Abstract
Deprescription is the process by which a physician supervises a patient’s withdrawal or dose reduction of a given medication due to side effects or diminishing efficacy. Prior studies on the process of deprescription have resulted in a number of models, two of which are used in this pilot study alongside the decision ladder to construct a novel analytic method. Initial findings indicate that this approach can offer unique insight into the deprescription process, particularly regarding the paths physicians take through the decision making process and when certain factors are most important. These early results are limited but lay the foundation for a rich variety of future work.
Keywords
Introduction
Polypharmacy, or the use of five or more medications, is a common risk factor for older adults (Linsky, Gellad, Linder, & Friedberg, 2019). More than 40% of U.S. adults 65 years or older are at risk of adverse consequences as a result of polypharmacy (Fried et al., 2014). While multiple medications are often necessary, the benefits must be weighed against possible side effects and diminishing efficacy. In response to this risk, there is a growing focus on better understanding of deprescription, or clinician-supervised withdrawal or dose-reduction of potentially harmful medications. Prior work has identified a web of factors that affect deprescription, which were categorized as patient, prescriber, or system factors (Linsky et al., 2019). Greater understanding of deprescription may lead to improved clinical tools, training, and guidelines to lead this practice.
This study was conducted as a pilot for an interdisciplinary project investigating multiple perspectives on the practices of deprescription at a collaborating medical practice. In order to justify the appropriateness of the methodology to be utilized in that effort, this pilot study analyzed the decision-making process employed by physicians during deprescription in nine video-recorded outpatient medical appointments. Physician and patient interactions around deprescription were evaluated through the lens of control task analysis (Vicente, 1999). This pilot, as well as the larger study, aims to answer two primary research questions: 1. What cognitive processes do physicians go through during the course of deprescription? 2. What factors effect the course of deprescription at each particular stage of the process?
Cognitive work analysis is a formative, multistep approach to analyzing a work domain at concentric levels of focus that outline all possibilities of action allowed within the domain constraints (Rasmussen, 1986). This study utilizes the second step of the process, control task analysis, which focuses on the transformation of specified inputs into outputs through a decision-making process (Vicente, 1999). That decision-making is modeled via decision ladders, which map out the possible steps of decision making. Tracing individual cases of decision-making onto that map allows for analysis of both rational decision-making and more naturalistic, heuristic-based approaches.
Methods
This study analyzed nine quality-assurance videos of patient visits at a primary care practice. The record review and analysis protocol received approval from the University at Buffalo Institutional Review Board.
Setting
The studied primary care practice was recruited through an interest group of healthcare professionals focused on older adult medication safety. The practice directors value continuing improvement and record selected patient visits for quality assurance and training purposes. The practice employs several physicians, physician assistants, and other specialists. The visits reviewed for this study included both in-person services and telemedicine. The research team was granted access to video recordings in a password-protected folder via the practice’s virtual private network. Video reviews and analyses were conducted in private settings to ensure confidentiality of health information.
Participants
The participants for this study included two primary care physicians and nine patients. The patients were all 65+ years of age and prescribed two or more medications, although further demographics were not captured. These criteria were selected because polypharmacy is a particular risk for older adults given their higher average number of medications and the potential for more negative side effects, such as falls, sedation, and confusion. The practice screened and gave the researchers access to recordings that met the inclusion criteria above.
Data Analysis
Wellness appointments were recorded by a medical scribe responsible for record-keeping of the patient visits. The recordings were screen recorded from the scribe’s point of view; hence the video footage consisted only of their note-taking, but the audio included the doctor and patient’s conversation throughout the visit. This study focused on observational analysis of these prerecorded videos provided by the collaborating practice. All videos used for analysis were first screened by the practice and most were trimmed to only include the portion of the visit that included deprescription discussions. Therefore, our analysis may not have captured the entirety of a patient visit.
The deprescription process in each appointment recording was analyzed through two previously published frameworks. Previous work outlined a five-step patient-centered deprescribing cycle (Reeve, Shakib, Hendrix, Roberts, & Wiese, 2014) consisting of 1) a comprehensive medication history; 2) identifying potentially inappropriate medications; 3) determining if medication can be ceased and prioritization; 4) planning and initiation of withdrawal; and 5) monitoring, support, and documentation. This cycle is consistent with a typical decision ladder (Vicente, 1999), which begins with situation assessment, compares potential options, acts, and evaluates outcomes. Therefore, we have overlaid Reeve et al.’s cycle onto a decision ladder (Fig. 1 – the red boxes are the steps in Reeve’s cycle) and mapped each deprescription event onto the resulting framework. Step 4 of the cycle was separated into determination and prioritization steps in order to better fit the form of the decision ladder.
A Deprescription Decision Ladder.
Another framework of interest categorizes factors that affect desprescription into patient, prescriber, and systemic factors (Linsky et al., 2019). As individual visits were mapped onto the proposed deprescription decision ladder, any pertinent factors present in that step were noted to allow cross-analysis between strategies and factors. A template to support systematic coding of these visit recordings was developed by the interdisciplinary team; after development, two of the authors analyzed the first video together to establish the coding process. All remaining videos were coded independently by those authors, who later met to establish final consensus on all codes for each appointment.
