Effectiveness of educational interventions in general practice for improving medication appropriateness and deprescribing in older adults: a systematic review protocol

Abstract

Introduction

Polypharmacy in older adults is a growing concern, particularly in general practice (GP), where general practitioners (GPs) are the main prescribers managing complex multimorbidity. While often necessary, polypharmacy increases the risk of potentially inappropriate prescribing (PIP), adverse drug events, hospitalisations and reduced quality of life. Although clinical medication reviews using specific tools are frequently employed to address these risks and guide a safe deprescribing process, the specific role and effectiveness of educational interventions—as a component of prescribers’ behavioural change—remain inconsistently evaluated. The objective of this review is to identify, describe and evaluate educational interventions targeting GPs, aimed at improving medication appropriateness and promoting deprescribing in older adults with polypharmacy in GP settings.

Methods and analysis

This systematic review protocol follows Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Protocols guidelines. Eligible study designs include randomised controlled trials (RCTs), cluster RCTs and quasi-experimental studies. Studies must target GPs, GP trainees or primary care physicians and report outcomes related to medication appropriateness, measured using validated tools (eg, Medication Appropriateness Index, Screening Tool of Older Person’s Prescriptions/Screening Tool to Alert to Right Treatment). Secondary outcomes include hospital admissions, quality of life, prescribing behaviour, medication-related harms and cost-effectiveness. In addition to evaluating effectiveness, we will characterise heterogeneity in educational content and objectives, duration/intensity, theoretical or pedagogical underpinnings, delivery format, implementation fidelity and contextual factors. A comprehensive search will be conducted in MEDLINE, EMBASE, CINAHL and CENTRAL without language or date restrictions. The systematic review will follow PRISMA 2020 guidelines for data synthesis, and if meta-analysis is not feasible, Synthesis Without Meta-analysis reporting guidelines will be used.

Ethics and dissemination

Findings will be disseminated through peer-reviewed publications and conference presentations. Results will inform the design of future educational strategies to optimise medication review and deprescribing practices in general practice and primary care, by identifying which approaches most effectively improve patient-centred outcomes and clarifying the role of educational components within complex, multicomponent interventions.

Protocol registration

https://www.crd.york.ac.uk/PROSPERO/view/CRD42025622443.

STRENGTHS AND LIMITATIONS OF THIS STUDY

• This systematic review will follow Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, and the protocol complies with PRISMA Protocols, ensuring methodological transparency and reproducibility.

• A comprehensive, peer-reviewed search strategy will be applied across multiple databases, with no date restrictions.

• Independent, duplicate screening and data extraction will minimise bias and improve data reliability.

• Educational components may be inconsistently reported across studies, limiting comparability and synthesis.

• Pre–post studies and qualitative designs will be excluded, which may restrict the scope of findings.

