We adhered to the Standards for Reporting Implementation Studies (StaRI) Statement (supplement 1) in this manuscript 33.
Implementation Mapping (IM) is an iterative 5-step process that guides the process of planning, selecting, and evaluating implementation strategies for health promotion interventions (Table 1) 34. Implementation Mapping is derived from Intervention Mapping, a broader framework for the systematic and theory-based development and implementation of multi-level health interventions. An implementation framework, instead of an intervention development framework, was used for this project since the EBPs (appropriate use of antiplatelet therapy and PPI gastroprotection) are clear – the gap is translating this knowledge into practice. This section describes how we executed each step of the IM process, which encompasses adoption, implementation, and maintenance.
Task 1: Conduct an implementation needs assessment and identify adopters and implementers
Identification of program adopters and formation of participatory planning group
The IM process took place at Michigan Medicine, a tertiary academic medical center in the Midwest of the United States, which will also be the site of the pilot trial. The implementation strategies were designed for delivery by the anticoagulation management service (AMS) to patients treated with warfarin, an anticoagulant. The clinical leaders of the AMS affirmed the value of addressing UGIB risk at the outset and a willingness to adopt and maintain the intervention developed. The AMS in this study is overseen by a board-certified cardiovascular physician, staffed by nurses, and provides remote services typically by phone; there are no face-to-face encounters.
There are over 3000 AMSs in the United States, typically staffed by nurses and/or pharmacists, and their overarching purpose is to ensure the safe and effective use of warfarin and other oral anticoagulants 35. Warfarin has a narrow therapeutic index and requires frequent monitoring of blood tests to ensure proper dosing; under-dosing increases the risk of thromboembolic phenomenon, whereas over-dosing increases the risk of hemorrhage. Some AMSs provide services in face-to-face visits, and others through remote visits, typically phone calls 36. AMSs often provide adjunctive services to ensure the safe use of warfarin, including education on foods and drugs likely to interfere with warfarin metabolism, and education on how to manage minor bleeding events (e.g., nosebleeds). However, AMSs typically do not undertake activities to minimize UGIB risk. In fact, anticoagulation clinics are not routinely staffed by physicians and as such cannot routinely discontinue or initiate medications. It was therefore determined at the outset that clinicians within the broader health system would be responsible for executing the EBPs (medication optimization).
To assist with all steps of the implementation mapping process, we convened a participatory planning group of stakeholders. The group included physicians from gastroenterology (1), cardiovascular medicine (1), and primary care (2), patients with anticoagulation clinic experience (2), a clinical pharmacist from the anticoagulation clinic (1), anticoagulation service nurses (2), a study coordinator (1), and a health IT specialist (1). The group met 19 times between July 2020 and October 2021. The group was involved in each step of the implementation process, including interpreting the results of the needs assessment, choosing implementation outcomes, selecting implementation strategies, developing implementation materials, and planning of the pilot trial.
Identification of Barriers and Facilitators to the Use of PPI Gastroprotection
We conducted semi-structured interviews with clinicians and patients to understand barriers and facilitators to medication optimization. Methods and results of the clinician interviews, which included cardiologists (n=4), primary care providers (n=5), vascular surgeons (n=3), and gastroenterologists (n=3), are described in a separate publication 14. This qualitative study had a particular focus on dynamics within and across medical specialties that affect use of PPI gastroprotection because patients on anticoagulant-antiplatelet therapy are often co-managed by multiple specialists (typically a cardiologist and PCP). Briefly, we found that while some cardiologists routinely prescribed gastroprotection, most PCPs, gastroenterologists, and vascular surgeons did not. Barriers to gastroprotection mainly related to (1) knowledge, (2) decision processes, and (3) professional role. Cardiologists tended to have greater knowledge of relevant guidelines on when and how to use PPI gastroprotection and were accustomed to managing antithrombotic medications to minimize bleeding risk; however, they often deferred prescribing PPI gastroprotection to other clinicians. For PCPs, prescribing gastroprotection was a low priority relative to other clinical demands; in fact, many PCPs were more focused on stopping PPIs for patients who were using them unnecessarily. In addition, PCPs were often reluctant to discontinue antithrombotic drugs in patients cared for by a cardiologist. We also found unclear delineation of responsibility for prescribing gastroprotection across specialties.
