Deprescribing medicines in older people living with multimorbidity and polypharmacy: the TAILOR evidence synthesis

Data on safety and impact of discontinuing medication

Advice for clinicians published by NICE,22 which it recognises as ‘guidelines not tramlines’ provides non-mandatory advice to inform, but not dictate, best practice. The limitations of guidelines for clinical practice are well recognised, for example in being ‘condition specific’ and ‘context blind’. 23 Guideline development has been criticised for using evidence that often excludes patients with multimorbidity,20 or for overlooking, or placing less weight on, evidence related to patients’ lived experiences of illness or treatment (e.g. theoretical or qualitative work). 24 Clinical practice for person-centred care inevitably involves working beyond guidelines, especially for people with multimorbidity, to reduce the risk of burden and iatrogenic harm. 25

In practice, guidelines are one source of data used by clinicians in the complex task of interpreting individual patient need,26–28 with their limitations well recognised. 29 Clinicians delivering tailored, whole-person care engage in a complex task of data collection, described by Donner-Banzhoff and Hertwig30 as inductive foraging, which draws on data from patient consultation, external data sources (including guidelines and evidence) and professional experience and expertise. 28

Access to data that helps clinicians in this process (to discuss the safety and impact of discontinuing medication with their patients and/or carers) is therefore key to tailored prescribing. 6 TAILOR addresses that gap through a scoping review to summarise the current available evidence on deprescribing in a form that makes a useful reference source for clinicians.

Framework for judging ‘best’ practice for tailored prescribing decisions

Tailored decisions require professional interpretation of multiple data sets, often in varied and varying circumstances, to generate individualised assessments of potential for risk and benefit. 14,28 Tailored prescribing may require the use of ‘clinically appropriate overrides’ of evidence-based guidance. 31 Outcomes of patient-centred care focus on clinical decisions that optimise health-related capacity for daily living, not simply condition- or medication-specific outcomes. 32–37

Defining and delivering best practice in tailored (de)prescribing can therefore be understood as a ‘wicked problem’: a complex (and often messy) problem that cannot be fixed because of incomplete, competing and changing requirements, but can be managed through iterative, adaptive and ongoing responses,38,39 resulting in solutions and outcomes that are better described as ‘better or worse’ rather than ‘right or wrong’. 40–42 Clinicians, patients and wider stakeholders, therefore, need a framework using which they can judge ‘better or worse’.

Medicines optimisation is the framework currently used to guide best practice, emphasising outcomes focused on clinical effectiveness, cost-effectiveness and minimising both harmful effects and waste. 43 However, it is known that patients define benefit from medicines differently from clinicians. What a medical perspective may describe as effective or optimal care may be experienced as burdensome by patients. 1,32,44 Evidence highlights that patients prioritise the impact of care (including medicines use) on their continued daily living33,44–47 over the management of disease. 48,49 Assessing best, or better, practice by adherence to guidelines will be insufficient.

Tailored care actively incorporating patients’ priorities and perspectives into clinical decision-making may produce varied outcomes depending on individual patient circumstances and priorities. Outcomes of tailored prescribing decisions may also not be immediately apparent. For example, a tailored decision to stop primary prevention medication may not produce any recordable effect for some time, if at all. Assessing best, or better, practice using simple outcome measures may not be sufficient.

The research demonstrates that if we are to address the identified barriers to tailored (de)prescribing of permission and supporting professional confidence in the knowledge work of complex decision-making, clinicians need additional tools to support judgement of better practice. 6 Clinicians describe needing ‘permission’ to work ‘beyond guidelines’6,28 and so seek a validated framework against which they can judge and defend these interpretations. 28 TAILOR addresses that gap through developing a realist programme theory that describes best practice for tailored deprescribing, and so provides critical framework for practitioners, patients and managers to judge the quality of tailored care. The TAILOR framework will also provide the additional evidence needed to address the identified organisational barriers to delivery and so describe practice redesign.

Our review of the literature, previous research and stakeholder engagement described a number of elements that may be important in developing this framework. These were incorporated into a draft programme theory used to inform the TAILOR realist review (Figure 1).

FIGURE 1.

Draft programme theory describing elements needed for individual tailoring of medicines.

In Chapter 2, we describe how these findings and observations shaped the design of the TAILOR evidence synthesis.

Justification for a scoping review

Preliminary searches undertaken in preparing our bid demonstrated that the current body of evidence on deprescribing is disparate with significant heterogeneity. Studies cover many topic areas (e.g. clinical problems and research methods used), although the volume of scholarship in each area appears to be relatively small. We concluded that standard systematic review methods (including meta-analysis) would not allow us to adequately describe and integrate the diverse literature in a way that met our goals to offer clinicians, patients and policy-makers resources to support safe and effective tailored deprescribing.

We therefore opted for a scoping review to identify, map and draw together data in a useable form. Scoping reviews are recognised to be the most appropriate methodology when reviewing the evidence on complex interventions, allowing for the variability and complexity of the intervention and evaluation methods. 50 We chose a scoping review to enable us to systematically distil, from a diverse literature, the data that would support clinical decision-making.

Justification for a realist review

Our second goal was to provide a robust, evidence-informed framework describing the key components needed to deliver person-centred (tailored) deprescribing. Our intention was that this framework provide clinicians with a model of ‘best practice’ to support them in their daily work, and a model that could explain the heterogeneity in the literature on deprescribing.

We recognised tailored deprescribing as a complex intervention. 1 An intervention is defined as complex (rather than complicated) because it consists of numerous components interacting in non-linear ways and is sensitive to context. 51 As discussed in Chapter 1, Addressing the problem: what we already know about deprescribing, addressing problematic polypharmacy needs a tailored approach to prescribing that recognises compromise between biomedical (condition-specific factors increasingly in the context of multimorbidity) and biographical (individual context, preferences and priorities) perspectives. 1 Decisions involve weighing up multiple factors that vary in themselves and through interaction with each other. 21 Tailored deprescribing is therefore an example of a complex intervention, in which controlled and uncontrollable variation is inevitable and the active ingredient(s) may behave differently in varying contexts and for different people. 52

The realist review methodology is particularly useful for understanding and illuminating the relationships and impact of the interaction between the components of a complex intervention. 53,54 Realist reviews ask ‘what works, for whom, in what circumstances, to what extent, how and why?’ and consider the interaction between context, mechanism and outcome [how particular contexts (e.g. people, practices) trigger or interfere with mechanisms to generate the observed outcomes]. 55 Realist reviews generate explanations about the mechanisms by which stopping medication may (or may not) achieve impact in different settings and within different subgroups.

We therefore opted to use a realist review methodology to address our second research question, providing clinicians with a framework that describes and explains what is needed to support ‘better’ practice. 41 Our intention was to provide a framework against which to ‘defend’ good practice, addressing the recognised barriers to tailored care of permission, professional confidence and performance management (see Table 1). We also anticipated that that framework could help explain the heterogeneity revealed in reviews of the literature.

Justification for a combined literature search

Although we identified a need for two distinct analytical approaches, our common goal was to deliver outputs that supported the clinical task of tailored deprescribing to address problematic polypharmacy. Our funders had stipulated a particular focus on an older population living with multimorbidity. Our initial proposal, therefore, was to use a combined search strategy to collect the initial data for analysis using both methodological approaches (Figure 2).

FIGURE 2.

Outlining the initial workflow for the TAILOR medication synthesis. WP, work package.

Finalising work plans

Kavalidou’s findings were discussed at an extended team meeting. We recognised that this data set provided the richness needed for a realist review. However, we were concerned that it would not allow us to meet our goal to provide a useful resource to clinicians from a scoping review. We therefore opted for a revised and refocused scoping review based on a revised search strategy.

The revised final work packages for TAILOR are shown in Table 2.

Joanne Reeve provided overview of, and support for, all work packages.

Detailing the research team

As described in our protocol (version 1.1, July 2019), we assembled a team of people to undertake this work including:

• core research team (project working group) – responsible for delivery of the work as detailed in Table 2

• academic advisory group – the additional co-authors of this report (see Chapter 10, Reviewing how we went about the work) responsible for overseeing the academic rigour of the research

• stakeholder group – consisting of end users of our work, responsible for ensuring that the research remains relevant and for supporting the dissemination activities.

Stage 1: setting the research question

The scoping review questions were agreed by the research team in collaboration with our stakeholder and advisory groups. The overarching research question was to identify what recent quantitative and qualitative evidence exists to support the safe, effective and acceptable stopping of medication in older people with multimorbidity and polypharmacy. We wanted to offer clinicians a resource (data) set to inform their clinical judgement when making tailored prescribing decisions. Our intention was therefore to produce a map of the current evidence base for deprescribing practice outlining what is being done, where and for what effect. Our map was also to describe the ongoing gaps in our knowledge: areas where clinical judgement is particularly necessary. We therefore described a focused set of subquestions for the scoping review:

• What research methods (study designs) have been used in the studies that focus on this topic? This offers clinicians an overview of what types of research have been done and where there are gaps (e.g. clinical trials with a biomedical outcome and/or intervention studies with a patient-centred outcome). It allows clinicians to judge the value and limitations of the reported TAILOR data set in relation to the specific clinical challenges they face.

• What clinical strategies, contexts and outcomes have been studied on this topic? This offers clinicians an overview of what types of clinical interventions have been studied, and where there are gaps. It allows clinicians to judge the value and limitations of the reported TAILOR data set in addressing the specific clinical challenges they face.

• What tools are available to support addressing problematic pharmacy in older people with multimorbidity and polypharmacy? This offers clinicians an overview of what tools for clinical practice exist and what the data tell us about the use and effectiveness of these tools.

Stage 2: search strategy

We used a comprehensive, broad and iterative approach to identify relevant literature. We conducted an initial exploratory search using search terms identified by the review team and PubMed PubReminer [URL: https://hgserver2.amc.nl/cgi-bin/miner/miner2.cgi (accessed 16 June 2021)] in MEDLINE (via Ovid). We then identified a set of key relevant studies identified in a recent scoping exercise undertaken by Kat Kavalidou (see Chapter 2, Refining the work plan). Free-text and thesaurus terms of MEDLINE records of the relevant key studies were analysed and the search strategy was amended to ensure that it captured all key relevant records. We conducted a sensitivity analysis on the search by comparing the retrieval of different search techniques (e.g. proximity operators, phrase searching and field searching) to develop a scoping search strategy that ensured the retrieval of all key relevant studies.

The exploratory search was then peer-reviewed by a second reviewer (RH). The following keywords formed the main structure of the search: A – multimorbidity terms combined with OR; B – polypharmacy terms combined with OR; C – deprescribing terms combined with OR; D – aged terms combined with OR. The initial findings suggested that not all relevant studies would be captured by combining A AND B AND C AND D; therefore, a multisearch combination approach was developed: search 1 – (A OR B) AND C AND D, Search 2 – A AND B AND C, Search 3 – (A OR B) AND C AND Qualitative terms. The results of search 1, search 2 and search 3 were combined with OR to obtain a single set of search results. The final version of the exploratory search was then translated into other databases (see Appendix 1 for full details of the search strategies).

We conducted comprehensive searches in MEDLINE, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Web of Science, EMBASE, The Cochrane Library [Cochrane Database of Systematic Reviews (CDSR), Cochrane Central Register of Controlled Trials (CENTRAL)], Joanna Briggs Institute Database of Systematic Reviews and Implementation Reports, Google (Google Inc., Mountain View, CA, USA) and Google Scholar (targeted searches for both Google sources). The search was limited to studies published in English between 2009 and 30 August 2019. The search was then updated on 23 June 2020 with an addition to the search of ‘five or more’ as a free-text term in polypharmacy concept (searches with this new term were also backdated to 2009 to capture any earlier studies that may have been missed in the initial search). An additional supplementary PubMed search was also conducted to ensure that online preprints were captured. An experienced information specialist (MM) conducted the searches. All searches were peer-reviewed by at least one other member of the review team. We also scanned through the reference lists of eligible articles to identify additional relevant studies. Finally, we conducted an abbreviated version of the CLUSTER search approach,60 using key relevant studies to identify sibling studies and additional relevant studies (via citation searching, lead author searching and project/tool searching).

Stage 3: selecting studies

Search results were downloaded into EndNote [Clarivate Analytics (formerly Thomson Reuters), Philadelphia, PA, USA], deduplicated and then uploaded into Covidence software (Melbourne, VIC, Australia) for screening. A two-stage screening process was conducted. First, all titles and abstracts were screened. Records that clearly met the inclusion criteria, or records for which it was not possible to tell from the title and abstract whether or not the study was relevant, were sent through to full-text screening. Full-texts were then screened against the eligibility criteria. One reviewer screened all records (MM) and a second reviewer (Gerlinde Pilkington, Yenal Dundar and Katherine Edwards) independently screened all records. Disagreements were resolved by a third reviewer (RH).

