STOPP/START interventions to improve medicines management for people aged 65 years and over: a realist synthesis

Potentially inappropriate medication in older adults

The prescription of potentially inappropriate medications to older people is highly prevalent, ranging from 12% in community-dwelling elderly people to 40% in nursing home residents in Europe and the USA. 14 Older people are particularly vulnerable to inappropriate prescribing because of their multiple drug regimens, comorbid conditions and age-associated physiological processes. 15 Drug-related problems and potentially inappropriate prescribing are highly prevalent among older adults and can exert a significant disease burden. In addition, they have been associated with adverse drug events (ADEs) (which includes ADRs and medication errors), leading to hospitalisation and death, and increased health resource utilisation. 16–18 A 2007 national population study in Ireland by Cahir et al. 19 estimated a cost of potentially inappropriate prescribing of > €38M.

Medicines in older people are considered appropriate when they have a clear evidence-based indication, are cost-effective and are well tolerated in the majority of the population. Although potentially inappropriate prescribing and ADRs are generally associated with overprescription, potentially inappropriate prescribing may also occur when a patient is not prescribed appropriate medication for the treatment and prevention of a disease or condition. 20 This might occur for many reasons, such as ageism, fear of adverse events, economic concerns and lack of prescribing knowledge. 21–23

Screening tools for medication optimisation in older adults

Detecting potentially inappropriate prescribing early may prevent ADEs and cost-effectiveness of medicines, which gives an opportunity for improving quality of care in older adults. 10,24 In addition, the quality of life of older people can be improved by discontinuing inappropriate medications. 25 Currently, prescribing and reviewing medications are largely based on clinicians’ clinical judgement for older patients. However, guidelines and systematically developed evidence-based tools have emerged in recent years to facilitate a comprehensive clinical approach to medication review. These typically consist of lists of appropriate medications for older people to predominantly avoid harmful prescriptions in the older population.

Adverse drug events and cost inefficiencies can be reduced by identifying potentially inappropriate prescribing. 10 Screening tools for supporting medicines optimisation of older adults have been developed in an attempt to address this. The most widely used are the Beers criteria for potentially inappropriate medication use in older adults and the combination of the Screening Tool of Older People’s Prescriptions (STOPP) and Screening Tool to Alert to the Right Treatment (START). Almost half of the drugs in the Beers criteria are unavailable for prescribers in Europe,26,27 several of the drugs are not contraindicated in older people, as per the British National Formulary,10 whereas other contraindicated drugs are omitted. 26 Given these shortcomings, the STOPP/START tools have recently received increased interest in European countries. 28

Development of STOPP/START tools

Two versions of STOPP/START were developed by a team of researchers at the University of Cork, Ireland. The first iteration. STOPP/START (version 1), was developed in 2008. 28 They are valid, reliable and comprehensive screening tools that enable the prescribing physician to appraise an older patient’s prescription drugs in the context of his/her concurrent diagnoses. To establish their content validity, a two-round Delphi survey was conducted with an 18-member expert panel in geriatric pharmacotherapy from academic centres in Ireland and the UK. Sixty-five of the 68 original STOPP criteria and all 22 of the original START criteria received consensus from the expert panel. Inter-rater reliability of the STOPP/START tools was also assessed by determining the kappa statistic and high levels were established between physicians and pharmacists.

Owing to the changing evidence base behind the initial version of STOPP/START tools, the licensing of important new drugs since 2008 and a long list of potentially inappropriate medications included in the first version, an updated version of the tools was deemed necessary. In addition, a number of STOPP/START criteria were no longer considered completely accurate or relevant, a small number (12 in total) of criteria were lacking in clinical importance or prevalence and there were some important criteria that were absent from STOPP/START version 1. The updated version (i.e. version 2) was developed based on an up-to-date literature review and consensus validation among a European panel of experts to reflect Europe-wide prescribing practices in the general population of older people (see Appendix 1).

Impact of STOPP/START tools

An updated recent Cochrane systematic review of the evidence base for interventions to improve the appropriate use of polypharmacy for older people concluded that it was unclear whether or not they could improve appropriate polypharmacy. 29 Although previous versions of the review highlighted STOPP/START as the basis for promising interventions,30 no specific examination was conducted of STOPP/START-based interventions.

A systematic review based on 13 randomised and non-randomised studies that was conducted up to 2012 using STOPP/START tools in community-dwelling, acute care and long-term care older patients found that its use reduced falls, hospital length of stay, care visits and medication cost. 31 A more recent systematic review of randomised clinical trials was conducted on the effectiveness of the STOPP/START tools. 12 This review identified four randomised controlled trials (RCTs). Two studies were judged as having a low risk of bias and two studies were judged as having a moderate to high risk of bias. Meta-analysis of the four RCTs found that the STOPP tool reduced potentially inappropriate medications rates, but study heterogeneity prevented the calculation of a meaningful statistical summary. There was evidence that using STOPP/START tools was associated with a reduction in falls, delirium episodes, hospital length of stay, care visits (primary and emergency) and medication costs, but there was no evidence of improvements in quality of life or mortality. The authors concluded that their use may be effective in improving prescribing quality, and in improving clinical, humanistic and economic outcomes. 12

Current National Institute for Health and Care Excellence (NICE) guidance on medicines optimisation includes the recommendation to consider using a screening tool for adults such as the STOPP/START to identify potential medicines related patient safety incidents in some groups of people, especially in relation to adults taking multiple medicines (polypharmacy), adults with chronic or long-term conditions and older people. 32 However, the evidence for this recommendation is limited. One of the research recommendations in the guideline is for research on the impact of using clinical decision support systems. The focus of the recommendation is specific for computerised systems. Even this particular application of the decision tools would benefit from knowledge on the mechanisms by which such interventions work more generally to optimise their implementation and to maximise their benefits.

Rationale for realist methodology

To date, and to the best of our knowledge, no attempt has been made to appraise evidence to explain how, when and why intervention based on the STOPP/START tools might improve medicines management in older people.

Systematic reviews can sometimes provide a picture of ‘outcome patterns’ and give indications of where the intervention was successful and where it was not. In systematic reviews, this variation is ignored and what matters is the mean effect across trials. The current evidence base for the STOPP/START tool, as reviewed by conventional methods, is limited for all types of population (especially frail older and community-living patients receiving primary care). Both a Cochrane review29 and a systematic review by Greenhalgh et al. 33 demonstrates the importance of eliciting with sufficient detail the mechanisms that lead to success of interventions and the context in which those interventions are conducted to appraise and eventually maximise their transferability. 33,34 Rather than simply assessing whether interventions based on STOPP/START tools are effective or cost-effective in a binary sense, review methods seek to explain how and why these tools are more or less effective in different circumstances or for different patient groups, thereby facilitating the development of optimal implementation strategies. This is particularly relevant at a time when improving care for older people with complex care needs, who frequently require complex medication regimens, is a priority for the NHS.

We can start to investigate the patterns in realist synthesis. Were there any contextual features that were common to the positive trials? How might we explain this? Does this give any hints as to possible mechanisms? This review is complementary to the two previous reviews,29,33 which did not have a realist approach. The review provides greater understanding and insight into how, when and why interventions work in practice, and the differences between settings, clinician types, patient groups, patient ages and shared decision-making, all of which are very important in terms of optimising implementation of the STOPP/START tools. We have also broadened our search to include studies from 2014 to 2019 and searched for evidence from non-RCTs and qualitative studies, unlike the other two reviews. In addition, we extracted the information from the relevant reports from the NHS.

