Can clinician champions reduce potentially inappropriate medications in people living with dementia? Study protocol for a cluster randomized trial

Accountable Care Organizations (ACOs) are designed to promote patient-centered care and reduce the use of unnecessary services, making them a natural setting to test the effectiveness of a clinician value champion intervention. Co-investigators with the Institute for Accountable Care (IAC), which is the research arm of the National Association of Accountable Care Organizations, (NAACOS), worked with NAACOS to recruit two ACOs for this study: United States Medical Management (USMM) and Oschner Health System. Clinicians with the Visiting Physicians Association contract with USMM to conduct home-based primary care from 39 offices across 13 states. Teams in each office consists of a provider/medical assistant dyad who is also supported by a patient care coordinator, a scheduler, and a nurse navigator who does outreach to more complex patients. Within each office, a medical team meets weekly to discuss patient cases and update their approach to patient care. Over 30% of the Medicare population they served in 2019 (approximately 6000 patients) had an established diagnosis of dementia (Table 1). Oschner Health System is in Louisiana and has 28 primary care clinic locations within their integrated group practice. Their Medicare ACO population had 31,525 enrollees in 2019 and within this population 1387 (4.4%) had a documented diagnosis of dementia (Table 1).

A schedule of enrollment, randomization and allocation, interventions, and assessments for the study is found in Table 2. The unit of randomization will be the primary care clinic, and outcomes for aims 1 and 2 will be assessed by comparing patients with a diagnosis of dementia who receive care in clinics assigned to the intervention arm to those who receive care in clinics assigned to the control arm of the study. We anticipate 32 out of 39 teams from USMM to meet eligibility and 26 out of 28 practices for Ochsner Health. Within each ACO, we will randomly select half of the practices for the intervention and the remaining practices will continue with usual care as controls. Prior to random assignment, we will stratify practices based on size using data for 2019–2020. Practices will be paired by number of unique beneficiaries treated (first and second largest practice, third and fourth largest, and so on) and then one member of each pair will be randomly selected via a random number generator to be a treatment site. Once randomized, a clinician champion will be recruited from each of the offices or clinics in the intervention arm of the study in partnership with the ACO leadership to participate in the clinician champion training program. Neither investigators nor participants will be blinded to allocation. The random allocation sequence and assignment to study arms will be done by the statistician (GR) at Brandeis University.

Each clinician champion will participate in one-hour training webinars over 6 months (January–June 2022) using the curriculum developed in the afore-mentioned RWJF Value Champions fellowship program. These sessions will be led by the PI (MP) who also was the Director of the RWJF value champion fellowship from which these materials were developed. The Learning Modules (Table 3) follow the same format and include a rationale, quotes from RWJF value champion fellows, learning objectives, required readings, discussion questions, application of the learnings to their project workbook, and additional resources. Each Learning Module is also accompanied by a Facilitator/Instructors Guide so that any health care organization can use these materials to train their own value champions. Near the end of the training webinars, each value champion will launch his/her own 12-month PIM de-prescribing project (April 2022–March 2023) using the Project Workbook as their guide. Following the 6 months of training clinician champions will participate in a monthly 1-h shared learning webinar to share successes, challenges, and brainstorm solutions during the 12 months when champions are doing the work to support reductions in the use of PIMs in PLWD.

For aims 1 and 2, de-identified Medicare claims data will be used to assess outcomes. Co-investigators from the Institute for Accountable Care (IAC) will access Medicare claims data in the CMS Virtual Research Data Center (VRDC) repository and claims provided to ACOs (Claim and Claim Line Feed File or CCLF files). Specifically, we will use quarterly Medicare clinician bills (Part B), institutional bills (Part A), and prescription drug event (PDE) data (Part D) for diagnostic and utilization information on Medicare beneficiaries and practices for 2019 and 2020. Data for 2021 and 2022, including prescription drug events data, clinician bills and institutional bills will be derived from the ACO’s CCLF files which are provided by CMS monthly. For aim #3, the five implementation outcomes will be assessed by using available project management data recorded by each value champion in their project workbook, specifically the milestones completed within and across each phase of work outlined in the workbook.

Following an intent-to-treat design, all Medicare beneficiaries meeting our dementia eligibility criteria will be included in the study, regardless of level of engagement with practice clinicians. Specific criteria include the following:

Medicare beneficiaries will be excluded if they have any months of Part C (Medicare Advantage) during study period since prevents us from seeing all of their care. All data will be available to all study investigators.