Stage 4: charting the data

Data were extracted on study design, population characteristics, intervention characteristics [using the Template for Intervention Description and Replication (TIDieR) framework],61 health inequalities [using the PROGnosis RESearch Strategy partneship+ (PROGRESS+) framework],62 and outcomes of interest. The template was piloted and all data were extracted by two reviewers (MM, Katherine Edwards) independently and cross-checked using Microsoft Access^® (Microsoft Corporation, Redmond, WA, USA).

Stage 5: collating and reporting

The results were synthesised to address the aims of the review (i.e. provide a map of the evidence in relation to the effects, safety and acceptability of interventions to support deprescribing in the elderly with multimorbidity and polypharmacy). A narrative descriptive approach to the synthesis was adopted to map the evidence on research methods, contexts, tools and outcomes used in deprescribing interventions for elderly people with multimorbidity and polypharmacy. Outcomes were categorised as effects, safety or acceptability. In addition, intervention outcome results were summarised as having a positive, negative or equivocal effect. A framework synthesis approach was adopted using the TIDieR framework61 to synthesise data on deprescribing intervention characteristics. TIDieR is a checklist designed to unpick complex intervention components; however, it is a generic checklist designed to be applied to all different types of complex interventions. Item 4 of the TIDieR framework (‘What procedures’; see Table 7) asks the reviewer to ‘Describe each of the procedures, activities, and/or processes used in the intervention, including any enabling or support activities’. 61 Given the complexity of the ‘procedures, activities and/or processes’ that we observed in the reported deprescribing studies, we expanded this TIDieR item using Reeve et al. ’s63 published framework detailing the expected elements of the deprescribing process. This allowed us to extract in greater detail and in a consistent manner the specific deprescribing processes from across multiple studies. Full details are described in Appendix 2, Table 23. In some studies, insufficient information was reported to rate an item as a full ‘yes’ and therefore a ‘partial yes’ was assigned. Based on the findings of the scoping review, a stage 1 logic model64 [i.e. a static (visual) model of components of the logic rather than the interactions/interdependencies] was developed to summarise the evidence in terms of population, intervention, context, outcome (PICO).

What research methods are being used in the studies on this topic?

Our first review question asked, ‘what type of research methods are used to explore deprescribing in this patient group?’. Our findings revealed variability in the study designs used, study populations and durations, and the definitions of multimorbidity applied.

What clinical strategies, contexts and outcomes have been studied on this topic?

Clinical strategies

We used the TIDieR framework61 supplemented by the Reeve et al. 63 framework for deprescribing to describe the clinical strategies used as interventions in these studies.

Owing to the complex nature of the deprescribing interventions employed, the TIDieR framework was found to be insufficient on its own to allow for a rich description of the deprescribing strategies. Specifically, this related to the lack of a detailed description of the deprescribing intervention components. Therefore, we used a novel approach in supplementing the TIDieR framework with Reeve et al. ’s63 deprescribing process framework (see Appendix 2, Table 23). Reeve et al. 63 described seven steps needed to support robust deprescribing practice: (1) a comprehensive medical history, (2) an assessment of risk/harm, (3) an identification of potentially inappropriate medicines, (4) a shared decision on whether or not to stop, (5) communication of a plan, (6) implementation and monitoring and (7) documenting the process.

The purpose of using both these frameworks to describe and assess included studies was twofold: first, to assess the quality of the reporting in deprescribing studies and, second, to identify specific intervention components and delivery modes of the deprescribing strategies to allow for an assessment of the replicability of the deprescribing strategies in practice. In using the two frameworks together, therefore, we can provide clinicians with a more detailed map of what deprescribing strategies are used and how they are used.

The extent to which individual studies (Table 7) and the studies collectively (Figure 6) reported on each of the TIDieR items is shown on the following pages (see also Appendix 2, Table 23, for further details of the frameworks).

TABLE 7 - Completeness of the description of the intervention by study and by TIDieR61 item

NR (light purple), not reported; P (orange), partial yes; Y (blue), yes.

Items 4a–g use the deprescribing framework by Reeve et al. 63

Study (first author and year)	1: brief name	2: why	3: what (materials)	4: what (procedures code)	4a: medication historya	4b: assess risk and patient factorsa	4c: identify inappropriate medicationsa	4d: shared decision-makinga	4e: planning, documentation and communicationa	4f: monitoring and supporta	4g: documentationa	5: who provided	6: how	7: where	8: when and how much	9: tailoring	10: modifications
Boersma 201966	Y	Y	Y	P	Y	P	Y	P	P	NR	NR	P	Y	Y	P	Y	Y
Caffiero 201767	Y	Y	NR	P	Y	Y	Y	Y	Y	NR	Y	Y	Y	Y	Y	Y	NR
Campins 201768	Y	Y	Y	P	NR	NR	Y	Y	P	P	P	P	P	Y	P	Y	NR
Chiarelli 202069	Y	Y	Y	P	Y	P	Y	NR	P	NR	Y	P	P	Y	Y	Y	NR
Curtin 202070	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	NR
Fried 201771	Y	Y	Y	P	Y	Y	Y	Y	Y	NR	NR	P	Y	Y	P	Y	NR
Köberlein-Neu 201672	Y	Y	NR	P	Y	Y	Y	Y	Y	Y	P	Y	Y	Y	Y	Y	NR
Komagamine 201773	Y	NR	Y	P	Y	Y	Y	Y	NR	P	P	P	P	Y	P	Y	NR
Komagamine 201874	Y	Y	NR	P	Y	P	Y	NR	NR	P	NR	P	P	Y	P	Y	NR
Martín Lesende 201380	Y	Y	Y	P	Y	P	Y	P	P	NR	NR	P	P	Y	P	Y	NR
McCarthy 201775	Y	Y	Y	P	Y	Y	Y	Y	P	P	P	P	P	Y	P	Y	Y
McDonald 201976	Y	Y	Y	P	Y	Y	Y	Y	Y	Y	Y	P	Y	Y	Y	Y	NR
Muth 201677	Y	Y	Y	P	Y	Y	Y	Y	P	NR	P	Y	Y	Y	Y	Y	Y
Muth 201878	Y	Y	Y	P	Y	Y	Y	Y	P	NR	P	Y	Y	Y	Y	Y	NR
Petersen 201879	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	P	P	Y	P	Y	NR
Potter 201981	Y	Y	NR	P	Y	Y	Y	Y	Y	Y	P	P	Y	Y	P	Y	NR
Russell 201982	Y	Y	NR	P	Y	NR	Y	Y	P	NR	P	P	P	Y	P	Y	NR
San-José 202083	Y	NR	Y	P	Y	Y	Y	NR	P	P	Y	Y	Y	Y	P	Y	NR
van Summeren 201784	Y	Y	Y	P	Y	P	Y	Y	Y	P	Y	Y	Y	Y	Y	Y	NR
Zechmann 201985	Y	Y	Y	P	Y	Y	Y	Y	P	NR	P	P	P	P	P	Y	NR

FIGURE 6.

Completeness of TIDieR items61 reported in included studies. a, Reeve et al. 63 framework.

A more detailed description of each criterion from the TIDieR framework is offered below, including items 11 and 12.

Item 1: brief name

All included studies (100%) reported the name of or a phrase that described the intervention. Eleven studies provided precise names for the intervention. The remaining studies provided a brief phrase or description.

Item 2: why (rationale, theoretical framework, goal)

Eighteen (90%) of the included studies provided the rationale for the intervention. None of the studies was explicit in reporting a named theory (e.g. theory of planned behaviour) to underpin their intervention. The rationale or underlying theories provided were largely based on the findings of previous research with reference to intervention components [e.g. academic detailing, medication review and specific tools (e.g. STOPP/START criteria) known to be effective], barriers to and facilitators of deprescribing (e.g. availability of health-care specialists with familiarity in managing polypharmacy in multimorbid populations and patient priorities) or setting (e.g. patients in hospital are seen as a captive audience and therefore more likely to be motivated to stop medications, which provides time for patients to discuss their options and the opportunity to observe patient outcomes closely).

Item 3: what (materials)

Fourteen (70%) studies reported using 20 different tools to guide the deprescribing process. Seven types of tools were identified: six studies (30%) used a clinical decision support system (CDSS), five (25%) studies used a criteria-led tool, one (5%) study used an algorithm, one (5%) used a CDSS plus criteria-led tool and one (5%) used an algorithm plus criteria-led tool (see What tools are available to support addressing problematic pharmacy in older people with multimorbidity and polypharmacy? and Table 27 in Report Supplementary Material 1 for details on the specific tools used). Twelve (60%) studies report using a single tool. Two (10%) studies used more than one tool to guide deprescribing. Six (30%) studies did not use a specific tool to guide the deprescribing process. Instead, pharmacists and clinicians were free to use any tool they wished or rely on their own expertise to propose and implement medication changes.

Item 4: what (procedures)

The seven elements of the deprescribing process as reported in Reeve et al. 63 were used to further define the procedures involved in the intervention. Only two (10%) studies reported on all seven items.

Item 4a: medication history

Nearly all studies (19/20, 95%) were explicit in detailing the taking of a patient medication history. One study, by Boersma et al. ,66 reported using a tool [Structured History taking of Medication use (SHiM)]66 to inform the medication review process.

Item 4b: assessment of risk of harm and benefit and individual patient factors

Eighteen (90%) studies recorded an assessment of risk of harm and benefit and/or patient factors. Two studies did not report an assessment of these factors in the deprescribing process. Two studies used a checklist-based pre-consultation interview tool, Medication-Monitoring-List (MediMoL), to assess risk and patient factors. 77,78

Item 4c: identify potentially inappropriate medications

All 20 (100%) studies recorded details of how potentially inappropriate medications were identified (see Chapter 3, Methods, for details on the tools used).

Item 4d: shared decision-making

Seventeen (85%) studies reported incorporating patient preferences into the deprescribing process. Seven studies (35%) incorporated patient preferences into the process before providers decided on the medications to deprescribe. 72,75,77–79,84,85 Of these, six were conducted in the primary care setting and one was conducted in a tertiary setting. Seven studies (35%) involved patient discussion at the end of the process, only after medications for deprescribing had been identified. 67,68,70,71,73,76,79 Three studies utilised tools to elicit patient preferences prior to the identification of medicines to be deprescribed; Muth et al. 77,78 used MediMoL, whereas van Summeren et al. 84 used the outcome prioritisation tool. 84 Patient preference was the focus of the deprescribing process in van Summeren et al. 84 In three studies it was unclear at what point in the deprescribing process the patient was involved.

Item 4e: plan tapering/withdrawal process with documentation and communication

Eight (40%) studies described planning the tapering or withdrawal process with documentation and communication among health-care professionals. Ten (50%) studies lacked information on the tapering and withdrawal process. Two (10%) studies did not state a plan for the tapering or withdrawal of medications.

Item 4f: conduct monitoring and support

Twelve (60%) studies detailed monitoring and/or support for patients following deprescribing. This involved the symptom and safety monitoring of patients (e.g. for adverse drug withdrawal events or disease relapse). Support offered included additional consultations and telephone follow-ups.

Item 4g: documentation

Sixteen (80%) studies described the process for documenting the outcome of deprescribing (e.g. dose reduced or medication ceased). Of these, seven (35%) describe sharing the documentation with all relevant health-care professionals.

Item 5: who provided

In more than half (n = 11, 55%) of the studies a physician led the deprescribing process (i.e. identified the medications to deprescribe). Of these, seven studies were general practitioner (GP)/primary care physician led and one was led by a specialist registrar in geriatric medicine. Five studies (25%) were pharmacist led and in four studies (20%) the deprescribing recommendations were made by the team involved in patient care. Three studies involved a single intervention provider (two GP led, one clinician led). The majority (n = 17, 85%) involved more than one provider in the deprescribing process. Pharmacists and specialist geriatric physicians were more likely to be involved in the deprescribing process in secondary care settings than in primary care settings. Table 27 (see Report Supplementary Material 1) provides more information on the personnel involved in the provision of the intervention.

Item 6: how

All studies (n = 20, 100%) detailed to some extent the mode of delivery of the deprescribing intervention. Multiple methods of delivery were reported involving face-to-face, online, telephone, electronic health record, fax and written modes of delivery. All studies (n = 20, 100%) employed an individual delivery format.

Item 7: where

Ten studies (50%) carried out the intervention in primary care settings, seven (35%) studies were set in secondary care settings, two (10%) studies were set in tertiary care settings and one (5%) study was set in a pharmacy call centre (Table 8).

TABLE 8 - Settings in which deprescribing interventions were delivered

Setting	Frequency (%)
Primary care	10 (50)
Secondary care	7 (35)
Tertiary care	2 (10)
Other (pharmacy call centre)	1 (5)

Item 8: when and how much

All studies (100%) described when the deprescribing intervention took place. Seven (35%) studies invited patients for a medication review. In seven (35%) studies patients were invited to participate upon or during hospital admission. Four (20%) studies invited patients who were attending another GP or outpatient appointment or were awaiting a primary care appointment. Two (10%) studies referred patients from primary care or hospital. Four (20%) studies reported the intervention as being delivered on a single occasion and one (5%) study offered an optional second consultation. In two studies a medication review was offered twice (once at hospital admission and again at discharge, and once after invitation for medication review and again at 6 months) and one study offered an annual medication review with quarterly targeted reviews. In 12 studies (60%) it was unclear how many times the intervention was delivered.