In contrast to a conventional systematic review of cost-effectiveness, in which it is typically presumed that empirical evidence about the costs and cost-effectiveness of an intervention solely exists in economic evaluations or comparative cost analyses, a realist review with a cost-effectiveness review question should take an integrated approach:

• Evidence about the supposed (i.e. programme theory) and actual costs and cost-effectiveness of interventions and comparators should be sought and captured in a wider range of study types, including economic evaluations, costing studies, effectiveness studies and other study types about the interventions or programmes of interest.

• It should be acknowledged that ‘effectiveness outcomes’ and ‘economic outcomes’ are not mutually exclusive categories. Most effectiveness outcomes, whether clinical or patient reported, are also resource committing and therefore cost affecting in some way (e.g. the treatment of ADEs and the reduction in medications taken).

• The underlying mechanisms of an intervention are the consequences of providing a particular combination of resources.

Both Pawson’s original conception of programme theory35 and more recent clarifications of the realist notion of mechanism36 clearly acknowledge that intervention mechanisms are fundamentally how people, providers and clients respond to particular (usually new) resources. As well as identifying the resource consumption required for intervention mechanisms to exist, realist review must also then identify the resource changes or commitments implicit in outcomes, and any resource implications of salient contexts.

Search strategy

We carried out searches of relevant bibliographic databases to identify theories and assumptions about how the STOPP/START tools are intended to work. The search aimed to identify position pieces, commentaries, letters, editorials and critical pieces on the STOPP/START tools wherein theories and assumptions about the tool would be discussed, but not necessarily tested. The search strategy was developed using MEDLINE (via Ovid) by an information specialist (SB) in consultation with the review team. A combination of indexing terms (e.g. medical subject headings in MEDLINE) and free-text terms (i.e. title and abstract terms in the bibliographic records) were used. We included search terms that described the STOPP/START tools combined with a published publication-type search filter that included search terms such as ‘comment’, ‘letter’, ‘opinion’, ‘views’ and ‘news or newspaper article’. 38 A publication date limit of 2008 to the date of the search was applied. No English-language filter was applied.

The search strategy was finalised in MEDLINE and translated for use in a selection of relevant bibliographic databases, including:

• Cumulative Index to Nursing and Allied Health Literature (CINAHL) (via EBSCOhost)

• EMBASE (via Ovid)

• Health Management information Consortium (HMIC) database (via Ovid).

The search strategies for all databases and the number of hits retrieved per database are reported in Appendix 2, Tables 8–11. Searches of all databases were carried out on 29 October 2018. The results were exported to EndNote X8 (Clarivate Analytics, Philadelphia, PA, USA) and de-duplicated using the automated de-duplication feature and manual checking.

Screening

Documents were selected based on relevance (i.e. whether or not data could contribute to theory building). A random sample of 10% of articles were selected, read, assessed and discussed by three reviewers (JBG, JG and JMV) using a preliminary set of inclusion/exclusion criteria (Box 1).

The remaining 90% of articles were completed by the same reviewers independently. However, a number of these necessitated discussion between the reviewers, as they were pivotal papers or difficult to understand/integrate. (In realist reviews, the study itself is rarely used as the unit of analysis. Instead, realist reviews may consider small sections of the primary study, e.g. the introduction or discussion sections, to test a very specific hypothesis about the relationships between CMOs.)39 We therefore refined our studies based on the knowledge we acquired about the theory development of the impact of the STOPP/START tools. At the end of this process, we retained the papers that provided the clearest examples of the ideas underpinning STOPP/START-based interventions.

Data extraction

The realist review method obtained information by note-taking and annotation rather than standardised data extraction (using a specific software or data extraction form), as used in a traditional systematic review. Documents were examined for data that contributed to theories on how an intervention was supposed to work, which were then highlighted, noted and given an appropriate label. Three researchers (JBG, JG and JMV) initially extracted information from a small set of documents to ensure consistency and then each of them extracted information separately for a subset of papers. The information was subsequently organised by one researcher (JBG). Quality appraisal is integrated into the synthesis narrative, rather than reported separately. Formal quality assessment was not undertaken as part of phase 1.

Synthesis and strategy for prioritisation and finalisation of the candidate programme theories

The final candidate theories were summarised through narrative synthesis, using text, summary tables and a single logic model to summarise individual papers/reports and draw insights across papers/reports.

The initial set of theories on how STOPP/START-based interventions are intended to work were informed by the following steps:

• searching and identifying programme theories from the literature

• qualitative interviews with experts in the field

• ongoing engagement with PPI, including two workshops to ground the study in real-life experience.

The initial list of draft theories from the literature, findings from the interviews and refined list of draft theories from an initial meeting with the Patient Advisory Group were presented to the project team in February 2019. Following productive discussions among the team members, a narrative of each draft theory was developed during the meeting. The group was asked to reflect on the triangulation of findings, to validate theories and to prioritise those appropriate for realist review. From working with the group, we were able to discuss and produce a flow chart with emerging examples of the causal links between CMOs. After working and reworking on the flow chart, we were able to develop the first draft of the single logic model (see Appendix 3, Figure 16). The draft model was circulated to the project team, and with the subsequent feedback we produced the candidate programme theories.

Interviews with experts

To prioritise the most important or explanatory programme theories, we consulted topic experts, NHS stakeholders and a patient group. Expert consultations were achieved through ongoing conversations with experts in the field on why the STOPP/START tools work, who they work for, in what circumstances and why. This involved semistructured interviews with experts to triangulate the emerging theories resulting from the literature review. Given the scope and the realist nature of these interviews (i.e. realist hypotheses are confirmed or abandoned not through saturation but rather through relevance and rigour), we expected that approximately 10 interviews would meet our needs, although the final number was confirmed by how interviews uncovered how emerging mechanistic and contextual factors might contribute to the outcome patterns emerging from the empirical literature. 35,40

We aimed to include and recruited clinicians with a special interest in medicines optimisation, drug safety and pharmacotherapy, experts involved in medicines optimisation at NHS RightCare (Leeds, UK), providers of care in primary and secondary care [e.g. geriatricians, general practitioners (GPs) and other clinicians], care home managers, and academics and those involved in developing education and guidance for older people. In addition, researchers involved in the development of the STOPP/START tools were recruited through Academic Health Science Networks, Clinical Research Networks, the Royal Pharmaceutical Society (RPS) and through co-applicant networks. Following the advice from the board, we considered inviting British Geriatrics Society (BGS) experts. We also considered inviting contributors to the NHS England Toolkit for General Practice in Supporting Older People Living with Frailty41 and the NICE Multimorbidity Guideline Committee. 42 Ethics approval for the interviews was granted by the University of Exeter Medical School Research Ethics Committee on 19 November 2018 (reference RG/CB/18/09/181).

Interviews were based on a pre-established stakeholder topic guide (see Appendix 4) that was developed collaboratively by three of the researchers (JBG, JG and JMV) and based on the emerging findings of the literature review. Interviews were planned to last up to 30 minutes and were conducted either face to face or by telephone at the participant’s convenience. Interviews were audio-recorded, transcribed and analysed thematically by two researchers (JG and IP)43 to expand and refine our programme theories. The purpose of the expert interviews was threefold: (1) to gain input on our list of candidate CMO configurations, (2) to identify additional CMO configurations and (3) to identify additional literature and/or relevant concepts that we may have missed. This was supported by further literature searches in this phase. Interviews were conducted after the patient advisory meeting and before completing candidate theory development.