To complement the clinicians’ perspective, we conducted 5 semi-structured interviews with patients who used antithrombotic medications and were not part of the participatory planning group. Details of the methods can be found in Supplement 2. Guided by the Theoretical Domains Framework (TDF) 37, we identified key areas of focus for patients related to knowledge of GI bleeding and risk factors, knowledge of PPIs and their use for bleeding prevention, clinician responsibility for use of bleeding prevention strategies, and barriers to PPI gastroprotection use. Across the board, patients had limited knowledge of upper GI bleeding risk factors, their personal risk, and of the role of PPIs for bleeding prevention. Patients had greater awareness of and concern about bruising or internal bleeding from other anatomic sites:
I have heard [of] internal bleeding potentially; I never thought of my stomach though. I always thought, like, organs.
Multiple patients believed that diet choices (e.g., spicy foods, excess coffee) would increase their risk of GI bleeding. Patients generally felt the doctor who prescribed their antithrombotic drugs should discuss bleeding risks and mitigation strategies with them. However, most found it acceptable if someone else from their care team, such as a pharmacist or nurse, reached out as long as care was being coordinated with their clinicians. While patients generally felt comfortable initiating a conversation about whether they might benefit from a PPI with their clinicians, some felt they did not have the requisite knowledge to do so.
I don’t think I had the knowledge before this interview to ask for [a proton pump inhibitor].
Barriers to PPI gastroprotection primarily related to medication costs, a desire to minimize the number of medications, and long waiting periods to see a specialist.
I want to be taking as few medications as I can, you know, just what I basically need to treat my condition. I don’t want to overdo it with things because I just – that’s just how I am.
Assessing Prevalence of Anticoagulant-Antiplatelet Therapy and PPI Gastroprotection
Using an electronic data extraction from the electronic health record (EHR) in August 2020, we identified 3,242 patients prescribed warfarin who had an encounter with the anticoagulation service in the prior 4 months. Among these, 949 (41.4%) also used aspirin and/or a thienopyridine, according to their medication lists. Of these 949 patients on anticoagulant-antiplatelet therapy, 658 (69.3%) were not using a PPI, which equated to 20.3% of all patients followed by the AMS. For patients using aspirin, the rate of PPI use was 29.2%, and for thienopyridine users, 46.8%. In a previous study of a similar population of patients using anticoagulant-antiplatelet therapy, we found the most appropriate medication optimization strategy was discontinuation of antiplatelet therapy for 56% of patients, while PPI was appropriate in the remaining 44% 12.
As part of task 1, we also conducted a literature review (see Supplement 3), which is summarized in the introduction, and evaluated the accuracy of the EHR medication list compared to self-reported use of antiplatelet therapy and PPIs (see Supplement 4), which demonstrated the ability to accurately identify candidates for medication optimization using electronic data.
Task 2: Formulation of change objectives
The participatory planning group next enumerated the steps necessary to accomplish medication optimization. These were formulated into “implementation outcomes'' specific to each adopter or implementer, tantamount to their implementation goals. Next, for each implementation outcome, we identified “performance objectives,” which are the sub-behaviors necessary to accomplish the implementation outcome. As an example, one implementation outcome was “Patients who are randomized to receive the patient activation intervention will review the activation guide and contact their clinician to discuss their GI bleeding risk and options to reduce this risk”, with performance objectives of “(1) The patient receives, reviews and comprehends materials sent by mail or portal, and (2) The patient contacts their clinician to discuss appropriate strategies for UGIB risk reduction”. See Supplement 5 for the full elaboration of implementation outcomes and performance objectives. In creating these tables, a critical insight was that some sub-behaviors could be performed by either clinicians (and their care teams) or by nurses in the AMS. Such activities included identifying clinical information necessary for decision-making and providing education to patients on changes in medications. This led us to formulate two sets of implementation outcomes and performance objectives, one that assigned a more active role to AMS nurses, and one that assigned a less active role. These differing sets of implementation outcomes led to the elaboration of two distinct clinician-facing intervention strategies – clinician notification (CN), and clinician notification + nurse facilitation (CN+NF), further described below.
While the group felt that ensuring evidence-based care was primarily the responsibility of clinicians, our literature review demonstrated many clinician-facing interventions have had limited success. Therefore, we developed a set of implementation outcomes related to patient engagement in medication optimization, based on the hypothesis that this approach would augment the effectiveness of clinician-facing interventions.