Item 9: tailoring

All 20 (100%) studies reported tailoring their interventions with decisions to deprescribe based on individual patient requirements as described under item 4d. Additional tailoring approaches described included protected time for individual consultations to discuss prescribing decisions and incorporation of individual patient medication-related problems. These were in addition to medications that physicians judged to be inappropriate or unnecessary but were sometimes continued owing to patients’ preference, and recommendations made regarding the need to simplify the regimen of patients with problems with adherence, compliance and poor social support.

Item 10: modifications

Three (15%) studies reported modifying the intervention. Boersma et al. 66 modified the intervention during the study by introducing consensus-based instructions to standardise the prescribing recommendations. Two pilot studies reported on making changes to the intervention after completion of the study to inform larger studies. Muth et al. 77 intensified the provider training and written CDSS manual. McCarthy et al. 75 made minor modifications to the training videos and medication review template to improve clarity of instruction and reduce repetition.

Item 11: adherence to the study recommendations

One (5%) study reported training pharmacists and home care specialists with a planned assessment of the training on 10 patients. Eleven studies (55%) reported that training of intervention providers was undertaken but planned assessment was not reported. Six studies planned assessment of patient adherence.

Item 12: outcome of training

None of the 11 studies that reported training intervention providers reported on the outcome of the training. Five of the six studies that planned to assess patient adherence reported on adherence outcomes.

Overall, in summary, our analysis revealed that studies offered clear accounts of the goals of the deprescribing interventions used and to whom they were offered (i.e. which patients were included). However, there was often less detail reported on who delivered the intervention and how (what specifically was done).

Using Reeve et al. ’s63 framework to further analyse details of the interventions revealed that most studies offered clear accounts of the assessment of patients leading to a decision to potentially deprescribe. Details on subsequent actions, including communication, documentation and planning, follow-up monitoring and support, and documentation of the clinical plan were all less clearly described.

Contexts

We sought to understand the context in which deprescribing interventions were being delivered by considering the clinical focus of the study (whether on stopping medication, or improving prescribing–medicines optimisation). We also examined the extent to which researchers considered the population context in which studies took place, through an examination of assessment of markers of inequalities.

Focus of the intervention

Deprescribing was the focus of the intervention in eight studies. The remaining 12 studies involved deprescribing as part of a wider medicines optimisation context (Table 9).

TABLE 9 - Focus of intervention: scoping review

Focus of the intervention	Frequency (%)
Deprescribing	8 (40)
Medicines optimisation	12 (60)

Inequalities

By assessing population characteristics using the PROGRESS+ framework,62 we sought to offer end users of our review an analysis of which populations the findings could be generalised to. Given the focus of our research question, all studies considered age and comorbidity inequalities. We assessed the extent to which other population contextual factors were also considered through an assessment of inequalities using the PROGRESS+ framework. This describes nine characteristics that stratify health opportunities and outcomes within populations, namely place, race/ethnicity, occupation, gender, religion, education, socioeconomic status, social capital and other.

Only two studies (10%) explicitly report a PROGRESS+ inequality, namely Medicare status, as being the focus of their study population.

PROGRESS+ inequalities collected

Nineteen studies (95%) collected baseline data from participants on characteristics described by the PROGRESS+ inequalities (Figure 7). All 19 studies collected data on gender. The one study that did not report baseline inequality characteristics reported pilot data within a study protocol. Other inequalities collected were age and comorbidity status (which reflected the target population) and Medicare status (i.e. health insurance).

FIGURE 7.

PROGRESS+ inequalities collected. SES, socioeconomic status.

PROGRESS+ inequalities analysed

Nine studies (45%) analysed data on PROGRESS+ inequalities (Figure 8). The most common inequality analysed was gender (nine studies). Other inequalities analysed were age and comorbidity status (which reflected the target population) and Medicare status (i.e health insurance). Inequality variables were mostly adjusted for in statistical analyses (e.g. through logistic regression models). However, only one study discussed the effect of population characteristics and inequalities on outcomes (Figure 9).

FIGURE 8.

PROGRESS+ inequalities data analysed. SES, socioeconomic status.

FIGURE 9.

PROGRESS+ inequalities data discussed. SES, socioeconomic status.

In summary, population-level contextual information in the form of key characteristics known to have an impact on inequalities was poorly reported and discussed across the studies. Reporting of clinical contextual information (deprescribing vs. medicines optimisation) was present in most studies, with approximately half looking at deprescribing.

What tools are available to support addressing problematic pharmacy in older people with multimorbidity and polypharmacy?

Studies reported a range of tools used to support deprescribing. A CDSS is a computer application designed to aid clinicians in making deprescribing decisions. It may incorporate algorithm- or criteria-led tools in its design. Algorithms are a set of rules or steps that guide deprescribing and are followed in a pre-determined way to lead to an outcome. Criteria-led decision-making involves the use of a list of criteria to consider when making decisions to deprescribe medications. Five studies69,72,77–79 reported the use of tools within a wider structured framework involving a process-driven approach to deprescribing, which details a set of rules or steps to be followed from patient identification through to discharge and follow-up.

Table 12 details the tools used in the deprescribing interventions. The seven studies employing a CDSS reported using seven different CDSSs. Many were created by the clinical team or study team for the purposes of the study, commonly drawing on previously published criteria or algorithms (see Table 12). In addition, five studies report using other tools to inform medication reviews. One used a tool to identify patient priorities and one study used a protocol for the withdrawal and reinstatement of drugs associated with potential for adverse drug withdrawal events. The most reported tool was the criteria-led STOPP/START tool.

Seven of the CDSS tools report incorporating criteria-led or algorithm tools or other frameworks (Table 13).

As described in Table 11, acceptability for all tools was poorly reported. Clinicians reported acceptability across all tools, but patient perceptions of acceptability were not generally recorded. Clinical safety concerns were described for secondary care studies not using any tool but inconsistently reported (positive or negative) in other contexts. Effectiveness reports were varied for all tools.

Reviewing the research methods used

To understand the utility of knowledge/evidence for clinical practice, we have to understand how it has been generated. The first part of our review examined the research methods used by studies in this area. As outlined in Chapter 1, and underlined by our patient partners, TAILOR sought to look in particular at a person-centred understanding of deprescribing. Here, therefore, we also consider the extent to which the published research methods support a whole-person understanding of deprescribing approaches.

Quantitative study designs were the most common approach used to exploring issues around effectiveness, safety and applicability, with the RCT as the most common method within that group. Around half of the studies were intervention studies, and half used observational designs (e.g. cohort). Interventional designs were associated with the consistent reporting of outcomes related to impact and safety. The studies therefore have the potential to provide clinicians with robust data on focused outcomes.

The patient’s voice was less clearly reported in the study set. For example, patient reports on acceptability were generally missing. Quantitative study designs can capture patient perspectives on treatments, but qualitative designs are often used to ensure that the patient’s voice is heard. Although we identified an abundance of qualitative evidence around deprescribing during our screening of studies, this work related to an understanding of deprescribing approaches in general, rather than an examination of the specific effects of an intervention. These studies were therefore excluded from our review at screening stage. Only one qualitative design met our inclusion criteria. 85 This reported that patients generally found a discussion about deprescribing to be acceptable, with conservative practice (hesitancy to change) and fragmented care both described as barriers to stopping medicines in practice.

Previous reviews32 have raised concerns about evidence relating to patients’ experiences of medications being overlooked in evidence synthesis work such as guideline preparation. Therefore, although our scoping review can provide importance clinical data supporting clinician interpretation, it misses an important wider area of study. This will be addressed in our realist review (see Chapters 6–8).

Our review noted that most studies were small to moderate in size with a short follow-up period (all < 1 year, and 30% ≤ 3 months). Study size and duration potentially affect the generalisability of findings to a clinical setting with divergent populations and continuity of care in effect.

We noted that studies on deprescribing in the elderly were not necessarily explicit in defining multimorbidity in their inclusion criteria (as evidenced by the large number of studies excluded for not meeting the multimorbidity criteria). This could be because there is difficulty in defining ‘multimorbidity’ or that there is an underlying assumption that anyone who is taking five or more drugs is considered to have multimorbidity (even though this may not strictly be true). This meant that a large number of studies were excluded purely on the basis that they did not report on the number of multimorbidities in the inclusion criteria. Again, this may have implications for clinicians using our findings. Multimorbidity is rarely a diagnosis/criterion used explicitly in clinical practice. Translating research findings based on strict clinical criteria to the practice context in which patient needs may be more uncertain is challenging. Our decision in applying a strict definition for the review was to ensure that we could offer clinicians a clear account of which patients our findings relate to, and which they do not.

In summary, our review findings offer clinicians a defined data set on the clinical outcomes related to deprescribing practice in a defined population of patients aged > 50 years living with two or more long-term conditions and taking five or more medicines per day. The TAILOR data set is, however, incomplete and does not provide insights into patient experience. This will be addressed by our realist review.

Reviewing the clinical interventions used

Our review next examined what type of interventions are being conducted, where and to what effect. By using a modified TIDieR framework, we offer a detailed account of what has been done in the published studies (see Table 7). Our outcomes data set (see Table 11) offers a visual overview of outcomes related to effectiveness, safety and acceptability.

The TIDieR analysis (see Table 7) and the logic model (see Figure 10) highlight the complexity and diversity of the deprescribing process. The included studies are heterogenous not only in terms of their populations, but also in terms of the intervention components, the types of outcomes assessed and the contexts within which deprescribing takes place. We discuss this further through consideration of our three foci of interest: clinical strategies used, context of the studies and outcomes assessment.

What tools are available to address problematic pharmacy in older people with multimorbidity and polypharmacy?

The review highlighted the diversity in the tools available for deprescribing, with studies evaluating a range of CDSSs, criteria-led guidance, algorithms and frameworks. In addition, the reported amount of training and level of experience required to deploy these tools also varied.

More importantly, the review findings highlight that although a variety of tools and guidance is available to provide decision support for deprescribing by identifying eligible patients and identifying the medications to deprescribe, the use of these tools alone is not enough to ensure a successful outcome. The lack of a clear direction of effectiveness impact by tool type (see Table 9) suggests that deprescribing is about more than just a set of tools: it is a patient-centred decision-making process that needs to recognise and understand diversity across populations, deprescribing processes, implementation and outcomes.

Methodological and topic-specific issues arising from our work

We experienced considerable challenges in identifying a focused data set for this review, related in part to problems in defining and reporting multimorbidity criteria in the title and abstracts of papers. We therefore had to amend our search strategy to merge two concepts relating to polypharmacy and multimorbidity into a single concept (multimorbidity or polypharmacy) to ensure that we were picking up key studies (a change needed/recognised by our sensitivity analysis). As a result, large numbers of studies were picked up and underwent full-text screening. Even then, the majority of studies failed to define what they meant by multimorbidity, or failed to include it as part of their inclusion criteria.

Similar challenges relate to the multiple definitions of both polypharmacy and deprescribing (e.g. ‘five or more’ and stopping medication, respectively). Therefore, we had to use an iterative approach to search strategy development that added in new terms when papers were found through supplemental searching that were not captured by the database searches.

If we are to meaningfully synthesise future work in this area, we need to pay further attention to how research and clinical teams are both describing their interventions, and the quality of reporting within published work.

Strengths and limitations of our work

One of the key strengths of this scoping review was the comprehensive, iterative approach used in the identification of studies. In a highly varied field of clinical practice and research, we describe a focused review of a carefully defined element of that practice. We can therefore be sure that the reported variability is a feature of the published research and not a limitation of our search strategy.

Use of the TIDieR framework61 to extract data on intervention components and the amendment of this to incorporate the Reeve et al. 63 framework allowed for greater detail of the reporting of deprescribing intervention processes than would have otherwise been possible. Our review therefore offers a clear account of what the deprescribing intervention was in each of our included studies. Again, we can therefore be confident that the reported variability reflects the reality of the approaches being used.

Limitations include the lack of quality assessment of included studies. This is not a requirement of a scoping review in which the goal is to map the field of study57 (and given that no statistical meta-analysis was planned). Applying a quality assessment standard may simply have limited our data set and our ability to comment meaningfully on the evidence base that we do have. Furthermore, we used strict inclusion criteria in defining multimorbidity, requiring that study participants had two or more long-term conditions. This potentially excluded many studies that would otherwise have met the remaining inclusion criteria. To address this issue we refined the population criteria to include studies that were not explicit in defining the number of multimorbidities, but that either reported their population as multimorbid, or gave a detailed report of their population that revealed it to be multimorbid. Strict application of this criterion enabled us to confidently offer a detailed analysis of a clearly defined data set. However, it flags two issues relevant to the application of findings to clinical practice. The first issue is about the quality of reporting of studies, that is how well study authors detail the population included and so allow clinicians to judge applicability to their own patients. The second issue is an ongoing concern about the clinical applicability of the concept of ‘multimorbidity’. In clinical practice, clinicians and patients will be discussing a given patient-centred concern (e.g. treatment burden, quality of life or specific symptoms) within a context of a patient living with multiple long-term conditions (multimorbidity), but it is rare that a clinician will be specifically ‘treating multimorbidity’. The careful definition of multimorbidity as an inclusion criterion in this review represents both an academic strength of our work and an applied limitation.