Search strategy

The search to identify studies to test the programme theory included searches of bibliographic databases, checking the reference lists of included studies, forwards citation searching of included studies and searching the websites of relevant organisations. The bibliographic database search was developed using MEDLINE (via Ovid) by an information specialist (SB) in consultation with the review team. We developed a highly sensitive search that identified all studies that mentioned the STOPP/START tools in the title or abstract fields of the databases searched. No publication-type filter or any other search terms were used. This approach was taken because the number of hits retrieved by searching for STOPP/START tools terminology was feasible to screen in full by two reviewers within the available time and resources. Therefore, the strategies were not designed to target the evidence for prioritised theories, but were very broad and aimed at identifying all empirical studies of interventions based on the use of STOPP/START tools. No indexing (e.g. medical subject headings in MEDLINE) was used, as we did not identify any indexing terms for the STOPP/START tools that were not also used for other prescribing tools. A publication date limit of 2008 to the date of the search was applied. No English-language filter was applied.

The final search strategy was finalised in MEDLINE and translated for use in a selection of relevant bibliographic databases. The full set of bibliographic databases searched included:

1. Cochrane Central Register of Controlled Trials (CENTRAL) (via Cochrane Library)

2. CINAHL (via EBSCOhost)

3. Cochrane Database of Systematic Reviews (CDSR) (via Cochrane Library)

4. EMBASE (via Ovid)

5. HMIC (via Ovid)

6. MEDLINE (via Ovid)

7. NHS Economic Evaluation Database (NHS EED) (via the Centre for Reviews and Dissemination website)

8. Social Policy & Practice (via Ovid).

The search strategies for all databases and the number of hits retrieved per database are reported in Appendix 5, Tables 12–19. Searches of all databases were carried out on 11 June 2019. The results were exported to EndNote X8 and de-duplicated using the automated de-duplication feature and manual checking.

The reference lists of included studies identified by the bibliographic database searches were manually inspected. Forward citation searching of included studies was carried out on 11 October 2019 using, primarily, Web of Science. If a study was not indexed in Web of Science, we used Scopus or Google Scholar (Google Inc., Mountain View, CA, USA) instead, depending on where the study was indexed. The results of forward citation searching were exported to EndNote X8 and de-duplicated before screening.

We searched the repository for grey literature (i.e. OpenGrey). The searches and number of hits are reported in Appendix 6, Table 20. URLs that linked to the search results were sent to the review team by e-mail for screening.

We additionally searched the websites of the following organisations on 11 October 2019 (see Appendix 6):

• Age UK (URL: www.bgs.org.uk/search/node)

• BGS (URL: www.bgs.org.uk).

Screening

In realist reviews, the study itself is rarely used as the unit of analysis. Instead, realist reviews may consider small sections of the primary study (e.g. the introduction or discussion sections, to test a very specific hypothesis about the relationships between CMOs). 39 We therefore assessed studies based on the knowledge we acquired about the theory development of the impact of the STOPP/START tools. Title and abstract screening were carried out by three reviewers (AD, IP and IRC) in Rayyan (Rayyan Software/Qatar Computing Research Institute, Doha, Qatar), with JMV resolving discrepancies (Box 2).

Data extraction

The realist review method obtains information by note-taking and annotation rather than standardised data extraction, as used in a traditional systematic review. Documents are examined for data that contribute to theories on how an intervention is supposed to work, which are then highlighted, noted and given an approximate label. The reviewer may make use of forms to assist the sifting, sorting and annotation of primary source materials, but do not take the form of a single standard list of questions, as used in a traditional systematic review. As such, we created a reference framework to guide extraction, which was intended to be a guide through the process, rather than to be applied in a routine one-size-fits-all manner. The extraction framework had three sections:

1. Standard information collected on all studies (e.g. year, author, title, study type, aims, outcomes, etc.).

2. Evidence of CMOs in studies to support the candidate programme theories.

3. Evidence in studies to refute/refine CMOs in the candidate programme theories.

The data extraction form was piloted by four reviewers (JMV, IP, JBG and AD) and approved by other project team members. The form was then used to extract the above information from all included studies by four reviewers (IP, JBG, AD and IRC). IP was responsible for co-ordinating and merging the data extracts from all the reviewers.

Quality appraisal

Quality assessment used the concept of rigour (i.e. whether or not the methods used to generate the relevant data are credible and trustworthy). Sources were classified as conceptually rich (thick) or weak (thin). This strategy has been found to be practical and useful in theory-driven synthesis, as it allows the reviewer to focus on the stronger sources of programme theories without excluding weaker sources that may make an important contribution. Quality appraisal from this perspective is integrated into the synthesis narrative, rather than reported separately.

Appraisal tools may also be used to determine rigour if appropriate. 44 We used the Critical Appraisal Skills Programme45 to further determine the rigour of the included studies. The importance of transparency in a realist review is similar to that of systematic reviews to ensure the validity, reliability and verifiability of findings and conclusions. 39,44 With the Critical Appraisal Skills Programme, study quality was considered low if half or less than half of the criteria were met, medium if more than half but less than three-quarters of the criteria were met, and high if most were met. 46 The scores were produced, summed and ranked high and medium quality for both randomised and non-randomised studies.

Data synthesis

Synthesis of the diverse sources of evidence was conducted through a structured process that included the following approaches:

1. Juxtaposition of sources of evidence (e.g. where evidence about the STOPP/START tools from one source enables insights into evidence about outcomes from another source).

2. Reconciling of sources of evidence (i.e. where results differ in apparently similar circumstances, further investigation is appropriate to find explanations of why these different results occurred).

3. Adjudication of sources of evidence (on the basis of methodological strengths or weaknesses). 47

4. Consolidation of sources of evidence (i.e. where outcomes differ in particular contexts, an explanation can be constructed of how and why these outcomes occur differently).

Based on the standardised information extracted for each study, each intervention was assigned to one or more mechanisms by consensus from two of the reviewers (IP and JMV).

For each causal link leading in the logic models, we summarised the number and quality of studies supporting, refuting or qualifying it. We first categorised the link’s evidential support in each study as providing:

• no evidence (where the study did not provide any empirical evidence on the link)

• positive evidence (where there were studies providing empirical evidence supporting the link in the direction predicted by the logic model)

• negative evidence (where there were studies providing empirical evidence on the link but not in the direction predicted by the logic model or the associations did not reach statistical significance).

We further categorised the strength of each causal link’s evidential support across all studies as:

• positive evidence [i.e. having a substantial majority of positive evidence studies (> 70%)]

• mixed evidence (i.e. having both positive and negative evidence studies and both categories exceeding 30% of studies)

• negative evidence [i.e. having a substantial majority of negative evidence studies (> 70%)]

• no evidence (i.e. all studies classified as providing ‘no evidence’).

How is the underlying problem conceptualised?

Chronic diseases and multiple medical conditions prevail in older people; therefore, there is a high prevalence of polypharmacy, which has been associated with negative health outcomes. 5 Lack of awareness of pharmacokinetics and pharmacodynamics changes in older people can contribute to inappropriate medicine use, resulting in adverse drug effects. Some of these reactions may be confused with progression of the given pathology or with some typical age-related syndromes. 48

The risks of potentially inappropriate medications outweigh their benefits, especially when a safer or more effective alternative therapy is available. The use of potentially inappropriate medications is considered a major problem among older people, and may contribute to increased risk of adverse drug effects and to developing drug–drug and drug–disease interactions. 19

Patients are often treated by different clinicians following either an outpatient visit or an acute admission, and the consultant may not be aware of the existing medication that a patient may be on. Most often, these health conditions are treated in general practice, which results in the attribution to GPs of the responsibility of whether or not prescribing has taken into account the patient’s clinical history.