Results
Decision Paths
Results revealed that the decision paths are mostly nonlinear and occur across multiple visits. Deprescription often involves a step-down process in which a patient is weaned off of a medication by lowering the dosage over time. We did not observe in any of the recordings a complete deprescribing process. Rather, the recordings viewed can be sorted into two categories: initial deprescriptions and follow-ups. We present three cases below to illustrate this distinction within our analysis framework.
Appointments that fall under initial deprescriptions came the closest to covering the full decision ladder process. They involved a doctor initiating a new conversation around some medication and planning its deprescription. Figure 2 presents the decision ladder of Video A, an example of this category.
Video A Decision Ladder: Initial deprescription example- linear.
Video A involved a patient reporting pain, anxiety, and significant difficulties sleeping. The physician indicated that a number of these symptoms may be a result of polypharmacy (step 1), and identified three potential medications for deprescription (step 2). Two of these medications were selected given their place on the Beers Criteria, a list of medications that should be avoided by older adults (The 2019 American Geriatrics Society Beers Criteria® Update Expert Panel, 2019). While the third drug is not included on the Beers Criteria, the physician suspected that it was responsible for the reported pain.
Video A stands as an interesting example given the process of prioritization between these three medications. First, the physician discussed the medications’ potential role in the side effects experienced by the patient. Then, the physician asked the patient which of the medications they would like to focus on first to begin the deprescription process (step 3). This interaction demonstrated patient-centered care and strategies that can be utilized to achieve patient buy-in for the deprescription process. The patient was hesitant to select any medication to deprescribe, but did express a desire to continue the medication prescribed as a sleep aid given their difficulties sleeping. The physician then decided to begin with the medication suspected of causing pain, an issue that was well controlled (step 4). The remaining deprescription process was straightforward—prescribed dosage was lowered (step 5), nonpharmaceutical options (e.g., physical therapy) were offered (step 6), and a follow up call from a social worker or pharmacist at the practice was scheduled (step 7).
An unusual feature of Video A was its linear nature. The three medications were discussed together while the doctor and patient ascended and descended the decision ladder through the deprescription process. This was not typical across the other videos in our sample. Nonlinear decision ladder mappings such as that presented in Fig. 3 were much more common. In Video B, the processes of planning, prioritization, and determining if medications could be ceased were much more cyclical and interwoven. Multiple medications were discussed at separate times during the visit, and the physician revisited various steps of the process for each independently. Space limitations preclude a detailed description of the specific conversations and decisions that constitute each of the steps (1 to 10, see Fig.3), but they were similar in nature to those in Fig.2 apart from their complex sequence.
Video B Decision Ladder: Initial deprescription example- nonlinear.
Follow-up appointments are best captured by Video C (see Fig. 4). These appointments only involved a brief revisiting of a prior deprescription conversation to see how well the patient has adapted to the new dose (step 1), adjust the deprescribing plan as needed (step 2), and arrange the next follow up (step 3). Five of the nine videos analyzed could be categorized as these simple follow-ups.
Video C Decision Ladder: Follow-up deprescription example.
While these visits offer less insight into the deprescription process, they do make it clear that this is a long-term, multi-visit process. Additionally, these videos provided insights regarding the factors that affect the deprescription process, as discussed below.
Factors Affecting Deprescription
Below we present factors identified in each step in the Reeve’s 5-step cycle. Factors are categorized as patient, prescriber, or systemic factors, following Linsky’s framework.
Medication history - During the initial medication review, the primary factors identified are the physician’s technical knowledge of medications and the patient’s capacity to share their medical history and current list of medications. One patient understated their medical history including prior diagnoses, requiring more intensive probing by the doctor to effectively review their medications. Another patient was open with their medical history, but they took their medications quite inconsistently and were not clear on what it was they were taking. Unwillingness or other barriers to sharing medical history and medication lists can interfere with a physician’s ability to assess a patient’s current state.
Identify potentially inappropriate medications - Identifying which medications might be inappropriate is dependent on a physician’s understanding of medication side effects and harms, supported by tools like the Beers Criteria. However, the success of the deprescription often depends on a patient’s biology as well, as patients hesitant to deprescribe often would point to their lack of side effects as a reason to not wean from a medication. Inversely, patients experiencing significant side effects were more enthusiastic about the deprescription process. While the physician would identify the medication as inappropriate in either case, a patient experiencing side effects does seem more likely to comprehend why a medication needs to be deprescribed. In another instance, the physician identified inappropriate medications because of the combination in which they were prescribed. This appeared to be driven by systemic factors as the medications were prescribed by different specialists without appropriate communication.
Determine if medications can be ceased - The process of determining whether a medication can safely be ceased appeared to be a balance between patients’ concerns and physicians’ knowledge and skill in identifying safer alternatives such as nonpharmaceutical treatments (e.g. sleep hygiene or counseling). Some patients were hesitant about the deprescription due to chronic pain, anxiety, sleep disorders, and out of habits or reliance on medications they had been on for years or decades. Physicians attempted to ameliorate these concerns through discussions of alternative treatments.
