Key messages regarding feasibility

  1. 1)

    What uncertainties existed regarding the feasibility?

    • The extent to which the implementation of TAPER was possible in a primary care setting for clinicians, participants, and researchers

    • The extent to which there appeared to be any evidence of a possible effect TAPER had on a range of outcome measures compared to usual care

  1. 2)

    What are the key feasibility findings?

    • TAPER is feasible to implement in a routine clinical practice setting in primary care.

    • A sufficiently high proportion of participants and clinicians were willing to engage in the intervention.

    • The direction of the effect on outcomes appears to favor TAPER versus usual care.

  1. 3)

    What are the implications of the feasibility findings for the design of the main study?

    • Challenges uncovered were those mitigated by adaptations in process and would not prevent the success of a larger randomized controlled trial.

    • There were trends in outcomes that suggested an effect is worthwhile testing in a large adequately powered randomized controlled trial.

Background

Polypharmacy is commonly defined as taking five or more long-term medications [1, 2]. Canadian older adults living in the community taking five or more medications have nearly doubled between the years of 1998–2008 from 13 to 27–30% [3,4,5]. Polypharmacy is associated with negative health outcomes in older adults, with increased risk of mobility-related functional decline, falls, hospitalizations, impaired cognition, and reduced quality of life [6,7,8,9,10]. Although polypharmacy is a reality of multimorbidity and drugs are beneficial for the management of symptoms in older adults, polypharmacy also contributes to the burden of treatment, and the balance of benefits and harms can change with time and aging [11]. This treatment burden may contribute to the association of polypharmacy with negative health outcomes. Medication adverse effects, medication interactions, complex regiments, and reduced medication adherence are factors that independently or in combination contribute to treatment burden and may also compromise the patient’s ability to cope [8, 10, 12,13,14,15]. For example, there is an increased risk of adverse drug reactions (ADRs) as the number of medications increases (2 medications have a 13% risk, 5 medications have a 58% risk, and 7 or more have a 82%) [12, 16]. According to a Canadian study, many ADRs are preventable [17].

There is increasing interest in processes to reduce polypharmacy; this involves reviewing a patient’s medications with the purpose of reducing the number and/or dose of medications, as well as the goal of reducing the harms and/or burden of polypharmacy [18]. These processes have generally been categorized as explicit (criteria-based tools) and implicit (judgement-based tools) [19]. Several explicit tools are available to guide clinicians when making decisions about deprescribing and/or identifying inappropriate medications (the Beers list, STOPP) [20, 21]. Numerous studies of various designs (retrospective, prospective, cohort, and randomized controlled trials) have shown such tools can predict a significant proportion of hospitalizations due to the adverse drug reactions [22,23,24,25,26], including in a primary care setting [27, 28]. To support a more individual focus in reducing particular drugs or classes, medication-specific deprescribing guidelines are available [29,30,31,32,33] such as those developed by the Bruyère Research Institute in Ottawa, Canada [34].

Despite the benefits of these types of approaches, many tools such as Beers and STOPP are only designed to flag those medications associated most commonly with drug-related problems in older adults. It is possible for a patient with polypharmacy to be taking no medications on these lists, be treated according to guideline-congruent care appropriate for single disease management, and yet experience multiple adverse effects related to their drug treatment [35, 36]. Furthermore, some tools do not constitute a patient-centered or patient-focused approach to care; they do not consider patient preferences. By including the patient’s voice (goals, priorities, and preferences for treatment) in situations where polypharmacy includes multiple medications that all may offer potential benefits, it is possible to help prioritize a medication list to reduce polypharmacy.