Role of the professionals in prescribing: who does what?

Professional boundaries play a key role in prescribing practices. There is an implicit assumption that clinicians with expertise in older people (e.g. geriatric pharmacotherapists) will be more aware of medications that are inappropriate for older people. There are different perspectives on the onus being on GPs or consultants to ensure that they prescribe drugs that do not lead to drug–drug interactions or ADEs, compared with pharmacists picking these issues up during a review of medications. Furthermore, if GPs identify these problems, it remains to be determined if and how GPs can change/challenge the prescribing of consultants, or reflect on their own prescribing practice or that of their colleagues. Finally, if and when pharmacists pick these up, it is unclear if (and, if so, how) pharmacists can challenge or modify prescribing by GPs or consultants.

These challenges operate within existing sets of expertise, relationships, professional boundaries and remits between GPs, consultants, pharmacists and nurse prescribers. These play out on a ‘macro/meso level’ in terms of national/professional norms that exist about the remit of different practitioners and the expertise of specialists compared with generalists, but is also subject to local variations in the ways in which these different clinicians work together. They are also likely to be influenced by the existing relationship and knowledge a GP or a consultant has with a patient (and therefore their knowledge about their history, etc.). These sets of relationships form an important background against which the STOPP/START tools are used.

Context of a STOPP/START medication review

To address these problems, current NICE guidance on medicines optimisation includes as a recommendation to consider using a screening tool for adults such as the STOPP/START to identify potential medicines related patient safety incidents in some groups of people, especially in relation to adults taking multiple medicines (polypharmacy) and adults with chronic or long-term conditions and older people. 32 Medication reviews can take several forms and vary in terms of their comprehensiveness, purpose and whether or not the patient is present during the review. Medication reviews could be carried out by the prescribers themselves or by an independent reviewer (usually a pharmacist). However, in the immediate context of the medication review, the process of the review might position them differently, either as someone reflecting on their own prescribing and making changes to their own practice, or as someone who is reviewing another clinician’s practice or decisions and therefore attempting to change their practice. Most of these patients are seen by a number of different physicians/prescribers, each of them with a partial knowledge of the patient, the patient’s problems and the medications. This makes it unlikely that anybody could use it for reviewing their own prescribing. The only exception (certainly at a stretch) would be the GP. This has a number of implications for thinking about the context in which the STOPP/START tools are used, how we conceptualise any programme theories and the kinds of higher-level theories we might draw on to think about it:

• It changes the nature of the intervention and the underlying assumptions about how it is intended to work (i.e. it changes the programme theories).

• It has implications for the broader theories that might come into play to think about the relevant contextual factors that might shape the intervention. For example, if it is about reviewing and making changes to someone else’s prescribing decisions or behaviour, theories about interprofessional boundaries come into play, and if it is about reviewing own practice then audit and feedback theories might come into play.

Some types of review are carried out in the absence of the patient (e.g. prescription reviews), whereas others are carried out with the patient present (e.g. a ‘clinical medication review’). Therefore, the type and nature of the medication review will also form an important context that might shape how STOPP/START tools work, especially in relation to some of the programme theories around patient-centred care. Several pieces of guidance advocate the use of the STOPP/START or other tools to assist in these processes. Exactly how they are intended to assist and, therefore, the underlying programme theories about their use vary.

What resources do STOPP/START tools offer?

The STOPP/START tools provide a comprehensive list of medications that clinicians should consider stopping in older people generally or, for some medications, in older people with certain conditions. In addition, they provide a list of medications that clinicians should consider starting if certain conditions are met.

Positive programme theories about how they are expected to work are as follows:

• The STOPP/START tools provide a systematic and structured way of carrying out the medication review process (resource). The assumption is that clinicians will use the tools to structure the way in which the review is carried out [i.e. the tools could either replace the usual way in which the task is carried out or the tools might somehow integrate the structure provided by the tool with usual practice (mechanism)]. The use of a more structured approach is expected to enable the clinician to identify a larger number of potentially inappropriate medications, ADRs or unmet need than if clinical judgement had been used alone (outcome).

• As the STOPP/START tools are evidence based, comprehensive and structured, the assumption is that they can be used with little need for clinical judgement (mechanism) and, therefore, can be used by a range of clinicians who may not be specialists (context) in older people to identify potentially inappropriate medication. This in turn increases the capacity for medication reviews to take place, rather than just to rely on clinicians who are specialists in older people.

• The identification of potentially inappropriate medications using the STOPP/START tools will prompt the clinician to start or guide discussion with the patient about whether they wish to stop or start particular medications and what their priorities are (mechanism), which is expected to result in care that is more patient centred (outcome). However, this is likely to depend on whether or not the patient is actually present at the review (context).

These theories envisage STOPP/START tools as screening tools and a means of identifying problems. However, to work this way, this assumes a number of contextual features being present or that certain mechanisms will trigger these features. These include the following:

• The clinician undertaking the review either knows about, or is able to access information about, the patient’s existing conditions and existing medication.

• The clinician undertaking the review agrees with the criteria set out in the tool (i.e. they do not clash with clinical judgement).

• The clinician undertaking the review has the time to go through the entire tool systematically.

• Clinicians will know how to act when potentially inappropriate medications have been identified and that the clinician will discuss this with the patient.

We identified a number of opposing theories about STOPP/START tools that are expected to work:

• The STOPP/START tools offer resources that enable clinicians to identify potential medication problems, it does not provide resources about the context of the individual patient in terms of how these criteria apply to specific patients. Therefore, the clinician needs to integrate their clinical judgement, information gathered from a discussion with the patient and the alerts from the STOPP/START tools (mechanism) to make a decision about whether to stop or start any medicines (outcome). This challenges the idea that the tool can be used without the need for clinical judgement and raises questions about the degree of specialist knowledge that is needed.

• Some authors have argued that one of the reasons why a clinician’s use of the tool may not enable them to identify as many potentially inappropriate medicines (outcome) is because clinicians do not use it ‘rigorously’, implying that clinicians ignore some of the alerts identified through the tool and therefore do not follow through on stopping medication when the tool guidance identifies it as potentially inappropriate (mechanism). This theory implies that the tool is intended to act not just as a means of identifying problems, but as a decision support tool whereby the tool acts as an instruction to decisions (i.e. the guidance should be followed, not just considered). This implies that clinical judgement is not required.

These two different positions highlight the contested nature of the tools and their role and purpose vis-à-vis clinical judgement and patient input. One set of theories position the tools as something that provides advice or alerts to clinicians, whereas the other set of theories position the tools as decision support tools:

1. Paper versions of the tools (resource/context) are time-consuming to use and require familiarity with the content of the tools (context). As many clinicians do not have time or are not familiar with their content (context), they do not use the tools or skim over them (mechanism) and this means that potentially inappropriate medications are missed (outcome).

2. Clinicians may not have all the information to hand (context) and so they are unable to use the tools to carry out the review with the full information (mechanism).

Potential solutions to these two problems are as follows:

• STOPP/START tools should be used electronically (context) and the clinician using the tool can draw on information in a patient’s electronic health record to carry out the review.

