Design

This descriptive qualitative study [20–22] included conducting and analyzing semi-structured one-on-one interviews with patients, their family caregivers, and healthcare providers that were involved in an assessment of possible prescribing cascades. The COREQ criteria were used to guide this report [23] (S1 Checklist).

Setting

Interviewees were recruited through physician referral from a Geriatric Day Hospital (GDH) program, two long-term care (LTC) homes in Ottawa, Ontario, and through self-referral across Ontario, Canada.

Researcher characteristics

The three investigators have a health care background (two PharmD–BF and LMM, and one RN/PhD—LJ); all three are affiliated with different research organizations. A pharmacist research assistant (BScPhm)–PH, conducted the majority of interviews, and a Masters-trained research associate, EG, with expertise in qualitative research supported data analysis. One pharmacist investigator, BF, also works clinically in the GDH program and only conducted interviews with research participants associated with the other recruitment sites. She and the research associate coded transcripts; all investigators reviewed code summaries and were involved in determining themes. All team members were female.

Participants, purposeful sampling and consent process

Patient participants (>65 years old; able to complete an interview in English) were identified using two approaches. First, for those from the GDH program and those in the two LTC homes, on-site physicians and pharmacists identified and approached people who may have experienced one or more prescribing cascades about their interest in the study. Site presentations and recruitment posters provided recruiting clinicians with a list of possible prescribing cascades (Fig 1). If interest was confirmed, the clinician provided the research team with a completed screening form specifying the suspected prescribing cascade(s) and providing contact information for the person. The research team first contacted the person by phone and then obtained written consent in a face-to-face appointment. Second, we offered a self-referral option for members of the public. These participants were recruited through advertising across Ontario using social media advertisements (Twitter, Facebook); newsletter announcements; and posters in pharmacies, family practice clinics, local community centres, retirement homes, and libraries. Electronic advertisements were distributed through email list servs of the RTOERO (https://rtoero.ca/about/), an organization representing >80 000 retired education workers (sample advertisement, S1 Appendix). Self-referring individuals were contacted by phone and oral consent was obtained and documented as per REB protocol.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. What are examples of prescribing cascades?

https://doi.org/10.1371/journal.pone.0272418.g001

Following the consent and interview process, patient participants were asked to provide contact information for and permission to contact a caregiver and/or healthcare provider(s) who could provide additional information about the prescribing cascade experience. These individuals were contacted separately by hand-delivered invitations (e.g., to pharmacies or family practices), by registered mail or by phone for consent (with signed consent form returned by mail or email).

The study was approved by the Bruyère Research Ethics Board (REB) (protocol #M16-17-009) and the University of Toronto Research Ethics Board (#36481).

Data collection

Interviews were conducted between July and December 2019 by the pharmacist research assistant (PH) and one pharmacist investigator (BF). Interviewees could opt for in person or telephone interviews. Only the participant and interviewer were present; there were no repeat interviews. Interview guides (patient, caregiver, healthcare provider versions, S2 Appendix) adapted from the pilot interviews [19], focused on factors associated with the development, identification and resolution of their suspected prescribing cascade(s); the resulting clinical impact; and participant insights regarding potential interventions to prevent, identify or resolve prescribing cascades. Interviewers completed memos following each interview detailing overall impressions and recommendations for questions of the other interviewees involved in the case. All interviews were audio recorded and transcribed verbatim. Transcripts were de-identified and verified for accuracy against recordings by the research associate, then paired with interviewer memos.

Data analysis

After the first three interviews had been transcribed (August 2019), the transcripts were distributed to the research team for familiarization, an important preliminary step in qualitative analysis when data collection and coding are fulfilled by different members of the research team [24, 25]. In our case, multi-disciplinary research team members were affiliated with different institutions in two cities, thus, early familiarization, and sharing of observational notes at monthly teleconference meetings, supported a coherent, collaborative analysis and introduced an element of reflexivity to the process [26]. Familiarization was on-going throughout the analysis process and all members of the research team participated regardless of whether they also participated in coding.

