Contexts and mechanisms relevant to General Practitioner (GP) based interventions to reduce adverse drug events (ADE) in community dwelling older adults: a rapid realist review

Background Older adults in Ireland are at increased risk of adverse drug events (ADE) due, in part, to increasing rates of polypharmacy. Interventions to reduce ADE in community dwelling older adults (CDOA) have had limited success, therefore, new approaches are required. A realist review uses a different lens to examine why and how interventions were supposed to work rather than if, they worked. A rapid realist review (RRR) is a more focused and accelerated version. The aim of this RRR is to identify and examine the contexts and mechanisms that play a role in the outcomes relevant to reducing ADE in CDOA in the GP setting that could inform the development of interventions in Ireland. Methods Six candidate theories (CT) were developed, based on knowledge of the field and recent literature, in relation to how interventions are expected to work. These formed the search strategy. Eighty full texts from 633 abstracts were reviewed, of which 27 were included. Snowballing added a further five articles, relevant policy documents increased the total number to 45. Data were extracted relevant to the theories under iteratively developed sub-themes using NVivo software. Results Of the six theories, three theories, relating to GP engagement in interventions, relevance of health policy documents for older adults, and shared decision-making, provided data to guide future interventions to reduce ADEs for CDOA in an Irish setting. There was insufficient data for two theories, a third was rejected as existing barriers in the Irish setting made it impractical to use. Conclusions To improve the success of Irish GP based interventions to reduce ADEs for CDOA, interventions must be relevant and easily applied in practice, supported by national policy and be adequately resourced. Future research is required to test our theories within a newly developed intervention.


Amendments from Version 1
The text has been updated to provide a clearer description of a realist review, what makes it different to a conventional systematic review, and why we opted to use this methodology; a systematic review provides evidence in relation to the efficacy of an intervention, but does not identify why the intervention worked, or not.There are notable differences in the methodology of conventional systematic reviews and realist reviews.A realist review is iterative and non-linear and much more likely to make use of grey literature rather than relying solely on articles in academic journals.In addition, the evidence in a realist review is appraised to assess the weight of its contribution to refining the developed theories.We also included a clearer definition of the term "theory" when used in realist reviews; in order for it to be a realist theory, it must refer to the context and mechanisms and how they relate to outcomes.
In addition, additional text was added to the Strengths and Limitations section in relation to the screening process and the limited use of two independent reviewers for some steps due to time limitations.To reduce the potential subjective impact of the approach used, the research team used discussions, pilots and detailed inclusion and exclusion criteria guidance documents, and a transparent record of the steps taken were recorded.Some minor corrections were made to Extended Data File 5 which listed the full text records sought for retrieval and assessed for eligibility.

Introduction
Adverse drug events (ADEs) are defined as 'harmful unintended consequences of medication usage including medical errors, side effects, adverse drug reactions (ADR) and overdoses' 1,2 .An alternative definition is "any untoward occurrence that may present during treatment with a pharmaceutical product but which does not necessarily have a causal relation to the treatment", which acknowledges that not all events are necessarily drug-related and that it is not always possible to ascribe causality 3 .We accepted any definition of ADEs in this review.Polypharmacy is most commonly defined as the concurrent use of ≥5 drugs [4][5][6] , which may contribute to increased levels of potentially inappropriate prescribing (PIP) [7][8][9] .
Older adults in Ireland are at risk of ADEs due in part to multimorbidity and the increasing rates of polypharmacy in this age group rising from 17.8% to 82.6% among those aged ≥65 years over the past 25 years 10 .Rates of hospital admissions due to ADR in Ireland for this population were 8.8% of which 57% were considered potentially avoidable 11 .These rates are similar to other international studies 12 .The likelihood of an ADE increases significantly with increasing exposure to PIP and contributes to the economic burden of healthcare 9 .The number of repeat prescriptions and patients' adherence to their medications have also been shown to be significantly associated with adverse health outcomes 13,14 .Consideration of strategies that may reduce the risk of ADEs is important.These include regular review of patients' medical conditions, medications, effects of treatment in conjunction with the use of World Health Organisation prescribing indicators, use of alert tools within electronic health records (EHR) and clear patient communication in relation to the benefits and risks and importance of correct adherence to the medication are all means used to prevent ADEs 15 .Another factor that may increase the risk of ADEs include a previous history of ADEs so clear documentation of such events in the patient record is important.
However, interventions to reduce the incidence of ADEs, or their causes, in older patients in primary care settings have had little or no success; a number of systematic reviews on the topic concluded that new approaches were required to reduce ADEs in older adults and patient-related outcomes should be assessed [15][16][17] .In order to develop a better understanding of how, why, when, where and for whom these interventions are effective or not, a closer examination of the data is required.
A systematic review provides evidence in relation to the efficacy of an intervention, but does not identify why the intervention worked, or not.There are notable differences in the methodology of conventional systematic reviews and realist reviews.A realist review is iterative and non-linear and much more likely to make use of grey literature rather than relying solely on articles in academic journals.In addition, the evidence in a realist review is appraised to assess the weight of its contribution to refining the developed theories".A realist approach is particularly suited to the synthesis of evidence about complex interventions as it uses a different lens to examine why and how the interventions are supposed to work rather than if, they worked [18][19][20][21][22] .Using a diverse range of evidence, including theoretical and empirical literature and involving key stakeholders in the process will increase the clarity and understanding of why an intervention succeeds or not 23 .
Realist reviews aim to identify what it is about interventions that generate change (i.e., the mechanisms) and under which circumstances the mechanisms are triggered (i.e., the contexts), which result in changes in the behaviour of the participants of the intervention (i.e., the outcome).These three elements, i) context, ii) mechanism and iii) outcome, are presented together as a statement or theory which attempts to describe what needs to happen for an intervention to work 24 .In order for it to be a realist theory, it must refer to the context and mechanisms and how they relate to outcomes 23 .Mechanisms can be divided into resource mechanisms and reasoning mechanisms.Resource mechanisms tend to be more concrete i.e. environmental, organisational or political, while reasoning mechanisms are more invisible responses to the context i.e. trust or confidence.We have differentiated between the two in this review where possible.By differentiating between them, it is hoped that the necessary resources can be more clearly identified.
The products of realist reviews are theories, often produced in the form of "if …. then" statements developed from one or more Context Mechanism Outcome Configurations (CMOCs) found in the available data that outlines the individual or collective responses to intervention strategies and resources.This methodology is supported by methodological guidance, publication standards and training materials for realist reviews 25 .
A rapid realist review (RRR) is a more focused and accelerated version of a full realist review which aims to produce theories in a time-sensitive way and that is useful to a specific audience about emerging issues, while preserving the core elements of realist methodology 26 .
The aim of this RRR is to identify and examine the contexts and mechanisms that play a role in preventing or increasing outcomes relevant to the reduction of ADEs in community dwelling older adults (CDOA) in the general practice setting that could inform the development of a successful intervention in the Irish community setting.Specifically, we examined evidence in the literature in relation to: 1.The mechanisms that cause GPs and their community dwelling older patients to respond to interventions to reduce the incidence of ADEs.
2. The contexts believed to influence whether different mechanisms produce their intended outcomes.
3. The circumstances in which interventions in GP practices to reduce ADEs in community dwelling older adults are most likely to be effective.

Methods
The protocol for this RRR was published on PROSPERO (CRD4202127757) in October 2021.
There are several stages to a RRR, some are similar to a conventional systematic review, but others are quite different (Table 1) 25 .As the process is an iterative one, the steps may not move in a linear fashion, at times steps were retraced or revisited.

Gathering an expert research team
The expert team included two authors from the 2020 systematic review of interventions to reduce ADE-related outcomes in older adults 17 , with expertise in data science and population health (KB, CC), an academic general practitioner (EW), a healthcare professional with a background in Realist methodology (CW), a pharmacist and PhD student (JH), and an information specialist (PM).