• It may be possible to design an algorithm based on the tools that carries out the review automatically (context) and simply presents the clinician with an alert about potentially inappropriate medications (resource). The clinician is then expected to accept and trust this alert and follow through on the advice accordingly (mechanism).

We might also consider a further set of theories a bit further down the implementation chain about how, when and why clinicians do actually make prescribing changes based on the use of the tool. This might depend on who uses the tool (e.g. a pharmacist), their remit to instigate changes and who actually makes the prescribing changes (e.g. GP, consultant) and the roles, remits and relationships between these different social actors. For example, if a pharmacist uses the tool to conduct a review, they may only be able to advise what changes might need to be made to either the GP or the consultant, who may choose to follow this advice or may choose to ignore it.

We further discussed the initial theories that arose from the literature and developed the succinct list of theories (Table 2).

Familiarity with STOPP/START tools

Some participants were very familiar with the STOPP/START tools and others had become familiar with it only recently. Participants reflected on how familiar different professional groups were with the STOPP/START tools. Some professionals (particularly GPs) seemed not to be familiar with the STOPP/START tools at all.

What are STOPP/START tools and what is their purpose?

Participants expressed mixed views about the role of STOPP/START tools in the decision-making process, such as whether patients should be involved in shared decision-making or if decisions should be solely the decision of the health-care professional. Some participants considered the STOPP/START tools as prompts and others considered the tools as checklists. In addition to supporting decision-making, some participants said that they can be used for improving prescribing appropriateness and some others said that they can be used as an education tool. Participants also discussed the need for using clinical knowledge and clinical judgement when applying the STOPP/START tools.

Target patients

Participants described which patients should be targeted, and proposed that those with multimorbidity and multiple medications might benefit most from the use of STOPP/START tools:

. . . but often, because of the cognitive problems with many older patients, they themselves may not be able to engage positively or negatively because of their cognitive problems. But that’s why we need to have a look at their medications to see if those medications are potentially causing cognitive problems or worsening the cognition, or if they’re going to survive long enough to derive any benefit from these drugs.

Interviewee 6, geriatrician

Appropriate setting

Participants expressed different views on the specific settings that STOPP/START tools would be applicable to, and considered primary care, hospitals, care homes and nursing homes as potential settings where they could be of value:

I very much feel STOPP/START feels like a hospital tool. It feels like the sort of tool you can use if you’ve got a fair bit of time with the patient, you are specifically seeing them for the review, and you’re going through the body system.

Interviewee 5, GP

In care homes, in – yeah in the elderly and the [inaudible 17.55] morbidity really because those are the ones that are most of risk of interactions and unnecessary drug side effects.

Interviewee 8, GP

One showing that in a nursing home environment that you can reduce the risk of falls the risk of hospitalisation over a 6-month period by intervening with STOPP and START criteria through multidisciplinary discussion and review of medications.

Interviewee 6, geriatrician

Professionals

Participants had mixed views about their role in the use of STOPP/START tools. Some participants said that they can be used by doctors (e.g. geriatricians or GPs), nurses [e.g. clinical nurse specialists, nurse consultants in geriatric medicine or practice nurses (diabetics/respiratory clinics)] or pharmacists (e.g. general practice pharmacists, hospital pharmacists or pharmacy technicians). Most of the participants proposed that a pharmacist as part of a multidisciplinary team and with access to the patient’s clinical history is the ideal professional to use STOPP/START tools.

It appears that the purpose of the tools is perceived by some as core to the scope of the practice of geriatricians. Some participants also believed that less experienced professionals could benefit more from using them than more experienced professionals, who may not benefit from using them at all. This may particularly be the case for geriatricians:

I think it can have a big impact. I mean the first example here is what we do so you know independent medication reviews – we use the patient notes and we have the patient there in a face-to-face encounter and we can make recommendations and follow them up so you know the STOPP/START interventions can have a huge impact on that as sort of guidance and I think, useful for people who are less experienced in medication review.

Interviewee 9, pharmacist

I think hospitals and geriatricians and, say, clinical nurse specialists, or nurse consultants in geriatric medicine, I think they have a role to play here as well. I think often GPs are very overburdened and they’re not geriatricians. GPs have to know about paediatrics, obstetrics, psychiatry, everything. I actually think that this kind of complex problem, perhaps geriatricians should have much more of a role and there should be much more of a demand for them.

Interviewee 6, geriatrician

I think the context of where you use it, I think contributes to the success of the tool because medication patient records are not available for example, at say the community pharmacy or in an emergency department or anything like that. I think there are issues around how the best places it’s got to be used or it can be used successfully.

Interviewee 10, pharmacist

Resources

Participants discussed various resources that would be needed for ensuring that STOPP/START tools would have the expected impact, including training, time, staff availability and costs.

Impact of the STOPP/START tools

Several participants expressed views about the degree of potential impact of using STOPP/START tools. Some participants were positive about their potential, whereas others reflected on several negative impacts.

Challenges, barriers and facilitators for using the tools

Several challenges, barriers and facilitators were identified, including effective patient engagement, the user-friendly nature of the tools (or lack thereof), availability of electronic formats, accessibility and integration of the tools into existing systems.

Suggestions for implementation

Most of the participants were very keen to have automated, computerised STOPP/START tools, but they also raised concerns about the problems with implementation. Some participants felt that there needs to be a more sophisticated system to prioritise actions resulting from using the tools:

I think there is always a risk of information or alert overload, so I think perhaps prioritising intervention, to say OK which are the most dangerous or top 10 most risky, instances of potentially inappropriate prescribing and in fact that’s what we are doing with this study that we are just in the process of completing, we have identified the top 10 most common, well not just top 10, we have ranked the STOPP/START criteria in order of prevalence in our population . . .

Interviewee 7, GP

Personalisation of the medication review

Personalisation of the medication review (Figure 2) refers to the degree to which the medication review processes and outcomes are tailored to each individual patient. STOPP/START tools are being promoted to support shared decision-making (taking into account patient preferences, experiences and expectations) (mechanism). For STOPP/START-based interventions to achieve the hypothesised improvement in patient awareness of the medication review process, adherence to the therapeutic plan, satisfaction with care, patient empowerment and quality of life, and to decrease patient burden and support a patient safety culture (outcomes), patients and/or carers need to be present and engaged in the process, and there should be sufficient time to carry out the review [i.e. variation in intervention delivery (context)].

FIGURE 2.

Logic model for personalisation.

The full set of related key variables (context and outcomes) was included in a logic model (see Figure 2). Note that, in Figures 2–4, the distance along the horizontal axis between any variable and the corresponding mechanism represents the expected degree of the association along the proximal–distal continuum. For example, individualisation of management is expected to trigger a cascade of effects that may result in improved patient empowerment, which can trigger an improvement in patient satisfaction, which can result in increased adherence and eventually modify the safety culture of patients. The link between individualisation of management and the latter outcome is expected to be more tenuous than between individualisation of management and patient empowerment.

Based on discussions and feedback from the PPI advisory group, we prioritised the following theories for phase 2:

• STOPP/START tools lead to greater adherence to pharmacological treatment when patients are present at their medication review.

• STOPP/START tools increase patient empowerment when patient perceptions and expectations are taken into account (elicited and acted on).

