An educational intervention to promote central nervous system-active deprescribing in dementia: A pilot study

Abstract

Background:

Central nervous system (CNS)-active polypharmacy (defined as concurrent exposure to three or more antidepressant, antipsychotic, antiseizure, benzodiazepine, opioid, or nonbenzodiazepine benzodiazepine receptor agonists) is associated with significant potential harms in persons living with dementia (PLWD).We conducted a pilot trial to assess a patient nudge intervention’s implementation feasibility and preliminary effectiveness to prompt deprescribing conversations between PLWD experiencing CNS-active polypharmacy and their primary care clinicians (‘clinicians’).

Methods:

We used the electronic health record to identify PLWD prescribed CNS-active polypharmacy in primary care clinics from two health systems. Clinics were assigned to intervention (n=10) or control (n=12), with PLWD in intervention clinics mailed an educational brochure to prompt discussion with clinicians about the appropriateness of their CNS-active regimen. We conducted chart reviews for evidence of documentation related to these medications and used the EHR to assess preliminary effectiveness 120 days after sending the brochure (e.g., number of CNS-active medications prescribed, change in total standardized daily dose [TSDD] of CNS-active medications, and change in prevalence of CNS-active polypharmacy). We interviewed 10 clinicians from intervention clinics to assess their perceptions about the acceptability of the intervention.

Results:

PLWD in the intervention group (n=61) and control group (n=68) had an average age of 72.4 years old (standard deviation [SD] 9.7), 62.8% were female, and 84.5% were white. We did not find any documented evidence of conversation related to CNS-active medications between PLWD who received the brochure and their primary care clinicians. After 120 days, there was no significant between-group difference in the mean number of CNS-active medications prescribed (−1.0 [SD 1.3] vs. −1.0 [SD 1.3]), mean TSDD (−1.6 [SD 6.0] vs. −1.3 [SD 5.8]), or the percentage of patients with CNS-active polypharmacy (52.6% vs. 50.4%). Interviews with clinicians suggested they were aware that combinations of CNS-active medications were not ideal; however, they reported inheriting patients who were already on these medications, and they did not have sufficient clinic time or access to safer alternatives to overcome patient hesitation to deprescribe.

Conclusions:

A direct-to-patient mailed educational brochure did not demonstrate feasibility in provoking deprescribing conversations between PLWD and clinicians or preliminary effectiveness in decreasing CNS-active polypharmacy.

This study was registered at ClinicalTrials.gov (NCT05628181) on November 16, 2022.

1. INTRODUCTION

Dementia affects approximately 10% of U.S. adults aged 65 and older and will affect more than 14 million Americans by 2060 [1]. Worldwide, the number of cases is projected to nearly triple to over 150 million people by 2050 [2]. People living with dementia (PLWD) have more emergency room visits, hospital admissions, and overall costs of care relative to other older adults of similar medical complexity [3, 4]. The combination of high healthcare use and the challenges introduced by functional and cognitive impairments make the effective delivery of care to this population a priority.

More than half of community-dwelling PLWD are concurrently prescribed at least five medications, while nearly 13% are prescribed 10 or more [5]. Furthermore, nearly 14% of community-dwelling PLWD are exposed to CNS-active polypharmacy, defined by the American Geriatrics Society Beers Criteria as concurrent exposure to three or more psychotropic or opioid medications [6, 7]. CNS-active polypharmacy is associated with an increased risk of drug-drug interactions, increased fall risk, impaired cognitive function, and increased likelihood of respiratory suppression or death in older adults [8, 9].

Given the risks associated with CNS-active polypharmacy, PLWD and their care partners—if fully informed—may choose to limit their exposure to such combinations. Indeed, the Eliminating Medications Through Patient Ownership of End Results (EMPOWER) trial demonstrated that a direct-to-patient educational intervention was effective in decreasing benzodiazepines in patients over the age of 65 [10]. In this pilot study, we expand on this by testing the feasibility and preliminary effectiveness of a pragmatic, direct-to-patient education intervention to prompt reductions in CNS-active polypharmacy among PLWD.

2. METHODS

This research was approved by the University of Michigan Institutional Review Board (IRB), which ceded oversight to Advarra, the IRB of Record for the National Institute on Aging IMPACT Collaboratory (#Pro00065204).