Developing the Candidate Theories (CTs)
Candidate theories were developed in relation to how an intervention is expected to work based on the research teams' assumptions, experience and knowledge of the field and on a brief search for evidence from recent literature.The research team developed eight candidate theories.

Reference Panel feedback and revision of the CTs
These candidate theories were reviewed by a reference panel of general practitioners (GP) in Ireland providing practical feedback using their local contextual knowledge.The reference panel completed an online survey on how well the theories were understood, how relevant they considered them and how feasible they would be to implement in an Irish setting (Extended Data File 1).The survey was emailed to academic GP colleagues, posted on Twitter with tags to GP organisations in Ireland and emailed to GP members of the Primary Care Clinical Trials Network Ireland group.
Details of the findings of the Reference Panel Survey are in Extended Data File 2. In summary, two of the eight theories were amalgamated, as they were judged to have overlapping themes: communication, providing information and shared decision-making.A theory relating to home visits was excluded as the respondents indicated that although home visits could be helpful, they were impractical in practice.The candidate theories were revised based on this feedback.The revised candidate theories are presented in Table 2.The six revised candidate theories formed the basis for the search strategy for the review.

Developing and undertaking the search strategy
Six electronic databases were searched (Ovid Medline, Embase, CINAHL, Web of Science, Cochrane Library and Lens).Four broad search strings were used, older adult, adverse drug event, primary healthcare and community dwelling.Details of the search terms and findings for each database are outlined in Extended Data File 3. Filters included a date limitation (Jan 2011 -September 2021) and articles in the English language only.

Screening, selecting and appraising of articles
The theories influenced the inclusion and exclusion criteria for the RRR search.A pilot screening of ten articles allowed some clarifications to be made to the inclusion and exclusion criteria.Conflicts were resolved by discussion and input from a third reviewer (KB).Table 3 outlines the broad inclusion and exclusion criteria used.A detailed inclusion and exclusion criteria Quality assessment (QA) of realist data is considered under the headings of relevance, rigour and richness 27,28 , defined in Table 4.A scoring system was developed to rate the articles.
Ten percent of the articles included for full text review had QA by two reviewers independently and disagreements were discussed and resolved (CW, JH).The refined QA process was applied by one reviewer (CW) to the remaining articles.Relevance, rigour and richness was scored for those articles that met all inclusion criteria at the full text stage review.Richness was reassessed at the data extraction stage.
The richness assessment at full text reading ensured only those with the most potential for providing rich data were included Revised Candidate Theories based on Expert Panel and Research Team Feedback (ranked in order of relevance)

Engagement of GPs in interventions
To engage GPs in interventions to reduce the levels of ADE in CDOA, the competing demands on their time, the complexity of their patients, and the barriers to changing or de-prescribing medications, must be addressed.

Health and clinical guidelines or policies
If guidelines or policies consider multi-morbidity, polypharmacy, and the commonly encountered adverse drug events that occur in older adults, then they will be more relevant to GPs who will be more likely to use them in practice thereby reducing the risk of ADE.

Continuity of Care
When CDOA have continuity of care, they feel more understood and supported and have increased trust in their GP, the GP will be more familiar with their patient's individual needs and confident when providing care thus improving medication management and reducing the risk of ADE.

Health Information Technology
If Health Information Technology, including summary electronic care records and clinically useful medication alert systems are available to GPs and are easy to use then GPs will feel more supported, informed and confident when prescribing or changing medications, thereby reducing the risk of ADE.

Shared Decision Making
When GPs communicate effectively, engage and support their patients and/ or carers in shared decision-making, there will be increased mutual trust and understanding about their illnesses and medications and patients will feel empowered, thereby reducing the risk of ADE.

Collaboration with Pharmacists
When GPs and pharmacists in primary care work together when caring for CDOA with polypharmacy, GPs will feel more supported, aware and confident in relation to their patients' individual needs resulting in more appropriate prescribing thereby reducing the risk of ADE.
English Language.
Participants: GPs, their nurses, patients, their informal carers and community pharmacists Setting: GP practices and any linked setting i.e. community pharmacy or patients home.
Focus on those countries where GPs have gate-keeping functions similar to Ireland unless the topics relate to human behaviour, support systems, beliefs, attitudes, opinion and perspectives that might be comparable to an Irish population.
Articles irrespective of study design, opinion pieces, policy or protocol Articles related to ADE i.e.PIP, Potentially inappropriate medication (PIM), Deprescribing, Reporting ADRs and Polypharmacy, where the GP or their staff have a role to play.

Nursing home or institutional settings
In-patient hospital settings

Articles reporting prevalence only
Studies about pharmacists or public health nurses that do not also include a role for GPs.

Studies about home visits Studies that use GP data only
Studies focused on patients < 65 years old Articles not immediately accessible via our own library (RCSI) or on open access.
in the RRR.Only those articles with a score of three or four for richness were included.

Extraction of the data
The included articles were imported into NVivo© and data extraction was carried out using this software by CW.Retroductive and abductive reasoning are used to make inferences in relation to how the data might be configured to explain how, why and in what context an intervention might work.The candidate theories formed the basis for the extraction process; sub themes, contexts, mechanisms, outcomes and some intervention details were extracted under a selection of codes.These codes provided an extraction template and were modified and developed throughout the process following familiarisation with the data.The final codebook can be found in Extended Data File 5.

Results
The details of the search and screening processes and findings are presented in the Prisma Flow Chart (Figure 1).After removal of duplicates, 606 articles and an additional 27 articles from other sources were screened for relevance by one reviewer (N=633) (CW).Of these, 145 broadly relevant articles were more carefully screened independently by two reviewers (CW and JH).Eighty full texts were assessed for eligibility using the agreed inclusion and exclusion criteria (Extended Data File 5).Twenty-seven of these were included in the final review.Forward chasing added a further five articles reporting more details on already included interventions.In addition, a selection of Policy or Guidance documents on medication safety, polypharmacy or multimorbidity, which were identified by the team as being relevant to the review (N=13), were also included.The final number of articles/documents included in this RRR was 45 (Extended Data File 6).
Six articles included patients under 65 years as they were considered to contain relevant rich realist data.Of these, two included patients ≥18 years in relation to their experiences of adverse drug reactions 42 or polypharmacy 43 , another three included patients ≥50 years [44][45][46] and one included patients ≥60 years 47 .
Of the remaining nine articles, four described the patients as living with multimorbidity [65][66][67][68] , four as elderly or old [69][70][71][72] and one did not provide details 73 .outlined below.Quotations from the articles to support the themes are presented in Table 5.

Theory 1: Engagement of GPs with interventions
This theory focused on the challenges of engaging GPs with any planned intervention to reduce the levels of ADEs in CDOA.
We hypothesised that the competing demands on their time, the complexity of their older patients, and the challenges associated with changing or deprescribing medications, must be addressed in order for the intervention to be accepted and implemented by them in a sustainable way.

Facilitators
Involving GPs in the design and implementation of interventions.It was generally accepted that involving GPs in the content and development of the intervention aided engagement and improved their practical implementation by identifying and minimising potential barriers 46,57,68,70,74,75 .The use of qualitative methods were thought to be useful at the development stage, many studies described focus groups and individual interviews with key stakeholders.
Having a plan that outlined how to implement an intervention helped.However, it was important that it should not impact excessively on existing processes in the practice or how GPs engage with their patients.Providing GPs with choices and options in relation to how they could adapt/tailor the intervention to suit their practice setting was highlighted.

Feedback and support.
Receiving follow-up support, seeing the benefits of their efforts, being given the opportunity to give and receive feedback were all seen as useful strategies in relation to successful implementation and sustainability of interventions.