Initial coding of the first three transcripts was conducted by two members of the research team (the research associate–EG, and one pharmacist investigator—BF) in August 2019. Both analysts independently applied an inductive coding strategy. Codes and supporting quotations were compared and discussed during a full team meeting during which proposed codes were reviewed for inter-coder and source data agreement; codes were correspondingly accepted, altered, merged, divided, or dropped. Based on this review, a working code book was created and the first three transcripts re-coded by two analysts; they met to discuss coding agreement and code clarity and refined the codebook by clustering codes and insights into categories (August-September 2019). The same two members of the research team subsequently completed all coding.

Coding was on-going with completed transcripts distributed to the analysts in batches approximately every two weeks (August–December 2019); new transcripts were divided between the two coders so only one coder was responsible for a given transcript. Every two weeks, coders reviewed one of the other coder’s completed transcripts to ensure consistency of code application and inhibit coder deviation from the established code book [27]; verifying agreement on coding served to improve the rigour of the coding process [26]. Following this review, the coders met to discuss any issues, ideas, or concerns, e.g. emergent or weak codes, disagreements in application of codes, and the codebook was updated accordingly with older codebooks being retained for verification purposes.

Transcript analysis was conducted using NVivo 12 [28], allowing for the disaggregation and sorting of coded data and facilitating inter-coder consistency and transparency. Following the completion of the coding process, node and sub-node summaries were reviewed by the full research team during a face-to-face analysis retreat (January 2020) to identify patterns, categories, and themes [29]. Coding summaries were completed, noting overlaps, redundancies, and potential inter-node relationships, and subsequently verified by the two coders, who then compiled a final analysis report for review by the team.

An audit trail includes raw data (recordings, transcripts, memos), codebook versions, coded transcripts, meeting minutes and reports filed by date.

1. Recognition of prescribing cascades is linked to awareness of medication side effects

While participating patients usually said they understood that a medication could be used to treat the side effect of another medication, they had not typically heard the term ‘prescribing cascade’ and did not apply it to their own medication experiences. Most patients and caregivers demonstrated a low level of knowledge and understanding of medications, usually recalling only a very general reason for the medication’s use and approximate duration. They reported they had not received information about side effects or, if they received information, it was not easily understandable, resulting in poor knowledge of potential side effects.

While many patients expressed frustration about this lack of knowledge about medications and their side effects and an interest in knowing more, few articulated strategies for gathering such information. Many tended to trust their physician to know how a medication might affect them:

“I believed right from childhood that the doctor, he knows more than me. So… I blindly took what the doctor prescribed me”. [P104]

However, they still expressed uncertainty at how knowledgeable physicians might be regarding medications, pointing to the narrow scope of some physician practice (e.g. specialists) and to their reliance on pharmacists to “figure it out”. [P104] Patients who related the highest amount of understanding of their medications described significant personal efforts to find information, including conducting independent research, as they sought alternatives to perceived issues and actively engaged with their healthcare providers, particularly their pharmacists.

Amongst physicians, awareness and knowledge of specific prescribing cascades was more prevalent but they did not describe a common approach to successfully identifying them in practice. The ability to recognize specific cascades was influenced by physicians’ training, practice experience and exposure to prescribing cascades:

“I work in an environment where we have patients with a significant degree of polypharmacy and I’ve worked there for some years now so I think it’s just generally on my radar and also on my background in care of the elderly means that I have a significant degree of training around polypharmacy and deprescribing”. [P103-MD409]

One physician stated that through involvement in this study, they were more aware of examples of prescribing cascades and this made it easier for them to recognize them:

“Up until the last year, I would have never even equated elevated blood sugar with hydrochlorothiazide”. [R202-MD401]

Physicians and pharmacists suggested that aspects of their formal education subsequently influenced their different approaches to assessing new symptoms and ability to recognize potential medication side effects and thereby potential prescribing cascades. Medical education was described as a disease-focused approach to symptom management:

“It’s not kind of in the rubric of the physician having not been formally trained in pharmacology, for example, to sort of immediately question, oh, could this symptom be caused by a medication. So, it’s more common I think and more usual for the physician to say, oh, what does this symptom represent in terms of a new medical condition and can I treat this?” [P101-MD406]

Pharmacy education was described as being aligned with a medication-focused approach with emphasis on linking symptoms to potential medication risks. These divergent educational paradigms were suggested as a source of physicians’ self-reported challenge in recognizing emergent signs and symptoms as potential side effects associated with causative drugs.

