INTRODUCTION

An estimated 11 million Americans are living with cognitive impairment,1, 2 a condition associated with increased disability, neuropsychiatric symptoms and health care costs, and decreased quality of life.3, 4 People with cognitive impairment have a high burden of coexisting medical conditions, making them likely to experience polypharmacy, potentially inappropriate medication (PIM) use, medication side effects, adverse drug events, and treatment burden.1, 5,6,7,8,9,10 For example, people with dementia take 5–10 medications, on average, of which 1–2 are prescribed for dementia and the remainder for coexisting, non-dementia illnesses.6 One-third of people with dementia take PIMs, in which the harms of medications may outweigh the benefits.6

For clinicians, achieving optimal medication use for older adults with cognitive impairment is challenging. Decisions about medication use in such patients often carry trade-offs between quality and length of life, or between different facets of quality of life. For example, many medications commonly used to treat incontinence, insomnia, and pain have anticholinergic properties11,which can cause dry mouth, constipation, falls, and delirium.12,13,14,15,16 Anticholinergic medications may also be associated with long-term, permanent cognitive decline.17 Given the prolonged and variable course of cognitive impairment, determining whether a medication to prevent or treat a coexisting condition is likely to result in benefits or harms during the person’s remaining lifespan is complex. Medications that were once appropriate may become inappropriate as cognition declines; exactly when this shift occurs may be difficult to determine. Lack of clinical trial data, guidelines, or decision support tools applicable to patients with cognitive impairment leaves clinicians forced to make prescribing decisions without much guidance.18

Optimizing medication use for people with cognitive impairment often involves deprescribing—tapering or stopping drugs to reduce inappropriate polypharmacy. Two recent reviews of deprescribing emphasized the importance of shared decision-making with patients and family caregivers.19, 20 When patients and caregivers engage in shared decision-making, they become better informed about risks and benefits and tend to choose fewer medicines.21 However, one barrier to optimizing prescribing may be clinician discomfort in discussing topics such as life expectancy and in elicitation of patient and caregiver preferences.18 This barrier may be heightened in people with cognitive impairment due to communication difficulties and weakened decision-making capacity.22 Furthermore, people with cognitive impairment are often accompanied to medical visits by family members or other companions, meaning that clinicians must be skilled at engaging and weighing multiple perspectives during the encounter.23

How clinicians communicate about medications during the clinical encounter may affect to what extent patients and family companions understand and participate in decision-making about medication use. This is particularly important for people with cognitive impairment because they face more complex decisions and greater risks from medications. Little is known about how clinicians can best facilitate shared decision-making about optimal medication use among patients with cognitive impairment and their companions. To begin developing a framework for fostering shared decision-making about medication use for people with cognitive impairment, the objective of this study was to characterize how primary care clinicians currently discuss medications during encounters with older adults with cognitive impairment and their family companions. We used qualitative methods to analyze medication discussions from audio recordings of patients’ primary care visits from the SAME Page Study, a randomized controlled trial to examine the effects of a patient-family agenda setting checklist on primary care visit communication among older adults with cognitive impairment.24

METHODS

Overall Study Design

We performed a qualitative content analysis of audio-recorded clinical encounters from the SAME Page trial. The trial did not intervene upon clinicians and did not address optimal prescribing and deprescribing. Further details are published elsewhere.24

Study Participants

SAME Page enrolled 93 patient–companion dyads at 2 primary care clinics and 1 hospital-based geriatrics clinic in Baltimore, Maryland. Eligible patients were 65 or older, English speaking, had at least 1 incorrect answer on a cognitive screener,25 and attended primary care visits with a relative or unpaid companion. Eligible clinicians were practicing physicians, nurse practitioners, or physician assistants at the 3 participating primary care practices. Fourteen clinicians met eligibility criteria and agreed to participate. The study was approved by the institutional review board of the Johns Hopkins Bloomberg School of Public Health.