Systematisation (of medication review processes)

Systematisation (of medication review processes) (Figure 3) refers to the degree to which the medication review is formalised and systematically implemented. STOPP/START tools support a standardised/systematic approach for medication reviews (mechanism). For STOPP/START tools to change professional and organisational culture, burden and costs, and for patients to be aware that a review has taken place (outcomes), there needs to be a delivery system (paper or computer based) that is linked to a patient’s medical records and information from previous reviews using STOPP/START tools. Information must also be shared between relevant professionals, teams and organisations [i.e. variation in intervention delivery (context)].

FIGURE 3.

Logic model for systematisation. IT, information technology.

Based on discussions and feedback from the PPI advisory group, we prioritised the following theories for phase 2:

• STOPP/START tools change professional and organisational cultures when there are computerised systems linked to medical records, access to previous STOPP/START tools data, and data-sharing takes place (with relevant professionals, teams and organisations).

• STOPP/START tools reduce professional and organisational burden in medication management when there is sufficient familiarity with the tool.

Implementation of evidence

Implementation of evidence (Figure 4) captures the degree to which an intervention oriented to support the medication review focuses on promoting the dissemination and use of the best-available evidence. Delivery of STOPP/START-based interventions is based on implementation of evidence (mechanism). For STOPP/START-based interventions to result in greater appropriateness of prescribing/deprescribing, reduction in adverse outcomes and hospitalisations, improvements in clinical outcomes, mortality and quality of life, along with having an impact on costs, health-care use, professional burden and organisational culture (outcomes), the supporting evidence base needs to be actively promoted by people in leadership roles and requires training of relevant clinicians (e.g. GPs, pharmacists, geriatricians or nurses) [i.e. variation in delivery of the intervention (context)].

FIGURE 4.

Logic model for the implementation of evidence.

Based on discussions and feedback from the PPI advisory group, we prioritised the following theories for phase 2:

• STOPP/START-based interventions are associated with increased appropriateness of prescribing/deprescribing when there are lower levels of familiarity with the evidence base (i.e. non-geriatricians, non-pharmacists and more junior professionals).

• STOPP/START-based interventions reduce adverse outcomes through appropriate prescribing/deprescribing.

• STOPP/START-based interventions can change professional culture in settings less familiar with the supporting evidence base.

Personalisation of the medication review

All the available evidence used in testing this mechanism relied on configurations including at least another mechanism, and in half of the studies both of the other two mechanisms. None of the included studies assessed interventions targeting this mechanism in isolation.

The single logic model for first candidate programme theory linked mechanism personalisation of care with the following specific outcomes: patient awareness, empowerment, satisfaction and adherence. Although as many as 16 studies1,3,4,6,8–10,12,19,21,31,33,35–37,39 reported an intervention relying on this mechanism, only two studies49,50 considered a relevant outcome in the personalisation logic model. Both of these studies evaluated interventions focusing on the use of both STOPP and START by pharmacists and were designed around both the personalisation and implementation of evidence mechanisms in non-randomised designs. The study by De Bock et al. 49 found positive evidence for improved patient satisfaction in a geriatrics hospital department, whereas the study by Twigg et al. 76 found positive evidence for improved adherence to medical treatment only in the intervention tested in community pharmacies (Figure 7).

FIGURE 7.

Outcomes linked to personalisation across all studies. Light blue shading indicates positive evidence.

The study by Coronado-Vázquez et al. 53 also provided evidence for an association between adherence and appropriateness. Although this was not part of the personalisation logic model, it would support the synergistic effect of interventions using both the mechanisms of personalisation and implementation of evidence.

The two prioritised theories for this mechanism were as follows:

1. STOPP/START tools lead to greater adherence to pharmacological treatment when patients are present at their medication review.

2. STOPP/START tools increase patient empowerment when patient perceptions and expectations are taken into account (elicited and acted on).

It was not possible to test either of these theories, as there was only one study assessing adherence and there was no study assessing patient empowerment.

Systematisation of the medication review process

The single logic model for first candidate programme theory linked mechanism systematisation of care with the following specific outcomes: professional and organisational culture, burden on patient and on professionals, and patient awareness of the medication review process. None of the studies of interventions using this mechanism examined the impact on these outcomes and we therefore could not find any evidence in support of the link between the proposed mechanism and any of the outcomes (Figure 8).

FIGURE 8.

Outcomes linked to standardisation across all studies. IT, information technology.

The two prioritised theories for this mechanism were as follows:

1. STOPP/START tools change professional and organisational cultures when there are computerised systems linked to medical records, access to previous STOPP/START tools data, and data-sharing takes place (with relevant professionals, teams and organisations).

2. STOPP/START tools reduce professional and organisational burden in medication management when there is sufficient familiarity with the tool.

It was not possible to test these theories, as there was no study assessing professional and organisational cultures nor professional and organisational burden in medication management.

In the study by Price et al. ,50 five physicians provided feedback on their experience with an intervention that focused on the delivery system (i.e. information technology, namely computer alerts). The feedback was oriented mostly at the evaluation of the proposed software solution. More general comments referred to the location on screen and the workflow, which were perceived to be barriers. STOPP alerts had been displayed in a separate location to simple drug alerts, which often meant that the user would need to tab between screens and refresh screens. Participants preferred a single location for all medication-related alerts, regardless of the logic behind those alerts.

Implementation of evidence

The programme theory related to implementation of evidence had a wider set of outcomes than the other mechanisms. More proximal outcomes included appropriateness of prescribing/deprescribing, burden on professionals and safety culture, which would mediate impact on intermediate outcomes [i.e. clinical outcomes, adverse outcomes (ADRs, falls and hospitalisations)], and more distal outcomes (i.e. health-care use, costs, quality of life and mortality).

Nineteen studies2,3,5–8,11–13,15,18–20,24,28,29,31,32,37 looked at using STOPP/START tools to improve appropriateness of prescribing/deprescribing (all of them, except for four, with mechanism configurations consistent with the logic model for implementation of evidence outcomes). Fourteen studies3,5,7,8,11–13,15,18–20,24,29,37 found positive evidence (14/19, 74%), whereas the five other studies4,11,12,26,27 provided negative evidence (Figure 9).

FIGURE 9.

Studies focusing on appropriateness of prescribing/deprescribing. Bold text, hospital; dark blue text, negative evidence; italicised text, primary care centre; light blue text, positive evidence; and regular text, nursing home.

Studies providing positive evidence had explored interventions using STOPP/START or only STOPP with a range of different configurations of mechanisms [the majority (14/19)3,5,6,8,11–13,15,19,24,29,31,32,37 relied on at least two mechanisms]. The studies had all been conducted in hospitals, except for two studies conducted in primary care17,28 and one study conducted in a nursing home,20 had used a range of research designs and were mostly considered as being of high quality.

No differences were found in these studies between impact on either appropriateness of prescribing or appropriateness of deprescribing. The five studies2,6,28,31,32 that found negative evidence had all been conducted in hospitals, used either STOPP/START or only STOPP, had implemented at least two mechanisms and also used both STOPP and START, had used both non-randomised and randomised designs, had medication appropriateness as the primary outcome and had similar quality ratings. The three studies2,17,20 that studied interventions targeting this mechanism in isolation provided mixed evidence in support of the link.

Three studies11,26,27 examined the impact of STOPP/START-based interventions on ADEs (Figure 10). None of them studied the implementation mechanism in isolation. They all relied on both the implementation of evidence and the systematisation mechanisms, but not the personalisation mechanism, and they all used a randomised design and were of high quality. All three studies found positive evidence of reduction in ADEs.

FIGURE 10.

Studies focusing on adverse outcomes: ADEs. Light blue text, positive evidence.