Most efforts undertaken to address polypharmacy have not explicitly considered patient preferences or priorities. We identified one approach, Systematic Tool to Reduce Inappropriate Prescribing (STRIP) [37], which identifies potentially inappropriate prescribing with the consideration of patient preferences. Inappropriate prescribing can arise when the risks of using a medication regiment outweigh the benefits [38]. For example, inappropriate prescribing can involve failing to use a safer alternative, omitting use of a beneficial or appropriate treatment, using an incorrect treatment regime such as dosage, or when a drug has significant interactions with another drug or patient’s comorbidities. A large randomized controlled trial is currently underway in Europe (OPERAM, clinicaltrials.gov) that aims to evaluate the implementation and effectiveness of STRIP within a hospital setting.

TAPER, team approach to polypharmacy evaluation and reduction, operationalizes a clinical pathway aimed to reduce polypharmacy. The theoretical basis for TAPER has been described elsewhere, but in summary, it is designed to address known barriers to deprescribing at the patient, provider, and system level as well as mapping to established models of care [39]. Like STRIP, this approach also considers patient preferences and priorities. However, TAPER has explicit consideration of priorities and preferences explicitly related to medications and individual medication experience and was developed for use within usual primary care setting workflow, with potential for adaptation to other contexts. Briefly, it is a model for addressing polypharmacy that involves the team of patient, pharmacist, and physician, who all bring particular expertise. TAPER uses sequentially linked consultations with both a pharmacist and physician. These consultations draw together the patient’s expertise and the effects of their medications on them, the clinicians’ expertise in medications, the context of their clinical state and circumstances (including multimorbidity), and their longitudinal relationship with the patient. TAPER is grounded in the idea of a “drug holiday” — the pathway is framed as a longitudinal structured “pause and monitor” process, with a patient-focused approach. Evidence-informed tools support this process, flagging potentially inappropriate medications as a “machine screen” and providing guidance around tapering and monitoring, and there are evidence summaries on risks and benefits in older adults. An underpinning secure digital platform (TaperMD) integrates these elements in a shared electronic record platform accessible by the pharmacist and primary care physician that also allows incorporation into clinicians’ existing individual record systems. The primary purpose of this study was to determine the feasibility of implementing TAPER in a primary care setting in patients 70 years of age or older and who are on five or more long-term medications, and the secondary purpose was to perform initial hypothesis testing.

Research questions and hypotheses

Nine research questions for this study are outlined in Table 1. The categorization of the feasibility sub-questions as process, resources, management and scientific in Table 1 are based on Thabane, Ma, Chu, Cheng, Ismaila, and Rios [40].

Table 1 Feasibility question, outcome, method of collection, pre-specified criteria for success
Full size table

It was hypothesized that process, resources, management and scientific indicators of feasibility will be demonstrated, with the identification of implementation challenges which could be mitigated in the design of a larger randomized controlled trial.

Methods

Study design and setting

We conducted a prospective 1:1 single-blinded randomized controlled feasibility trial. At the end of the study, the control group was offered the intervention. The findings of this trial will help in conducting a larger scale randomized controlled trial (clinicaltrials.gov, no. NCT02942927). This study’s outcome measures of interest for hypothesis testing are registered at clinicaltrials.gov (no. NCT02562352). The study was carried out in Hamilton, Ontario, by the McMaster University, Department of Family Medicine at the McMaster Family Health Team (MFHT). Family Health Teams are primary care organizations that formally link physicians and a variety of healthcare professionals together [41]. Patients who were 70 years or older, rostered with physicians who are part of the McMaster University Sentinel and Information Collaboration Practice-Based Research Network (MUSIC), and who were taking 5 or more long-term medications were eligible to participate in the study.

Participants and participant recruitment

All 31 family physicians and 3 clinical pharmacists at MFHT were invited to participate. Patient participants were recruited from the McMaster University Sentinel and Information Collaboration Practice-Based Research Network (MUSIC) network through the already established Health TAPESTRY program (Health Teams Advancing Patient Experience: Strengthening Quality) [42, 43]. Participants were eligible if they were 70 years of age or older and taking five or more long-term medications at the time of their initial assessment. Participants had to be willing to try medication discontinuation. Participants were excluded if they had a recent comprehensive medication review (within 6 months), had inadequate English or cognitive skills to understand and respond to the surveys, or had a terminal illness or other circumstances that would preclude them from a 13-month study period. We aimed to recruit 30–40 participants as we felt this was adequate to test the processes required for a larger RCT and collect adequate data to access the proposed tools.