Implications for future work

Our review demonstrates through rigorous analysis of the literature something that is commonly recognised within clinical practice: deprescribing is a complex process. Our findings demonstrate that we need to understand deprescribing not as a technical process akin to a single diagnostic test for which the mechanisms of action are easily understood. Instead, we need to recognise it as a complex process (potentially part of a wider complex intervention of tailored prescribing practice) in which the mechanisms of action are not yet fully understood. Clinical judgement remains paramount in the rigorous application of this process. We consider the implications of this in the light of our realist findings in Chapter 10.

Step 1: clarifying the scope and constructing a more refined initial programme theory

A realist review begins with an initial ‘draft’ theory of how any intervention is understood to work, also known as a programme theory. Our prestudy draft programme theory is shown in Figure 11 and was further developed in stakeholder discussion in the first 2 months of the project. To develop our initial programme theory further we held a half-day face-to-face meeting at which four members of the project team (JR, JK, GW and DM) discussed and debated the processes and assumptions behind deprescribing. We drew on our content expertise for this process as well as our initial programme theory. At the end of this meeting we developed a more refined programme theory that set out the important concepts we needed to consider in our realist review, as well as putative inter-relationships between these processed. This more refined programme theory was circulated to the wider project team by e-mail and further refined based on their feedback (Figure 12).

FIGURE 11.

Prestudy draft programme theory: elements needed for tailored prescribing.

FIGURE 12.

Refined version of the initial programme theory.

This more refined initial programme theory informed the realist review in the following ways:

• Searching – we developed our searches so that they would capture the concepts found within our programme theory.

• Data analysis – the concepts contained within the programme theory informed our sense making of the data. The programme theory also provided a means for us to organise our emerging context–mechanism–outcome configurations (CMOCs).

As the review progressed, we made further gradual refinements to the programme theory when relevant data emerged. During the review, we focused our CMOC development on parts of the programme theory that we judged to be the most important in providing explanations of deprescribing.

Step 2: searching for the evidence

Our search strategy describes a comprehensive, structured approach to identifying relevant literature from as many relevant sources as possible on the complex intervention that is stopping medication (in the context of individual tailoring of medication use). Petticrew50 explains that a search strategy for a review of a complex intervention needs to adopt broader eligibility criteria than those used in traditional systematic reviews, going beyond PICO to include context, processes and theory (mechanisms of action). Similarly, Peters et al. 57 propose that scoping reviews need to also consider populations (i.e. types of participants), context and ‘concepts’ (the interventions being examined and the outcomes used to assess their success). We combined these approaches to describe the search eligibility criteria we used for the realist review in the TAILOR study (Table 14).

The criteria described in Table 14 were used to produce a detailed search strategy in conjunction with our information specialist (Dr Nia Roberts). The initial search was tested and refined to the Ovid MEDLINE, Cochrane Library and EMBASE databases by analysing words contained in the title, abstract, and index terms of identified studies.

The refined search terms were then applied to the following databases: Ovid MEDLINE, EMBASE, CINAHL, The Cochrane Library [including CENTRAL and the Database of Abstracts of Reviews of Effects (DARE)], Cochrane Effective Practice and Organisation of Care (EPOC) Group Specialised Register, Campbell Collaboration Library of Systematic Reviews, Joanna Briggs Institute Database of Systematic Reviews and Implementation Reports, PsycInfo, Allied and Complementary Medicine Database (AMED) and CAB Abstracts (see Appendix 4 for details of our search strategy).

We also searched:

• Trial registries – https://clinicaltrials.gov/ and https://apps.who.int/trialsearch/.

• Grey literature – Google (Google Inc., Mountain View, CA, USA) and Google Scholar websites and websites of relevant stakeholders {including Royal College of General Practitioners Bright Ideas, National Clinical Guideline Centre, Royal Pharmaceutical Society and conference abstracts [e.g. Prescribing and Research in Medicines Management (PRIMM)]}. We also used personal communications to contact experts in the field who may have been able to signpost us to further relevant information.

We also used ‘pearling’, in which we examined the reference list of finally included relevant studies to identify additional documents.

In the scoping review, we described a search strategy to systematically identify the current literature on the approaches used, impact, safety and acceptability of interventions for stopping medication. The realist review method also specifically examines the mechanism of action of an intervention in different contexts. We therefore identified that two additional search elements would be needed for this work package:

1. ‘Sister papers’ (qualitative studies, process evaluations, etc.) for any studies identified in the above search,57 along with purposive searching to find relevant data that would enable us to develop and then confirm, refute or refine (‘test’) aspects of the draft programme theory.

2. For each theory area in our draft programme theory, we generated a sequence of search questions. For example, Figure 1 highlights ‘sense making’ as a concept in our draft programme theory. Emerging questions might include ‘what impact does the interaction between individual (patient and professional) beliefs and values and setting have on individual tailoring of medicines?’ From this, we drew up a series of specific search terms: a systematic search strategy that sought to identify research (as ‘data’) related to the targeted programme theories. This searching captured the additional relevant data necessary for our developing programme theory that were not captured within existing specific studies of stopping medications.

In addition to the formal searches for the realist review, we also drew on the search conducted for the scoping review described in Chapter 4. The 20 studies identified for inclusion in the scoping review were also screened for eligibility to be included in the realist review. Furthermore, we also screened qualitative studies identified by the scoping review search for their eligibility to be included in the realist review, as these were likely to contain rich information relevant for the development of the programme theory.

Step 3: selecting articles

Our inclusion and exclusion criteria for the review were identified based on our research questions, draft programme theory and discussion in the team. Our inclusion and exclusion criteria are outlined in Table 15.

Criteria were applied to the data set in a first phase of screening conducted at title and abstract level by Amadea Turk and a random sample of 10% of these was reviewed independently by Kamal Mahtani and Geoff Wong to help ensure that the criteria were applied consistently, and disagreements were resolved through discussion.

From these, the selection of full-text documents primarily focused on the extent to which the articles included data that could contribute to the development and refinement of the programme theory. Documents were assessed on their relevance (whether or not they contributed to the development of the programme theory) and rigour (whether or not the data contained in the documents were trustworthy). 106 Documents that did not include mention of involvement from patients in the deprescribing/medication management process were deemed to be of low relevance to our research question, which explored individually tailored approaches to medication management, and were therefore excluded from the review.

Steps 4 and 5: extracting, organising and synthesising evidence

Included full text documents were uploaded into NVivo (QSR international, Warrington, UK) a qualitative data software tool supporting analysis, and study characteristics were recorded in a Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) file. The coding of relevant extracts from documents was largely inductive, although consideration of concepts within the initial programme theory enabled a degree of deductive coding, as did discussions with members of the project team and stakeholder groups.

The initial stages of coding focused on the conceptual level and classified content into broad descriptive categories. This initial process helped us manage the data as well as make sense of the landscape of the literature, and helped us make decisions about whether or not we had captured enough data to further develop and refine the programme theory. During this first stage of coding, data were not immediately categorised into contexts, mechanisms and outcomes, but instead the focus was on looking at the data with an open mind to understand key issues. The data within these broad categories were then reread and, when needed, recoded and reclassified. Once this conceptual-level coding was completed, we started to consider whether or not these categories and the subcategories within them included sections relating to contexts, mechanisms and outcomes.

The development of CMOCs began by considering an outcome and using interpretations of the data to develop explanations of how specific contexts might have triggered different mechanisms to produce the outcome. A list of potential CMOCs was created by Amadea Turk and then shared and discussed with Geoff Wong, Joanne Reeve and Kamal Mahtani as well as with our patient and public involvement (PPI) partners. Developing CMOCs were then incorporated into the refined programme theory. Diagrams of partial theories or subsections of the programme theory were created to help guide and illustrate our findings. This process continued iteratively until we were able to develop CMOCs that explained what we judged were the most important parts of the programme theory.

The CMOCs were considered to have sufficient explanatory value when they met the key criteria for programme theory coherence. These criteria included:

• consilience – when the CMOC was able to account for as much of the possible data related to that CMOC

• simplicity: when the theory/CMOC was simple and did not have to have special (or ‘ad hoc’) assumptions made to explain data

• analogy: when the theory/CMOC fitted in with what we currently know/and or substantive theory.

Organisational/system-level factors

Realist analysis of organisational/system-level factors

Our realist analysis of organisational and system-level factors is summarised in Figure 14 below, with further details in Table 16.

FIGURE 14.

Partial programme theory CMOCs 1–5: influence of organisational/system-level factors.

TABLE 16 - Context–mechanism–outcome configurations for influence of organisational/system-level factors

ACF, advanced care facilities; C, context; M, mechanism; O, outcome.

Note

Data in Table 16 and subsequent CMOC tables are from the stated references. Quotations have been edited to < 50 words to meet with reporting standards.

CMOC	Description	Supporting data
1	In the absence of applicable deprescribing guidelines and evidence (C), health-care professionals may feel that they cannot make justifiable decisions regarding medication changes (O) because they do not feel that these decisions are supported by the system (M)	The GP quoted as follows: we can’t justify that you can just stop [prescribing] it. We can’t really do that because of how our treatment criteria look, where there aren’t any defined criteria for when we can stopNixon and Vendelø173GPs generally felt insufficiently supported by the guidelines in their efforts to treat hypertension in older people . . . I would really like to have a guideline that states: in elderly you have pay attention to this, this and thisvan Middelaar et al.212
2	When health-care professionals feel that they cannot make justifiable decisions that are supported by guidelines (C) they may be reluctant to make changes to medications (O) because they are afraid of negative consequences (M)	Physicians reported comfort in deprescribing preventive medication, but fewer were comfortable with deprescribing guideline-recommended therapeutic medications in patients with poor life expectancy. One explanation may be the fear of adverse withdrawal effects, which were also mentioned by physicians as a potential factor that prevented them from deprescribingDjatche et al.132Although most ‘guidelines’ were not proven in older people, particularly in the very old, those with co-morbidity, dementia, frailty, and limited life-expectancy, doctors are afraid of lawsuits and of the patient/family reaction if they do not follow all experts’ recommendationsGarfinkel138The Australian GPs were overwhelmingly negative about aged care and expressed dissatisfaction at the financial reimbursement provided for ACF services. Their attitudes towards deprescribing for ACF residents were influenced by concerns of blame in the case of negative health outcomesBolmsö et al.121
3	When health-care professionals are not supported by incentive and policy structures (C) they may not be able to take the time necessary for complex medication management processes (O) and they may be reluctant to make changes (O) because they do not feel supported to do so (M)	Inability to maintain follow-up to support a gradual process of deprescribing was a major frustration. One participant cited administrative rules as an impediment to follow-up: I think that the 2-year window [for reimbursement] makes it difficult to follow-up, especially for the complex patients that need that stepwise approachAnderson et al.117Patients who are most in need of a home medication review and most complex were considered less likely to receive equally detailed reviews because pharmacists seemed unwilling to substantially extend the review duration without additional remunerationMc Namara et al.166
4	When health-care professionals cannot access information about a patient’s medication regimen (C) they do not have an accurate understanding of the medication regimen (O) because they do not understand the patient’s history (M)	GPs found that there were discrepancies between the systems [across sectors] and that they were not properly informed about the changes made at the hospital . . . ‘Often we are not informed about the changes. It is us, the GPs, that must try and figure it all out, that isn’t easy’Laursen et al.160There were mixed views on the quality and extent of documentation of medicines in patient records . . . ‘The problem, many times there is no documentation about the medication. Most people write documentation. Some don’t write it. Some write incoherent handwriting’Al Shemeili et al.115All groups discussed health system structure concerns including limited funding and incomplete medical and medication histories. Medication histories rarely detail why and when medications were initiated and which prescriber is responsible for ongoing monitoringTurner et al.208
5	When health-care professionals are unsure about whose responsibility medication management is (C) they may struggle to engage in making medication changes (O) because they do not feel that they have ownership over the process (M)	Some GPs felt responsible towards overseeing patients’ medicines. Others were insecure about their task and desired ‘a clear line of responsibility’Bokhof et al.120Some physicians were of the opinion that they dealt with the management and control of their specialist condition only, and while this may involve an element of polypharmacy in the use of several medicines, they considered polypharmacy to be the responsibility of othersAl Shemeili et al.115Whilst an oncologist may be in a good position to judge prognosis and whether a medication is appropriate, he or she may defer to a patient’s GP regarding such medications which are usually initiated in primary care. Thus neither party may feel that they have ownership of the problemCashman et al.122

Health-care professional factors

Health-care professionals’ ability to engage in deprescribing is shaped by a number of individual and interpersonal factors including their skills and experience, their professional etiquette, and the amount of time and headspace they have available to engage in a complex decision-making process.