Data Collection and Analysis

Patient diagnoses and medications were abstracted from the electronic medical record. The Mini-Mental State Examination (MMSE) was administered by research staff at the enrollment visit. Clinician characteristics were assessed from surveys administered at the time of informed consent. Dyads in the intervention arm completed a self-administered checklist in the waiting room to establish a shared visit agenda. Dyads in the control group waited for their visit as usual. Patients and companions also completed a 2-week follow-up questionnaire, which included questions about medication adherence and communication with clinicians, both in general and specifically relating to prescription medications. Response categories for the communication question were poor, fair, good, and excellent. Visits were audio-recorded and transcribed verbatim. All 93 recordings were included in the qualitative analysis.

We developed a preliminary coding scheme after an initial reading of 6 transcripts by a multidisciplinary team (A.R.G., D.E., C.M.B., and J.L.W.) with expertise in geriatric medicine, deprescribing, family caregiving, dementia, and qualitative research. Four investigators (A.R.G., M.C., D.S., and A.P.) refined the coding scheme by reading 12 additional transcripts. Coding was subsequently performed by these 4 team members using Atlas.ti, version 8 (ATLAS.ti Scientific Software Development). Each transcript was independently read by 3 coders. The constant comparative approach was used to qualitatively analyze the content of the transcripts.26, 27 Open coding allowed inductive identification of new themes in addition to deductive coding28 guided by shared decision-making frameworks.29, 30 At weekly meetings, the transcripts were reviewed and assessed for the emergence of new ideas or themes. Differences were reconciled by consensus until 100% agreement was reached. Revisions to the coding scheme were applied to all previously coded transcripts. Content analysis generated major themes and subthemes. Visits occurred between August 2016 and August 2017. Data analysis was performed between June 4 and July 27, 2018.

RESULTS

Clinician Characteristics

Clinicians were on average 49.3 years of age (SD, 11.1; range, 29–62). About half reported practicing 15 or more years (n = 8; Table 1). Clinicians were trained in family practice (n = 6), internal medicine (n = 3), or were nurse practitioners or physician assistants (n = 5). Most (n = 5 of 6) clinicians with geriatrics training practiced at the geriatrics clinic.

Table 1 Characteristics of Primary Care Clinics, Clinicians, and Study Participants
Full size table

Patient and Companion Characteristics

Patients were on average 79.9 years of age (7.6). About half of enrolled patients (49.5%) had a diagnosis of dementia, cognitive impairment, or symptoms of cognitive impairment recorded in their electronic health record. We will hereafter refer to this population as having cognitive impairment, a category which ranges from mild to severe. Patients’ average MMSE score was 21.6 (6.8). Most companions were female (75.3%) and were spouses (39.8%) or adult children (55.8%) of patients; few (5.4%) were friends or other relatives.

Nature of the Encounters

The visits lasted a mean of 25.2 min (SD, 13.8; range, 6.01–57.0). The majority of patients (52%) and companions (63%) who responded to the question about quality of medication communication with the clinician rated it as excellent (Appendix Table 4). Our qualitative content analysis clustered into three major categories, or themes, and associated subcategories. Below, we describe each theme; selected representative quotes are presented in the tables.

Theme 1: Key principles of optimal prescribing discussed by clinicians

We found numerous examples of clinicians introducing patients and companions to key principles of optimal prescribing for older adults with cognitive impairment. These principles are presented in Table 2, along with representative quotes. Clinicians talked of scaling back the intensity of certain medications and other interventions as a routine part of care in order to avoid treatment burden and improve quality of life, as in this encounter with an 82-year-old patient with a MMSE score of 26 who routinely saw four different physicians and took multiple medications that required monitoring (digoxin, warfarin, and levothyroxine); the patient’s health status was further complicated by “sundowning”:

Table 2 Key Principles of Optimal Prescribing Discussed by Clinicians
Full size table

Clinician: One of the things that I like to do is to try and simplify a person’s life and I do that by simplifying medications where I can and…also…by trying to simplify doctors that you see.