Five studies11,15,35,37,38 looked at using either STOPP/START tools or STOPP to reduce falls (Figure 11). None of them studied the implementation mechanism in isolation. Two studies took place in nursing homes, two studies took place in hospitals and one study took place in a community pharmacy. The studies relied on both the implementation of evidence and the systematisation mechanism, but not the personalisation mechanism, and used a range of research designs. Three of the studies11,15,35 provided positive evidence (60%), whereas the other two studies37,38 provided negative evidence for this link. There were no differences in research design and quality between studies providing positive evidence and those providing negative evidence, except for all negative studies focusing on falls as a secondary rather than a primary outcome.

FIGURE 11.

Studies focusing on adverse outcomes: falls. Dark blue text, negative evidence; light blue text, positive evidence; and underlined text, community pharmacy.

Eight studies4,9,11,12,15,26,27,37 examined the impact of using either STOPP/START tools or only STOPP on health-care utilisation (all eight studies used mechanism configurations consistent with the logic model for this outcome, except for one study, and none studied the implementation mechanism in isolation) (Figure 12). The studies used a range of randomised and non-randomised research designs. The studies had all been conducted in hospitals [with a single exception (nursing home)] and all of them were considered to be of high quality (with a single exception). The majority of studies4,9,11,12,26,27 provided negative evidence (6/8, 75%). One of the studies that provided positive evidence was conducted in a nursing home15 and another study37 was considered to be of medium quality. Both of these studies used the same configuration of mechanisms, whereas the four other studies4,11,26,27 provided negative evidence.

FIGURE 12.

Studies focusing on health services utilisation. Dark blue text, negative evidence; light blue text, positive evidence; bold text, hospital; regular text, nursing home.

Four studies11,25,35,37 examined the impact of STOPP/START-based interventions on quality of life (all of them using a mechanism configuration consistent with the logic model for the outcome, except for one study) (Figure 13). Two of the studies11,37 took place in hospitals, one study took place in a primary care centre25 and one study took place in a community pharmacy. 35 Three studies25,35,37 provided positive evidence for this link and one study11 provided negative evidence. Two studies35,37 provided positive evidence for STOPP/START-based interventions using non-randomised designs, whereas the third study25 was a RCT that had evaluated only STOPP. All four studies11,25,35,37 were considered to be of medium quality. The only study11 providing negative evidence was a single-site RCT considered to be of high quality.

FIGURE 13.

Studies focusing on quality of life. Dark blue text, negative evidence; italicised text, primary care centre; light blue text, positive evidence; and underlined text, community pharmacy.

Five studies9,12,27,37,38 looked at using either STOPP/START tools or STOPP to reduce mortality, all of them as a secondary outcome and all of them using configurations aligned with the logic model (except for one study). None of the studies considered the implementation mechanism in isolation (Figure 14). The studies had been conducted in hospitals, except for one study38 that was conducted in a nursing home. The studies used both randomised and non-randomised designs and were considered to be of mostly high quality. All five studies9,12,27,37,38 provided negative evidence for the impact of STOPP/START-based interventions on mortality.

FIGURE 14.

Studies focusing on mortality. Dark blue text, negative evidence; bold text, hospital; and regular text, nursing home.

The evidence on the economic impact of STOPP/START-based interventions is reviewed separately in Chapter 6.

In summary, we were able to confirm pathways from the proposed mechanisms to appropriateness of prescribing/deprescribing, ADEs and quality of life. We found mixed evidence for the impact on falls and we found consistent evidence of lack of impact on health-care use and mortality (Figure 15).

FIGURE 15.

Outcomes linked to implementation of evidence across all studies. Light blue shading indicates positive evidence; orange shading indicates negative evidence; light orange shading indicates mixed evidence; and white indicates no evidence available.

The three prioritised theories corresponding to this mechanism were subsequently tested.

Hospitals

Twenty-five studies49,52,54–58,60,64,66–70,72–74,77,78,80,82,83,85,87 were conducted in hospitals. The most common type of intervention was for a pharmacist to administer the STOPP/START tools and for a physician to implement recommendations.

None of these studies assessed outcomes linked to the personalisation mechanism (i.e. patient empowerment, satisfaction, adherence and safety culture), neither did they assess outcomes linked to the systematisation mechanism (i.e. professional and organisational safety culture, professional and organisational burden and patient awareness of the medication review). All the evidence provided in these studies focused on outcomes linked to the implementation of evidence mechanism. They provided positive evidence of improvement in medication appropriateness (10/14)5,8,9,11–13,19,24,29,37 and reduction in ADEs (3/3),11,26,27 mixed evidence on improvements in falls (1/2)11 and quality of life (1/2),37 and negative evidence of any positive impact on the use of health services (6/7),4,9,11,12,26,27 and mortality (4/4). 9,27,37,38

Nursing homes

Nine studies59,61,62,65,75,79,81,84,88 were conducted in nursing homes. The studies took place in Spain (n = 4),61,79,84,88 Norway (n = 1),59 France (n = 1),62 Serbia (n = 1),65 Portugal (n = 1)75 and the Netherlands (n = 1). 81 There was a fairly even split in the administration of STOPP/START tools by pharmacists and physicians, with one study61 utilising automatic administration via software. As with studies carried out in hospital settings, there was a trend for recommendations to be implemented by physicians, although it was more common for multidisciplinary teams to be involved.

Studies in nursing homes provided positive evidence of the impact on prescribing/deprescribing appropriateness (2/2)15,20 and service utilisation (1/1),15 mixed evidence of positive impact on falls (1/2) and quality of life (1/2), and negative evidence of any positive impact on mortality (1/1). 38

Primary care

Five studies50,51,53,63,86 were conducted in primary care centres. The studies took place in Spain (n = 3),51,53,86 Canada (n = 1)50 and France (n = 1). 63 Typically, GPs both administered STOPP/START tools and implemented the information/recommendations. A study by Campins et al. 51 was the exception, in that an experienced clinical pharmacist evaluated all drugs prescribed using the STOPP/START tools, then discussed recommendations with the physician, which were then discussed with the patient. The studies by Gibert et al. 63 and Naveiro-Rilo et al. 86 were notable in that GPs were given specific training in using the STOPP/START or STOPP tools. In the Naveiro-Rilo et al. 86 study, doctors were trained on the use of the tools over two sessions.

Studies in primary care provided positive evidence of the impact on quality of life (1/1)25 and mixed evidence of positive impact on prescribing/deprescribing appropriateness (1/2).

Community pharmacies

Only one study76 was conducted in a community pharmacy, which was notable for being the only one carried out in the UK. The STOPP/START tools were administered by a pharmacist in an intervention that targeted all three mechanisms, and a physician implemented the recommendations. The study found positive evidence for improved adherence to medical treatment and quality of life, and a reduction in falls. 76

Comparative studies

In a single study56 that included multiple settings (primarily hospitals but also nursing homes) and evaluated a pharmacist-led intervention, acceptance of the interventions was higher in nursing homes than in hospitals.

Personalisation

• Confirmation of a positive impact on patient satisfaction (hospital) and patient adherence (community pharmacies).

• Neither confirmation nor refutation that STOPP/START-based interventions lead to greater adherence to pharmacological treatment when patients are present at their medication review (no evidence).

• Neither confirmation nor refutation of a positive impact on patient empowerment and safety culture of patients, or, accordingly, that STOPP/START-based interventions increase patient empowerment when patient perceptions and expectations are taken into account (elicited and acted on) (no evidence).