Our sampling plan anticipates enrolling at least 30 of the 39 practices in USMM and 26 of the 28 practices in Ochsner. Based on Medicare enrollments of 39,000 members in USMM with a 20% dementia rate in 2019, and 10,000 members in Ochsner with approximately 10% of members diagnosed with dementia om 2019, we should have at least 2000 patients per treatment group for USMM and 500 patients per treatment group for Ochsner. For our primary outcome, with a follow-up period of 1 year and with mediation possession rates calculated per patient each quarter, we will have approximately 4 MPR observations per patient or 8000 observations per group for USMM analyses and 2000 per group for Ochsner. Even after assuming significant correlation across time within patient and modest correlation between patients within a practice, which will reduce effective sample sizes by 20%, we should have the equivalent of 6400 observations per treatment group for our USMM analyses and 1600 observations per treatment group for Ochsner. The expectation of our intervention is that for each medication class, it will produce clinically relevant decreases in mean possession rates of 10% of a standard deviation. If the effect size is this large for a medication class and using a 5% level of significance, our Ochsner sample (with effective size of at least 1600 per group) will have 80% power to discern a significant difference, while our larger USMM sample (with effective size of at least 6400 per group) will have almost 100% power. Indeed, our USMM sample will have 80% power to find a significant difference, even if the effect size is only 5% of a standard deviation. Ethics approval for this study was obtained through the IMPACT Collaboratory’s central institutional review board, Advarra, on September 8, 2021.

The NIA IMPACT Collaboratory’s Data Safety Monitoring Board will provide oversight and review. The data safety monitoring plan includes monthly adverse event reporting by each clinician champion for patients with dementia seen in clinics randomized to the intervention arm. All adverse events that are serious and unexpected will be reported to the IMPACT Collaboratory Regulatory and Data Team Leader, NIA IMPACT Collaboratory Project Officer, and the project’s Safety Officer within 48 h of the study’s knowledge of SAE. Since adverse event reporting is not available for patients in the control arm, we will provide the board with quarterly reports comparing all emergency department visits, inpatient admissions and mortality among patients with dementia who are seen in clinics randomized to the intervention arm of the study and those seen in clinics randomized to the control arm of the study.

The Institutional Review Board (IRB) for the NIA IMPACT Collaboratory, Advarra, reviewed and approved the study protocol on September 8, 2021. A waiver of consent was approved for use of Medicare claims data to conduct the analyses for aims 1 and 2 was approved. Clinician champions were provided a study information sheet at the start of the intervention (January 2022) with information about their participation and how to withdraw from the study. Any subsequent protocol modifications will be submitted to the Advarra IRB.

In their commentary on unpacking de-implementation, Norton and Chambers encouraged the field of implementation science to consider “critically important distinctions between the implementation of new, evidence-based interventions, and the de-implementation of currently delivered inappropriate interventions.” [20]. Although there is a strong rationale for the use of a clinician champion as a de-implementation strategy [24], few studies have employed them to address low-value care, although Colla and colleagues identified strategies that champions might employ such as clinician education and provider feedback as effective [34]. Santos and colleagues recently completed a systematic review of the effectiveness of a clinician champion when implementing an innovation into health care settings [35]. Only one of the 35 studies included in the review was focused on de-implementing a low-value care service, that of urinary catheters in pediatric patients. However, it is possible that the search strategy focus on implementation for this systematic review did not identify studies of de-implementation. This embedded pragmatic trial will add to the literature and expand our understanding of the effectiveness of a clinician champion as a strategy to de-implement care for which the potential for harm is greater than the potential for benefit: potentially inappropriate medications prescribed to people living with dementia.

This study will also allow health care systems and investigators to gain real-world experience in developing clinicians who can be champions of high-value care within their setting. Leaders in both participating ACOs have expressed an interest in spreading this intervention to all primary care clinic locations within their ACO and to other members of National Association of Accountable Care Organizations. Both findings from the trial and lessons learned from preparing clinician champions will be disseminated through publications in peer-reviewed publications, presentations at national conferences, and for interested stakeholders such as ACO leadership.

It is possible that clinician champions may leave or relocate their clinical practice during the study. We will guard against this be identifying a primary and secondary clinician champion within each clinic. Information shared by the clinician champions within their own clinical setting in the intervention arm of the study may be inadvertently shared with clinicians who provide care in clinics randomized to the control arm of the study. Finally, there may be clinics with a small sample of patients with dementia, which may adversely impact our ability to draw conclusions from the results.

Value champions who complete the study will be prepared to serve as mentors of clinician value champions within their own organizations and across other ACOs after the conclusion of the study. The clinician champion training is generic and can be applied to any low-value care service and the content of the curriculum, along with the project workbook, have been deliberately designed so that any health system might use them to train their own cohort of clinician champions in the future.