Assessing Anticoagulation Management and Shared Decision-Making Documentation From Providers Participating in the SUPPORT-AF II Study.

BACKGROUND
A previously tested intervention featured educational outreach with modified academic detailing (AD) to increase anticoagulation use in patients with atrial fibrillation. Currently, this study compares providers receiving and not receiving AD in terms of inclusion of AD educational topics and shared decision-making elements in documentation.


METHODS
Physicians reviewed themes discussed with providers during AD and evaluated charts for evidence of shared decision-making. Frequencies of documentation of individual items for providers receiving AD versus non-AD providers were compared. To understand baseline documentation practices of AD providers, encounters of AD providers before their AD participation were randomly selected.


RESULTS
There were 113 eligible encounters in the four months after AD-36 from AD providers and 77 from non-AD providers. Thirty-five encounters were identified from AD providers before participating in the intervention. Providers infrequently documented many reviewed items (% documenting): anticoagulation mentioned (44%), multiple options for anticoagulation (5%), CHA2DS2-VASc score (11%), bleeding risk factors (2%). Compared with non-AD providers, AD providers had statistically significant higher percentages for the following items: mention of anticoagulation (64% versus 35%), stroke risk (11% versus 0%), anticoagulation benefits (8% versus 0%), and patient involvement (17% versus 0%). There was no improvement, however, for AD providers compared with baseline documentation percentages.


DISCUSSION
Providers infrequently documented important items in anticoagulation management and shared decision-making. AD participation did not improve documentation. Improving adoption of AD educational items may require more prolonged interaction with providers. Improving shared decision-making may require an intervention more focused on it and its documentation.

A nticoagulation therapy is the cornerstone for stroke prevention in patients with atrial fibrillation. However, only about half of eligible patients receive anticoagulation in USbased, ambulatory registries. 1 There are multiple factors that likely contribute to the gap in anticoagulation use. Health care providers often overestimate risks and seriousness of bleeding from anticoagulation and underestimate risks of stroke from anticoagulation undertreatment in their atrial fibrillation patients. 1 At the same time, patients may be reluctant to initiate or continue anticoagulation due to real or perceived risks of side effects, burdens from monitoring, and financial costs. 2 Continuing medical education, such as traditional lectures with slide presentations and/or dissemination of treatment guidelines like those from the American Heart Association (AHA), have not proven effective in reducing the anticoagulation treatment gap. 3 A cluster randomized controlled trial was recently completed comparing increase in anticoagulation use for providers receiving electronic messaging and educational outreach in the form of academic detailing (AD) with standard provider decision-making without additional educational resources. 4 AD is the process by which a clinical educator trained in the techniques of social marketing meets with a provider one-on-one to discuss medical evidence and recommends specific changes to improve patient care and outcomes. It is an alternative method of delivering education to providers. AD has been shown to successfully improve provider adherence to guidelines in a variety of circumstances including for blood pressure management of patients at high risk for cardiovascular disease, 5 appropriate use of antibiotics among hospitalists, 6 and anticoagulation prescription for uptake of DVT prophylaxis. 7 Approximately one-third of providers participated in both parts of the intervention, whereas another third received only electronic messaging. There was no statistically significant effect on anticoagulation use for the combined messaging + AD intervention. Comparing documentation of encounters conducted by AD providers with those conducted by non-AD providers may shed light into anticoagulation management and the use of elements of shared decision-making.
Giving health care providers up-to-date evidence about use of anticoagulation may help to reduce undertreatment. In addition, giving providers information about direct oral anticoagulants may enable them to overcome patient hesitation to use anticoagulation based on previous poor experience/ dissatisfaction with use of warfarin. In light of the complexity of the decision and the strong influence of patient preferences and knowledge of successful anticoagulation treatment, the American Heart Association/American College of Cardiology Atrial Fibrillation treatment guidelines recommend that health care providers use shared decision-making when advising atrial fibrillation patients about the risks and benefits of oral anticoagulation. 8 Without full patient understanding of the risks and benefits, patients may fail to realize the value of the therapies providers suggest. Therefore, a comprehensive review was conducted of the documentation of patient encounters occurring up to four months after completion of AD. The intent was to understand how often providers addressed anticoagulation for atrial fibrillation in their documentation and how often their documentation reflected the various topics covered during AD sessions. Their documentation was also reviewed for evidence of shared decision-making elements, including mention of options for treatment, mention of the risk of stroke, risk of bleeding, and evidence of involvement of the patient in discussion about anticoagulation. It was expected that the documentation of providers participating in AD would reflect higher levels of anticoagulation prescription as well as more descriptive and more patient-involved conversations about anticoagulation.