Allocation and randomization

Participants were randomly allocated to either the intervention or control group (1:1 ratio) using variable block sizes of 2, 4, 6, or 8 through REDCap (Research Electronic Data Capture) [2, 44], a secure web-based software that can be used for both randomization and data collection and management. The randomization sequence was generated and maintained by the Biostatistics Unit at the St. Joseph’s Healthcare Hamilton.

Blinding process

Participants were not blinded to their group allocation. Participant blinding in this study was neither necessary nor practical as the focus is feasibility of the effectiveness of reducing medications rather than a trial of the pharmacological effect of a drug. The family physicians and pharmacists were masked to allocation as they were not aware of whether the appointment they were completing was for an intervention or 6-month waitlist control participant. Procedures were executed to ensure that the participants did not accidently unblind the researcher completing the outcome assessments. The effectiveness of the blinding process was evaluated at the completion of the study.

Procedures

To boost potential enrollment, participants who were eligible were contacted by their family doctor (and not the research team directly) by mail with an invitation letter outlining the study. Participants returned a prepaid postage envelope to the study team indicating their interest in the study. Those who expressed their interest were then contacted by the study team to be screened for eligibility and to formally go through the consent process. After consent, a research data collection session was booked according to group allocation. The session was anticipated to take 1–2 h and was done either at the patients’ home or at the research facility. Outcome measures were collected by a researcher at baseline and then again after 6 months. Follow-up symptom assessments were also conducted at 1 week, 3 months, and 6 months by a researcher over the phone. The control group received the usual standard of care. After all data was collected at 6 months, control participants were offered the intervention. Baseline recruitment began in November 2016 and ended in December 2017. Six-month follow-up collection started in May 2017 and ended in June 2018.

Intervention

The intervention operationalized a clinical pathway (TAPER) aimed to reduce polypharmacy. It involved a cooperative team-based structure for a complete medication review by the pharmacist and the physician aimed at reducing medication burden. The approach collected foremost patient’s priorities, preferences, and experience of their medications. It then used explicit evidence and tools to automatically screen for and flag potential inappropriate medications or combinations through use of an integrated e-tool “machine screen” within the pathway. The objective was to combine this range of existing evidence, any available specialized tools, provide evidence for discontinuation management, and integrate this with the patient’s preferences to develop a collaborative, longitudinal plan as a “pause and monitor” trial of medication discontinuation. Information was entered, stored, and shared via an online platform, TaperMD, which provided a shared platform for recording and teamwork between pharmacist and physician. All clinicians were provided with training on TaperMD with an initial overview tutorial of around 30 min and then a “ride-along” with a researcher specifically available at the consultation time during use with the first patient to answer any questions. The pharmacists were provided with additional 1-h, in-person training on TaperMD. Video tutorials on each section of the tool were also available on an internal YouTube channel. A summary of the pathway is found in Fig. 1, following a description of each step below.

Fig. 1
figure 1

Structured medication discontinuation clinical pathway

Full size image

Step 1: collection of participant information at baseline

The research assistant interviewed the participant about their goals, preferences, and priorities, recorded medication side effect risk factors, perceived medication side effects, and entered an initial current list of medications using the dispensing pharmacy list, information in the electronic medical record, and the participant into TaperMD.