Next, we identified the determinants (barriers) most relevant to each implementation outcome, informed by the prior qualitative interviews with patients and clinicians. For each implementation outcome, we then crossed the performance objectives with the salient barriers to create a “matrix of change objectives,” which show “what has to change in this determinant to bring about the performance objective” 34. The full set of matrices can be found in Supplement 5. Based on our prior qualitative work, we focused on the determinants of knowledge, professional role, and memory/attention/decision processes for clinicians. For patients, we focused on the determinants of knowledge, skills (e.g., self-efficacy to engage clinicians), and environmental context and resources (e.g., ability to afford medications).
Tasks 3 and 4: Selection of theoretical methods, development of intervention strategies, and production of implementation protocols and materials
In this step, we selected the implementation strategies to influence the determinants/barriers. We relied on multiple sources to identify relevant candidate strategies. These included the Expert Recommendations for Implementing Change (ERIC) compilation of implementation strategies38; the Behavior Change Wheel 39, which was made easier by the fact that it explicitly links intervention strategies to barriers included in the TDF (see Step 1); a taxonomy of intervention strategies that has been developed for use with Intervention Mapping 40; and the findings of our literature review. Candidate strategies were discussed in meetings of the participatory planning group to identify those with the greatest potential and best fit to the setting.
All implementation strategies were designed to be delivered by AMS staff, in keeping with the pragmatic design of the study. Of note, the strategies aim to facilitate appropriate use of the EBPs by clinicians; the AMS and its staff could not directly prescribe or discontinue patients’ medications. The theoretical methods selected for each of the implementation interventions are discussed below. For each of the theories used in selecting implementation interventions and translating them into practical interventions, we took into account the pre-conditions under which theory was likely to be effective (“parameters” in the parlance of Implementation Mapping) 40.
Altogether, the implementation strategies developed are responsive to the Fogg Behavioral model, which identifies three pre-conditions necessary to trigger a target behavior: motivation, ability, and a prompt 41. The logic model by which the activities within each of the clinician- and patient-facing intervention strategies compel clinician prescribing behavior are shown in Figure 1.
Clinician-facing implementation interventions
“Clinician notification” (CN): Because clinician knowledge was a prominent barrier to medication optimization, the group elected to notify clinicians in the form of a message sent through the EHR for relevant patients (ERIC strategy: “remind clinicians”). The message identifies the patient as high risk for upper GI bleeding, provides an estimate of upper GI bleeding risk based on population averages (2.8% per year), describes the medication optimization strategies, and contains an intranet link to guideline summaries on appropriate use and duration of antiplatelet therapy in patients using warfarin, compiled as part of the AEGIS project (see Supplement 6; ERIC strategy: “develop and distribute educational materials”). The text of the notification message is shown in Supplement 7. The message also contains a list of EHR “smart texts” that were developed, consisting of templated language that clinicians could quickly include in EHR communications with office staff and other clinicians about medication optimization. An asynchronous notification was chosen instead of an interruptive alert because alerts are more disruptive to workflow, and because they would only be triggered when a clinician was already in a patient’s medical record, which might delay the time to medication optimization.
“Clinician notification with nurse facilitation” (CN+NF): This implementation strategy is similar to CN but incorporates additional features to streamline appropriate decision-making based on our finding that time was a major barrier to medication optimization. This strategy leverages the clinical expertise of the AMS nursing staff in three ways: First, they review the medical record of the patient to identify possible indications for antiplatelet therapy (e.g., “CAD with percutaneous coronary intervention [PCI]”), information which is then included in the notification message (ERIC strategy: “Facilitate relay of clinical data to providers”). Second, they provide a concise and tailored summary of guidance on appropriate use of antiplatelet therapy (e.g., “if PCI within the past 12 months, continue antiplatelet therapy, otherwise stop antiplatelet therapy”), avoiding the need for clinicians to review the full 2-page guidance summary on the intranet. Third, once the clinician has decided on the appropriate medication optimization strategy, the message indicates that the nurse can enter order for the PPI and pend it for the clinician’s signature (when relevant), and relay the recommendation to the patient, along with appropriate patient education. Together, these steps would allow the clinician to make an instantaneous decision on medication optimization while minimizing cognitive workload. The content of both the CN and CN+NF messages were reviewed on multiple occasions with the members of the participatory planning group.