Population and Setting
Cardiology and primary care providers who were caring for patients with an active atrial fibrillation diagnosis not on anticoagulation were identified within the large health care system affiliated with a medical school in central Massachusetts. At the beginning of the intervention in July 2018, the anticoagulation rate within the health system was 71.9%. Experts have suggested 80% as a target benchmark to aim for anticoagulation percentage. 9 To be eligible, providers had to be taking care of at least five patients with atrial fibrillation who had an elevated stroke risk (ie, CHA 2 DS 2 -VASc score $2 10 ) and were not currently on anticoagulation using the repository of patients available from the electronic medical record (EMR). Providers were randomized in a 2.5:1 ratio to an electronic profiling and messaging combined with AD intervention. In the intervention, providers were emailed monthly reports of their anticoagulation percentage relative to peers for atrial fibrillation patients with elevated stroke risk (CHA 2 DS 2 -VASc $ 2). EMRbased messages were also sent shortly before an appointment with an anticoagulation-eligible but untreated atrial fibrillation patient. The study team also offered to meet with intervention providers using an AD approach. The AD portion of the intervention was therefore voluntary.

AD Intervention
The AD approach included two live sessions spaced 1 to 2 months apart. Two sessions were selected anticipating that it would be difficult to secure participation if further sessions were requested. In addition, given the overall constraint of eight months for study duration, AD sessions needed to be completed within four months to permit at least four months of time after completion of all components to assess if AD led to an increase in anticoagulation use. For each session, conducted either in person or remotely by web conferencing software, reviewed topics were chosen by the provider from a list of 10 general topics contained in a Microsoft PowerPoint slide set. Although typical PowerPoint presentations do not provide ability to easily move from one topic to the other, a "menu slide" was designed of all the topics, which permitted navigation directly to that topic without having to go through slides from other topics. At the end of each topic, a link was created to return to the menu slide allowing the provider being detailed to choose from remaining topics. This permitted engaging providers and making the format of the sessions interactive. Coverage of a fixed number of topics was not required but rather as many topics as the provider selected were covered within the target amount of time. Topics included the role of aspirin in preventing strokes, anticoagulation use in the elderly, use of stroke and bleeding risk calculators, resuming anticoagulation after bleeding or fall, dual and triple antithrombotic therapy, treating a patient with depression and/or cognitive impairment, direct oral anticoagulant information, common scenarios where anticoagulation decision-making is complex, resources available to providers at the institution, and general resources. A faculty general internist from the investigative team (A.K.) conducted the AD sessions with PCPs, whereas a faculty cardiologist from the investigative team (D.D.M.) conducted the sessions with cardiology providers. The first sessions lasted between 25.05 and 57.83 minutes, on average lasting 39.7 minutes (see Appendix A for presentation material, Supplemental Digital Content 1, http:// links.lww.com/JCEHP/A77).
The study team followed up the first meeting by supplying providers with any evidence or articles that were discussed during the AD session. One email was sent reminding the provider about the discussion and to prompt them to try to implement any practice changes that they had agreed to try after Assessing Anticoagulation Management Documentation the first session. The second session was shorter and included a recap of the first session and dialog about the provider's experience with anticoagulation management for patients seen in the interim between first and second sessions. The second sessions lasted 2.35 to 38.90 minutes, on average lasting 18.0 minutes. This study was approved by the University of Massachusetts Medical School's IRB. Written informed consent was obtained from providers participating in AD sessions.