Realist analysis of health-care professional factors

Our realist analysis of health-care professional factors is summarised in Figure 15 below, with details in Table 17.

FIGURE 15.

Partial programme theory CMOCs 6–11: influence of health-care professional factors.

TABLE 17 - Context–mechanism–outcome configurations for influence of health-care professional-level factors

C, context; FP, family physician; M, mechanism; O, outcome.

CMOC	Description	Supporting data
6	When a health-care professional has previous experience deprescribing medication (C) they are more likely to feel able to deprescribe (O) because they know what to do and expect (M)	Best guesses were also required because ‘you don’t have guidelines for every situation — there are times when you just have to make a decision as best you can’ (GP6). GPs relied heavily on their prior knowledge and experience of the patient in this processSinnott et al.199Later on now in my career, I’ve taken on a different approach. I understand that managing polypharmacy is an art as much as it is a science. You have to balance quality of life, risks and benefits, when prescribing medications. I don’t feel the need to fix everythingHernandez148Negative experiences reinforced a tendency to opt for the status quo, whereas positive or neutral experiences fostered open-mindedness toward deprescribing . . . ‘As you get older, you realize that is not really true because you have done it so many times and they have not had a stroke the next week’ (GP4)Anderson et al.117
7	When health-care professionals feel that they do not have the necessary skills and knowledge to manage medicines in older adults (C) they are less likely to make changes to patients’ medicine regimes (O) because they are not confident in their ability to make good decisions (M)	‘I’m not clever enough to have all the statistics in my head to be able to say, well, that Statin is stopping all that absolute relative . . . It would be a great help [with deprescribing] to have further training and to meet with GPs in the same situation’Bolmsjö et al.121Often, the GPs needed to discuss the patient’s treatment because they did not feel they had the knowledge or skills to make correct therapeutic decisions. As one GP stated, ‘A specialized treatment belongs at the hospital, where the specialist can use their expertise’Laursen et al.160
8	When medicines have been prescribed by a specialist (C), other health-care professionals from other specialties may be reluctant to make changes to patients’ medicine regimens (O) because they do not feel that they have the knowledge to make a safe decision (M)	Often they will stop these days and just go back to one so I would question the dipyridamole but not necessarily stop it. Looks like the cardiologist has prescribed dipyridamole so I guess we would accept that (GP6)Ailabouni et al.113Physicians may be reluctant to review or alter decisions that were made by experts from other specialties, or to deviate from recommended therapeutic guidelines that were derived from younger populationsDjatche et al.132Prescription by a specialist or other practitioner is a factor identified in this survey that inhibits many FPs from deprescribing medications . . . this may be due to lack of confidence in their own knowledge and experience or being unclear about the indication for the medication chosen by the specialistHarriman et al.146
9	When medicines have been prescribed by another health-care professional (C), health-care professionals may be reluctant to make changes to patients’ medicines (O) because they are worried about damaging relationships with the original prescriber as well as between the original prescriber and the patient (M)	GPs discussed being intimidated by specialist physicians for deprescribing medications they initiated, with one recounting being ‘scorned by a colleague’. Furthermore, GPs expressed disappointment when deprescribed medications were restarted by a specialist physician or in hospital. These factors have also been identified in previous researchTurner et al.208External factors GPs were reluctant to discontinue medication prescribed by other medical specialists without contacting them. Contacting the specialist to change medication, however, took additional effort and GPs feared that it would be difficult to reach a consensus as the specialists often have a different viewpointRieckert et al.186There was also a reluctance to ‘interfere’ with other healthcare providers’ prescribing driven by fear of disturbing therapeutic relationships, hesitation to contradict prescribing by other healthcare providersMcNamara et al.166
10	When health-care professionals do not have dedicated time (C) they may be less likely to make changes to patients’ medications (O) because they do not have the emotional and cognitive capacity to consider complex issues (M)	It is important to consider a patient’s goals of care, prognosis, and functional level when considering which medications are potentially inappropriate. The process of deprescribing also requires special attention and timePruskowski and Handler183Participants emphasized, consistent with the importance of sense-making, that communication should go beyond checks on understanding of information to communicating the benefits of any changes and being responsive to patient concerns. Both patients and professionals agreed this would require dedicated, protected time to enable issues to be exploredKnowles et al.155
11	When health-care professionals do not have time (C) they may find it difficult to fully consider a patient’s care goals (O) because they are forced to prioritise what they spend their time on (M)	There is time taken away during a visit because of a variety of screenings that must take place during a patient assessment, such as asking questions about falls, depression, and abuse. This is impossible to achieve in our health care system, which demands high efficiency and throughputChen and Buonano125Finally, there is limited time in which these complex shared decision-making conversations can take place. Thus, if medications are not causing a noticeable problem, it is often easier to just continue themMcGrath et al.165Key barriers to engagement related to the practical constraints of workload placing limits on the necessary time and ‘head space’ needed to engage with this complex form of clinical practice. ‘limited by time, caseload and so lack of mental capacity’ (GP) ‘I barely get through the day reacting’ (GP)Reeve et al.6

Patient-level factors

Many studies75,79,114,143,162 included in the review suggested that, in principle, patients are open-minded about deprescribing and may be willing to discontinue one or more medications that are considered to be ‘inappropriate’ or unnecessary. However, patients’ willingness to engage with, and consider, deprescribing is shaped by the relationship patients have with their medicines and the value they place on them, as well as the involvement from their families and carers.

Realist analysis of patient-level factors

Our realist analysis of patient-level factors is summarised in Figure 16, with details in Table 18.

FIGURE 16.

Partial programme theory CMOCs 12–19: influence of patient-level factors.

TABLE 18 - Context–mechanism–outcome configurations for influence of patient-level factors

BP, blood pressure; C, context; M, mechanism; O, outcome.

CMOC	Description	Supporting data
12	When patients believe that medicines are a sign of good care (C) doctors may be reluctant to consider deprescribing (O) because explaining and justifying any deprescribing is more emotionally and cognitively taxing (M) than not doing this	Patients and providers alike possess psychological connections to medications. This may be because medications are the most visible form of health carePruskowski and Handler183Moreover, in the absence of ‘low-hanging fruit’ or a clear trigger to cease therapy, deprescribing, compared with initiating therapy, appears a riskier, less certain, and more cognitively and socially demanding process, with minimal decision supportAnderson et al.117GPs had specific difficulties talking to multimorbid patients about stopping medications; they feared this could be interpreted by the patient as a withdrawal of care and potentially damage the doctor–patient relationshipSinnott et al.200
13	When patients believe that their medicines are providing them with benefits (C) doctors may find it difficult to discuss deprescribing (O) because explaining and justifying any deprescribing is more emotionally taxing (M) than not doing this	Factors influencing GPs’ deprescribing were beliefs concerning patients; patients have no problem with polypharmacy; patients may interpret a proposal to stop preventive medication as a sign of having been given up on; and confronting the patient with a discussion of life expectancy vs. quality of life is ‘not done’Schuling et al.195Primary care physicians have also acknowledged worry about discussing life expectancy and that patients will feel their care is being reduced or ‘downgraded’McGrath et al.165GPs acknowledged that in the terminal phase it would be rational to discontinue anti-hypertensive medication. However, they often hesitated to take this step to avoid the impression they were giving up on the patient or unnecessarily deprive them of a sense of being in control with adequate BP measurementsvan Middelaar et al.212
14	When patients believe that a medicine might be working or will work in the future (C) they are likely to want to continue taking it (O) because they hope that they are doing something to help their condition (M)	Patients feel an improvement when they start taking the drug or hope for a future improvement and for that reason they do not want to discontinue it. Some patients think that they are doing something that can help their condition and in doing so they feel reassuredGonçalves144Patients and family members sometimes cling to the hope of future effectiveness of a treatment, especially in the case of medications like donepezil for dementiaMcGrath et al.165
15	When patients believe that their medicines are keeping them alive (C) health-care professionals may find it difficult to discuss deprescribing (O) because they do not want their patients to feel that they have abandoned them (M)	They also considered the reaction of the patients, who might have come to value their medicines or feel that deprescribing is a sign of abandonmentBokhof and Junius-Walker120It is important to anticipate barriers to deprescribing and to discuss these with patients and carers. Barriers may include psychological discomfort when ceasing a medication they have been taking for many years, or feeling their situation is hopeless since medications for chronic diseases are being ceasedHardy and Hilmer145Moreover, GPs are reluctant to initiate a discussion about stopping medication because they are concerned that patients may interpret this as a sign of being given up on. People may then get the feeling, ‘Don’t I count anymore, am I not important?’Schuling et al.195
16	When patients view medicines as prolonging their lives (C) they may be reluctant to stop taking them (O) because they view deprescribing as a sign that they are not worth keeping alive any more (M)	Patients may raise difficult questions that the doctor may wish to avoid, for example: ‘am I not worth treating anymore?’, ‘I was told I should take this for the rest of my life, does this mean I am going to die?’, ‘won’t I get ill without the tablets?’Cashman et al.122Still other physicians voice concerns that patients will feel the physician is ‘giving up on them’ or ‘leading them to quicker deaths’Harriman et al.146
17	When patients believe that medicines are providing them with benefits (C) they may be reluctant to discontinue them (O) because they are afraid of negative consequences (M)	For some participants, a complex drug regimen was the only means through which they could gain equilibrium, relief from distressing symptoms, or a sense of having a ‘normal’ life (though this varied in degree of success and setbacks)Townsend et al.207The overarching pattern of ‘preserving self’ was a surprising and clear finding. Taking medication was closely tied to self-identity and manifested in various ways, described in the ensuing sections. Taking multiple medications was significant and personalVandermause et al.213Mrs. D derives an important sense of empowerment from her supplement use that should be respected by her physiciansPitkälä et al.181
18	When families or carers perceive medicines to have a benefit for the patient (C) health-care professionals may be reluctant to consider deprescribing (O) because they feel pressured not to do so (M)	Several GPs talked about the challenge of keeping patients on potentially unnecessary medication at the urging of family membersJansen et al.152Ms L: The family is all guilt-ridden and they tell themselves that they have to keep dear old dad alive . . . and his caregivers face choices about using medications that may increase his longevity but negatively affect his quality of lifeSteinman and Hanlon203GPs discussed exercising caution with initiating medication changes, particularly where they assumed a resident’s family had strong expectations of medicines keeping their relative alive. ‘We really need to be in a situation where we’re educating relatives about what is realistic, it’s very hard to initiate the discussion with relatives’Palagyi et al.179
19	When families/carers are involved in a patient’s health-care (C) patients may be more able to engage in decision-making about their medicines (O) because they feel supported by them (M)	Families can facilitate exchange of information and encourage patient engagement in their healthcare. Studies have found that office visits in which the older adult patient was accompanied by a companion who prompted their involvement were 4.5 times more likely to be involved in decision making than their counterpartsHernandez148However, it was also evident that daily routine medication-work can and does go beyond the self, with network members being involved or called upon selectively to provide ad-hoc and/or regular support in the performance of a particular type(s) of workCheraghi-Sohi et al.126This patient also mentioned relatives as an important means of support. As they are more used to the internet, her daughters look up information for her. In addition, she discusses her disease and treatment with her daughters, who motivate her to go to the doctor and address difficult topicsSchöpf et al.194

Shared decision-making

Shared decision-making was widely discussed in the included documents as an important strategy in the management of problematic polypharmacy and deprescribing. Shared decision-making allows health-care professionals and patients to make collaborative decisions about treatment priorities. This model of care recognises the patient experience and embodied learning that equips them with the knowledge to make decisions regarding their treatments.

By drawing on shared expertise, patients and health-care professionals may be able to navigate some of the complexities and uncertainties associated with the deprescribing process by establishing treatment priorities and situating changes within the context of patients’ lives and understanding of their medicines (CMOCs 12–19).

As described in CMOCs 12–18, withdrawal of medications may be perceived by patients as also being a withdrawal of care, which in turn makes them reluctant to consider deprescribing and also makes it difficult for health-care professionals to discuss that option with them. By engaging in shared decision-making, health-care professionals are likely to become more aware of patients’ beliefs about their medicines and their goals of care, making them more likely to achieve patient-centred outcomes (CMOC21). However, it is worth noting that these beliefs may not always be in line with the recommendations made by healthcare professionals. 85

Involving patients in the decision-making process by making them aware of the risks and benefits, as well as drawing on their expertise (CMOC 20), also allows health-care professionals to share the responsibility for deprescribing and can help them make decisions that are defendable (CMOC 22), therefore helping to address some of the issues laid out in CMOCs 1 and 2.

Our realist analysis of shared decision-making factors is summarised in Figure 17, with details in Table 19.

FIGURE 17.

Partial programme theory CMOCs 20–22: shared decision-making.