In the future, we intend to use two additional ERIC strategies after any modifications to the interventions based on the pilot study -- conduct educational meetings with providers and inform local opinion leaders.
Patient activation guide
We developed a written activation guide intended to prepare and motivate patients to have an informed discussion with their clinicians about medication optimization and to initiate that discussion if necessary (ERIC strategy of “prepare patients/consumers to be active participants”). The guide was developed for delivery through the electronic patient portal or by mail. We considered the option of having nurses call patients to deliver the intervention verbally, but this would have been prohibitively time-consuming for the clinic staff.
In selecting theories for developing the patient activation guide, we mainly considered using the Health Beliefs Model (HBM) and Self-Determination Theory (SDT), given their prominence in health promotion programs. We identified the role of fear appeals as an important difference between these theories. The HBM model would have led us to emphasize the risk and seriousness of upper GI bleeding. However, evidence suggests that fear appeals may only be effective when patients have high efficacy to enact change 42. Because we believe many patients lack the efficacy to engage clinicians around medication optimization, we opted to use self-determination theory (SDT) instead 43,44. In recognition of the important role of autonomy in SDT, we avoided use of strongly directive language (for example, “You should speak with your doctor”) in favor of more suggestive language (“You may consider speaking with your doctor…”) and presented the two EBPs for medication optimization as “options.” The guide also acknowledged patients' need for control of the outcome by providing specific language they could use to discuss medication optimization with their clinicians and by providing reassurance that clinicians welcomed their input. Other theoretical methods used in the patient brochure included modelling (vignettes of patients who successfully engaged their clinicians), and persuasive communication (e.g., appeals to medical authority) 45.
The patient activation guide was designed in collaboration with a graphic designer, using Microsoft PowerPoint. The layout and flow were initially modelled off of an activation guide used in a prior successful intervention that activated patients to reduce use of benzodiazepines 46–48. However, all content for our guide was newly developed. We incorporated principles of patient communication and plain-language as recommended by the Centers for Medicare and Medicaid Services (CMS) 49 and the Centers for Disease Control and Prevention (CDC)50. We also included features to enhance ease of information processing 51. For example, we used informative section headings and pop-out effects, selectively bolding the most important information throughout the guide. The guide was repeatedly revised to ensure text coherence, with logical and clear connections from one section (or sentence) to the next. Few graphic elements were included because they were found to add little information content (e.g., images of pill bottles).
Between March and July 2021, the guide was pre-tested with 12 patients, all of whom self-reported use of an oral anticoagulant. Interviews were completed over Zoom (Zoom Video Communications, Inc. Version: 5.7.0), using a using a think-aloud, user-centered design approach 52. Details of the approach and the interview guide can be found in Supplement 8. After every 2-3 interviews, the guide was revised based on feedback. We also sought feedback in multiple research forums at the University of Michigan, including the Center for Bioethics and Social Sciences in Medicine (CBSSM) working group and the Michigan Program on Value Enhancement (MPrOVE) Innovation to Implementation (i2i) Lab. The guide was also reviewed multiple times by the participatory planning group.
The final 8-page guide was titled “Did you know a simple change in the medications you take could reduce your risk of bleeding?” (Supplement 9). One page each provided information about anti-thrombotic drugs, proton pump inhibitors, and the two strategies for medication optimization. We also included two patient vignettes, one for each of the medication optimization strategies, accompanied by images of patients; an FAQ section on gastrointestinal bleeding; and specific instructions and language on how to discuss medication optimization with a clinician. The guide was accompanied by a cover letter addressed to the patient that explained its purpose and was signed by the AMS’s Medical Director.
Implementation strategies, protocols, and information systems to support AMS staff
For the patient and clinician-level interventions, nursing protocols were developed (see Supplement 10). Both protocols specified that the notification should always be sent to a cardiologist for patients seen by one within the prior year (with rare exceptions), and otherwise be sent to the clinician of record for the anticoagulation clinic, to whom routine communications are sent. This decision reflects an insight gained from our qualitative interviews that many primary care providers felt lack of ownership of medication optimization for patients who also saw a cardiologist, whom they often preferred make decisions related to antiplatelet therapy and gastroprotection.