Chart Review
Outpatient encounters were identified of patients being seen by AD and non-AD providers in the four months after AD. Four physician reviewers were trained to conduct chart reviews. Each physician reviewer had completed at least one year of internal medicine residency. None of the physicians were aware of the treatment assignment of providers (ie, they were blinded). During the training process, the four resident physicians reviewed charts until their responses were consistent with those of a faculty author (A.K.). Reviewers then each reviewed between 25 and 35 charts of the final analysis set. Reviewers responded to several items included in a chart review form developed for the study. In the form, physicians confirmed the patient was not on anticoagulation at the time of the encounter. Reviewers were also asked to determine whether the visit was a regular, comprehensive visit versus a visit scheduled because of acute illness or other narrow purpose. Reviewers then evaluated if the patient started anticoagulation within 3 months of the visit. Next, reviewers assessed whether providers discussed any of the themes mentioned during AD sessions. Finally, reviewers evaluated charts for evidence of shared decision-making, including mention of options for treatment, mention of the risk of stroke such as documentation of the CHA 2 DS 2 -VASc score, risk of bleeding, and evidence of involvement of the patient in discussion about anticoagulation.
In addition to examining charts of encounters occurring after providers received AD, charts were reviewed for this group of providers before beginning the intervention to gain additional insight into the impact of the intervention given that the providers who agreed to meet for AD were potentially different in terms of diligence with documentation compared with other providers.

Data Analysis
The frequency of documentation was compared of individual items in providers receiving AD compared with non-AD providers. The frequency of each of the items was compared in visits conducted by AD providers before versus after intervention. For all comparisons, the Fisher exact test was used because of the relatively small numbers for analysis of comparisons. A comparison was considered statistically significant for a P value < .05. All analyses were conducted in Microsoft Excel 2016 (Microsoft, Inc, Redmond, WA).

RESULTS
One hundred thirteen patients were identified as having encounters with 62 providers after the intervention. Of these 113 encounters, 22% of the time, patients started anticoagulation within 3 months of the index visit. Although more than 90% of the encounters were scheduled/comprehensive visits, atrial fibrillation was mentioned only 57% of the time. Linked to this finding is that providers discussed anticoagulation in only 44% of visits. Moreover, providers did not document having discussed multiple options with patients (only 5% of the time). They also infrequently documented the CHA 2 DS 2 -VASc score (11% of the time). Similarly, providers infrequently documented discussion of the resumption of anticoagulation after bleeding (although it was not established how often a patient had a bleeding history for which the resumption of anticoagulation was relevant). They infrequently documented special topics (5% of the time) that were part of the AD intervention.
Of the 113 encounters, 36 of these were patients seeing one of 18 providers who received AD; 77 encounters included patients seeing one of 44 non-AD providers. AD providers were 50% female, and on average, 20.4 years had elapsed since their medical or nursing school graduation. On average, 26.6 years had elapsed since medical or nursing school graduation for non-AD providers, and they were 25% female.
There were several differences in documentation between encounters of AD and non-AD providers. Most notably, in 64% of encounters with AD providers, there was mention of anticoagulation vs. only 35% of encounters with non-AD providers (P value .005). In addition, AD providers more frequently documented the risk of stroke (11% versus 0% P value .009). They also more frequently discussed the benefits of anticoagulation and more frequently made reference to patient involvement in the discussions, although in general, both groups infrequently documented these items (Table 1 for further details).
Thirty-five visits were identified for patients seeing one of 17 AD providers before the provider received the intervention for comparison with documentation of AD providers after the intervention; 12 of these providers were also among the set of 18 AD providers contributing encounters in the post-intervention phase.
When comparing the frequency of documentation of most items for AD providers before versus after intervention, there were multiple differences. Before intervention, AD providers more frequently documented multiple anticoagulation options (29% versus 6% P value .012) compared with post-intervention AD providers. Similarly, before intervention, AD providers more frequently documented discussing anticoagulation with their patients (40% versus 17% P value .037). Before intervention, they documented a trend toward (as opposed to a statistically significant difference) more patient involvement in the discussion (37% versus 17% P value .064). By contrast, post-intervention AD providers more frequently documented risk of stroke (11% versus 0% P value .115), more frequently addressed bleeding risk factors (6% versus 0% P value .493), and more frequently documented one of the special topics from AD compared with pre-intervention AD providers (14% versus 3% P value .199).