Clinicians further laid the groundwork for conversations about deprescribing by explaining that a person’s needs, and the balance of benefits and harms of medications, change with aging and cognitive impairment. They often prioritized function and quality of life over the traditional, disease-based approach to care, particularly for older adults with more severe chronic conditions or poorer cognitive status. They explained the concept of diminishing returns—i.e., when the risk of a negative health outcome, such as a cardiovascular event, can be reduced in several different ways, the benefit of each successive intervention is modest and further attenuated by iatrogenic harm. Clinicians helped patients and companions understand that medications usually do not “fix” every symptom and that they carry risk. They also recognized the uncertainty of the medical evidence applicable to older adults with dementia and coexisting conditions, thereby acknowledging the preference-sensitive nature of medication decisions.

Theme 2: Clinician strategies to establish partnership with patients and companions

Clinicians used a variety of approaches to develop partnerships with patients and companions and to foster shared decision-making about deprescribing: eliciting the patient or companion perspective; providing reassurance that deprescribing is not a withdrawal of appropriate care; and providing reassurance about ongoing care and monitoring during the deprescribing process. For example, in this encounter with a 78-year-old man (MMSE 26), the clinician first made the patient aware that his anticholinergic medication could be contributing to his recurrent falls, and then sought the patient’s perspective on stopping the medication:

Now we have an event that is happening [i.e., falls], we need to modify. Down the road, if I want to take you off of the amitriptyline and replace it with something for neuropathy, would you be okay with that?

The clinician also emphasized that he did not intend to withdraw appropriate care:

I want to get to know you a little bit more…And then work together on minimizing your medications. Understanding that I need to take care of your neuropathy, I need to make sure you're sleeping okay, and I need to make sure your blood pressure is okay. So I will not sacrifice what needs to be done, but I will try to minimize the amount of medicine that we use.

Phrases such as “Would you be okay with that” and “I want to get to know you” foster elicitation of the patient’s priorities. Clinicians also reassured patients and companions that they would monitor closely for returning symptoms after stopping a medication, as in this encounter with an 81-year-old patient (MMSE 26): “I think you’re going to do fine without [amlodipine]. We’ll check you again down the…short road.”

Theme 3: Factors complicating optimal prescribing

Clinicians had to balance the principles of optimal prescribing against other considerations. These considerations are presented in Table 3. Factors complicating optimal prescribing included the following: distressing behavioral and psychological symptoms of dementia; patient or companion expectations of receiving a prescription; discordance between clinician recommendations and patient or companion preferences; the challenges of triadic communication, particularly involving people with cognitive impairment; and the complexity of the patient’s health status, missing records, or technical or logistical issues—which sometimes left little time for discussions about optimal prescribing or deprescribing, even when patients or companions explicitly mentioned concerns about medication appropriateness or treatment burden. For example, medication reconciliation took up a large portion of some visits, particularly when the patient was accompanied to the visit by a companion who was not a primary caregiver.

Table 3 Factors Complicating Optimal Prescribing
Full size table

Clinicians struggled with balancing disease-based guidelines against quality of life, treatment burden, and side effects, as in this interaction between a clinician and the companion of an 83-year-old woman with a MMSE score of 12:

Clinician: The lisinopril and the metoprolol both are good for her heart and make it pump better. So I’d rather have that and put up with a little bit of a problem than have her going into congestive heart failure. But if [her dizziness] gets worse we can always take that lisinopril [and] drop it down to 2.5 [mg].

Clinicians had to navigate between patients and companions in their communication, which further complicated attempts to discuss optimal prescribing. With patients who had more advanced cognitive impairment and spoke little, clinicians tried to address them directly, to the extent possible, and determine what mattered most to the patient, as in the following interaction with a 74-year-old patient with a MMSE score of 12:

Companion: A lot of times… she sits there and cries… in pain and doesn't ask for it… And I have to kind of ask her sometimes and then I get frustrated...