To enable anticoagulation staff to carry out the implementation strategies, we developed a quality monitoring system (ERIC strategy), in the form of an EPIC workbench report. Using Boolean logic and EHR data elements, the electronic report identifies patients each day who may benefit from medication optimization. In addition, it summarizes relevant medical data, and allows rapid identification of clinicians to whom the clinician-level interventions should be targeted and tracking of delivery of the interventions. By creating a new role for nurses in ensuring medication optimization, these interventions employ the ERIC strategy of “revise professional roles”.
For both the CN and CN+NF messages and the patient activation guide, “smart-text” (templated messages that automatically pull in available encoded data elements) were created that would allow the anticoagulation staff to quickly send the materials through the EHR.
Pilot Trial Design (Implementation Mapping Task 5)
We will carry out a pragmatic pilot implementation trial with a cluster- and individual-level randomized factorial design that will primarily assess the feasibility of completing a larger trial with a similar design in the future, and the feasibility of delivering the implementation strategies as intended.
Study setting and eligibility
The study will take place at Michigan Medicine, as described above. Nurses from the AMS will deliver the implementation strategies to eligible patients and the clinicians who care for them. Because our study design is multi-level, the study’s inclusion criteria rely on the concept of clinician-patient dyads; a dyad links each patient to a single clinician (hereafter referred to as a “target clinician”). This will result in a one-to-many-relationship between target clinicians and patients. For patients who have seen a cardiologist (with the exception of electrophysiologists) in the prior year, the target clinician is defined as that clinician. Otherwise, the target clinician is defined as the clinician of record with the anticoagulation service. Every patient treated by the anticoagulation service has a designated clinician of record, who may be in any medical specialty, is the prescriber of the warfarin, and receives all routine communications from the AMS. This manner of identifying target clinicians, which gives precedence to cardiologists, is based on our qualitative analysis, as described above. Unlike other cardiologists, cardiologists in the subspecialty of electrophysiology were not given precedence as target clinicians because members of the participatory planning group felt that electrophysiologists were less accustomed to routinely managing anti-platelet therapy and gastroprotection.
Detailed eligibility criteria for clinicians and patients are shown in Table 2. In brief, clinicians will be eligible if they are a target clinician for at least one patient treated by the anticoagulation service. Patients are eligible for participation if they are followed by the AMS, are included in a clinician-patient dyad that contains a participating target clinician, are candidates for medication optimization (i.e., use warfarin and an antiplatelet drug without a PPI), and do not have an allergy or intolerance to a PPI. Eligible patients will be identified using the reporting workbench, described above.
Patient Eligibility Criteria
• Enrollment with the Michigan Medicine anticoagulation monitoring service
• Currently prescribed warfarin with anticipated use for ≥90 days on day 1 of trial enrollment, according to the EHR medication list
• Currently prescribed an antiplatelet drug (aspirin, clopidogrel, ticagrelor, or prasugrel) according to the EHR medication list
• Age less than 18
• Prescribed a PPI
• Documented intolerance or allergy to PPI use
• Left ventricular assist device
• Heart transplant
Clinician Eligibility Criteria
• Practicing cardiologists at Michigan Medicine who in the prior year had a face-to-face or virtual visit with a patient who meets eligibility criteria for this study
• Practicing clinicians in any specialty who are designated as the clinician of record with the anticoagulation clinic for a patient who meets eligibility criteria
• Cardiologists specializing in electrophysiology who are not the designated clinician of record for at least one patient meeting eligibility criteria
|EHR=Electronic health record.
Study design, sample size, and aims
The design of the pilot study will be a miniature version of the optimization study that is anticipated as phase 2 of MOST (Figure 2). This will be a pilot pragmatic cluster- and individual-level randomized controlled factorial implementation trial. A stratified random sample of target clinicians, consisting of 6 clinicians in cardiology and 6 clinicians from other specialties, will be selected for participation. These target clinicians will be randomized 1:1 to receive either CN or CN+NF for all patients for whom they are the designated target clinician. Cluster randomization at the clinician level was employed to avoid having clinicians receive both types of clinician-level intervention, which might have resulted in uncertain expectations of the AMS. Separately, patients will be individually randomized to receive the activation guide or to patient communication as usual. We anticipate 50 patients cared for by the selected target clinicians will be included in this pilot trial.