Step 2: clinical pharmacist consultation

The clinical pharmacist completed a comprehensive review of the participant’s medications (gathered by the research assistant from the patient) and medical conditions (from the electronic medical record system or from the patient themselves) in an appointment with the participant. The pharmacist reviewed medications that were entered by the research assistant in TaperMD, along with the medications on the patients’ dispensed list, and created a reconciled, current medication list. TaperMD automatically created the “machine screen” which flags potential inappropriate medications, which could be suitable for discontinuation or dose reduction. These tools and lists highlight potentially inappropriate medicines (and reasons) in older adults in a simultaneous multidrug view consistent with a multimorbidity approach. Specific medication dimensions flagged include standard interaction checking, potentially inappropriate medicines in older adults drawn from assessment of a wide range of jurisdiction specific lists [19], drugs contributing to anticholinergic burden score, QT prolonging drug burden, hypotensive drug burden, serotonergic drug burde,n and deprescribing guidelines and algorithms where these are available. The screens included are described in detail in Additional file 5. Informed by this information, clinical judgement and knowledge and the patient’s priorities, a list of suitable recommendations for appropriate discontinuation, were made. An initial plan was developed and stored in TaperMD for family physician review.

Step 3: family physician consultation

Within approximately 1–2 weeks following the pharmacist consultation, the family physician met with the participant for an extended appointment. The family physician reviewed the pharmacist recommendations and patient priorities and the reports of medication effects prior to the appointment. Using the same principles, and the pharmacist’s suggestions, at the consultation with the patient, a discussion of the final assessment of the medications suitable for a trial of “pause and monitor” occurs, and a plan was made. The physician validated or adjusted recommendations within the pharmacist’s initial plan. The final plan includes information about the following: what will be discontinued, what will be monitored, who will monitor (patient pharmacist or physician), how often and when, and when would medication restarting be considered.

Clinical monitoring

Participants attended monitoring visits after the initial plan, as clinically indicated by the particular drugs selected, during the “pause and monitor” phase. Planned monitoring was recorded in TaperMD.

Data collection

Demographic information (collected at baseline) and research outcomes (listed below and collected at baseline and 6 months) were collected by the research assistant. Participant characteristics (age, gender, and income) were self-reported. The Charlson Comorbidity Index [45] was used to assess burden associated with chronic conditions via chart audit [46]. Medication-related information (beliefs about medications [47], current medication list with indication for prescribing [48] (was also collected via patient self-report, chart audit, and using pharmacy data.

Feasibility outcomes

We considered four categories of feasibility questions to determine whether a larger randomized trial would be feasible [39]. Specifically, process outcomes were included to assess the feasibility of the steps that need to take place for a successful main trial, resource outcomes were collected to assess time and budget challenges for the main trial, management outcomes to determine challenges of human or data management, and scientific outcomes were considered to assess intervention safety and outcome variance. Each feasibility research question, feasibility outcome, the method of collection, and pre-specified threshold for success have been outlined (Table 1).

Proposed outcome measures for process, performance, and hypothesis testing

Primary outcome measures

The primary outcomes were the number of prescribed and prescribable medications, at 6 months after baseline. For intervention patients, this information was collected 6 months after the patient met with the physician (step 3), and for control patients, it was collected 6 months after the baseline study visit appointment with the researcher (step 1). The medications were gathered from the medication list in TaperMD, which was the reconciled medication list done by the study pharmacist at baseline and at 6 months and recorded changes made to medications during the intervention period. Medications were categorized as either prescribed, prescribable, or non-prescribable. Prescribed medications include all schedule 1 medications (medications requiring a prescription) [49] and any medications that were dispensed through a prescription or where a prescription was found in the electronic medical record (EMR). Prescribable medications are defined as schedule 2, 3, or unscheduled [49] medications that a medical physician or nurse practitioner could reasonably prescribe (e.g., vitamin D, vitamin B12, calcium carbonate, acetaminophen, ibuprofen) but are purchased over the counter rather than dispensed through a prescription.