Continuity of care and development of trust

Continuity of care, understood as ‘The extent to which a person experiences an ongoing relationship with a clinical team or member of a clinical team and the co-ordinated clinical care that progresses smoothly as the patient moves between different parts of the health service,’219 featured prominently across the studies included in our review. Three main types of continuity of care were identified in the literature. 220 The first is informational continuity, which refers to the use of information on past events and personal circumstances to make patient-centred decisions about care. The second is management continuity, which involves a consistent and coherent approach to the management of a patient’s changing needs. Finally, relational continuity, which refers to the ongoing relationship between a patient and one or more health providers. Continuity can help contribute to the building of trust by providing the opportunity to amass cumulative experiences of trustworthy behaviour and establish norms of co-operation and reciprocity. 221 This trust may also contribute to the effectiveness of medical care. 222

Within the context of medication management and deprescribing, siloed care and difficulty accessing up-to-date patient information can influence whether or not health-care professionals make changes to patients’ medications (CMOC 4). Lack of continuity, be it informational, management or relational, can damage the trust patients have in health-care professionals (CMOC 23). Relational continuity can help ensure that patients feel like they are being managed by a professional that knows them and their situation personally, and is therefore tailoring recommendations to them and knows what is in their best interest (CMOC 26). In turn, patients may then be more willing to consider recommendations for medication change made by that professional (CMOCs 24 and 25). Furthermore, management continuity can help reassure health-care professionals and patients that any potential harms of medication changes will be managed and may make them more likely to consider making these changes (CMOC 27).

As described in CMOCs 23–27, continuity of care may help build patient trust, which may make deprescribing less emotionally taxing for the health-care professional (CMOCs 4, 12 and 13). If patients trust health-care professionals, they may be more likely to listen when their beliefs about their medications are challenged and be more likely to take recommendations on board (CMOCs 14–18).

By engendering trust, continuity of care (as described in CMOCs 23–27) may also help facilitate shared-decision-making (CMOCs 20–22).

Our realist analysis of continuity of care is summarised in Figure 18, with details in Table 20.

FIGURE 18.

Partial programme theory CMOCs 23–27: continuity of care.

Monitoring

Deprescribing and medication management in general are complex interventions comprising a number of interactions between components, different groups, variable outcomes and a number of uncertainties (CMOCs 1–19).

Putting a monitoring process in place following a decision to make a change to a patient’s medication regimen may help to alleviate some of the concerns relating to the fear of negative consequences of withdrawing medications (CMOCs 15–17) and worries around deprescribing symbolising the withdrawal of care. Monitoring does this by reassuring both health-care professionals and patients that potential harms will be managed (CMOCs 28 and 29, and also related to CMOC 27), providing an opportunity for patient perspectives, which may change over time, to be taken into account (CMOC 31) and allowing patient feedback to further inform the deprescribing process (CMOC 30). Monitoring may also help to contribute to the continuity of care by providing opportunities for management and relational continuity (CMOC 27).

Our realist analysis of monitoring is summarised in Figure 19 below with details in Table 21 following.

FIGURE 19.

Partial programme theory CMOCs 28–31: monitoring.

Multidisciplinary teams

Working in multidisciplinary teams to make treatment decisions for patients was another commonly cited intervention strategy for the management of polypharmacy across the literature. 118,198,201

Health-care professionals may sometimes feel that they do not have the specialist skills or experience necessary to make complex decisions about patients’ medications (CMOCs 5–7), and although additional training may help to increase their knowledge and confidence, working in multidisciplinary teams may allow them to draw on the expertise and experience of colleagues (CMOCs 31 and 32). Being able to discuss complex cases with colleagues may help to reassure health-care professionals that the changes they plan to make to a patient’s medications are safe and may give them the support necessary to help them feel confident about these decisions (CMOCs 31 and 32). It may also help them feel that they are making defendable decisions in the absence of adequate guidelines (see CMOC 1).

Working in multidisciplinary teams may also contribute to continuity of care (CMOC 33) by encouraging informational continuity (countering effects of CMOC 23), thereby helping to increase patient trust (CMOC 25). Working collaboratively also allows health-care professionals to share the responsibility and workload associated with deprescribing (CMOC 34), thereby helping to mediate some of the challenges imposed by limited time (CMOCs 10 and 11).

Our realist analysis of multidisciplinary collaboration is summarised in Figure 20 below with further details in Table 22 following.

FIGURE 20.

Partial programme theory CMOCs 32–34: multidisciplinary collaboration.

Enabling infrastructure

Managing patients living with complex multimorbidity, including managing (de)prescribing, is an inherently uncertain process. Our review consistently identified fear of harm as a barrier to change, including deprescribing. For clinicians and patients already juggling the challenges of managing complex (and uncertain) multimorbidity, overcoming the ‘inertia’ associated with fear of change may be one task too many. If we are to ask clinicians to take on the work of managing the uncertainty of tailored (de)prescribing, we need to provide them with supportive guidance to do so. Current guidance describes the principle of person-centred care as ‘good practice’, with policy documents recognising tailored care as good practice. 1,7,8,43 However, our analysis suggests that these documents, to date, do not provide adequate permission to counteract the fear described by clinicians.

Existing guidance describes the steps within the consultation that support good practice around deprescribing. 63 But CMOCs 1–5 highlight that organisational context also affects professionals’ confidence and ability to deprescribe. Supportive guidance, therefore, needs to offer more than a consultation model/set of consultation steps, and describe at an organisational/system level a framework offering ‘safe boundaries’ for uncertain practice. This may include clarity on whose responsibility is tailored (de)prescribing (CMOC 5), recommendations on time and resource allocation to this complex task (supporting the work by prioritising time for it – CMOCs 10 and 11), and guidance on how feedback is offered to support and reinforce best practice.

Supportive guidance may also be important for legitimising the deprescribing role. Health-care professionals taking on the responsibility of deprescribing may be described as having ‘boundary spanning’ roles. 223 Boundary spanners are agents that relate practices in one field to practices in another by negotiating the meaning and terms of the relationship between them. 224 They enable the translation, co-ordination and alignment of different perspectives and practices. 223 Health-care professionals taking on the responsibility of tailoring medications need to negotiate and reconcile different sources of information (see Conservation of resource theory and loss aversion bias) and perspectives (CMOCs 12–17), as well as develop and co-ordinate a plan to manage these changes (CMOCs 28–31).

For boundary spanning roles to be realised and to take hold, they need to be seen as legitimate within the broader system around them. 223 By clarifying whose responsibility medication tailoring and (de)prescribing is and by allocating sufficient time and resource to this complex task, supportive guidance may act to formalise the deprescribing role, making it legitimate and accepted within the wider system, and therefore granting health-care professionals ‘permission’ to undertake deprescribing.

A key source of support for health-care may also come from taking a multidisciplinary approach to managing patients with complex medications. Our analysis shows that being able to work with colleagues to manage deprescribing may allow health-care professionals to draw on the expertise of their colleagues (CMOCs 31 and 32) and help share the workload (CMOC34) and responsibility.

Access to high-quality, relevant data

Our review highlighted that access to data matters (CMOC 4). This was perhaps a surprising finding given that health systems routinely collect large numbers of data. However, our review highlighted that it is contextual data that professionals need – an understanding of history and context (CMOCs 15–17) to support the complex, and sometimes uncertain, tailored decision-making (CMOC 4). In the general practice setting, these data have perhaps traditionally been held in the head of the practitioner (often a GP) who knows the patient. In a world of extended clinical teams, the data highlight the importance of ensuring access for all to these data. These data may also help support shared decision-making (CMOCs 20–2), as well as provide informational continuity (CMOC 25), which may be key in supporting patient trust (CMOCs 23 and 25).

Access to data also applies when dealing with (and potentially reversing) decisions that have been made by other professionals (CMOC 8). Our findings highlight that when communicating between teams/sectors, we need to convey more than what medicines have been started, but also the contextual information on why – what impact is anticipated, how is this to be judged and what conversation has been had with the patient. The TAILOR findings explain why we need to look differently at what and how data are generated for, and used in, practice to support tailored (de)prescribing decisions.

Shared understanding of meaning and purpose of medicines

Prescribing decisions need to stem from tailored understanding and explanation of a patient’s illness in context – through an understanding of the patient’s values and beliefs about the role of their medicines within their wider health care and daily living (CMOCs 12–18). Shared decision-making is a process that can achieve such tailoring (CMOCs 20–2). Tailored explanations convey to the patient that this professional understands them (CMOC 24) and is acting in their best interests (CMOC 26).

Tailored explanations require the professional to have a good understanding of the individual patient context (see Access to high-quality, relevant data), and the value and importance of medicines to the patient and carer (CMOCs 12–16 and 17–19).

Tailored explanations require the elimination of conflicting information/advice from different members of an extended clinical team to avoid the effect of conflict undermining the patient’s trust in their health-care professionals and the advice being offered (CMOC 23). This requires attention to sharing data (see Access to high-quality, relevant data) and to recognising the team-wide nature of care for patients with complex comorbidities (see Trial and learn).

Trial and learn

The data describe the professional inertia associated with the uncertainty and complexity of deprescribing (CMOC 7) and fear of negative consequences (CMOCs 15–17). A trial and learn process incorporating small incremental changes to medicines (CMOC 28) and a harms minimisation process (CMOC 29) with follow-up and continuity (CMOCs 27, 30 and 31) that enables patient perspectives to be heard (CMOCs 30 and 31) may enhance patient trust (CMOC 31) and increase chances of patients considering medication changes and patient-centred outcomes being achieved.

Trust

The need for trust is implicit in all that health services and health professionals do, and particularly so in situations of vulnerability such as managing complexity, uncertainty and new relationships. 225–227 But what that involves, beyond good communication skills and a good relationship between patient and professional, is rarely articulated. Our analysis highlights key elements including:

• professional trust (confidence) in the professional’s own decisions, for example stemming from guidance supportive of this form of practice, and shared responsibility for decisions (with patients and other members of the team) (CMOCs 1–5, 7 and 8)

• patient trust in the professional through a sense that this decision is made in their best interests, generated through production of tailored explanations (CMOCs 12–18 and 23–7)

• the importance of building consistency of care across a team, for example to minimise conflicting advice and also because team decision-making supports professionals in taking complex decisions in situations of uncertainty (CMOCs 32–4)

• the importance of planned follow-up to review (CMOC 27), and if necessary amend, decisions made based on feedback (CMOC 30) – an approach that supports harms minimisation (CMOC 29) as well as trust.

In summary, we can thus describe a more finalised and revised programme theory describing the contextual and consultation factors needed to support tailored (de)prescribing. Our final analysis recognises that tackling problematic polypharmacy needs a tailored approach to (de)prescribing. Tailored prescribing commonly involves beyond-protocol decision-making, a complex process that involves emotional and cognitive (knowledge) work for clinician and patient alike. The impact of this cognitive and emotional load can contribute to inertia: a failure to implement tailored decisions even if recognised. Our final programme theory therefore describes and explains the components needed to enable tailored prescribing through addressing this cognitive and emotional load (Figures 21 and 22).

FIGURE 21.

TAILOR programme theory: core elements needed to support effective tailored prescribing.

FIGURE 22.

Detailed final programme theory: delivering TAILORED deprescribing. Numbers in brackets refer to CMOCs.

Conservation of resource theory and loss aversion bias

Conservation of resource theory is a theory of motivation that posits that individuals are motivated to protect their current resources and acquire new ones. 228 A resource can be loosely defined as anything that an individual believes can help them attain their goals and can include objects, states, conditions, relationships, social support and other things that people value. 229 The value of resources varies between individuals and is tied to the experiences and the realities in which they live. 229 This basic principle of conservation and acquisition lead to other principles within the theory, such as the primacy of resource loss and the idea that it is more psychologically damaging for individuals to lose resources than it is beneficial for them to gain the resources that they have lost. 229

Deprescribing involves a change from the status quo and the removal or loss of medications (resources) that patients see as having value in their lives (CMOCs 12–18). Conservation of resource theory, and especially the primacy of resource loss, may help to explain the prominent phenomena reported in the literature of inertia experienced by health-care professionals and patients when needing to consider deprescribing.

Discussions with our PPI contributors (see Chapter 9) led to the suggestion that it may be more helpful to change the framing of deprescribing (stopping something) to ‘re-prescribing’ (reviewing medication that may lead to multiple outcomes, including starting, stopping or maintaining the status quo). Given the primacy of resource loss highlighted above, reframing deprescribing, which by definition entails the removal/loss of something, to something more neutral, ‘re-prescribing’, which does not necessarily entail a loss to resources, may help overcome this inertia. This may be a potentially fruitful area for future research.

Social support theory

Social support refers to the broad process through which social relationships promote health and well-being. 230 Information that leads individuals to believe that they are cared for, held in esteem and belong to a network of communication is a form of social support that can help individuals cope and adapt in times of stress. 231

Deprescribing often involves dealing with high levels of complexity and uncertainty, which can be psychologically and emotionally taxing for health-care professionals (CMOCs 12 and 13) as well as patients (CMOCs 16 and 17). Across our review, a number of mechanisms identified relate to health-care professionals and patients needing to feel supported and understood for desired outcomes to be achieved (CMOCs 1, 2, 19, 21, 31, 33 and 34).