The pilot trial has a 2x2 factorial design. This design was selected because in a future optimization trial, it will allow us to compare simultaneously the effectiveness of CN vs. CN+NF, and the patient activation guide vs. usual care. However, for the purposes of the pilot study, the primary aims are to (1) evaluate the feasibility of recruiting patients to participate in study assessments as part of a randomized quality improvement trial; (2) evaluate the feasibility of delivering the patient and clinician intervention components as intended; and (3) evaluate multiple additional implementation outcomes for each of the intervention components.
Because this is a pilot intervention study, if challenges with the feasibility of intervention delivery or with study procedures are identified, they may be modified during the trial, with appropriate IRB approval.
The research team will conduct interviews with patients, clinicians, and anticoagulation nurses by telephone or Zoom videoconference (Zoom Video Communications, Inc.). Interviews will be recorded with participants’ permission. The contents of these interviews, focused mostly on feasibility and acceptability, are described below. For interviews with patients, clinicians, and anticoagulation nurses, questions related to acceptability were developed based on the seven constructs contained within the Theoretical Framework of Acceptability 53.
Interviews with patients
Patient interview 1 is a structured interview that will be conducted between week 5-8 and contain questions related to medication use at the time of randomization and at the time of the call, and the patient’s engagement with the activation tool (for patients randomized to receive it). This assessment, which is anticipated to last 5-10 minutes, was scheduled as such because we hypothesize that most medication optimization will occur within the first month following the interventions.
Patient interview 2 is a semi-structured interview that will be conducted with the subset of patients who were either randomized to receive the activation guide or who in assessment 1 reported that they had communicated with their clinicians about medication optimization. The interview will be scheduled for shortly after patient interview 1. The interview will explore how patients received the activation guide (mail or portal), initial perceptions, acceptability, understanding of the intent, perceived clarity, relevance to their health, whether they read through the guide, and recall of the main message. Images from the activation guide may be shared with patients over Zoom to facilitate recall. The interview will next explore the role that the guide played in any medication changes, specifically whether patients communicated with their clinicians about medication optimization, when it occurred relative to intervention, the mode of communication (portal messages, phone calls, in-person discussion), who initiated the communication, patients’ perceptions of the communication or discussion, including perceived receptiveness of the clinician, and the usefulness of the scripted questions provided in the activation guide. The interview will also explore ways in which the guide might have been improved. We will also administer a validated four-item acceptability measure (Acceptability of Intervention Measure [AIM]) 54.
Patient interview 3 will be a brief structured interview that pertains to current medication use, and 30-day adherence to anti-platelet therapy and/or PPI using a validated three-item self-report for medication adherence 55. It will occur at week 9-12.
Interviews with clinicians
All clinicians who participated in the study will be invited to participate in a one-time semi-structured qualitative interview, which will take place at 5-10 weeks after the last patient in the clinician-cluster entered the study. The interview will start with a broad exploration of clinicians’ perceptions of and experiences receiving CN or CN+NF, including any action taken in response to the message, the elapsed time between intervention and medication optimization, and involvement of the patient and/or other clinicians in medical decision making. Next, the interview will cover the acceptability of the implementation strategies, fit with the clinician’s workflow, and features that were particularly well liked or that could have been improved. Clinicians who received CN+NF will be asked about the perceived accuracy of the clinical information provided by the anticoagulation staff and the usefulness of order entry and patient education performed by the anticoagulation clinic staff.
For clinicians who did not recall receiving the notification, the interview will include a description of the intervention, after which the clinician’s hypothetical impressions of the implementation strategies will be sought. The interview will also explore steps that might have increased the likelihood of encountering the message (for example, delivery of the message to a different folder in the EHR inbox). For clinicians who had at least one patient randomized to receive the activation guide, the interview will explore clinicians’ opinions of engaging patients in this way, and any specific experiences with patients who had received the guide. Clinicians will also complete the AIM 54.
Interviews with nurses
AMS nurses who delivered the interventions will participate in semi-structured interviews at the conclusion of the trial. The interviews will cover acceptability, feasibility, including the process of extracting relevant clinical information for patients randomized to CN+NF, barriers encountered, perceptions of the informatics tools used, and the amount of time required to deliver the interventions. We will also administer the four-item Intervention Appropriateness Measure, the four-item Feasibility of Intervention Measure, and the AIM 54.