled trial (AF-ALERT)’†, by G. Piazza et al., on page 1086. Care for atrial fibrillation (AF) and its complications leads to substantial utilization of healthcare resources and creates a financial burden, primarily driven by the occurrence of stroke. While AF-related strokes are particularly devastating, numerous trials have demonstrated that oral anticoagulation (OAC) is highly effective at reducing the risk of stroke and all-cause mortality.1 Nevertheless, implementation of OAC in the routine clinical care of AF patients has been disappointing. There are multiple pitfalls that can impair the appropriate prescription and utilization of OAC for stroke prevention in patients with AF (Figure 1A). Numerous studies have documented underutilization of OAC in patients with AF at risk for stroke, and across multiple healthcare settings and patient populations.2 , 3 While rates of OAC may be affected by analytical methodologies (e.g. failure to capture appropriate OAC contraindications),4 , 5 as well as variations in local practice patterns, it is clear that a large proportion of AF patients with indications for OAC never receive treatment. This is a significant missed opportunity to improve outcomes in patients with AF, using proven, available therapies. While there have been many randomized clinical trials assessing the efficacy and safety of pharmacological and non-pharmacological therapies to prevent stroke in patients with AF, there are relatively few studies demonstrating effective interventions to improve the proportion of eligible patients who actually receive these treatments. (A) Pathway and pitfalls of achieving optimal stroke prevention in atrial fibrillation. (B) Impact of electronic health record best-practice alerts on rates of oral anticoagulation. The SUPPORT AF study (left of the dotted line) data provide rates of oral anticoagulation before and after the electronic health record intervention. The randomized studies (right of the dotted line) provide rates of oral anticoagulation in the electronic health record alert and control groups. In this issue of the European Heart Journal, Piazza and colleagues present the results of the Alert-Based Computerized Decision Support for High-Risk Hospitalized Patients with Atrial Fibrillation Not Prescribed Anticoagulation (AF-ALERT) trial.6 The investigators randomized clinicians caring for 458 inpatients with AF to an active, best-practice advisory (BPA) through the electronic health record (EHR), or to usual care (no BPA). The active BPA led to improved inpatient prescription of OAC, but also improved rates of outpatient OAC use at 90 days post-discharge (28% vs. 17%). Moreover, the patients in the BPA group were observed to have a lower odds of death, myocardial infarction, stroke, transient ischaemic attack, or a systemic embolic event at 90 days (11% vs. 22%). The authors should be commended for testing the effectiveness of this simple intervention, instead of simply saddling clinicians with more ‘clicks without evidence.’ One of the components that sets the current trial apart from other, similar studies, is the link to clinical outcomes. While some (but not all)7 , 8 EHR-based interventions for OAC have demonstrated improvement in prescription rates (Figure 1B),9 the evidence for improved clinical outcomes has been limited. Furthermore, the improved outcomes in AF-ALERT suggest that there was substantial compliance and persistence of therapy. This is important as the clinical community has struggled with high rates of OAC discontinuation, particularly with vitamin K antagonist oral anticoagulants (e.g. warfarin). Up to half of all OAC users will discontinue therapy at 1 year and, while the present study did not look that far, the data do support the idea that inpatient hospitalization provides valuable opportunities to improve chronic, outpatient treatment—a paradigm also observed in the treatment of heart failure.10 Nevertheless, there are shortcomings to the AF-ALERT trial. For example, the authors nicely describe the rationale for randomizing at the clinician level (by provider identification number), yet this methodology has limitations as well. First, as noted in the manuscript, it yielded an imbalance in the number of patients assigned to each group. Secondly, it allows for individual provider characteristics to vary and possibly impact treatment decisions; there may be patients assigned to BPA clinicians who are inherently more likely to prescribe OAC, whereas another entire group of patients may not get alerts and may simultaneously be treated by a clinician less likely to use OAC. In short, clinician-level randomization makes it more difficult to isolate the effect of the BPA from clinician-specific variation in OAC utilization for inpatients with AF. Importantly, we are not given any characteristics about the clinician groups, and whether randomization yielded balance across their characteristics—a similar prior study by the AF-ALERT investigators instead randomized the BPA at the patient level,11 and prior studies of AF have shown significant influence of clinician characteristics on use of OAC for stroke prevention in AF.12 Additional limitations include the use of EHR and/or administrative sources for case identification and risk stratification—emerging data suggest that there may be significant limitations to these methods (though they remain widely used).13 This may be important, as appropriate patient identification and risk stratification are vital to the effective implementation of any BPA. Finally, while AF-ALERT is a step in the right direction, the trial also shows us how far we still have to go, as three-quarters of the BPA group did not receive treatment despite the intervention. The following are among the important considerations raised by the AF-ALERT trial with respect to using BPAs to improve care. (i) Should they only target interventions supported by the highest evidence and national guidelines? (ii) Should a randomized trial with clinical outcomes, such as AF-ALERT, be the benchmark required for implementation of a BPA? (iii) What are the most effective implementation strategies for BPAs, in order to minimize clinician complacency and maximize impact? The latter may be the most important in the quest to leverage EHRs to improve clinical outcomes. As technology has become ubiquitous in our daily lives, legions of ‘user experience’ scientists are employed across industries in order to ensure the best possible experience—nowhere is this more important than in the use of EHRs, where superfluous clicks and burdensome user processes increase clinician burnout and limit the power of EHRs to improve care.14 Streamlining BPAs to target the most relevant patients, in the most important clinical settings, and with engaged clinicians, will reduce frustrating, unnecessary clicks and potentially increase the value and efficiency in the system. Research reported in this publication was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under Award Number K23HL143156 (to B.A.S.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Conflict of interest: B.A.S. reports no conflicts. J.P.P. receives grants for clinical research from Abbott, the American Heart Association, Boston Scientific, Gilead, Janssen Pharmaceuticals, NHLBI, and Philips; and serves as a consultant to Abbott, Allergan, ARCA Biopharma, Biotronik, Boston Scientific, Johnson & Johnson, LivaNova, Medtronic, Milestone, Oliver Wyman Health, Sanofi, Philips, and Up-to-Date. The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.