Introduction

Medication appropriateness in older adults is a major public health concern, especially in high-income countries where increased life expectancy and multimorbidity—defined as two or more long-term chronic conditions—lead to prolonged polypharmacy, commonly defined as the concurrent use of ≥4 medications.^{1 2} While polypharmacy may be necessary to manage multiple conditions, alleviate symptoms or slow disease progression, concerns are rising that many older adults with multimorbidity are using an inappropriately high number of medications, where intended benefits are outweighed by cumulative harm.^{3 4} Polypharmacy affects a large proportion of older adults, with a prevalence of 36.2% in Europe⁵ and 44.1% in the USA⁶ among those aged ≥65 years. Notably, in the USA, medication review and deprescribing interventions have been estimated to reduce hospitalisation costs due to adverse drug events by up to US$750 000 annually.⁷ However, the proportion of polypharmacy cases that can be considered clinically appropriate versus inappropriate remains largely unclear.⁸ Nevertheless, even when clinically justifiable, polypharmacy is often linked to medication-related harms,^{9 10} including non-adherence,^{11 12} increased morbidity (eg, frailty, nutritional deficits, functional decline),13,15 hospitalisations and mortality.^{16 17} This underscores a recognised dilemma: even the most rational and evidence-based prescribing decisions can carry significant safety concerns in the context of polypharmacy.^{18 19} In fact, many efforts are underway to refine the concept of polypharmacy, shifting the focus from the mere number of medications to considerations of clinical appropriateness, therapeutic effects and patient outcomes. More clinically relevant and patient-oriented definitions are now emerging,^{4 20} incorporating factors such as medication classes and comorbidities to better identify the drug-disease clusters most strongly associated with adverse outcomes within a polypharmacy regimen.^{9 19} In this evolving framework, ensuring that polypharmacy is both safe and effective, healthcare professionals must routinely review prescriptions, check interactions, discontinue unnecessary drugs, adjust dosages or introduce safer alternatives—a process referred to as deprescribing.^{21 22} A key structured approach through which deprescribing is implemented in practice is the clinical medication review (CMR), consistently recognised as a core component of interventions addressing polypharmacy in older adults. CMR may be conducted by pharmacists, general practitioners (GPs) or specialists, nurses or multidisciplinary teams and can range from basic medication reconciliation to comprehensive geriatric assessments.^{19 23 24} To ensure consistency and clinical relevance, these reviews often rely on validated tools that help identify inappropriate prescribing practices and guide deprescribing decisions. Polypharmacy increases the likelihood of potentially inappropriate prescribing (PIP), a concept that includes both potentially inappropriate medications, defined as those with an unfavorable risk-to-benefit ratio in many older adults, and potential prescribing omissions, defined as the failure to prescribe medications that are clinically indicated for disease treatment or prevention. 25,27 To systematically address PIP within CMRs, a range of evidence-based instruments is available,^{19 28 29} including judgement-based instruments such as the Medication Appropriateness Index (MAI)³⁰ and criteria-based tools like the Screening Tool of Older Person’s Prescriptions/Screening Tool to Alert to Right Treatment (STOPP/START) criteria³¹ or Beers list.³² GPs are responsible for the majority of prescriptions in older adults, placing them at the forefront of efforts to ensure safe and appropriate CMRs in this population.33,36 Medication decisions in older patients are rarely straightforward; each prescription must be assessed not only in isolation but also in light of the patient’s clinical history, comorbidities, personal context and evolving therapeutic priorities.^{8 18 19 37} While numerous systematic reviews have examined interventions aimed at reducing inappropriate prescribing and promoting medication appropriateness in older adults with polypharmacy,^{19 23 24 38 39} the overall quality of evidence remains limited. This is largely due to heterogeneity in study designs, small sample sizes, short follow-up durations and inconsistent reporting of theoretical underpinnings, implementation strategies and patient involvement.^{18 21} These medication appropriateness interventions are often aided by additional tools and strategies, including prescribing criteria, clinical decision support systems and educational programmes. Among these, educational interventions targeting healthcare professionals—particularly GPs—have shown promise in improving prescribing behaviours and facilitating deprescribing.⁴⁰ However, such interventions remain under-investigated—particularly within general practice (GP) settings.33,36

Moreover, when they are studied, critical details regarding their design, delivery methods and theoretical foundations are often poorly reported,^{19 35 40} limiting the ability to evaluate, replicate and scale successful models. Through a systematic identification and analysis of existing interventions, this review will provide a structured and comprehensive synthesis of educational strategies implemented in GP to improve safer prescribing in older adults. By examining how medication appropriateness and deprescribing are defined, operationalised and measured within these interventions, the review will identify current gaps in reporting, design and outcome consistency. These insights will inform the development of more robust, evidence-based educational interventions to support prescribers’ behavioural change, promoting more appropriate deprescribing in primary care.

Objective

This paper presents the protocol for a systematic review of randomised controlled trials (RCTs) and quasi-experimental studies evaluating educational interventions/clinical training sessions in GP aimed at improving medication appropriateness and supporting deprescribing in older adults. The systematic review objectives are as follows:

1. To identify and describe educational and training interventions delivered in GP that aim to improve medication appropriateness and support deprescribing in older adults.

2. To characterise the heterogeneity of these interventions across key dimensions, including educational content and learning objectives, duration and intensity, theoretical or pedagogical framework, delivery format (eg, workshops, online modules, audit-feedback, blended approaches), implementation fidelity and contextual factors.

3. To evaluate to what extent these educational interventions have demonstrated to be effective in improving prescribing appropriateness and deprescribing practices, as assessed by validated outcome measures.

Methods and analysis

Study design and registry

The systematic review will adhere to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.⁴¹ The protocol was registered on the PROSPERO database (registration number: CRD42025622443). This systematic review protocol follows the PRISMA Protocols guidelines.⁴²

Review question and PICO

Specific research questions are as follows: (1) how are educational/training interventions to improve medication appropriateness and deprescribing in older adults modelled and described in GP settings? (2) how do these interventions conceptualise, measure and operationalise the medication appropriateness process? This review follows the PICO framework (Population, Intervention, Control, Outcome) shown in table 1.