Challenges
Research element.The research element of interventions resulted in additional pressures on GPs and the research team in relation to patient enrolment and consent.
Beyond the design and implementation of interventions, other contexts that challenged the success of interventions were identified; how these challenges could be overcome were  "Another issue was that they feared the use of a check-list aiming at standardizing or structuring the conversation would impede the individual care for the patient."Jäger 2017.
Feedback and Support "Some GPs appreciated that the implementation action plan helped them to raise awareness and to motivate the practice staff for change.'This helped us a lot.The motivation of the staff was stronger and as you can see we have realised most of the issues we have elaborated.That was most helpful (GP).' " Jäger 2017b "Less intensive feedback on prescribing behaviour is generally not sufficient to impact on prescribing practices" Clyne 2016b Research element "Overall, patient identification and recruitment was reported as being quite onerous and was considered "…the only graft" (GP16, intervention practice) involved in participating in the study by many of the GPs."Clyne 2016b Time "Practice nurses or other multidisciplinary team members can contribute in specific ways, including undertaking target assessment of chronic disease and psychological or functional capacity assessments that can support doctor and patient shared decision making."(Wallace 2015) "The risk patients come any way, at least once per year for the check-up (…) and I think you can combine this very well" Jäger 2017 "Many chronic conditions, which were once managed in secondary care, have now become the remit of general practice.Adequate resourcing of primary care is of paramount importance in ensuring that this workload can be safely and effectively managed."ICGP 2020 Case Complexity "Respondents were also confident in their prescription of the identified PIMs and reported being comfortable with continued prescription."Voight 2016 "I mean it [medication review] is time consuming which will be the biggest challenge… it's nearly a bit of detective work going on, through the notes, trying to work out how did somebody on 16 items get onto some of these drugs."P4 Clyne 2013 "As was recognised by several physicians in the focus group discussion, risk is determined by more than just the crude number of medications being taken by a patient.A more nuanced approach to determining risk of ADR involves considering a combination of pharmacological, physiological and environmental determinants."Ridge 2019 Patient complexity (e.g.polypharmacy, multimorbidity), as well as prescriber complexity (e.g.multiple prescribers, poor communication, restricted autonomy) were all identified as factors contributing to a complex prescribing environment where PIP could occur Clyne 2016a Isolation "A key theme was GPs' sense of professional isolation in the management of multimorbid patients.This emanated from the interplay between four aspects of the management of patients with multimorbidity: (i) the disorganisation and fragmentation of healthcare between primary and secondary care, (ii) the inadequacy of guidelines and evidence-based medicine for multimorbidity, (iii) challenges in delivering patient-centred, rather than disease-focused, care and (iv) barriers to shared decision-making.""Although NICE full evidence summaries do provide information on the risks and benefits of treatment, few clinicians will have the time or expertise to read and interpret these documents, and the information is not consistently presented to facilitate comparison."Hughes 2013 "For example, GPs adopted a passive approach to medication management due to their uncertainty (lack of psychological capability) about which medications were most valuable in patients with multimorbidity, especially given the absence of satisfactory guidelines in this field" Sinnott 2015b Professional Judgement "I don't think that there is any good way to make that decision [regarding risks and benefits] other than your own clinical gut instinct or intuition."Fried 2011 "Although attempts are under way to improve the attentiveness of guidelines to multimorbidity, they will not be able to cover all eventualities in multimorbidity and some professional judgement will always be required" Sinnott 2015a "This leads to a situation where every individual recommendation made by a guideline may be rational and evidence based, but the sum of all recommendations in an individual is not."Wallace 2015 Theory 3:

Continuity of Care
Changes to practice "Changes in the delivery of general practice service have reduced the provision of continuity of care.Patients value continuity, with over 80% of older patients (aged ≥75 years) in a recent UK survey reporting a preference for seeing a particular doctor in their general practice."Wallace 2015 "Our findings suggest that fragmentation of care between multiple prescribers results in poor communication of up-to-date patient medication information" Clyne 2016a "The communications space for shared care in community settings has receded and needs to be enlarged with concomitant improvements in communications technology, process and protocol to support effective multidisciplinary working."Rodgers 2014 Theory 4: Health Information Technology Information Technology "Using the potential of information technology and data will help bridge the gaps between care services and enable people who use these services have access to their health care information, all of which can help optimise the use of medicines.

NICE Medicines optimisation 2016
Concomitant improvements in communications technology, process and protocol are urgently required to offset potentially serious risks to patient safety.Rogers 2014 Alert Tools "The use of a decision framework to identify (PIMs) for an individual could prove superior to lists of "drugs to avoid," Anderson 2020.
"In the interviews, the main barrier for using the tools was that they were not integrated into the practice software."Jäger 2017b "To be effective, such tools need be applicable in routine clinical practice, not only in a research environment.However, as our study indicates, this gap may not have yet been successfully bridged in primary care."Clyne 2016a "The answer to the problem does not seem to lie in mono-causal pharmaco-centered approaches or practical helps/tools."Pohontsch-2017 "While some GPs appreciated the list, others had a more negative view, because they felt (severely) restricted in their freedom to choose medications.Rather than having a blacklist "banning" certain medications, they would prefer a whitelist indicating which medications can be safely used for elderly patients."The training helped the healthcare professionals understand how shared decision making differed from their current ways of working, by helping them improve their communication of risk and the way they explore what matters to patients.Some clinicians reported changing their view from we do this already to we could do this better."NICE 2019 Use of SDM "The law now requires healthcare professionals to take reasonable care to ensure that the patient is aware of any material risks involved in any recommended treatment, and of any reasonable alternative or variant treatments."NICE 2019 "The qualitative evaluation of the pilot study indicated that GPs were very positive about both their experience and the patients' feedback of the review process, and GPs were motivated to alter their prescribing practice: 'O ya, and she was delighted, I stopped some "Patients should be empowered to manage their own health and be provided with the necessary skills and supports to do so."HSE Framework 2020 "Two factors were negatively associated with a good knowledge of the purpose of medications.The first factor was polypharmacy:...The second factor was receiving help with drug management: the more help, the lower the odds of good knowledge" Hoisnard 2018 "Almost none of our patients was able to use this tablet themselves.I think the medical assistant did it with them and read it to them or showed it to them (GP")

Case complexity.
A number of the studies reported that GPs are often not in agreement with each other or with the guidelines in relation to the medications that are appropriate to prescribe for older patients with multimorbidity.There was evidence that GPs acknowledged or recognised that potentially inappropriate prescribing occurs in practice 44,58 .However, regardless of whether this awareness of PIP was raised following an intervention, or already existed, GPs did not believe there was always a need for change.The GP's in-depth knowledge of the patient and their consideration of the importance of the GP-patient relationship supported their confidence in their individual decisions.GPs were confident about making "informed decisions" to use medications that had potentially adverse effects once these were balanced against treatment benefits, in the patient's best interest 54,62,73 .
Professional isolation.Professional isolation was identified as a factor for GPs.To address this, using existing relationships for example, informal peer support and relationships with Community Pharmacists could be means to providing support.Collaboration with Community Pharmacists and other GPs (which was the limited focus of this RRR), not only for their additional knowledge, but also to 'share the onus of responsibility', were identified as supportive strategies 66,68,76 .
Incentivisation.Structured medicines reviews (SMRs) with a focus on optimising appropriate polypharmacy may not be considered core work by Irish GPs 57 .Other international studies also reflected this sentiment 53,73 .Reimbursement systems that include SMRs, which acknowledged the level of expertise and time involved, or external monitoring of patient medications by pharmacists, were identified as possible incentives 57,66 .
Other contexts that may influence the engagement of GPs in interventions that aim to reduce ADEs will be covered under the other more specific theories.
The data support Theory 1 and provide some individual Context -Mechanism -Outcome configurations (CMOCs) that could be applied to the design, implementation and sustainability of interventions to engage GPs in the reduction of ADEs in CDOAs in the Irish GP setting (Table 7).