Prioritization - We observed fewer factors being discussed in the context of prioritizing medications for deprescription than other categories. Typically, the severity of side effects appears to drive the prioritization process, but as previously mentioned (Video A), there was one instance of patient participation in prioritization. We hypothesize that much of this step occurs unilaterally by the doctor and not being communicated. Factors influencing prioritization will be an emphasis of our planned future work via follow-up interviews with the physicians.
Plan and initiate withdrawal - The planned strategy to deprescribe consisted of a slow, staged decrease in dosage over time in all nine videos reviewed. This was generally supported by the recommendation of some nonpharmaceutical treatment such as improved sleep hygiene, physical therapy, or counseling. The primary factors affecting success of the deprescription at this point were the availability of these services and the patient’s willingness to utilize them. In a number of instances, patients were hesitant to see a counselor, potentially due to stigma regarding mental healthcare, and wanted to instead continue to depend on medications to minimize their anxiety. Reluctance to follow a plan of deprescription may also be driven by patient experiences, particularly long-term usage of the medication in question.
Monitor, support, and documentation - While factors previously mentioned did appear in the monitor, support, and documentation step, including preference for nonpharmaceutical treatment and patient hesitancy, a factor of special note was the systemic resources available to this practice. This practice has both social workers and a pharmacy team on site who were often involved in the planned follow up of a deprescription. The social workers were often utilized to help coordinate other external services and resources, as well as to support patients in navigating their emotional reactions to the changing care. The pharmacy team supported more technical issues and monitored side effects. The multidisciplinary nature of this practice seems to be a powerful aid in the practice of deprescription.
Discussion
This study served as a pilot to investigate the analytic potential of combining two models of deprescription with the decision ladder framework. It was apparent from the analysis that Reeve’s cycle of deprescription (Reeve et al., 2014) neatly captures all of the steps involved in the recorded deprescribing discussions provided by our collaborating practice, adding further support for their model. Mapping a number of deprescription events onto that model (adapted to fit the decision ladder) allowed us to identify the ways in which initial deprescription visits covered all or nearly all of the steps of the cycle, often in a nonlinear fashion, while follow up appointments may cycle only through the planning and monitoring stages of the model. Analysis of more deprescription events, particularly from other physicians and practices, would allow for a more rigorous understanding of these dynamics, and our team will continue to analyze appointments with this methodology as they are provided by the collaborating practice.
The decision ladder is used to better understand what conditions allow practitioners to make cognitive shortcuts, such as jumping from situation analysis directly to enacting a plan when prototypical situations are identified (Vicente, 1999). Some shortcuts were noted in the recordings, particularly skipping over “determine if medication can be ceased” and “prioritization”. These skips represent cases in which a particular medication was obvious and apparent as a target for deprescription to the physician. However, this sample is too small to provide strong evidence for how these jumps may be better supported. Additionally, it is unclear from this work whether those skips would be beneficial in this domain. While time pressures might encourage these shortcuts in decision-making, skipping these steps may be deleterious by reducing a patient’s involvement in their own care and reducing the potential benefits that can accrue from shared decision-making. Further research on patient perspectives on the deprescription process will be necessary to better understand this phenomenon.
This analysis has also supported further investigation into the deprescription factors outlined in Linsky et al. (2019). Generally, the patient’s medication induced side effects, or lack thereof, were a strong factor in determining how hesitant they would be in the process. The two physicians in this study demonstrated reliance on knowledge and availability of safer (especially nonpharmaceutical) treatments to mitigate patients’ symptoms while deprescribing, but some patients were hesitant with these options due to past experiences or beliefs. The two physicians studied appeared to have strong rational knowledge of Beers Criteria and other factors that allowed them to quickly identify problematic medications, but other physicians may require more support in this identification process. Additional system resources, particularly social workers and pharmacists, were often involved in the monitoring and support process; further research into patient experiences will be necessary to determine the efficacy of such support.
The team will conduct analysis of additional recorded deprescriptions using the current approach to allow for a rigorous understanding of prototypical cases and special circumstances. Future work will aim to examine these cases through additional lenses beyond human factors to consider how communications, pharmacy, social work, and anthropological approaches might contribute to an understanding of deprescription.
These analyses will also be supplemented by interviews with physicians and patients. It is crucial that patient perspectives are better understood, as their buy-in is key to deprescribing. Physician interviews will allow for further investigation into individual differences, prioritization processes, and various strategies to encourage patient involvement.
Conclusion
Deprescription is an important field of study as polypharmacy becomes increasingly commonplace, particularly among older adults. This paper presented a pilot study analyzing nine videos of deprescription events using a new method that builds upon previously published studies. The decision ladder and Reeve’s cycle of deprescription provided a framework onto which each event was mapped, while Linksy’s factors enabled examination of the details present at each step of that process. Findings for this study are limited by sample size but encourage further use of this method to understand deprescription and how it might be improved.