Healthcare providers relied on patients reporting emergent symptoms and indicating a relationship to a new medication; when this did not occur, they were less likely to investigate the possibility of side effects and more likely to treat a new sign or symptom as an emerging medical condition. The likelihood of recognizing symptoms as potential medication-related side effects decreased over time i.e., health care providers were less likely to consider that new side effects can arise from long-term medications:

“I would say at the initial onset, absolutely, we would be paying closer attention as to whether we were seeing side effects. But once it’s been onboard for quite some time, it kind of escapes the mind and probably doesn’t come up as frequently.” [R201-MD401]

Overall, an awareness of the potential for side effects of medications, knowledge about the types of side effects that medications can cause, and an intentional monitoring approach is needed by both patients and their healthcare providers to as a starting point for recognizing, and then investigating and managing prescribing cascades.

2. Investigation and management of prescribing cascades is simultaneous and iterative

The experiences of our interviewees demonstrated that investigation and management of prescribing cascades are not linear or discrete processes. These processes are inextricably linked, often occurring simultaneously. Significant knowledge about medication use and trialing periods of deprescribing (i.e., reducing doses or stopping a medication that might be causing a side effect) are needed to confirm hypotheses about causation.

To investigate whether a prescribing cascade has occurred, healthcare providers described starting with the “time-consuming” [P102-MD408] process of gathering a history of medication use. This included identifying the sequence of events (i.e., timing of medications and appearance of signs/symptoms) and the reasons particular medications are being used or were prescribed. To accomplish this, they spoke about needing to use detective work, find clues and “decipher” [R201-MD401] the information uncovered:

“People have been on things for years or they’re not really sure why they’re taking something… you’re not even sure who the original prescriber was… you’re kind of teasing information together based on consults you can get your hands on and whatever comes from the GP”. [R201-MD401]

Challenges included limited, or lost, history in both paper and electronic charts, patient’s difficulty in recalling information (particularly when experiencing cognitive impairment) and uncertainty arising when medications have multiple possible reasons for use:

“People probably did have prescribing cascades, but we don’t know the timeline or we don’t understand the chronology of how drugs were prescribed to be able to say with any degree of certainty that this in fact [happened].” [R201-MD401]

Further uncertainty about the original reason for a medication hampered clinicians’ ability to assess its effectiveness. This made it difficult to assess the relative risks and benefits of continuing or stopping a medication, an integral component of investigating and managing a possible prescribing cascade.

Even when the original reason a medication was prescribed was relatively clear, its’ ongoing reason(s) for use were often less clear, with one physician describing a particular case as “kind of blurry.” [R201-MD 401] Without an understanding of ongoing indications or timing of other medications, it was difficult to ascertain whether a prescribing cascade has occurred; this was further complicated when a long-standing medication could be associated with a symptom or side effect that has more recently emerged.

Healthcare providers spoke about how prescribing cascades do not necessarily follow a classic linear ‘medication causes side effect for which second medication is used’ scenario.

“It’s hard to identify a pure prescribing cascade in the sense of saying, this medication was started at this point in time, it caused this side effect and in response to that, this other medication was prescribed to address the side effect. It takes time…to be able to say categorically that [drug B] was prescribed to attend or to address the side effect of [drug A] it’s hard to get down to that level.” [P102-MD406]

Additionally, common or vague symptoms can sometimes be caused by multiple drugs making it difficult to sort out which medications were most likely contributing.