Theory 2: Health and clinical guidelines or policies
This theory considers the relevance of medication safety or prescribing policy or guidelines for older adults with multimorbidity or polypharmacy.We theorised that if guidelines or policy documents considered multimorbidity, polypharmacy, and the commonly encountered adverse drug events that occur in older adults, then they would be more relevant to GPs who would then be more likely to use them in practice thereby reducing the risk of ADEs.

Relevance
Clinical guidelines tend to focus on single diseases and usually exclude older populations or those with multimorbidity.When guidelines do refer to older patients, they tend to be general statements about considering individual drug characteristics and age-adjusted doses of medications.Hughes et al in his 2013 review of UK NICE clinical guidelines stated that one of the major challenges facing clinical guidelines is accounting for multimorbidity 65 .There was consensus in the articles reviewed that current clinical guidelines with their single disease focus act as a driver for polypharmacy in older multimorbid patients and do not providing guidance on how best to prioritise recommendations 31,59,64,65,67,72,76 .
The lack of relevance of clinical guidelines for older patients with multimorbidity was shown to have negatively influenced the level of proactive involvement of GPs in relation to managing the medications of these patients 68 .
However, more recently a NICE clinical guideline focusing on multimorbidity has highlighted issues to consider when caring for patients with multimorbidity and includes risk factors associated with the highest risk of experiencing ADEs.These patients could be targeted for structured medication reviews in general practice 35 .

Professional judgement
A number of articles addressed the dilemma for GPs when confronted with clinical guidelines that do not work for their patient 54,62,64,67,70,72,73,76 .The GPs used their judgement and accepted less stringent levels of disease control, stratified risk and benefits for individual diseases, prioritised the patients' quality of life and preferences or modify guidelines in anticipation of adverse effects.
GPs felt justified in using their clinical judgement stating reasons such as the exclusion of older patients or those with multimorbidity in clinical trials underpinning clinical guidelines and the cumulative risk of polypharmacy if all guideline recommendations were followed.In addition, GPs used clinical and laboratory monitoring (e.g.blood tests, asking patients about side effects) as a means to support their decisions.
Sinnott et al. (2015) highlighted the relative autonomy experienced by GPs in the Irish healthcare system with respect to chronic disease management, which allowed them to use their judgement to a greater extent than GPs in other countries with tighter frameworks such as the UK 67 .

Using policy and guidelines in practice
Table 8 outlines what the data suggest in relation to the content of clinical guidelines and policy documents and the contexts that may encourage GPs caring for CDOA to use them in practice.

Irish guidelines, policies and strategies
GPs in Ireland must follow the legislative requirements, local and national clinical guidelines, and professional standards when prescribing medications.
The Health Information and Quality Authority (HIQA) and The National Clinical Effectiveness Committee (NCEC) provide the framework for developing these clinical guidelines.The various disease specific guidelines provide Summary of recommendations and Summary of Good Practice Points documents, which are available to GPs, although they do not target them specifically; they acknowledge the professional judgement of GPs when the individual guideline recommendations are not appropriate or are declined by an individual patient.
The HIQA "Guidance on Medicines management" focuses predominately on patients in residential care, but is relevant to patients in a community setting 29 .It recommends medicines management, which includes assessing, supplying, prescribing, dispensing, administering, reviewing and assisting people with their medicines, to reduce ADEs.
The Irish College of General Practitioners' Quality and Safety in Practice Committee produces quick reference guidance, which can be accessed by their members.In "Medication Review -A Guide for GPs" they recommend that high-risk patients are identified and offered regular medication reviews.They also highlight the need for adequate resourcing to support the conduct of structured medication reviews in general practice 31 .
The recently published National Framework for the Integrated Prevention and Management of Chronic Disease in Ireland 2020-2025 strategy is likely to have a significant impact on the workload of GPs in the community 30 .
One element already being rolled out in Ireland is the Health Services Executive (HSE) Structured Chronic Disease Management (CDM) Programme, currently available to patients eligible for free GP care through a general medical services card who meet other criteria e.g.certain chronic medical conditions e.g.Type 2 diabetes, asthma, cardiovascular disease.Carry out an assessment within the practice against the recommendations to identify gaps in current practice.

k)
Measure and record improvements and feedback this to staff and patients.
A clinical data repository is proposed in the National Framework, that will gather information electronically on demographics, diagnoses, clinical examination results, diagnostic results and life-style risk factors on patients from GPs.
This National framework, if successfully implemented, goes some way to providing the elements outlined from the data in relation to the content of guidelines and the contexts that may encourage GPs caring for CDOA to use them in practice if they are properly resourced.To account for multimorbidity and polypharmacy, the theory might be amended to include respect for the GPs' professional judgement, adequate resourcing to support the conduct of structured medication reviews, which are often complex and time-consuming and an electronic system of data collection and audit/feedback.The CMOCs for this theory are outlined in Table 6.

Theory 3: Continuity of care
This theory focused on the relationship between the patient and the GP and continuity of care.It hypothesised that when CDOA have continuity of care, they feel more understood and supported and have increased trust in their GP, the GP will be more familiar with their patient's individual needs and confident when providing care thus improving medication management and reducing the risk of ADEs.
A smaller number of articles addressed the issue of continuity of care; while they all acknowledged its importance, most also acknowledged the changing work environment that has reduced the level of continuity and increased the risk of fragmentation of care 53,56,66,74,76 .
Improving the level and quality of communication between healthcare professionals and assigning a specific GP to patients identified as complex were the limited proposed solutions to this issue 53 .The National Framework strategy aims to support improved sharing of patient health records and speaks about co-ordination of care but does not specifically refer to continuity of care.
In summary, while continuity of care was acknowledged as being important, there was insufficient evidence in the included articles to support or reject this theory.

Theory 4: Health information technology
This theory focused on the use of health information technology (HIT) to manage medications.We also collected data in relation to the use of any other forms of technology likely to have an impact on patient care and ADEs.
We hypothesised that if HIT, including electronic summary care medical records and clinically useful medication alert systems, are available to GPs and are easy to use then GPs will feel more supported, informed and confident when prescribing or changing medications, thereby reducing the risk of ADEs.

Use of technology
Use of technology in general was felt to be a positive addition to improving medication safety, however the data related to its proposed use in the future rather than any current or past experience with it 34,53 .
When using an electronic alert system it was considered important to have it integrated into the general practice software for ease of use.Commonly handheld computers or tablets were used due to their mobility, robustness, simplicity of use, and low needs for technical support 45,71 .GPs used electronic and hardcopy versions of the tools either alone or together.
It was acknowledged that the alert tool needed to be specific to the country in which it was being used due to countryspecific differences with respect to drug approval, prescribing practices, and treatment guidelines 54,55 .However, even when considerable effort was put into developing a specific tool for a specific setting, this did not guarantee successful implementation 58 .Some level of frustration and confusion was expressed by researchers attempting to address the multiple barriers to their use, as their interventions failed to show impact 58,64 .
GPs were often sceptical in relation to the value of alert tools.
Reasons for the scepticism included: tools not considering multimorbidity or polypharmacy, not designed for use with older patients, not easily used in daily practice, creating too many irrelevant alerts, too time consuming or not trusted to be up to date 36,45,48,54,62,71 .
Positive features that were identified or suggested in relation to making alert tools more acceptable in practice included: focusing on a select number of high-risk drugs, provision of clinical validation and intelligent alternative recommendations for alerts, allowing for clinical judgement, acknowledging the complexity of prescribing for older adults, allowing for patients' preference, provision of training to ensure correct use of the system, understanding of its limitations and providing reminders to review existing medications or follow-up on pathology 36,52-54,57,62,66,71,74 .
In summary, there were limited data in relation to the use of electronic summary care records to reduce ADEs for CDOA.
There was some evidence in relation to the use of alert tools.However, although a number of articles reported on interventions that had attempted to addresses the issues identified in our theory, accessibility and ease of use of alert tools, there appeared to be many other factors that needed to be taken into account.Training and building trust in the system, providing alternative recommendations, acknowledging the complexities and personal choices of older patients and the judgement of GPs are contexts and mechanisms that would need to be added to this theory and a new search for this specific data undertaken.