Clinician: [Addressing patient] Would it make it better if we just make it that you get [acetaminophen]… a couple of times a day so you don’t even have to ask?... I hate to add it on to the other drugs… Miss [Patient], is there anything that you can think of that I can do to help to make your life better?

Clinicians were particularly challenged to engage patients who had poor insight due to cognitive impairment, or when there was discord between patients and companions, as in the following interaction with a 66-year-old patient with an MMSE score of 13:

Companion: It’s still like kind of sporadic where there’s confusion with pictures as actual people being in the house.

Patient: ...I beg to differ...

Clinician: [Addressing companion] How many times do you notice, like, a week?

Companion: Maybe once or twice a day...

Patient: [Addressing companion] It really infuriates me when you don’t listen to me and you jump to so many conclusions... I’m not crazy. I’m not stupid...

Clinician: [Addressing patient] He just wants to let us know just so we can make sure that you’re not getting any more infections or nothing else is happening.

In such interactions, clinicians often had to obtain history from the companion and then assure the patient that they were not being undermined to avoid upsetting them.

DISCUSSION

In this qualitative, observational study, we characterized primary care clinicians’ approaches to discussing medication use during encounters with cognitively impaired older adults and their family companions. Despite high rates of polypharmacy and PIM use, few investigations have detailed the elements of communication about medication use in clinical care. To our knowledge, this is the first study to use audio-recordings of actual conversations between clinicians and older adults with cognitive impairment and their companions to describe the current state of communication about medication use.

Deprescribing protocols have been developed20, 32, 33; however, a barrier to their implementation is that clinicians find it challenging to elicit patient preferences and goals and to discuss concepts such as disease trajectory, estimated lifespan and ambiguity about benefits and harms—particularly given high levels of clinical complexity and time limitations.34, 35 We present examples in which clinicians introduced deprescribing to patients and companions during primary care encounters, with the goal of beginning to develop a framework for fostering shared decision-making about medication use for people with cognitive impairment. Clinicians discussed concepts such as the uncertainty of the evidence for many medications, particularly for people with cognitive impairment and other chronic conditions—a critical part of informed decision-making. We also observed clinicians asking patients and companions what health problems and symptoms they wanted to prioritize—treating the person, rather than the disease—and making it clear that deprescribing recommendations were designed to improve quality of life, not because the patient was no longer “worth treating.”19 If confirmed in different types of populations, our results suggest that it is possible for clinicians to facilitate conversations about deprescribing as part of the routine care of people with cognitive impairment.

Previous research has shown that patients and companions may not recognize the potential harms of medication use or be aware that deprescribing is possible. For example, a recent study found that most caregivers of people with dementia believed antipsychotic medications used to treat behavioral and psychological symptoms of dementia were safe, despite these drugs’ well-documented potential for severe side effects.36 Other studies have found older adults and companions to be open to the idea of deprescribing if their clinician recommends it,37, 38 but often feel that they are unable to obtain information about medications or make their preferences heard.39, 40 By analyzing the content of conversations about medication use, our study offers additional insight into key language clinicians can use to initiate discussions about deprescribing, as well as barriers they face in doing so. For example, it would appear from these transcripts that clinicians often struggle to deprescribe preventive medications in people with cognitive impairment, in part because of the difficulty of extrapolating risk–benefit data to individual patients, and in part because prescribing is seen as a positive action, while deprescribing is not.19 Risk prediction tools available at the point of care would enable clinicians to calculate how much each recommended medication (e.g., ACE inhibitor and beta-blocker for a patient with systolic heart failure) contributes to absolute risk reduction and whether the benefit of using both medications justifies the potential harms for patients with cognitive impairment, who are at increased risk of adverse drug effects.31 Existing risk prediction calculators41 and decision aids21 generally do not incorporate cognitive impairment and may perform poorly for older adults with multiple competing risks of mortality.42 Future research is needed to generate evidence to inform the development of risk prediction calculators applicable to people with cognitive impairment.