Creating a supportive infrastructure as described in Enabling infrastructure may provide health-care professionals with the support and resources they need to be able to cope and manage when dealing with the uncertainties and complexities of deprescribing.

Providing data (see Access to high-quality, relevant data), tailored explanations (see Shared understanding of meaning and purpose of medicines) and building trust (see Trust) may enable patients to feel supported by reassuring them that health-care professionals understand and value their goals and needs.

Therapeutic relationships

Positive relationships between health-care professionals and patients have been shown to improve patient satisfaction and professional fulfilment, increase compliance with prescribed medication, help save time and reduce the number of complaints from patients. 232

Our analysis has shown that relationships between patients and health-care professionals may also be central to supporting tailored deprescribing (CMOCs 25 and 26). The social support (see Social support theory) provided by these relationships may be key to working through the uncertainties deprescribing entails,117 perhaps mediated through a process of generating and maintaining trust. The importance of this relationship to both patients and health-care professionals may, however, sometimes also act as a barrier to deprescribing. When considering making changes to prescriptions, health-care professionals may be hesitant to make these changes because of the fear of damaging the therapeutic relationship (and trust) between the patient and the original prescriber (CMOC 9). Furthermore, deprescribing a medication that a health-care professional knows a patient values, even though it may be inappropriate, may be especially difficult in the light of the possibility of damaging their relationship with their patient. 160

Given the centrality of therapeutic relationships and the potential dual role they play as both facilitators of and barriers to deprescribing, shared decision-making (in which both health-care professionals and patients actively participate, share information and reach consensus) may be key to negotiating some of these possible tensions to reach desired outcomes (CMOCs 20–2). Shared decision-making has long been recognised as a principle of best practice in the NHS. 233 Our findings explain why it matters (how it contributes to meaningful outcomes), and also highlights the need for supportive infrastructure (the potential need for organisational change) for this to be effective.

Trust has been described as one of the most important components of well-functioning relationships,234 and may also contribute to the effectiveness of medical care. 60 Continuity of care has been shown to be associated with higher levels of trust between patients and healthcare professionals. Our analysis has emphasised the importance of all forms of continuity (relational, informational and management) for the building of trust necessary for successful engagement in deprescribing (CMOCs 23–6).

Inviting a patient co-applicant on the study team

Ed Ranson was a participant in a previous study47 we ran to understand personal experiences of living with long-term conditions and the implication for health care. Ed has personal experience of long-term health issues and polypharmacy, and of seeking to improve care through research as a lay member of a team.

Ed Ranson played three key roles:

1. supporting the core team in project planning, including PPI engagement activities – Ed has commented on and amended project plans at all stages, from the original application through to reporting and dissemination plans

2. as a member of the extended project team, taking part in 6-monthly review meetings with the full team to consider data collection and analysis

3. as lead for PPI activities, working in partnership with Joanne Reeve in leading the design of planned PPI meetings and the review and revision of PPI activities and engagement with comments from our second PPI support (Michelle Dickenson; see Acknowledgements).

Planned patient and public involvement stakeholder events

Ed Ranson and Joanne Reeve planned for, and invited lay members to, four attempted PPI meetings in March and August 2019 and February and October 2020. Flyers were prepared and distributed through stakeholder networks (including practice patient participation groups, local GP networks, ‘generic’ PPI networks within host universities and personal contacts) as well as via social media. For example, Joanna Reeve attended a lay health-care network meeting at Castle Hill Hospital in Hull in August 2019 to introduce the study and invite people to join the next planned PPI meeting.

Ed Ranson supported the writing, and revision, of all invitations. For example, he advised that including the word ‘deprescribing’ in the event title was not likely to encourage engagement and that we needed to focus on patient experience of using medicines. Later versions instead described the project as being about ‘making medicines work better for you’. Ed was directly engaged in contacting his own patient participation group network, and provided advice to Joanne Reeve when invitations were unsuccessful.

In the light of the difficulties we experienced, we got in touch with PPI contacts from previous studies. Michelle Dickenson offered to support the project through reviewing and providing feedback on written documents.

Presentation of collated findings and draft report to patient and public involvement members

We were unsuccessful in recruiting members of the public to join our final (virtual) PPI meeting to discuss draft findings (October 2020). We therefore modified our plans to two actions. We held a virtual meeting with Ed Ranson, Joanne Reeve and Amadea Turk to present the realist review findings to Ed Ranson. A copy of the presentation and notes from the meeting were subsequently sent to Michelle Dickenson for further review and comment. Ed Ranson and Joanne Reeve met to discuss Michelle Dickenson’s reflections.

Co-authorship of this section of the report

Reflections on the opportunities and challenges of PPI engagement within TAILOR have been discussed at core team and extended team meetings throughout the project. These were collated in the preparation of this report. Joanne Reeve drafted this section, with Ed Ranson editing and offering comment.

Challenges of recruiting to face-to-face patient and public involvement meetings – recognising the importance of meaning

The challenge in recruiting new PPI members to the project has led us to reflect on the importance of meaning – reflections that have also informed the realist review work.

We arranged four PPI meetings over the duration of the project, refining the invitation and design of the meeting each time to seek to improve engagement. In August 2019, Joanne Reeve met with a Hull-based PPI group to discuss the work. People at that meeting told us that tailoring of medicines/care was an area of interest. We were approached by one potential participant after that meeting, but he indicated that his particular interest was in the use of genomics to tailor care. When advised that this was not the focus of our meeting, he withdrew.

Ed Ranson supported us to reflect on our recruitment challenges as viewed from the patient-centred focus of our study. For example, he proposed that we change the wording of our invitation flyers. In 2019, we invited people to come to talk with us about potential ‘burden’ of using medicines, to help us understand how to ‘safely’ tailor medicines. Burden and safety were both concepts identified from previous research and conversations. When people did not take up our invitations, Ed suggested that our invitations were not yet tapping in to the issues that were important to potential participants. This led us to change the invitations in 2020 to invite people to discuss ‘making medicines work for you’.

Our realist review identified the need to recognise the meaning and value of medicines to individuals when thinking about deprescribing (see Chapter 7). While reflecting on our recruitment challenges, we have recognised how our invitations to participants may have misrepresented their understanding of the meaning of medicines. Our invitations placed a (evidence-informed) focus on burden. Our PPI partners encouraged us to reflect on ‘a prescription . . . [as] the currency of the doctor–patient transaction’, in which any proposed changes are ‘likely to be seen as being motivated by the desire to save money’ (Michelle Dickenson, 2020, personal communication). With hindsight, we needed a different approach to establish shared meaning in inviting PPI engagement with our research.

In the light of our experiences, Ed Ranson has proposed the need to change from talking about deprescribing (stopping something) to ‘re-prescribing’ (reviewing medication, which may lead to multiple outcomes including starting, stopping or maintaining the status quo) (see details in Conservation of resource theory and loss aversion bias).

Shifting outcomes – the importance of trust

The importance of establishing and maintaining trust when considering changes in prescribing was a key finding from our data analysis (see Figure 22). The issue of trust also underpins our reflections on the importance of establishing shared meaning when engaging PPI partners. Trust was therefore recognised as key to successful engagement in discussions about prescribing, whether in the clinical context or in research conversations about how to improve practice.

Trust is recognised in our analysis as a key element in the successful delivery of tailored prescribing decisions. Reflections from our PPI discussions also considered whether establishing or strengthening trust can be seen as an outcome of good practice around prescribing, as well as research into prescribing practice.

As highlighted in our examination of an extensive literature in our scoping review, previous research on (de)prescribing interventions has focused on the impact on medication and biomedical outcomes (e.g. changes in the number or types of medication used, or effect on measures of illness or risk). Both our analysis and our experience of seeking to engage with PPI partners in the process of doing the research, have highlighted that other factors matter. We need to recognise what medicines mean to individuals in the context of their daily life. A patient must be able to trust that a clinician, or research team, understands that.

Academic Advisory Group

The realities of pressures on, and availability of, all parties led to some modifications of these groups. Competing priorities for members of the Project Working Group not involved in detailed day-to-day generation and analysis of data meant that their input to the project was refocused as academic advisory roles. Some original members of our AAG changed roles (T Fahey), one relocated to another country (T Walley) and time pressures limited engagement of others. Our academic teams therefore consisted of the core team (JR, MM, RH, AT, GW and KM) that was responsible for day-to-day delivery of the work, and a revised AAG (DL, DM, JK and RB – all listed as co-authors – along with Dr Nia Roberts, an information specialist providing input into the search strategies). The core team met monthly to ensure cross review between the two reviews. The AAG originally planned to meet quarterly. As work progressed, we recognised a need for flexibility to optimise the use of AAG members’ time, as well as to ensure timely input into the emerging work. The AAG therefore met on five occasions at key stages in the project:

• set-up – reviewing the draft programme theory and focus for the scoping review (May and September 2018)

• data collection – reviewing and revising the data collection process (June 2019)

• interpretation – reviewing and revising the data analysis (January 2020)

• integration – reviewing the integration of findings (March 2020).

The AAG were also actively involved in the preparation of the initial draft final report (submitted December 2020) as co-authors of the work presented. This work was undertaken remotely because of the COVID-19 pandemic.

Stakeholder Group

Competing pressures and changes in roles also led to changes in the makeup and working of our Stakeholder Group. We originally intended that this group would meet twice per year. The focus for discussion would be on the relevance of the research for end users, and on supporting dissemination activities. Mirroring our experiences with PPI (see Chapter 9), we experienced some difficulties in recruiting people to attend stakeholder meetings. Feedback from our first event clarified that busy professionals needed to prioritise time for work with more ‘tangible outputs’ than research at the set-up stages.

However, we were able to hold two stakeholder meetings during the project. A mixed audience of clinicians, NHS managers and clinical academics joined us to review and feed back on our progress and emerging observations. Nine external partners, along with co-applicants and researchers from the TAILOR team, met face to face in Birmingham in March 2019. Stakeholders reviewed and commented on the proposed direction of the research, and guided our development of search strategies. In September 2020, eight external partners (including four new members) met with the team using a virtual platform. At this meeting we reviewed the draft results of both analyses and invited stakeholder discussion on the interpretation (i.e. meaning and value) of the findings. Our stakeholder meetings contributed to ensuring that our work remained grounded in the needs of the end-users of our research, namely clinicians and managers. We also held a third stakeholder meeting in December 2021, following the submission of our draft final report. We shared the key findings from TAILOR, discussing their contribution to the challenges posed in the National Overprescribing Review. 237 Stakeholders fed into our described dissemination work, as discussed further in Dissemination activities: continuing our work to optimise the impact of TAILOR.

Scoping review

Our scoping review identified a broad and complicated field of research. We therefore opted to conduct an in-depth look at a clearly delineated part of that picture, focusing on research that explicitly examined deprescribing effects in populations living with defined multimorbidity. Our focused inclusion/exclusion criteria led us to identify just 20 papers from a large and diverse field. However, even within this dataset, there was considerable variation in what was done, to whom and with what effect.

Our analysis revealed that deprescribing under ‘research conditions’ mapped well to expert guidance on the steps needed for good clinical practice (see Table 7). 63 When reported, interventions were generally safe (see Table 11) with an observed ‘pocket’ of negative outcomes on safety in four studies of deprescribing conducted in secondary care without the use of a defined tool or framework. Interventions were commonly reported as acceptable to clinicians, although with fewer data available on acceptability to patients. Effect outcomes were variable across our data set, a perhaps unsurprising finding given the variability in interventions used and context of practice, along with methodological issues of studies with mainly small to moderate sample sizes and short follow-up periods. However, there was evidence of a positive impact on prescribing behaviour, although there was less clear evidence of clinical effect. Similar observations of hard-to-interpret variability in the outcomes dataset was described for a systematic (Cochrane) review of the range of interventions being used to enhance appropriate polypharmacy. 105 However in our review, although a wide range of indicators was used, of those reported in Table 11, positive impacts were seen for 444 reported outcomes (upward arrows), 328 reported negative impact (downward arrows), 84 reported no change (side-to-side arrow), and for 208 the effect could not be interpreted (dots).

We can therefore conclude that our map of the evidence offers clinicians evidence-informed support for the safety, clinician acceptability and potential effectiveness of deprescribing approaches that demonstrate structured approaches to deprescribing decisions (conclusion 1).

Our scoping review was not designed to offer clinicians specific details on the absolute benefits/risks associated with specific deprescribing decisions (e.g. stopping drug X in condition Z produces outcomes A, B or C) but rather to consider whether such work might be possible or desirable to do. In conducting this scoping review, we have taken a detailed look at an extensive body of literature. Our observations lead us to believe that attempts to undertake the detailed subanalyses needed to generate a deprescribing dataset would experience similar challenges of hard-to-interpret variability recognised in the Cochrane review discussed above. 105 We are not confident that the data available for such a piece of work (namely the conducted/published research studies to date) would support generation of meaningful synthesis or meta-analysis because of the considerable heterogeneity of clinical and research methods used.