2.1. Study Design

We conducted a two-arm pilot study in primary care clinics of two academic health systems [System A (n=5) and System B (n=17)] in southeast Michigan (USA). We used the electronic health record (EHR) to identify patients aged 50 and older with at least one internal or family medicine encounter between October 1, 2021, and September 30, 2022. Then, in those clinics, we identified patients diagnosed with dementia or cognitive impairment using ICD-10 codes and terms from the patient’s EHR problem list (Tables S1 and S2 in the supplementary material) using encounters between October 1, 2017, and September 30, 2022 [11]. Among these PLWD, we then used outpatient prescription orders to identify those prescribed CNS-active polypharmacy between August 17, 2022, and September 30, 2022 (i.e., 45 days before the intervention). Exposure to CNS-active polypharmacy was defined as having a supply of three or more unique medications from the following classes: antidepressant, antipsychotic, antiseizure, benzodiazepine, opioid, or nonbenzodiazepine benzodiazepine receptor agonists (“Z drugs”) [7]. Primary care clinicians (‘clinician’) were provided a list of eligible patients and were given an opportunity to withdraw any patients they believed were inappropriate for the intervention (e.g., they determined that all the medications were indicated); clinicians opted out a total of two patients in each health system (Figure 1).

Figure 1.

Consort diagram for pilot intervention to reduce CNS-active polypharmacy

2.2. Intervention

As previously described [13], we developed a patient-facing educational brochure addressing prescribed combinations of psychotropic and opioid medications, informed by prior interventions developed and tested in the EMPOWER and OPTIMIZE trials [10, 12]. The brochure included facts about common side effects of CNS-active medications, how medications may need to be adjusted as our bodies age, and the importance of talking with your doctor or pharmacist to determine if any medications could be reduced or stopped. It mentioned some of the most commonly prescribed CNS-active medications (i.e., duloxetine, trazodone, gabapentin, mirtazapine, and bupropion) and included a QR code with a complete list of relevant medications. The brochure was serially revised with input from three focus groups of PLWD (n=12) and care partners such as a PLWD’s spouse, partner, or other family member (n=10) (Figure S1 in the supplementary material) [13].

Each clinic was assigned to be an intervention (n=10) or control (n=12) clinic to recruit 60 participants per arm. PLWD who received care at an intervention clinic were sent the educational brochure using the mailing address identified in the EHR. PLWD who received care at a control clinic did not receive an educational brochure and continued to receive care as usual.

2.3. Outcomes

After the educational brochure was mailed to the PLWD, we measured whether a CNS-active polypharmacy discussion was documented in the clinic EHR (implementation feasibility) and the change in CNS-active polypharmacy (preliminary effectiveness) within 120 days. We also invited clinicians at the intervention clinics to participate in a brief interview to gauge their acceptability of this intervention.

2.4. Data from Electronic Health Record

We collected the patient’s demographic characteristics (i.e., age, sex, and race) and medication-related characteristics (i.e., total number of medications and the name and total daily dose of each CNS-active medication). We used descriptive statistics to characterize patients overall and by treatment group and to identify the top 10 CNS-active medications during the baseline period. At baseline and during the last 45 days of the follow-up period, we measured the following outcomes: 1) total standardized daily dose (TSDD) of CNS-active medications (primary outcome), 2) number of unique CNS-active medications, and 3) presence of CNS-active polypharmacy (yes/no). To calculate TSDD, we summed the available supply of each CNS-active medication during the 45-day baseline period, divided each CNS-active medication by the minimum effective geriatric daily dose (MEGDD) for that medication, and summed this amount for all CNS-active medications, and then divided by 45 days [14]. For example, if a patient was prescribed a 45-day supply of citalopram 20 mg daily (MEGDD = 10 mg daily), gabapentin 300 mg three times daily (MEGDD = 900 mg daily), and zolpidem 5 mg daily (MEGDD = 5 mg daily), their TSDD would be 4.0. We used linear regression to examine the change in the continuous dependent variables (1) TSDD (primary outcome) and (2) the number of unique CNS-active medications by treatment group (independent variable). We used logistic regression to assess differences between CNS-active polypharmacy (yes/no) in the treatment and control groups from baseline to the follow-up period. For each model, we computed and applied inverse probability of censoring weights to account for six patients who died before the end of follow-up. Robust standard errors accounted for the clustering of patients within the clinics.

2.5. Data from Clinician Interviews

To gauge the acceptability of the intervention, we interviewed clinicians from intervention clinics using a standardized nine-item questionnaire after the intervention period was complete. Clinicians were recruited through recruitment emails from Principal Investigators and Study Coordinators. Interviews were conducted by telephone, and the discussion was audio recorded and transcribed for accuracy. All clinicians provided verbal consent prior to the interview. The guide included questions about whether PLWD came to their appointments wanting to discuss the intervention material related to CNS-active medications, how the clinicians made prescribing decisions for CNS-active medications for PLWD, and perceived barriers to reducing the use of such medications (Table S3 in the supplementary material). Following the interviews, two members of the study team (NB and MT) used an inductive rapid analysis approach to summarize interview notes and transcripts to identify overarching themes [15, 16]. Throughout the analysis, research team members with different backgrounds reviewed the data and participated in team meetings to reach a consensus on coding discrepancies, findings, and interpretation of the data. The research team also reflected on perceptions and potential biases to ensure the credibility of the analysis.