Systematisation

• Neither confirmation nor refutation of a positive impact on professional and organisational culture, burden on patient/professional, or patient awareness of the medication review process (no evidence).

• Accordingly, neither confirmation nor refutation that STOPP/START-based interventions change professional and organisational cultures when there are computerised systems linked to medical records, access to previous STOPP/START data and data-sharing takes place (with relevant professionals, teams and organisations). Likewise, neither confirmation nor refutation that STOPP/START-based interventions reduce professional and organisational burden in medication management when there is sufficient familiarity with the tool (no evidence).

Implementation of evidence

• Confirmation of a positive impact on appropriateness of prescribing/deprescribing (hospitals, nursing homes, primary care), ADEs (hospitals) and quality of life (hospitals, nursing homes, community pharmacy).

• Confirmation that STOPP/START-based interventions reduce adverse outcomes through appropriate prescribing/deprescribing (hospitals, nursing homes).

• Neither confirmation nor refutation for a positive impact on falls (nursing homes).

• Refutation of a positive impact on health-care use (hospitals) and mortality (hospitals, nursing homes).

• Refutation that STOPP/START-based interventions are associated with increased appropriateness of prescribing/deprescribing when there are lower levels of familiarity with the evidence base (non-geriatricians, non-pharmacists and more junior professionals) (hospitals, nursing homes, primary care).

We did not find any evidence that studies targeting more mechanisms achieved better outcomes than those targeting fewer, nor was any other configuration linked with better outcomes.

Interventions

It would be worth replicating the clinical effectiveness and cost-effectiveness study76 of the intervention delivered through community pharmacies in more settings. It was the only study76 tested in the UK, one of the few studies that simultaneously considered the three proposed mechanisms of action and one of the studies that demonstrated better results in the use of STOPP/START tools. The nature, scope and target population of the intervention are all particularly relevant to primary care UK economies and offer promising potential in addressing the needs of frail and potentially pre-frail people.

Study design

Studies of the effectiveness of STOPP/START tools have mainly selected outcomes related to appropriateness of prescribing. There has been little exploration of other relevant and longer-term outcomes, such as patient-reported activation, experiences and outcomes, medication concordance or patient safety culture. We have also observed a mismatch in the vast majority of studies between the mechanisms targeted by the interventions and outcomes selection. Studies should be conducted that explore the impact of the tools on outcomes, such as the ones outlined above, and should carefully select outcomes so as to maximise alignment with the hypothesised mechanisms. This would provide more valuable evidence both for testing the single logic model and for informing policy on the potential benefits of routine use of the tools.

We found very limited research using qualitative research to support the evaluation of STOPP/START-based interventions. Qualitative information on experiences of implementing, delivering or being involved in such interventions (e.g. as a health professional, patient or carer) would be a significant addition to the literature, and would provide key information for further testing and refining the proposed logic model.

Economic evaluation

Although savings may have been demonstrated, these were often achieved in the context of uncosted or unreliably costed interventions and short-term capturing of a relatively limited set of patient-relevant outcomes. More high-quality economic evaluations and more comprehensive and fully reported costing studies are required alongside any future clinical effectiveness evaluations. As well as more fully reporting the resources required, and the outcomes and resource impacts produced, studies should endeavour to explain what drives the particular mix and levels of resource use and cost differences compared with usual care. Only when more studies do this better will more valuable realist, explanatory syntheses of cost-effectiveness research be possible. 93

Primary care

Studies of the effectiveness of STOPP/START tools in PRIMARY CARE consistently evaluated interventions in which the primary care physician both used the tools and implemented the recommendations, whereas hospital-based studies very frequently involved a number of different professionals with a more distributed allocation of tasks. Primary care has first contact and continuity functions within the health-care system and is the source of the vast majority of chronic prescriptions, thereby providing the natural environment for ensuring continuous optimisation of drug therapies. The further development of the evidence base for the relative merits of different approaches for STOPP/START implementation in primary care would inform health policy for the delivery of safe and effective care. RCTs comparing relevant outcomes for different alternatives would be particularly well suited for developing this evidence base. Of particular interest would be GP-supported enhancement of the previously discussed community pharmacy intervention. 76

Reporting

The studies identified included little information on context. There is a need for studies of the evaluation of interventions to include greater details on context to facilitate adequate integration of the evidence. The development of consensus-based reporting standards for promoting focused and structured reporting of context for extending existing reporting guidelines, specifically to those relevant to the evaluation on interventions [mainly Consolidated Standards of Reporting Trials (CONSORT), Standards for Quality Improvement Reporting Excellence (SQUIRE) and Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)], would be an important step forward.

Contacts networks

We will showcase our findings through existing contacts networks (including Applied Research Collaborations and Academic Health Sciences Networks) and also to the RPS through a dissemination event (see Dissemination event).

Dissemination through and to patients and members of the public

The Plain English summary has been created with the help of members of the Patient Advisory Group in the last Patient Advisory Team meeting (see Appendix 12, Table 23). We will offer this summary to relevant organisations (including Age UK) and networks where the summary can be fed into pamphlets, newsletters, websites and blogs.

Website

We have already set up and maintain a project webpage [URL: http://medicine.exeter.ac.uk/research/healthresearch/healthservicesandpolicy/projects/qualitysafety/stoppstart/ (accessed 4 November 2020)] on the University of Exeter Medical School website to raise awareness of the project and to publish findings following the completion of the report. This link has been shared to the project team and the scientific advisory board’s affiliated universities. We plan to highlight the project findings on the websites of the institutions of the co-applicants following final submission of the report:

• Leeds (URL: www.leeds.ac.uk/)

• Bristol (URL: www.bristol.ac.uk/)

• Queen’s University Belfast (URL: www.qub.ac.uk/).

Social networks

We will use social networks such as Twitter (www.twitter.com, Twitter, Inc., San Francisco, CA, USA) to showcase our findings to a wide range of audiences following completion of the report using the hashtag #STOPPSTART. In particular, we will use the accounts @josemvalderas (553 followers), @UoEAPEx (357 followers) and @uemshspr (185 followers).

Peer-reviewed academic journals

At the time of publication, we are intending to submit a manuscript for publication in an open access peer-reviewed academic journal; we have started drafting the manuscript for the journal Age and Ageing. Alternative journals will cover pharmacy, general practice, internal medicine and/or geriatrics.

Conference and policy presentations

Conference and policy presentations will be disseminated to audiences of pharmacists and pharmacologists, general practice and hospital-based clinicians, voluntary agencies and policy- and decision-makers. We have disseminated the results of phase 1 at the Annual Society for Academic Primary Care conference held at the University of Exeter in July 2019. An abstract has also been accepted for the International Conference for Realist Research, Evaluation and Synthesis 2021 to be held in Dublin, Ireland.

Search strategy

STOPP.

Search strategy

STOPP.

Personalised

• Opportunity for the patient to give input on the medications they take.

• Discussion about the possible implications of the changes to medicines and what to look out for.

• Having a follow-up after changes are made.

• Face-to-face discussion where the patient is involved in discussion about what changes to make.

• The changes made are agreed with the patient.

• The aim of making changes should be about patient outcomes rather than financial outcomes.

Not personalised

• Discussion happens without the patient, and then they are asked to agree (or not) with changes.

• When the patient agrees to proposed changes without being part of discussion about them.

A subsequent meeting focused on the results, Plain English summary and dissemination of the results (see Chapter 8 and Appendix 12 for details on input on dissemination).