DISCUSSION
Low rates of documentation were found of several elements of anticoagulation management and shared decision-making. This included anticoagulation management items-timelines for resuming anticoagulation after bleeding and special topics we discussed in AD sessions. There was also a low rate of documentation of shared decision-making items including stroke risk, bleeding risk, and mention of multiple options for treatment. Although AD providers more frequently documented these elements compared with non-AD providers, this did not appear to reflect an improvement compared with review of documentation written by AD providers before intervention.
In the context of stroke prevention for atrial fibrillation, previous investigators assessed written feedback and AD sessions delivered among 66 providers over a 12-month period through telephone. They found no significant difference in anticoagulation prescription rates among 1116 atrial fibrillation patients treated by providers receiving the intervention. 11 Similar results were observed in the SUPPORT-AF II study which is described above. To the best of the study team's knowledge, no one has previously assessed the impact of AD on quality of documentation of anticoagulation management. Provider documentation often correlates with quality of patient care, particularly regarding continuity of care between providers. 12 In the 2011 AMIA policy meeting, participants discussed ways to foster improved documentation practices with the goal of reducing health care costs and improving quality. 13 Therefore, assessing documentation of anticoagulation management is an important step in evaluating and potentially improving anticoagulation management.
The main implication of these findings is that AD delivered using this approach of two meetings plus interim correspondence and provider support did not predict an increase in the frequency of documentation of elements of anticoagulation management, including shared decision-making. Improving the documentation of these items may require more prolonged interaction with pro-viders. In addition, documentation of shared decision-making items may require an intervention more focused on documentation. Although professional societies have made recommendations to encourage shared decision-making with regard to anticoagulation for atrial fibrillation, 8,14 to the best of the study team's knowledge, they have not disseminated document templates/ checklists for shared decision-making content or the organization of the content.

Study Limitations
There are limitations to the reported findings. There is an unclear correlation between documentation of discussion with patients compared with what was actually discussed. There were no standardized guidelines for provider documentation, and many providers may only have recorded final management steps, rather than all topics that were discussed with patients. Audio recordings would provide complete information but were not part of the original scope of the study. Future research is welcomed to look at the conversations occurring between patients and providers and examining the correlation with provider documentation. In addition, patients cared for by AD and non-AD providers may have differed significantly in some characteristics that were not explored or could not have discerned from the small sample. Finally, all electronic actions of providers were not examined. Providers may

Frequency of Shared Decision-Making Elements and Other Characteristics in Visits Conducted by Providers Receiving Full and Partial Intervention Compared With Controls
After AD Intervention Before AD Intervention have accessed resources such as patient educational materials available through the EMR without having explicitly documented having provided this to patients. Future research may want to examine these electronic actions when scoring the extent of provider shared decision-making. Identifying barriers to documentation and designing an intervention specifically to address that would potentially improve documentation.

CONCLUSION
In conclusion, both AD and non-AD providers infrequently documented several important factors in anticoagulation management and shared decision-making. There was no detected association between participation in the AD intervention and improvement in the documentation of these elements, although these patient groups may have been different in untracked ways. Based on these findings, it is likely this method of AD alone is not sufficient to increase the frequency of documentation of these elements. For improving documentation of shared decision-making, future research is encouraged using a more focused intervention and the development of templates to facilitate shared decision-making documentation.
Lessons for Practice n Providers significantly underprescribe anticoagulation for patients with atrial fibrillation. Current guidelines recommend providers engage in shared decision-making with patients to start them on anticoagulation when indicated.
n Academic detailing in the form of voluntary participation in two live sessions about selected anticoagulation topics did not have a significant impact on evidence of shared decision-making and other anticoagulation management elements in provider charts.
n Different forms of academic detailing, perhaps directly addressing shared decision-making and documentation of these actions, should be tested in provider populations to address this treatment gap.