Secondary outcome measures

Secondary outcomes included number of non-prescribable medications (defined as schedule 3 or unscheduled [49] medications that would not be prescribed by physicians or nurse practitioners but are purchased over the counter at the discretion of the patient, e.g., naturopathic preparations, homeopathic preparations, multivitamins), medication dose changes, quality of life, psychological distress, cognition, fatigue, nutritional status, physical functional capacity, falls, adverse events, healthcare utilization, and patient enablement. All were collected at baseline and at 6 months (Table 2). All outcomes were assessed using validated measures or custom self-report forms, or data was extracted from the EMR. Note, we also assessed the utility of the Flinders Fatigue Scale (to assess fatigue) in a few participants [50].

Table 2 Secondary outcome measures for assessment of process and performance
Full size table

During each study visit, the researcher recorded any potential serious adverse events. We used a separate researcher as unblinding was possible. If present, a local serious adverse event form was completed and reported to the family physician. Potential side effects that were unmasked as result of the study were also collected from patient and clinical reports. Minor and serious adverse events were categorized using the FDA criteria. We did not collect or report on when medications were tapered or the order of tapering in this study, although the information was available in the patient’s EMR and TaperMD profile.

One week and 3- and 6-month follow-ups

Follow-up was done by telephone at 1 week and 3 and 6 months by a research assistant to capture any changes in side effects or symptoms (either positive or negative) and to record healthcare visits. In an open-ended question, patients were asked if they have experienced any worsening of, or improvement in, any side effects or symptoms. They were also asked how many times they visited their family physician or the emergency room since the last study appointment. This researcher was different to the researcher recording outcome data to maintain blinding as it was possible that unblinding could occur in symptom description. Serious adverse events were also extrapolated from these conversations; a detailed assessment of any adverse event was also recorded, and addressed by the Principal Investigator, and the patient’s usual primary care clinical setting.

Sample size

The sample size for this study was based primarily on feasibility considerations [40, 60]. We aimed to recruit n = 36 participants (n = 18 per group). This is in line with general guidance for sample size for pilot trials aimed at assessing feasibility [53, 54].

Statistical analysis

Results from this trial are reported in accordance with the CONSORT statements for pilot and feasibility studies (Additional file 2, [60]) and TIDieR checklist (Additional File 3, [55]). Baseline participant characteristics were reported in terms of mean (standard deviation) or median (first quartile, third quartile), depending on the distribution, for continuous variables and count (percentage) for categorical variables. We used descriptive statistics such as, count (percentage) to analyze the feasibility outcomes. The primary analyses were performed using intention-to-treat approach. Multiple imputation method was used to impute the missing data. In total, 5 datasets were generated and the pooled estimates were reported.

The primary outcomes, the number of medications (prescribed and prescribable) and number of prescribed medications, were analyzed using the Poisson regression with the treatment group as the covariate. The rate ratios (RRs) along with 95% confidence intervals were reported. The secondary continuous outcomes (shown in Table 2) were analyzed using the linear regression with treatment group as the covariate. The mean differences (MDs) along with 95% confidence interval were reported. All statistical tests were two sided at the level of significance 0.05. The data were analyzed using statistical software R version 3.5.1 [56]. Sensitivity analyses were performed to assess the robustness of the results of the primary analyses using per-protocol approach.

Results

Participant characteristics by group are reported in Table 3, and the results of the feasibility outcomes are presented in Tables 4, 5, 6 and 7, and the flow of participants through the trial is described (Fig. 2). Of note, we demonstrated success in all categories of feasibility where thresholds were pre-specified. We describe results for each feasibility question below, followed by results of emergent evidence of the potential effectiveness of TAPER.