Theory 5: Shared decision making
Shared decision-making (SDM) involves two-way communication between GPs and their patients.GPs share the best available evidence to patients and their carers so that they can consider all the options available to make informed decisions regarding their healthcare.It often includes medication review, deprescribing and discussions around adherence to recommendations 38 .
This theory considered the contexts around SDM and what influence they have on its success.We examined evidence to support or reject the theory that if GPs communicate effectively, engage and support their patients and/ or carers in SDM, there will be increased mutual trust and understanding about their illnesses and medications and patients will feel empowered, thereby reducing the risk of ADEs.All but four of the included articles provide data relating to this theory 29,32,48,65 .The contexts relevant to GPs and patients are outlined separately.

GPs perspectives
Training and skills in SDM.It was recognised that GPs required SDM training focusing on eliciting patient preferences, using decision-making tools, incorporating SDM into routine practice, communicating risks and benefits of treatments, broaching and discussing care plans, quality of life and end of life issues 34,38,45,54,72 .
The types of training described included role-play, emphasis on practical skills, the opportunity to challenge and discuss embedded attitudes, online courses and the use of peer support 38,66,68 .
Facilitators.One facilitator identified to support the implementation of SDM in general practice included having it as a policy recommendation; making it routine practice, such as the "no decision about me, without me" strategy in the UK 34 .
Providing GPs with the opportunities to experience SDM and receive affirmation of its usefulness and effectiveness had the potential to increase engagement in the process, provide empowerment, encouragement, motivation and build confidence and competence 55 .
A positive relationship between the GP and their patient and the resulting mutual trust between them was identified by many as being a major facilitator for SDM 56,62,63,66,67 .
The GP's trust in the patient related to their belief in the accuracy of the information the patients communicated to them, in relation to symptoms, adherence etc. and the patient's ability to understand the information they were sharing with them, fearing the information would only serve to confuse and worry them 31,41,45.Challenges.The main challenge for implementing SDM identified in the articles was the time consuming element of SDM in an environment where consultation times were short, the workload extensive and, as already outlined in Theory 1 in relation to SMR, there was a lack of incentivisation 31,50,55,58 .
Some of the suggestions for overcoming this challenge included identifying those patients most at risk or adopting extended consultations for particularly complex patients 31,35,39,45,52,76 .
The complexities of the patient, of prescribing and managing third party prescriptions, left GPs feeling overwhelmed, frustrated and lacking in confidence in relation to SDM, and its associated elements of medication review and deprescribing 46,50,58,59,64 .

Patient perspectives
Patient knowledge/education.The patient's awareness of and level of interest in their medications, how to use them and treatment options varied and was influenced by their level of polypharmacy and their need for the support of others 51,53,60,61 .
Methods to provide information to patients included information leaflets, use of technology; patient friendly interactive training tools or' serious gaming' approaches and posters to act as reminders about medication use.There were mixed results in relation to their use.One intervention using technology had referenced research supporting the use of iPads/tablets for patients with dementia but found the patients in their study were unable to use them [44][45][46] .Information leaflets were seen as useful by the GPs but not always valued by patients 45,55 .
Providing health education to patients so that they can play their role in SDM may require structural changes to the health services in relation to funding and access 30,38 .

Patient preferences.
In general, patients regarded their Quality of Life (QoL) as being more important than mortality 70 .Patients' decisions regarding their treatment was influenced by their emotions, treatment goals and willingness to experiment.Patients' willingness, and ability to be involved in decision making varied widely in the included articles 30,31,33,36,38,41,55,70,72 .
This theory focused on the role of the GP in SDM process.However, the data suggest that for SDM to be successful, the role of the patient must be equally supported.
The theory should be amended thus; if GPs and their patients and/ or carers are supported to engage and communicate effectively in SDM, there will be increased mutual trust, empowerment, awareness and understanding about their illnesses and medications, thereby reducing the risk of ADEs.The CMOCs for this theory are outlined in Table 6.

Theory 6: Collaboration with community pharmacists
This theory focused on collaboration with Community Pharmacists only.It proposed that when GPs and pharmacists in primary care work together when caring for CDOA with polypharmacy, GPs will feel more supported, aware and confident in relation to their patients' individual health needs resulting in more appropriate prescribing, thereby reducing the risk of ADEs.

Facilitators
Respect.The data showed that, in general, GPs valued and respected the input of pharmacists for medication review or advice in relation to prescribing 44,63,66,71,72,76 .This existing positive perception is likely to enhance any collaboration between them.

Supportive relationships.
Developing and maintaining a positive relationship between the GP and the community pharmacist was regarded as an important prerequisite to collaboration, this included a willingness to share patient information.It was also identified as a means to reduce the sense of isolation felt by GPs and a way to share the responsibility of caring for complex patients 57,58,63,68 .This collaboration was perceived to facilitate patient centred care and medication management, and to reduce fragmentation of care 44,45,47,53,63,66,74 .
Patient trust.Patients were most likely to ask either their GP or their pharmacist if they needed information or had concerns about their medicines.They trusted and had generally positive relationships with their pharmacists 42,56 .

Challenges
Responsibility.A lack of clarity in relation to roles and responsibilities for medication management, particularly because of fragmentation or lack of continuity of care was reported in UK settings and identified as a potential gap in the integrated care process 53 .Irish GPs have also indicated that they do not consider SMR to be their core responsibility 57 .
Workload.The existing workload of community pharmacists may limit the role they can currently play 50,68 .
Collaboration in the Irish setting.While existing Irish policy, guidance and strategy documents acknowledge the importance of collaboration in general and specifically between GPs and pharmacists, there is currently no formal arrangement and existing shortages of community pharmacists limit the impact of informal collaboration 30,32,35,68 .Research is required in the Irish setting to provide further evidence.
The data support this theory in relation to positive and respectful relationships and providing support and increased drug knowledge to GPs.However, a number of barriers must be surmounted to facilitate a more formal and structured approach to medicines management between GPs and pharmacists in the community in the Irish setting.There is a need to identify which HCP is responsible for conducting a structured medication review, with sufficient resourcing to support this, and improved communication and HIT must all be in place before this collaboration can be included in any future intervention to reduce ADEs for CDOA.For this reason, the theory was rejected by the team.