3. RESULTS

3.1. Patient Demographics

PLWD (n=129) were an average of 72.4 years old (SD 9.7), 62.8% of patients were women, and 84.5% were non-white (Table 1). On average, PLWD were prescribed 9.5 (SD 4.9) unique medications during the last 45 days of the baseline year. There were 61 PLWD from 10 intervention clinics and 68 PLWD from 12 control clinics. At baseline, PLWD in the treatment and control clinics were prescribed 3.6 (standard deviation [SD] 0.8) and 3.5 (SD 0.7) unique CNS-active medications, respectively, with baseline TSDDs of 9.0 (SD 7.2) and 9.0 (SD 7.2). The most prescribed CNS-active medication was gabapentin (43.4% overall) in both groups, followed by duloxetine in treatment clinic patients (37.7%), and no notably higher use of a particular CNS-active medication in control clinic patients.

Table 1.

Characteristics of persons living with dementia experiencing central nervous system-active polypharmacy by treatment group^a

Characteristic	Overall N=129b	Treatment Group N=61	Control Group N=68
Age in years (SD)	72.4 (9.7)	71 (8.4)	73.7 (10.7)
Sex
Female	81 (62.8)	40 (65.6)	41 (60.3)
Male	48 (37.2)	21 (34.4)	27 (39.7)
Race
White	109 (84.5)	56 (91.8)	53 (77.9)
Black	15 (11.6)	4 (6.6)	11 (16.2)
Asian	3 (2.3)	0	3 (4.4)
Other	1 (0.8)	0	1 (1.5)
Missing	1 (0.8)	1 (1.6)	0
Total unique medications per participant at baseline (SD)	9.5 (4.9)	9.4 (4.5)	9.5 (5.2)
Number of participants per health system
Health system A	80 (62.0)	37 (60.7)	43 (63.2)
Health system B	49 (38.0)	24 (39.3)	25 (36.8)
CNS medications at baselinec (SD)
TSDD	9.0 (7.2)	9.0 (7.2)	9.0 (7.2)
CNS-active medicationsd	3.5 (0.8)	3.6 (0.8)	3.5 (0.7)
Top CNS-active medications
Gabapentin	56 (43.4)	27 (44.3)	29 (42.6)
Duloxetine	35 (27.1)	23 (37.7)	12 (17.6)
Trazodone	25 (19.4)	12 (19.7)	13 (19.1)
Bupropion	24 (18.6)	12 (19.7)	12 (17.6)
Sertraline	24 (18.6)	11 (18.0)	13 (19.1)
Oxycodone	19 (14.7)	10 (16.4)	9 (13.2)
Quetiapine	19 (14.7)	9 (14.8)	10 (14.7)
Escitalopram	18 (14.0)	10 (16.4)	8 (11.8)
Hydrocodone	17 (13.2)	8 (13.1)	9 (13.2)
Mirtazapine	16 (12.4)	9 (14.8)	7 (10.3)

SD, standard deviation; TSDD, total standardized daily dose; CNS, central nervous system

^aData are presented as N (%) unless otherwise stated

^bIncludes n=6 who died during follow-up

^cDuring the last 45 days of the baseline year, i.e., the baseline period

^dMedian (quartile 1, quartile 3) for overall, treatment, and control was 3 (3, 4).

3.2. Implementation Feasibility

All PLWD in the intervention arm (n=61) were mailed brochures, of which 1 (0.7%) was returned due to an incorrect address. During the 120-day intervention period, there were no documented conversations related to the educational brochure or CNS-active medications.

3.3. Preliminary Effectiveness

When comparing patients in intervention and control clinics, there was no between-group difference in change in CNS-active TSDD from baseline (−1.6 [standard deviation (SD) 6.0] vs. −1.3 [5.8]; treatment effect estimate [TEE] = −0.5, 95% confidence interval [CI] −2.5, 1.5). We did not find a between-group difference in change in the number of CNS active medications from baseline (−1.0 [intervention] vs. −1.0 [control]; TEE −0.01, 95% CI −0.6, 0.5) nor a change in the percentage of patients with CNS polypharmacy following the intervention (52.6% [intervention] vs. 50.4% [control]; TEE odds ratio = 1.0, 95% CI 0.5, 1.9).