Our data do not enable us to comment on encounters in which neither the clinician nor the patient or companion raised the issue of medication appropriateness or deprescribing. However, we identified missed opportunities to discuss optimal prescribing, even when patients or companions explicitly mentioned concerns about medication appropriateness or treatment burden. Cognitive impairment is one of the most challenging illnesses that primary care clinicians address because of the behavioral and psychological complications and because of structural barriers.43 This complexity, compounded by logistical obstacles (e.g., locating records from other clinicians or sites of care and reconciling medication lists), often left little time for discussions about optimal prescribing. For example, the transcripts revealed that medication reconciliation was time-consuming and complicated, particularly when companions were poorly equipped to clarify medication regimens; this is in line with a previous study that found that 96% of primary care provider medication lists contained discrepancies compared with what the patient was actually taking.44 In other cases, such as the example involving the companion of the 100-year-old patient with a MMSE score of 9, clinicians attempted to deprescribe (or to avoid starting a PIM) but the interactions went astray. Even if a clinician recognizes a medication as potentially inappropriate or a candidate for withdrawal, talking about deprescribing is difficult, especially in the context of cognitive impairment, with its prolonged and uncertain trajectory. It involves eliciting patients’ goals and preferences, conveying complex benefit–harm information, and broaching the subject of illness trajectory without being perceived as giving up on the patient.18, 35 Future efforts should focus on providing clinicians with frameworks or conversation guides to facilitate shared decision-making about medications. Such guides can be modeled after existing resources developed for advance care planning.45 Additional studies are needed to elicit patient and caregiver preferences regarding clinician communication about deprescribing and to examine how such communications are received and interpreted. These are important next steps before designing deprescribing conversation guides for clinicians as well as patient-directed educational materials. Patient-directed educational materials are essential to foster shared decision-making around medication use, and there is strong evidence that such materials work.46, 47 In addition, further research is needed to understand how other health care providers, such as pharmacists and nurses, can contribute to deprescribing.48 Lastly, strategies are needed to help clinicians overcome complexity and technical or logistical hurdles during visits for patients with cognitive impairment and leave time for important discussions about medication appropriateness.49

This study has several limitations. Each of the examples we present reflects a single visit; a clinician may not have addressed deprescribing but may have done so during a previous interaction. We are only able to comment on how clinicians discussed deprescribing, not on situations in which they did not discuss it. The visits were recorded as part of the SAME Page trial, which may have affected whether medications were discussed but not how. One of the items on the SAME Page checklist was “managing or taking medications.” Overall, 29% of patients and 33% of companions identified this as a visit priority. It is possible that inclusion of this item on the checklist encouraged patients and companions to discuss medications during the visits. It is also possible that there was a carryover effect for clinicians, such that they were more likely to discuss medications with patients and companions in both groups because exposure to patients in the intervention group altered their behavior. We also do not know how many of these conversations resulted in a drug being discontinued or not prescribed unless it was stated during the conversation. The study was conducted in the Baltimore metropolitan area and two of the clinics are affiliated with Johns Hopkins University (although the clinicians do not have academic appointments); therefore, the results may not generalize to other locations. The clinicians knew they were being recorded, which may have affected their behavior. Finally, we strived for objectivity but assessment of the elements of communication is inherently subjective. Certain quotes could be classified in multiple themes. Each transcript was independently read by 3 coders, with disagreements resolved through discussion.

CONCLUSION

Achieving optimal medication use for older adults with cognitive impairment is challenging.1, 5,6,7,8,9,10 How best to foster informed decision-making about optimal medication use in older adults with cognitive impairment is unknown. Based on these transcripts, we identified many instances in which primary care clinicians incorporated deprescribing into routine clinical care. These examples can be used to help create deprescribing conversation guides for clinicians, as well as patient-directed educational materials. Yet, we also identified numerous challenges that prevented clinicians from working together with patients and companions to optimize prescribing. Future research should develop and test deprescribing interventions that seek to overcome these barriers for the growing population of older adults with cognitive impairment.