It is also unclear whether or not clinicians would find such a resource useful in their daily practice. Datasets describing absolute benefits associated with the use of named medicines in specified patient groups already exist [e.g. the Scottish polypharmacy guidelines previously discussed4,5 and the Database of Treatment Effects linked to the NICE guideline on multimorbidity assessment and treatment (NG56)]. 238 However, we have been unable to identify any evaluation studies describing if and how such datasets are being used by clinicians and patients in practice. Our realist review offers insight into why this might be the case. Our realist review findings describe the complexity of the tailored decision-making process, with outcome data playing a limited role in the broader knowledge work of practice.

The findings of our scoping review, therefore, resonate with our realist review in highlighting the need to recognise deprescribing as a complex intervention (see Figure 10). Our analysis leads us to conclude that the currently available data (published studies) do not readily support the production of a ‘deprescribing outcomes data set’ owing to significant heterogeneity in both the conduct and reporting of the studies to date. We suggest that academics and research funders would need to consider the development and use of both a core outcome set and clearly defined reporting guidelines to achieve this outcome. Our critical observations in undertaking both reviews lead us to question the utility to clinicians, and, therefore, patients, of such work. Discussions at our final stakeholder meeting in December 2021 supported this stance.

In the final dissemination stage of TAILOR, we will therefore use the presentation of our findings to ask clinicians and patients whether or not a dataset on the absolute benefit and harm associated with specific deprescribing decisions would change their decision-making practice (conclusion 2).

Our analysis highlighted two key challenges for the research community to consider in generating evidence to support patient-centred clinical practice. First, we recognised the need for research that recognises, and examines, deprescribing in context (see Chapter 5, Implications for future work). Our review highlights why deprescribing cannot be researched as a linear, single intervention but requires the use of methodologies to evaluate complex interventions. Second, our review highlighted the challenges in synthesising data (whether as a clinician or researcher) from such a fragmented research base. In the absence of a coherent and co-ordinated map describing what research questions are needed in the context of current practice, using clear agreed definitions and measures, we will continue to generate a dataset that is hard to interpret meaningfully. Again, the Cochrane review of measures to support appropriate polypharmacy highlights this point. 105 We propose the need to consider if and how we might address this issue through new (potentially international) thinking on decision-making on how research is generated and prioritised. Our PPI partners also recognised the importance of this issue in highlighting the need for patient partners involved in research to also have the ‘contextual understanding’ of the research necessary to support development of a coherent body of understanding (see Chapter 9, Challenges of recruiting to face-to-face patient and public involvement meetings – recognising the importance of meaning).

A common theme across this discussion is the importance of recognising the need for research in context. To optimise the impact of research on complex health care, knowledge generation cannot be understood solely as a ring-fenced task to be done and then translated into practice. Evans and Scarbrough239 challenged the research community to consider the benefits as well as risks of generating knowledge in context, blurring the roles between clinicians and researchers. As Green240 described, if we want evidence-based practice, we need to generate practice-based evidence.

Our review recognises the importance of generating practice-based evidence for complex health care, and raises questions for the research community about how we best achieve that (conclusion 3).

Realist review

Our realist review critically examined a complex body of research to understand the mechanisms behind the outcomes reported in the scoping review, and so generate a new theory describing tailored deprescribing practice. We sought to describe how tailored prescribing happens, and to explain the variability in practice by understanding for whom and in what context practice occurs.

Our realist review recognised the significant cognitive (intellectual) and emotional load involved in the knowledge work of producing tailored explanations and decisions about medication use, working ‘beyond guidelines’, managing uncertainty, and maintaining continuity of approach and trust across a team and across an extended timeline. Our analysis identified that the complexity of this work can contribute to inertia, with both patient and practitioner maintaining a prescribing status quo even when it was recognised as not ideal. Our programme theory described a number of components needed to manage/reduce this load and so support tailored deprescribing.

Specifically, our analysis described four key concepts necessary to support strong tailored deprescribing practice:

1. an enabling infrastructure that provides clear guidance on, and support for, professional responsibilities; enables multidisciplinary working; and supports continuity of consistent care

2. consistent access to high-quality, relevant (notably contextual) data

3. support for development and maintenance of tailored explanations – a shared understanding of the meaning, purpose and impact of medicine

4. attention to generating and maintaining trust through monitoring and continuity that supports a mutual ‘trial and learn’ approach.

Our concepts recognise that deprescribing is a complex intervention: an interpretive practice28 that occurs in the interaction between patient and practitioner to generate a tailored understanding of priorities (including the meaning and value of medicines) and possibilities (exploring what is known – data – on the potential impact of use or discontinuation of medicines in the context of an individual’s conditions and circumstances – contextual data). It is the generation of a tailored explanation of medicines use in context that is necessary for effective care, and needed also to support and maintain the trust that is necessary to sustain management of complex health-care needs and so optimise outcomes. Trust is also enhanced and maintained by planned follow-up to review, evaluate and, if necessary, amend decisions.

Tailored deprescribing therefore relies on both the interaction between clinician and patient and the context in which the interaction occurs. Context includes the recognition that health care is delivered by multiple professionals in multiple settings. Continuity of approach across teams is vital to avoid conflicting explanations, and so the undermining of trust. In addition, we recognise that both clinician and patient need external resources that offer support for the complex task they undertake, including how to recognise if and when things are going well, or could be ‘better or worse’ (an infrastructure of support).

The TAILOR realist synthesis therefore provides a framework for ‘better’ tailored deprescribing as an interpretive practice based on the four key elements of the need for an enabling infrastructure for person-centred health care; consistent access to high-quality, person-centred (including contextual) data; supporting the generation of tailored explanation and so shared understanding of medicines use; and continuing review to enable mutual learning and so the development and maintenance of trust (conclusion 4).

Our framework resonates with existing descriptions of best practice (e.g. Reeve et al. 63) in recognising the need to explicitly consider patient priorities for care, and interpret potential harm and benefit of decisions in individual contexts. However, our work extends these existing frameworks in two ways, first, by highlighting the need to also recognise the importance of the perceived value and meaning of medicines to patients (see Chapter 7, Perceived value of medicines) and explicitly include these in decisions about their health-care needs and medication use. The impact of stopping medicines, from a patient perspective, may not relate to the effect on a given clinical condition.

Second, we recognise development and maintenance of trust as key and necessary components of ‘better’ practice. Recognising the importance of trust as both an element and outcome in prescribing practice potentially requires a redesign of health systems and health-care practices in areas such as data management and workflow planning. Data systems need to ensure consistent access to the contextual data (both biomedical and biographical) needed to support understanding, and so trust, between clinician and patient. Care models need to be designed to provide adequate time for discussions and robust follow-up arrangements.

Our analysis therefore highlights the complexity of the knowledge work involved in tailored prescribing, and suggests that tailored deprescribing requires models of health care in which all parties have consistent access to the resources needed to generate, implement and review tailored, shared decisions about medication use. Our work provides data highlighting the importance of professional roles (including permission to deliver tailored, ‘beyond guideline’ care, and clarity of roles across diverse and changing teams/communities of practice); enhanced resources for practice (including extended professional skills, consistent access to high-quality data, and prioritisation of protected time for review and shared learning); and the review of expected outcomes for the delivery of quality person-centred care.

Our TAILOR deprescribing framework extends existing models of good practice by recognising the need to consider the potential impact of prescribing decisions beyond biomedical or pharmacological effects by demonstrating the need to include organisational/contextual factors in models of better practice (conclusion 5).

Our findings have implications for the development of clinical guidance and research activities.

Our findings describe why an infrastructure of support is necessary for person-centred prescribing that offers ‘permission’ as well as the resources necessary for practice. We now seek to work with clinicians and patients to translate the principle described in our analysis into a clear description of what is needed on the ground. We have incorporated the principles outlined in Figure 22 into the logic model for a new research Programme Grant looking at deprescribing of sleep medication for people living with dementia. 241 During the dissemination stages of the TAILOR project, we will share our robustly generated, evidence-informed framework (guidance) on better practice with stakeholders and ask (how) could this model support the changes in practice needed if we are to consistently deliver person-centred prescribing practice? We will invite stakeholders, and especially patients (see Chapter 9), to help us to consider how best to share our findings to stimulate recognition of, and action on, the changes needed.

In the dissemination stages of TAILOR, we will continue our work with stakeholders to help us optimise the impact of our work through translating our findings into resources for front-line practice (conclusion 6).

Current guidelines and guidance largely focus on the management of specific (single) conditions (e.g. epilepsy or hypertension) or interventions (e.g. interventional radiology), and less commonly on complex conditions or interventions (e.g. multimorbidity or medicines optimisation). TAILOR has focused instead on an outcome of care (tailored, person-centred outcomes) albeit in the context of a given intervention (deprescribing). Although delivering person-centred care is an NHS England policy priority,242 survey data highlight that success in delivery of person-centred care has been declining in some areas of care. 243 Our findings raise a question about an opportunity for future outcomes-focused guidance on ‘better’ delivery of person-centred prescribing as part of developing an ‘infrastructure of support’.

Our findings therefore raise the question as to whether or not there is an opportunity for new NICE guidance on person-centred care using tailored prescribing as one example (conclusion 7).

Our scoping review described a diverse set of outcomes being used to assess the impact of deprescribing initiatives. Our realist review findings help to explain the variability in practice in recognising the key importance of offering tailored explanations. ‘Better’ outcomes of practice may, therefore, vary considerably between different individuals when priorities, meaning and context are actively included in considerations on what care is needed and why. We discussed the potential value of work to generate outcome sets for deprescribing research in Chapter 5, Generating a reference set for clinicians. Findings from our realist review offer additional insights in to what these outcome sets could consider and include. Specifically, we query the need for outcome sets to include measures of delivery of tailored (meaningful) explanation and trust. Current measures for both exist,244,245 although our data highlight that they may need to be adapted for the tailored prescribing context.

Our realist review also adds further insights to our discussion on the need to support the generation of practice-based evidence (see Scoping review). The data/analysis demonstrates that deprescribing is a complex intervention. Research methodologies designed for the study of complex interventions have been published;99 these include the robust development of the theoretical frameworks needed to support complex interventions, as well as the co-delivered implementation methods needed to evaluate impact.

Our findings highlight the importance of recognising person-centred health care, including deprescribing, as a complex intervention needing robust practice-based evidence to support delivery of quality care. The methodological implications of these observations should inform future research funding and prioritisation setting (conclusion 8).

Working with clinicians and policy-makers

We have been sharing the preliminary findings of our work at clinical educational meetings including the Avoiding Harm event at the Royal College of Physicians (November 2019),21 and within the Humber Coast and Vale CATALYST programme for new-to-practice GPs (URL: www.hyms.ac.uk/catalyst; accessed 11 March 2022).

In our original bid, we described a plan to hold an 1-day professional learning event, in collaboration with the Royal College of General Practitioners. In the light of the changes resulting from the ongoing pandemic, we have negotiated a change in that plan. We are now in the process of developing a massive online open course (MOOC) to make our findings widely accessible as a learning resource to clinicians. We will use the feedback from the MOOC and our other professional development activities to inform ongoing discussions with policy-makers about future practice in this area.

The MOOC will be launched in summer 2022, following discussion with our stakeholder group.

Working with patients

We are producing a short video for patients highlighting the key findings from our work. We had originally planned for this to be shown publicly (e.g. in GP practice waiting rooms) but are revising our ideas in the light of the current pandemic. We now aim to produce a short animation for social media sharing the findings of our work and highlighting the opportunity of engaging with research. Ed Ranson as our PPI lead will continue to offer support for these discussions.

Working with researchers and research funders

The TAILOR protocol describes an anticipated output of a submitted application for funding to support implementation research using the developed TAILOR model. This was discussed at our stakeholder event in December 2021, which highlighted a need and an opportunity to integrate recent policy and practice changes into the emerging plans.

However, the work from this review has already supported an additional successful application for funding. Tailored prescribing is a rapidly developing field of interest for the clinical and academic community. During this project, including through our stakeholder work, we have developed new collaborations with partners focusing on deprescribing in specific contexts. Joanne Reeve and Geoff Wong are co-applicants on a successful NIHR Programme Grants for Applied Research application looking at tailored sleep management for people living with dementia, which will include work to both avoid initiating medicines and deprescribe hypnotics (starting February 2022). 241 The TAILOR programme theory (see Figure 22) has informed the logic model for this work, with the practice model to be developed and refined using co-design methods. Other bids are also in preparation.

We are also preparing a briefing paper highlighting the implications/recommendations for future research calls and activities identified from our work. We will share our briefing paper with other researchers through our networks including the Society for Academic Primary Care network, the British Pharmacological Society and the Royal Pharmaceutical Society, and so invite commentary and contribution from other projects that have raised similar issues. In this way we seek to generate a discipline-wide report on implications for future research that we will take to key research funding bodies and policy-makers.