“Which drug am I going to choose that’s causing nausea? It’s probably a combination of the 10 drugs that they’re on.” [R201-MD401]

Not being able to stop each of the potentially contributing medications made it difficult to say a prescribing cascade had happened with certainty. Complexity was further increased when several medications could cause a constellation of different types of symptoms which might be treated by different types of other medications.

“I grouped together the tremors, the sweating, and the nausea as a constellation of symptoms to which those five drugs might be contributing.” [P101-PH407]

And, it was also hard to know if a medication was worsening a pre-existing medical condition already being treated by a drug.

“Some people might say it’s not a prescribing cascade then because the second drug was started before the first drug. But the question is, are the drugs like trazodone, dimenhydrinate, baclofen and cetirizine, are they all making the urinary retention worse? I’m not sure how to answer that.” [P102-PH407]

Physicians described their inclination to first establish a differential diagnosis of medical conditions, which typically does not include specific medication-related causes, to address and manage a particular sign or symptom. This often included referral to specialists who did not connect symptoms to medications as potential causes but carried out many tests and investigations. One patient related that:

“Nobody ever linked the adverse drug reactions to a medication, period. What they did was they tried to figure out other things… I just went for a lot of tests. I saw four cardiologists in that time.” [P108]

Pharmacists were more likely to question whether a medication could be causing the sign or symptom and to recommend deprescribing as both an investigation and management strategy.

When there were several potential prescribing cascades where more than one drug might be contributing to a sign or symptom, or where there multiple different prescribing cascades, there was a need to prioritize which potentially causative drugs to reduce first. This required consideration of the risks and benefits of each drug, goals of care, the possibility of needing to increase or add a different medication to achieve those goals, and the need for a step-wise approach with monitoring for resolution or worsening of symptoms. One pharmacist related the potential complexity of this situation:

“You’ve got the pseudoephedrine, the venlafaxine, the ibuprofen and the diclofenac, all of which can be contributing to high blood pressure…if we could reduce the venlafaxine, stop the diclofenac, try to minimize the ibuprofen and reduce the Tylenol Sinus®, we could see if the blood pressure got better. And she was taking two drugs to treat her hypertension.” [P103-PH407]

Situations in which a patient’s function seemed stable, or the risk-benefit of continuing or deprescribing a medication was unclear led to paralysis in making changes:

“When you see someone’s function stable and not deteriorating, you don’t have as much of an impetus to do something about it. Yeah, like not rocking the boat. Everything is status quo. There’s a big reluctance to stop things as an inheriting physician because you’re like, well, what if it was really doing something super important?” [R201-MD401]

Patients themselves were rarely able to describe the benefits of a medication or its adverse effects, and when they could, their balancing of these effects sometimes shifted over time, making it important to continually seek their input about changed preferences:

“I didn’t like all the side effects, but it helps me with the pain, you know? By that time, I just cared about my pain. I didn’t care about the side effects, but now, because I thought that the sweating and all the things together that I feel like I’m dying. It was because of the medications. Then, I started worrying about it. But before, I didn’t, I just cared about the pain”. [P101]

It was clear that it was hard to know who should be ultimately responsible for making changes:

“No one is kind of taking responsibility to actually resolve problems that are hinted at people might argue, well, it’s not the gastroenterologist’s place to stop a diabetes medication. But if he’s investigating the side effect of the medication, then maybe it is the specialist’s place to do that.” [P109-PH407]

Changing medications required conversations with, and agreement from, patients and amongst healthcare providers. Decisions were influenced by patient priorities and their feelings about how well the medication was working and the impact of side effects. Physician input was also essential in weighing benefit and risk:

“When there are four or five drugs that could be contributing to this one symptom, it helps to have the physician perspective on which drugs it might be safer to decrease first because of the likelihood that they might be being effective in some way.” [P101-PH407]

Conversations about these decisions were more challenging when there was cognitive impairment or limited understanding of medications, or when patients were reluctant to make changes. Gaining buy-in from patients involved being honest but not alarmist about side effects and potential risks and appealing to the idea of reducing medications:

“When you sort of explain that some symptoms can be caused by medication and that if we take the tact of reducing meds, it will mean less of a pill burden and helps in a more positive response.” [P101-MD406]

Overall, the processes for investigating and managing prescribing cascades are interconnected. Deprescribing potentially causative medications, as well as those treating side effects, is typically necessary to increase certainty that a prescribing cascade had occurred. Carrying out these processes is challenging as information to weigh risk/benefit of each medication is often lacking, people may not consider or know that a medication may be causing a side effect and taking action to investigate or manage a prescribing cascade requires knowledge, a willingness to be responsible and to facilitate communication amongst patients and prescribers.

3. Prevention of prescribing cascades requires intentional strategies to help people anticipate and recognize medication side effects

People spoke about strategies that could be used to prevent prescribing cascades. On the part of prescribers and pharmacists, this would involve knowing about the side effects of medications and common prescribing cascades; incorporating a ‘can this be caused by a drug’ process “Could this symptom be caused by one or more of the medications that they’re taking…?” [P101-MD406] into their assessment of new signs or symptoms; taking steps to reduce potentially causative drugs before ordering a test or adding another medication: “Should we be thinking about deprescribing as a means of treating the symptom?” [P101-MD406] and monitoring for the improvement in these signs or symptoms.

For medications being added to reduce the likelihood of another drug’s potential harm, an assessment of whether the initial drug’s ongoing benefit) is critical before a medication is added for prevention. For example, when deciding whether pantoprazole should be added to reduce the risk of bleeding for a patient taking multiple agents that increased bleeding risk:

“One of the questions I had for her cardiologist is how long does she need to be on two anti-platelet agents? Because what we would rather do if we could is reduce her bleeding risk by stopping some of the drugs that increase bleeding risk, rather than giving a fourth drug to reduce the risk from the first three.” [P109-MD409]

Interviewees indicated that current practice does not typically include consideration of deprescribing with one pharmacist stating:

“it’s very common to have patients who have had multiple consultant visits to investigate a problem. And then we decide to take a chance on stopping a drug that we think might be causing the problem, and we stop the drug and the problem gets better. But no consultant has ever suggested that.” [P103-PH407]

Computerized alerts, medication reviews, enhanced monitoring for side effects, including patients in monitoring for side effects, and thinking outside their own specialty area about symptoms that may be caused by a drug were raised as potential strategies for prevention:

“I think in the context of maybe another prescribing cascade, if there is a very common-known side-effect, how do you alert someone to identify that side-effect before just treating the side-effect and thinking about the original drug.” [R202-MD401]

However, these were seen to be potentially limited by alert fatigue: “When you alert people to too many things, then you just start ignoring the alerts.” [R202-MD401], lack of time and funding for medication reviews, lack of knowledge about side effects and reduced suspicion of side effects when they develop long after medications are initially prescribed:

“Sometimes, people develop side-effects to treatments they have been on for years, way down the road. In my mind, I don’t always equate that to a drug right away, because in my mind I say, well, you tolerated for so long, you must be tolerating it. This can’t necessarily be related. But maybe in fact it actually is.” [R202-MD401]

Several healthcare provider interviewees discussed central roles for patients in knowing their medications, identifying medication side effects and working with prescribers to prevent prescribing cascades

“The idea is to empower people to understand that they’ve got a responsibility to collect and share this information” [P101-PH407].

However, it was recognized that this strategy requires knowledge and confidence on the part of the patient to feel empowered to ask questions: one patient expressed that “there is a huge need for consumers to know that they can ask questions” [P108] Their caregiver agreed, stating that:

“I still feel that a lot of people don’t understand that one of their medications could be causing their problem, and that it isn’t a health issue that’s causing their problem”. [P108-CG308]

In summary, preventive strategies largely rely on the ability to anticipate and recognize symptoms as potential side effects of medications, to evaluate ongoing benefit of medications and to consider deprescribing before adding another medication.