Table 3 Participant characteristics
Full size table
Table 4 Process feasibility results: enrollment and recruitment
Full size table
Table 5 Process feasibility results: eligibility, data collection, and unblinding
Full size table
Table 6 Resource feasibility outcomes
Full size table
Table 7 Management feasibility outcomes
Full size table
Fig. 2
figure 2

CONSORT participant flow diagram

Full size image

Process feasibility research questions

Responses generally came in within the first 2 weeks after mailing patients. We assessed 85 participants for eligibility, and 39 (46%) of these were enrolled into the study and subsequently randomized. Two participants (5%) withdrew from the trial (1 from the intervention group and 1 from the control group), and 3 (8%) were lost to follow-up (2 from the intervention group and 1 from the control group). Two participants were excluded from the intention-to-treat statistical analysis as they were under study inclusion criteria for age. All three pharmacists and 31 physicians who were invited to participate in this study agreed to take part. Overall, the structure of the intervention was considered efficient and fit with normal clinical workflow based on clinical user reports; the family physicians took no longer than the 30 min allotted and could sometimes be completed in less time. Furthermore, some participants did not actually start the deprescribing plan right after the physician appointment and did not implement recommendations as they believed there was a further step. In response to this, the process was adjusted to include a pharmacist check-in call with the patient a week after the physician appointment to reinforce the instructions of the plan and respond to any questions. Six instances of unblinding of the outcome assessor occurred. All the recruitment and randomization process outcomes met or exceeded the threshold for success stated a priori (Tables 4 and 5).

We trialled two quality-of-life scores, and the general feedback was that the WHODAS was longer and more cumbersome to administer in this group [57]. Similarly, the Pittsburgh sleep scale [58] was lengthy considering it was not a primary outcome domain, and so we identified and tested a single item for sleep quality in the latter part of the trial, which was more practical to administer [62]. The Stanford Self-Efficacy Scale for Managing Chronic Disease and the Flinders Fatigue scale [50, 61] were abandoned early on as it was felt by both participants and study personnel to be too lengthy and difficult to administer.

Resource feasibility questions

The processes for data collection fit with our pre-specified criteria for success. Fifty-nine percent of data collection visits were performed at the participant’s home, and none of these visits required a travel time of longer than 30 min one-way. Data collection appointments were able to be kept to less than 2 h long in most cases, with baseline visits taking 1.5–2 h and the 6-month visit 1–1.5 h. There were 2 instances where baseline data collection took up to 2.5 h. In response, we adjusted processes, allowing participants to break data collection appointments into multiple sessions if they became fatigued. Also pertinent to this, as outlined later in the results, section several questionnaires were felt by researchers and participants to be too taxing for the yield. All the data collection resource outcomes met or exceeded the threshold for success stated a priori (Table 6).

Management feasibility questions

We identified several data management/entry errors that operational changes in data management would mitigate in a larger trial. Data management and entry errors included accidental exclusion of survey questions from the database and problems with launching and syncing of the database. We were unable to calculate scores for the role-emotional domain as there was misalignment between our database and SF36-V1 response options for these questions when the database was programed, and bodily pain domain as these questions were omitted from the database, and thus, this data was not collected.

Database launching and syncing errors were mitigated by utilizing paper surveys as a backup, and exclusion of questions will be corrected by ensuring the database is correctly set up prior to implementation of a larger trial. Three additional outcome measures were also added into the study protocol for testing midway through the trial: hand grip strength, the healthcare utilization questionnaire, and the 15D quality-of-life sleep question, so these outcomes were not measured for most participants at baseline, and also not measured for several participants at 6 months. All the data management outcomes met or exceeded the threshold for success stated a priori (Table 7).

Scientific feasibility questions

Collecting data on potential outcome measures allowed us to see the range and variance in the target group and to assess and compare tools both in their variance and in the practicality of their administration in this group. The detailed results of these outcomes can be found in Tables 8 and 9, Tables 10–14 in Additional file 1, and results are presented visually in Figs. 3–7 in Additional file 4. We found no substantive floor or ceiling effects. With such a small sample size, higher standard deviations are to be expected and should be reduced in a large-scale trial. We found nothing to preclude use of these instruments in a larger trial, except for a few based strictly on feasibility in terms of process (Tables 8, 9, 10).