Discussion
Three theories were supported by data (Theory 1, Theory 2, and Theory 5), one theory was supported to some extent (Theory 6) and two theories lacked sufficient data to make any firm conclusions (Theory 3 and Theory 4).The theories, relating to GP engagement in interventions (Theory 1), relevance of health and clinical guidelines or policies (Theory T2), and shared decision-making (Theory 5), provided some context-mechanismoutcome configurations that can be used to guide future interventions to reduce ADEs for CDOA in an Irish setting by highlighting the facilitators and barriers to success.
Involving GPs in the design and implementation of interventions that target them was identified as a facilitator by ensuring that the intervention was relevant and easy to apply in practice (Theory 1).However, insufficient internal or external support in the form of health policy or legislation, funding, incentivisation, feedback and follow-up were likely to hamper success.A SR of Australian GPs involvement in interventions to improve patient management by Bernardes et al. (2019)  reported on the challenges of engaging GPs in interventions and concluded similarly that communicating directly with GPs, exchanging ideas and ensuring that the topics were relevant and useful to them, improved their engagement.Feedback was also identified as useful 77 .
Clinical guidelines and policies that account for multimorbidity and polypharmacy, respect GPs' professional judgement and that have adequately resourced and monitored recommendations, are more likely to be used by GPs (Theory 2).Recent Irish policy and guidelines, such as the National Framework for the Integrated Prevention and Management of Chronic Disease in Ireland 2020-2025 strategy, if successfully implemented, could support these requirements.A number of other studies have identified challenges in relation to the use of clinical guidelines by GPs; mistrust, being overloaded with information, lack of respect for their expertise and loss of autonomy to make their own decisions, as barriers, but acknowledge that knowing recommended practice is helpful 77,78 .
The role of both the GP and the patient in the SDM process must be equally supported for it to have an impact on reducing ADEs in the community (Theory 5).The need for tailored training for both stakeholders to acquire the skills for this complex interaction was identified, as were the challenges already identified in Theory 1 and Theory 2. The complexity of learning and using the skills of SDM for both the healthcare professional and the patient has been the topic of a number of recent publications, these publications are likely to assist in understanding what is required to support GPs and their patients in this process 79 .
Theory 1 and Theory 2 are dependent on each other, Theory 5 could be tested independently, once the contexts identified in Theory 1 and Theory 2 are in place (Figure 2).
Collaboration with community pharmacists was supported to some extent by the data (Theory 6), but existing barriers in the Irish settings made it impractical.There is currently no formal arrangement in Ireland for this type of collaboration and existing shortages of community pharmacists limit the impact of informal collaboration.The role of clinical pharmacists as part of the primary care team, similar to those introduced in the UK, and tested in Ireland in a pilot study shows some promise 80 .
There were insufficient data for two theories to make any firm conclusions (Theory 3 and Theory 4).There were insufficient data in relation to the impact of continuity of care for either the patient or their GP in relation to ADEs (Theory 3).In addition, the changing work environment for GPs in the community was highlighted, resulting in reduced levels of continuity of care.
Theory 4 focused on electronic care records and medication alert systems and evidence was in relation to the future rather than any current or past experience.There was some evidence in relation to the use of alert tools.Our theory focused on the accessibility and ease of use of alert tools; however, the data suggested that many other factors need to be taken into account including access to training, building trust in the system, providing alternative recommendations, acknowledging the complexities and personal choices of older patients and the judgement of GPs.A complete revision of this theory and new search for relevant data is proposed based on these findings.

Strengths and limitations
The strength of this review is the additional insight it has provided in relation to past interventions, and their application specifically to an Irish context based on interventions limited to countries where GPs have gate-keeping functions.The date limitation (2011 -2021) has also ensured the findings were relevant to current practice.
Limitations to this review are that the resulting final theories relied on the articles identified in our search; there was a lack of evidence to assess two theories.This may imply a lack of data on these topics or that a more in-depth search of the literature is required.The time limitations for this rapid review also meant that although two reviewers screened all relevant articles by title and abstract, there was only one reviewer for f the full text reading of the articles, 10% of the excluded articles at this stage were reviewed by the 2 nd reviewer.To reduce the potential subjective impact of this approach, team discussions, pilots and detailed inclusion and exclusion criteria guidance documents were developed, and a transparent record of the steps taken were recorded.Quality appraisal was based on the research teams' subjective judgements.In realist reviews, reasoning mechanisms are identified that might trigger positive or negative actions within the contexts.In reporting the CMOCs in this review, the mechanisms are presented in their positive form, but they could equally be working in their negative form if the resource mechanisms are not in place.
The range of outcomes measured in the interventions tended towards the conventional, i.e. changes to numbers of hospitalisations or medicines prescribed.Patient-reported and systemrelated outcomes provided greater opportunity to understand what works for whom and under what circumstances, how and why.

Implications for practice
In order to improve the success of interventions to reduce ADEs for CDOA GPs must be involved in the design and implementation of the interventions to ensure their engagement.The interventions must be relevant and easily applied in practice, supported by national policy and be adequately resourced.
In addition, tailored education for both GPs and patients is required to support SDM.

Conclusion
Three theories with their related CMOCs, can be applied to the design and implementation of future interventions to reduce ADEs for CDOA in the Irish GP setting, based on this RRR.Future research is required to test our theories within an intervention.

Extended data
Zenodo: Realist-Review-of-ADEs-in-CDOA, https://doi.org/10.5281/zenodo.10328046 81 Peer review: This article addresses and allows new insides into a highly relevant topic.The authors clearly stated objectives and rationale for their RRR and provided rich details of the methods in the main text and in additional web-supplements.The conclusions are in general adequately supported by the results presented in the review.The manuscript is well-written and concise.
My major concern is about the usage of the term "theory", as it is applied to models (such as Continuity of care), principles (such as Shared Decision Making), processes (such as the engagement of certain health professionals), decision support (such as guidelines), and tools (such as Health information technology), and what does it implies to the application of the realist review methods.For example, the following quotations at table 5 were assigned to health information technology ("theory 4"): "Using the potential of information technology and data will help bridge the gaps between care services and enable people who use these services have access to their health care information, all of which can help optimise the use of medicines.NICE Medicines optimisation 2016

○
Concomitant improvements in communications technology, process and protocol are ○ urgently required to offset potentially serious risks to patient safety.Rogers 2014" It would be worth a discussion if quotations would provide data to support "theory 3" (continuity of care or more specifically, informational CoC).This example highlights, that some of the "theories" can be tools or components of others.
It would perhaps help if the term "theory" were avoided and replaced by another, neutral term (eg, rationale for components or mechanism/mode of action?) and if the authors explained why they treat care models, principles, tools etc. equally and discuss the strengths and limitations of this approach.
In addition, I have some minor amendments, which may help to improve the manuscript: 1. Introduction, first para: the authors present two different definitions of ADEs.It would be helpful, if they would state, which definition they used.

Inclusion/exclusion criteria:
The inclusion criterion "Focus on those countries where GPs have gate-keeping functions similar to Ireland unless the topics relate to human behaviour, support systems, beliefs, attitudes, opinion and perspectives that might be comparable to an Irish population." is difficult to understand.For example, German GPs don't have a gate-keeping role.However, German studies formed a relevant body of included studies.Why did the authors include them (and why did they exclude others)?
○ Why did the authors include "opinion pieces"?
○ Figure 1: the box about records excluded shows "Reviews ( 27)".It is not clear, why these reviews were excluded, as the publication type was not presented as an exclusion criterion per se.○