3.3. Clinician Interviews

Four primary themes emerged during the clinician interviews (n=10) (Table 3). First, clinicians often inherit PLWD who are experiencing CNS-active polypharmacy from older physicians who have recently retired. These physicians almost universally believed that the problem of overprescribing CNS-active medications is improving over time as a function of changes in clinical practice, with more judicious prescribing. Second, clinicians perceived PLWD or their care partners as being reluctant to consider deprescribing, particularly as it related to opioids or benzodiazepines. Third, clinicians did not have adequate time during the clinic to discuss deprescribing CNS-active polypharmacy medications. Finally, clinicians believe there was a lack of timely access to safer pharmacologic or non-pharmacologic alternative treatments.

Table 3.

Representative quotes from clinicians at intervention clinics (n=10)

Theme	Representative quotes
Central nervous system-active polypharmacy is an inherited problem that is improving over time	“I think [the] bigger issue is… with patients that come to me that were seeing older physicians that have retired because [over-prescribing CNS-active medications] was more of the norm 20 years ago but now…there are other options.” (Clinician #6) “Part of this will get better…as generations of doctors are trained to be more cautious with [overprescribing]...Usually, it’s patients that have been with their doctor for 30 years and [there] was just a different prescribing practice that people were trained in [back then].” (Clinician #3)
Persons living with dementia are reluctant to deprescribe	“A lot of times [patients] are not interested [in deprescribing CNS-active medications]. They feel like… it’ll be worse trying to come off of it or be off of it.” (Clinician #5) “Barrier number one is the reluctance of the patient… They say “Oh, no, I need it. I've been on it. I cannot survive without it.” (Clinician #8)
There is not enough time for deprescribing conversations in clinic	“But that conversation [on deprescribing] takes time… This is one of 100 things I need to talk about with them today.” (Clinician #4) “We are very limited on time… I think that’s the biggest limiting factor. We just simply cannot do all of these things all at once.” (Clinician #8)
Access to safer pharmacologic and non-pharmacologic treatments is needed	“We need more alternatives, and not only do we need more alternatives, but we need to have quick access.” (Clinician #2)

4. DISCUSSION

This pragmatic pilot trial implemented a direct-to-patient education intervention to prompt the deprescribing of psychotropic and opioid medications in PLWD experiencing CNS-active polypharmacy. Though the current study was supported by participating clinic physicians who identified CNS-active polypharmacy as a problem, the intervention did not demonstrate implementation feasibility, as there was no EHR documentation of CNS-active medication conversations between PLWD and their clinicians during the 120-day intervention period.

The presumed intervention mechanism was that the PLWD or care partner would discuss CNS-active polypharmacy with their clinician after receiving the educational brochure. However, there was no documentation that these conversations occurred. In addition, no clinicians indicated that their patients discussed receiving the intervention material. Given the absence of evidence of implementation feasibility, it is unsurprising that we did not demonstrate preliminary effectiveness, given there was no difference in the number of CNS-active medications or TSDD when considering both the intervention and control groups. However, we acknowledge that this pilot was not designed (i.e., powered) to demonstrate statistically significant effectiveness.

Overall, the clinicians interviewed agreed that CNS-active polypharmacy is a clinical problem worth addressing; however, they identified barriers to deprescribing in their practice consistent with prior work [17]. One common challenge they identified was that medications were often initiated by other prescribers, leading to inadequate information about the decision-making process when starting the medication (e.g., the rationale for this medication and alternatives that were considered) [18]. Clinicians also acknowledged the limited time to discuss important health topics, including deprescribing, during appointments. While this is a commonly reported systems-level barrier, it may be particularly challenging when conducting appointments with PLWD and their care partners [19, 20]. Furthermore, clinicians believed that some older adults may not be receptive to deprescribing. This aligns with research indicating that older adults vary in their openness towards deprescribing medications and interest in actively engaging in the deprescribing decision-making process [21–23].