Table 8 Number of prescribed, prescribable, and non-prescribable medication at 6 months by group
Full size table
Table 9 Number of medication changes in intervention arm patients (n = 18) during the 6-month study period
Full size table
Table 10 Feasibility questions and the classification of the feasibility sub-questions
Full size table

Emergent evidence of potential effectiveness

While we were not assessing for significant differences between groups at 6 months as this study was not designed with adequate power to do this, and a number of adjustments to process were made, we examined the data for signals around the direction of effect of outcome measures (Figs. 3–7, Additional file 4). Our results do show that most of the outcome measures signal a direction towards the effectiveness of TAPER compared to usual care, with the exception of the mental health domain of the SF-36, patient enablement, MMSE, and grip strength. There were no meaningful differences between groups for serious adverse events, and the intervention itself was not associated with any serious adverse events (Tables 12 and 13, Additional file 1).

Sensitivity analysis

The results of the sensitivity analyses using per-protocol approach are provided in the additional information file. The overall conclusion for all the outcomes was similar to the primary analyses.

Discussion

We examined the extent to which TAPER is feasible to implement in a primary care setting and if there is any emerging evidence of the direction of effect of TAPER compared to usual care on a range of outcomes. We found support that the trial met all pre-specified thresholds for success across all feasibility indicators. Notably, 100% of invited pharmacists and family physicians participated, 46% of patients assessed for eligibility were enrolled, and the number of participants who withdrew was low, with reasons unrelated to the study (e.g., surgery). These results exceeded our thresholds for success and give us confidence for a larger trial where uptake and participation are essential.

Operationally, research procedures were relatively efficient. The number of participants lost to follow-up was low, and the primary outcome could still be ascertained in those who were unavailable for the full outcome assessment. Furthermore, instances of unblinding, which were balanced between staff versus participant, triggered unblinding, and challenges identified are those easily remediated in a larger trial with more staff and with clear role assignment among the research team. Travel time (10–30 min) and time to complete data collection were reasonable (1–2.5 h); generally, only one visit was required to collect data at baseline or follow-up. However, in few instances, more time or splitting collection up over two visits was required. These strategies were particularly relevant for those participants whose primary language was not English or had hearing or vision problems. Data management issues relating to using Microsoft Access software to enter some of the scales and entering in multiple and remote locations were significant enough to search for alternate software to use for a future larger randomized controlled trial. The intervention itself was not associated with any serious adverse events, confirming findings from a previous study [59].

With a few exceptions, we also found evidence to support our hypothesized group differences. Even in this small sample size, the groups were reasonably similar in demographic makeup, with a slightly higher Charlson comorbidity burden and a slightly lower number of prescribed medications in the intervention versus control group at baseline. Data were inadequate to make meaningful conclusions about any emergent evidence of potential effectiveness on healthcare utilization, with emergency department visits or hospitalizations occurring too infrequently in such a small sample. Furthermore, meaningful conclusions about cognition are also challenging given that most people in the study had MMSE scores between 26 and 30. This is not surprising given that we excluded people who did not have adequate cognitive skills to understand and respond to the surveys. We did however find that the outcome measures we included in this trial showed no evidence of floor or ceiling effects and generally performed such that with a larger sample and adequate power; we expect to be able to make conclusions about the effectiveness of TAPER on these outcomes. Only 5 participants, out of 37, had missing data in this pilot trial. We implemented multiple imputation approach to impute these missing data based on this small sample size. Thus, this multiple imputation had some impact on the precision of the estimated treatment effect.

Together, our results provide the support needed to proceed with a full randomized controlled trial, with some modifications, so on the basis of these data, we made the decision to proceed with a larger RCT. These findings allowed us to shorten the study data collection by providing a sound basis for refining the number and choice of outcome measures, reduce the risks of unblinding of outcome assessors, and seek new data management software. The signals of effect also support proceeding with formal testing of the hypothesis that TAPER will reduce medications, and that the negative associations of polypharmacy with health outcomes may be at least partly reversible if this is achieved.