Selection:
It would be helpful to present the connection between the results and the supporting evidence.For example, table 5 presents a quotation about patient preferences ("Their health goals focus more on quality of life than on extending their lives" Schuliing 2012).4. Quality Assessment: In Figure 1, "records linked to included interventions (n=5)" and "relevant policy documents/guidelines (n=13)" seem to bypass the quality assessment, while they are partly assessed in Extended Data File 6 -please check for inconsistencies.Why is the quality assessment of the guidelines/policy documents "N/A"?What does mean the criterion "Richness at extraction"(in addition to "richness") in Extended Data File 6?
5. Discussion of limitations and implications: the authors name a few limitations with regard to their methods.How do the authors assess the effect of their restrictions (e.g.publication language and period, quality feature "richness" as exclusion criterion) and limited resources (screening, data extraction and coding could mostly be done by only one person, the panel survey could not be tested a priori) on the results?Would there be a need for further research to support the results of the RRR?
as Continuity of care), principles (such as Shared Decision Making), processes (such as the engagement of certain health professionals), decision support (such as guidelines), and tools (such as Health information technology), and what does it implies to the application of the realist review methods.For example, the following quotations at table 5 were assigned to health information technology ("theory 4"): • "Using the potential of information technology and data will help bridge the gaps between care services and enable people who use these services have access to their health care information, all of which can help optimise the use of medicines.NICE Medicines optimisation 2016 • Concomitant improvements in communications technology, process and protocol are urgently required to offset potentially serious risks to patient safety.Rogers 2014" It would be worth a discussion if quotations would provide data to support "theory 3" (continuity of care or more specifically, informational CoC).This example highlights, that some of the "theories" can be tools or components of others.It would perhaps help if the term "theory" were avoided and replaced by another, neutral term (eg, rationale for components or mechanism/mode of action?) and if the authors explained why they treat care models, principles, tools etc. equally and discuss the strengths and limitations of this approach.
Thank you for your thoughts on the use of the term "theory" in our rapid realist review.The term theory when used in realist methodology refers to informal theory and is not the same as formal theory.We have stated in the introduction that the products of realist reviews are called "theories", often in the form of "if …. then" statements, developed from the Contexts, Mechanisms and Outcomes found in the available data that outlines the individual or collective responses to intervention strategies and resources.The "theory" therefore in realist reviews is an attempt to describe what needs to happen for an intervention to work.We have added an additional sentence to further explain what a realist theory is "In order for it to be a realist theory, it must refer to the context and mechanisms and how they relate to outcomes (Hunter et al 2022."Wong in his article "Making theory from knowledge syntheses useful for public health.Int J Public Health.2018;63(5):555-6."states that there are many definitions of theory, with a simple definition being an attempt to organize the facts and that a theory must be supported by facts or data.The two quotations you picked out from table 5 in relation to Theory 4 were examples from the included articles to support that theory which was focused on the use of health information technology to manage medications and stated "If Health Information Technology, including summary electronic care records and clinically useful medication alert systems are available to GPs and are easy to use then GPs will feel more supported, informed and confident when prescribing or changing medications, thereby reducing the risk of ADE." Theory 3 was about continuity of care and stated, "When community dwelling older adults have continuity of care, they feel more understood and supported and have increased trust in their GP, the GP will be more familiar with their patient's individual needs and confident when providing care thus improving medication management and reducing the risk of ADE".It was focused on the relationship between the patient and did not include the use of technology.Although it is possible that the same data can be used to support more than one theory, we don't see that this is the case in this instance.As realist theories are the cornerstone of realist reviews, we do not see this as a limitation, but do acknowledge that there are some limitations to this rapid realist review, as there was a lack of data to assess two of the theories put forward.
2. Introduction, first para: the authors present two different definitions of ADEs.It would be helpful, if they would state, which definition they used.
Thank you for highlighting this, our intention was to use as wide a definition of ADEs as possible when screening the articles, we have added a line to explain this.This paragraph will now read: Adverse drug events (ADEs) are defined as 'harmful unintended consequences of medication usage including medical errors, side effects, adverse drug reactions (ADR) and overdoses' 1,2 .An alternative definition is "any untoward occurrence that may present during treatment with a pharmaceutical product but which does not necessarily have a causal relation to the treatment", which acknowledges that not all events are necessarily drug-related and that it is not always possible to ascribe causality 3 .We accepted any definition of ADEs in this review.

Inclusion/exclusion criteria:
The inclusion criterion "Focus on those countries where GPs have gate-keeping functions similar to Ireland unless the topics relate to human behaviour, support systems, beliefs, attitudes, opinion and perspectives that might be comparable to an Irish population." is difficult to understand.For example, German GPs don't have a gate-keeping role.However, German studies formed a relevant body of included studies.Why did the authors include them (and why did they exclude others)?
We did allow for the inclusion of articles from countries with GPs with non-gate keeping functions if the topics relate to human behaviour, support systems, beliefs, attitudes, opinion and perspectives that might be comparable to an Irish population.We also outlined in Extended Data file 4 more details in relation to the inclusion and exclusion criteria.Although we focused on those countries where GPs have gate-keeping functions similar to Ireland: which included UK, Australia, New Zealand and the Netherlands, we also included Canada and USA, if the focus of the article was on family practice only and studies from other countries if the topics relate to human behaviour, support systems, beliefs, attitudes, opinion and perspectives that might be comparable to an Irish population.

Why did the authors include "opinion pieces"?
This is normal practice for realist reviews, as the perspectives of informed individuals may be a rich source of relevant data. 5. Figure 1: the box about records excluded shows "Reviews ( 27)".It is not clear, why these reviews were excluded, as the publication type was not presented as an exclusion criterion per se.
As review articles, they may have included some of the individual articles already included in this review.Review articles also tend to summarise the findings, whereas we were most interested in the individual findings from each article.We did clarify in Extended data file 4 that relevant Systematic, Scoping or Literature Reviews would be included at the title and abstract screening stage and tagged as "Review Articles" and assessed at this stage for studies of interest.
6. Selection: It would be helpful to present the connection between the results and the supporting evidence.For example, table 5 presents a quotation about patient preferences ("Their health goals focus more on quality of life than on extending their lives" Schuliing 2012).The study by Schuliing was about the view of Dutch GPs.Readers could misinterpret the quotation as patient's view (while patients' preferences widely vary -between persons but also within persons over time, also in CDOA: eg, https://pubmed.ncbi.nlm.nih.gov/31843855/1,https://pubmed.ncbi.nlm.nih.gov/33383951/2).This quotation related to the theory about shared decision making which stated, "When GPs communicate effectively, engage and support their patients and/ or carers in shared decision-making, there will be increased mutual trust and understanding about their illnesses and medications and patients will feel empowered, thereby reducing the risk of ADE".All of the quotations in this section relate to GPs, but this one came under the heading of patient preferences, and as you say could have been clearer that it was the opinion of the GPs.We have added a clarification as we have done in some other quotations with [ ] brackets."Their [the patients'] health goals focus more on quality of life than on extending their lives" Schuliing 2012 7. I'd suggest presenting a "baseline table" of included documents in the main text.
We were limited to the number of tables we could include in the main text, and so provided these details in the extended data files.
8. Please check the supplemental documents for inconsistencies: eg.Amos et al. and Bregnhøj et al. were classified as included in Extended data File 5 but are not presented in the final list of included studies (Extended Data File 6).
Thank you for bringing this to our attention, these two articles along with 6 others were considered suitable for inclusion based on the inclusion criteria.However, they were subsequently excluded based on their quality rating for Richness.We have added an additional column to the Extended Data File 5 to identify these articles.9. Quality Assessment: In Figure 1, "records linked to included interventions (n=5)" and "relevant policy documents/guidelines (n=13)" seem to bypass the quality assessment, while they are partly assessed in Extended Data File 6 -please check for inconsistencies.Why is the quality assessment of the guidelines/policy documents "N/A"?What does mean the criterion "Richness at extraction"(in addition to "richness") in Extended Data File 6?
We did not feel the need to quality assess the policy or guideline documents as these are what is available to GPs to inform their practice and regardless of quality needed to be taken into account.The 27 articles included at full text review stage met the inclusion criteria, were assessed for relevance, rigour and richness, and scored either 3 or 4 on a four-point scale for richness.The five articles linked to articles already included were automatically included as they were regarded as having the same level of relevance and rigour as their linked articles and provided additional details which added to the richness of the data.These 32 articles were reassessed for richness at the extraction stage as more careful assessment was possible and it became obvious that the original score may have been too high for some articles as the data was extracted for each theory, the revised scores are available in Extended data file 6 under the heading "Richness at extraction".Six of the articles were downgraded to a richness rating of 2, which means that they supplied limited data of interest, but were quick to extract it and could add weight to findings.10. Discussion of limitations and implications: the authors name a few limitations with regard to their methods.How do the authors assess the effect of their restrictions (e.g.publication language and period, quality feature "richness" as exclusion criterion) and limited resources (screening, data extraction and coding could mostly be done by only one person, the panel survey could not be tested a priori) on the results?Would there be a need for further research to support the results of the RRR?
The restrictions you list are relevant to conventional systematic reviews but do not apply to the same extend to realist reviews.We have added some text to the introduction to highlight that the two methodologies are different."A systematic review provides evidence in relation to the efficacy of an intervention, but does not identify why the intervention worked, or not.There are notable differences in the methodology of conventional systematic reviews and realist reviews.A realist review is iterative and non-linear and much more likely to make use of grey literature rather than relying solely on articles in academic journals.In addition, the evidence in a realist review is appraised to assess the weight of its contribution to refining the developed theories."It is not unusual for a realist review to be limited to the language of the researchers as all the nuances of the data need to be clearly understood.The restriction of the review to articles published between 2011 and 2021 was seen as a positive, ensuring the findings were relevant to current practice.The section on strengths and limitations does acknowledge that the time available to undertake this rapid realist review may have impacted on the range of articles included.Decisions regarding how to focus the review based on the time available again are an accepted process in realist reviews.We chose to limit the review to those articles with the richest data to get the most from the time available.The need to have two independent screeners for all the articles is not essential, as long as the main screener/extractor is knowledgeable about the topic of the realist review, has spent time with the rest of the research team and the panel of experts agreeing the theories in advance, which formed the basis of the extraction process and recorded in a transparent way the steps taken.However, we agree, this should be stated in the limitations.We had added a sentence to this effect "The time limitations for this rapid review also meant that although two reviewers screened all relevant articles by title and abstract, there was only one reviewer for the full text reading of the articles, 10% of the excluded articles at this stage were reviewed by the 2nd reviewer.To reduce the potential subjective impact of this approach, team discussions, pilots and detailed inclusion and exclusion criteria guidance documents were developed, and a transparent record of the steps taken were recorded." Yes, there is a need for further research.The product of any realist review are theories based on available evidence, but are simply theories, which must be tested in future research to agree, refine or reject them.This would involve applying the theories to the design and implementation of an intervention and evaluating the results.We state this in the conclusions.Thank you for the invitation to review this rapid realist review, which aims to identify and examine the contexts and mechanisms involved in outcomes relevant to the reduction of adverse drug events in community dwelling older adults in the general practice setting.This is a very important and complex topic, and one for which context is often overlooked.
The paper represents a tremendous amount of work and is very well written.The background clearly outlines why the topic is important, and why a rapid realist review is the most appropriate methodology to use to provide answers to the research question.
Robust realist methodology is used throughout, with input from stakeholders i.e. general practitioners to develop theories.The iterative nature of theory development is particularly valuable as the studies identified during the review were used to modify these as necessary.
Results are presented clearly and logically.Guidance is provided for those wanting to develop interventions in general practice to reduce the occurrence of adverse drug events in community dwelling older adults.The results presented are contextualized for the Irish setting, but are of international relevance and importance.Indeed, the results presented could be extrapolated and applied to develop interventions in general practice focusing on other clinical areas and patient groups.The paper concludes that easily applied and well-resourced interventions, supported by national policy, could enhance the success of interventions aimed at reducing adverse drug events for community dwelling older adults in primary care in Ireland.

Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes
Are sufficient details of the methods and analysis provided to allow replication by others?

Is the statistical analysis and its interpretation appropriate? Yes
Are the conclusions drawn adequately supported by the results presented in the review?Yes Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Intervention development, theory, appropriate prescribing, polypharmacy I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
decision-making could help GPs to elicit patient preferences.Schuling 2012

Competing Interests:
No competing interests Reviewer Report 07 September 2022 https://doi.org/10.21956/hrbopenres.14833.r32547© 2022 Ryan C.This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Cristin Ryan School of Pharmacy and Pharmaceutical Sciences, Trinity College Dublin, Dublin, Ireland

Table 5 . Quotations from the articles to support the themes.
The MRC Framework] ensured that the intervention was developed using the best available evidence, was acceptable to GPs and feasible to deliver in the clinical setting."Clyne 2013 her other medications because she was in front of me and I had a bit of time to do it.' P5." Clyne 2013GPs also often criticize specialists' lack of a holistic or geriatric view on elderly patients.Compared with the GP, they know much less about the patients concerning comorbidities, established medications or other specifics (e.g.medication sensitivity, changed metabolism) and may, therefore, consider risks and benefits less.Pohontsch 2017 "[GPs] feelings with regard to their management of the problem [Deprescribing] ranged from moderate optimism to something close to despair."Schuling 2012 Patient Knowledge/ Education "… it's [information leaflet] a good way of helping people, it's a good negotiating thing, here's the information…" P6" Clyne 2013

Table 6 . Facilitators to making the most of GPs time. Facilitators to making the most of GPs time
Making the most of existing opportunities to implement new processes, i.e. disease management programmes or yearly check-ups Identifying the patients most at risk Involving, delegating or expanding the role of other staff members (i.e.practice care assistants, receptionists, nurses) the intervention simple Keeping any training local, short, and in workshop style, with opportunities to share experiences Practical, easy to use resources and tools Use of technology/standardised software packages to record and share data

Table 8 . Recommendations for contents and implementation of guideline and policy documents. Recommendations for contents and implementation of guideline and policy documents
h) Set out additional costs or savings i)Identify a "practice champion" to motivate others to use the documents j) . Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).thecontexts and mechanisms that play a role in reducing adverse drug events (ADE) in community dwelling older adults (CDOA) in the GP setting that could inform the development of interventions in Ireland.Six candidate theories in relation to how interventions are expected to work were based on knowledge of the field and recent literature including a feedback from a reference panel of seven Irish GPs formed the search strategy.In an iterative approach, screening and refining of inclusion and exclusion criteria took place resulting in a detailed operationalization of broad inclusion and exclusion criteria with regard to participants, setting, and objective.Restrictions were applied with publication date (2022 to 2021), publication language (English only) and accessibility (either open access or RCSI database availability).The selection of articles was conducted by one reviewer with a quality control of 20% of the excluded articles by a second reviewer.Quality assessment with regard to relevance, rigour and richness was refined after a rating of 10% of the included articles by two reviewers and conducted by one reviewer for all articles.Poor richness (scoring less than three points at a four-point Likert scale) led to the exclusion of an article.The extracted data on subthemes, contexts, mechanisms, outcomes and some intervention details were based on the candidate theories and linked to codes that allowed explaining how, why and in which context an intervention might work, which was also expressed through context-mechanism-outcome configurations (CMOCs).In total, 27 articles (resulting from the systematic search in six electronic databases) plus 18 documents achieved from additional (hand-)searches were included which provided rich data for 3/6 pre-specified theories (THEORY 1: Engagement of GPs in interventions, THEORY 2: Health and clinical guidelines or policies, and THEORY 5: Shared decision making), insufficient data for Continuity of care (theory 3) and Health information technology (theory 4) and only some data for a Collaboration with community pharmacists (theory 6).Theory 6 was rejected as existing barriers in the Irish setting made it impractical to use.The authors concluded that interventions need to be relevant, easy to apply in practice, supported by national policy and adequately resourced to improve the success of GP-based interventions to reduce ADEs in CDOA in Ireland.
The study by Schuliing was about the view of Dutch GPs.Readers could misinterpret the quotation as patient's view (while patients' preferences widely vary -between persons but also within persons over time, also in CDOA: eg, https://pubmed.ncbi.nlm.nih.gov/31843855/ 1 , https://pubmed.ncbi.nlm.nih.gov/33383951/ 2 ).I'd suggest presenting a "baseline table" of included documents in the main text.Please check the supplemental documents for inconsistencies: eg.Amos et al. and Bregnhøj et al. were classified as included in Extended data File 5 but are not presented in the final list of included studies (Extended Data File 6). ○