Given the time constraints during medical appointments, one alternative that has been explored is population-based education about deprescribing. The OPTIMIZE pragmatic cluster randomized trial concluded that an educational brochure addressing polypharmacy via deprescribing among PLWD who were prescribed five or more medications did not significantly reduce the number of long-term medications or potentially inappropriate medications [24]. In contrast, in the EMPOWER cluster randomized trial, 27% of older adults who received an eight page educational booklet that included a stepwise tapering recommendation stopped their benzodiazepine at six months compared with 5% of older adults in the control group [10]. Similar to the EMPOWER trial, we developed our educational brochure using the social constructivist learning and self-efficacy theory to encourage PLWD to take action to address their CNS-active polypharmacy. However, given that we were targeting multiple types of CNS-active medications, we encouraged PLWD or their care partners to contact their prescriber to discuss their idea of stopping a medication rather than providing a tapering plan. In contrast, older adults participating in the EMPOWER trial received information that allowed them to self-taper their benzodiazepine. In this study, PLWD may have experienced barriers to acting on the educational brochure, such as difficulty scheduling the visit, the cost of an additional medical appointment, or forgetting to bring up the information during their medical appointment. Furthermore, older adults who are reluctant to deprescribe may value more detailed information about replacement medications, tapering strategies, and reassurance that the medication can be restarted if needed [25].

In our study, clinicians shared that PLWD may have reservations about stopping medications such as benzodiazepines or opioids. This aligns with a recent nationally representative survey, which found that older adults may be less interested in stopping medications to manage symptoms (e.g., pain) as opposed to preventing future health problems (e.g., HMG-CoA reductase inhibitor to decrease the risk of a cardiovascular event) [26]. CNS-active medications are typically prescribed to PLWD to address current symptoms, which may pose a challenge to deprescribing. For example, some older adults have reported a strong desire to continue their long-term benzodiazepine use because of fear of withdrawal effects or worsening health symptoms if the medication is discontinued [27, 28].

The primary limitation of this pilot study was that it was conducted with a small number of PLWD at two health systems. Furthermore, clinics were assigned to the intervention or control arm in collaboration with the primary care clinical leadership, as it was believed that some clinics would experience barriers to participation. Our assessment of implementation feasibility relied on clinical documentation rather than observed encounters; it is possible that some PLWD or their care partners brought up concerns about CNS-active medications during a medical appointment, but it was not documented in the EHR due to time constraints and therefore was not captured in this study. The follow-up period was 120 days, compared to 180 days in the EMPOWER and OPTIMIZE trials, which provided less time for medication-related conversations during the intervention window. Finally, sole reliance on a mailed brochure in this population without additional prompts to the clinician, PLWD, or care partner may contribute to the lack of effect, which is consistent with two larger studies of similar approach published after this work was underway [10, 29].

5. CONCLUSION

The direct-to-patient mailed educational brochure did not demonstrate feasibility in provoking deprescribing conversations between PLWD and clinicians or preliminary effectiveness in decreasing CNS-active polypharmacy. Educational interventions may not be sufficient to overcome the challenges of clinical practice, particular for patients with cognitive impairment. Additional research is needed to identify and test alternative patient-centered strategies or combinations of strategies that could feasibly be implemented in routine clinical practice to encourage deprescribing of potentially inappropriate CNS-active medications among PLWD.

Table 2.

Preliminary intervention effectiveness^a

Outcome	Treatment Group	Control Group	Treatment Effect Estimate (95% CI)b
Change in TSDD, mean (SD)	−1.6 (6.0)	−1.3 (5.8)	−0.5 (−2.5, 1.5)
Change in CNS-active medications, mean (SD)	−1.0 (1.3)	−1.0 (1.3)	−0.01 (−0.6, 0.5)
CNS polypharmacy present, %	52.6	50.4	1.0 (0.5, 1.9)

CI, confidence interval; TSDD, total standardized daily dose; CNS, central nervous system-active.

^aEstimates and models have been weighted using inverse probability of censoring weights to account for death (n=6 died); robust standard errors were used to account for the clustering of patients within the clinic.

^bThe model for change in TSDD from baseline includes treatment group and TSDD at baseline; the model for change in number of CNS-active medications from baseline includes treatment group and number of CNS-active medications at baseline; the model for CNS polypharmacy during follow-up includes treatment group and number of CNS-active medications at baseline. The odds ratio is the treatment effect estimate reported for CNS polypharmacy during follow-up.

Key Points:

• Persons living with dementia (PLWD) were mailed an educational brochure designed to provoke conversations with their primary care clinicians (‘clinicians’) about deprescribing CNS-active medications (intervention) or received no brochure (control).

• Clinicians did not document any conversations related to CNS-active polypharmacy with any PLWD.

• There were no differences in CNS-active medication prescribing among PLWD in the intervention group compared to the control group.

• The direct-to-patient mailed educational brochure did not demonstrate feasibility in provoking deprescribing conversations between PLWD and clinicians or preliminary effectiveness in decreasing CNS-active polypharmacy.

III. Availability of data and material:

This study was conducted with a waiver of informed consent for the main intervention, so participants did not consent to the sharing of their data. Therefore, because of privacy and confidentiality regulations, data are not available on request. The interview guide is available in the supplemental appendix.