Table 1. PICO framework (Population, Intervention, Control, Outcome).

Population	General practitioners
Intervention	Educational/training interventions
Control	Regular practice/standard care
Outcome(s)	Improve medication appropriateness and deprescription on potentially inappropriate drugs in older people (≥65 years) with at least one long-term condition who were receiving polypharmacy (≥4 regular medicines).

Eligibility criteria

Study designs

Only RCTs, cluster RCTs and quasi-experimental controlled studies will be included.

Participants

Eligible studies must focus on GPs, primary care physicians or GP trainees, either alone or in collaboration with other health professionals. We will exclude studies involving only healthcare professionals other than GPs, as well as those focused on undergraduate student education.

Intervention

Only studies that investigate clinical training or educational programmes, either as standalone or as part of a multicomponent intervention, which focus specifically on medication appropriateness and pharmacological deprescribing in older adults (≥65 years) will be eligible.

For this review, educational interventions are defined as structured activities designed to support improvements in GPs’ prescribing behaviour. Eligible interventions will include formal training programmes (eg, workshops, seminars, lectures, online courses or blended formats); audit-feedback and decision-support tools approaches, provided they are primarily and explicitly educational (eg, accompanied by guidance, explanation or instruction intended to support change in GPs’ prescribing practice) or only if at least one educational or training component is present (eg, structured training session, guided tutorial or manual, feedback reports accompanied by interpretative guidance). Otherwise, they will be excluded as non-educational. To promote detailed description of interventions and enable replication, we will extract detailed information on interventions using a piloted, standardised data extraction form aligned with the Template for Intervention Description and Replication (TIDieR) checklist.⁴³ Intervention characteristics will be coded across predefined domains, including educational content and objectives, duration and intensity, theoretical or pedagogical framework, delivery format, implementation fidelity and contextual or system-level factors (online supplemental appendix 2). To facilitate comparison of educational strategies and investigate potential implementation challenges, the Theoretical Domains Framework⁴⁴ will be applied as a coding framework for behaviour change to identify which determinants of GPs’ prescribing behaviour (eg, knowledge, skills, beliefs about capabilities, beliefs about consequences, social/professional role, reinforcement, environmental context/resources, intentions, goals, memory/attention/decision processes, emotion, optimism, social influences, behavioural regulation) are addressed within the educational components. We will exclude studies in which the intervention addresses non-pharmacological prescribing appropriateness (eg, diagnostic tests, imaging or screening procedures), those enrolling fewer than 10 older adults and those focused exclusively on deprescribing in palliative care settings. Interventions in which education is absent or incidental, such as stand-alone administrative tools or decision-support systems without a training or feedback element, will also be excluded.

Comparator(s)/control

Only studies in which the comparator is a GP control group receiving (eg, usual training/standard care, an active control or being placed on a waiting list). Standard care/usual training is defined as the routine educational activities or professional development normally available to GPs in the study setting (eg, continuing medical education, guideline updates, in-service meetings). We do not assume standardisation across settings. For each study, we will extract and report the content, frequency, duration, delivery format, provider and whether participation was mandatory or optional, aligning with TIDieR checklist items⁴³ where available. When details are unclear, items will be coded as ‘not reported’ and considered in sensitivity analyses. Active comparator is defined as an alternative educational strategy (eg, different curriculum, format or intensity) or another implementation approach against which the index intervention is compared. The same TIDieR-aligned details will be extracted to enable like-for-like comparison. Wait-list control is defined as groups scheduled to receive the index intervention after follow-up. We avoid the ambiguous label ‘no intervention’, since GPs typically receive some form of ongoing training within standard care; where authors use this term, we will reclassify the comparator under the categories above based on reported features.

Outcomes and prioritisation

The primary outcome of interest for this review will be medication appropriateness, assessed using judgement-based and/or criterion-based tools. These may include established instruments such as MAI, the STOPP/START criteria or other validated tools, provided they explicitly aim to evaluate prescribing appropriateness in older adults. This outcome directly addresses the core question of this review: whether educational interventions for GPs can improve prescribing quality in older adults with multimorbidity and polypharmacy. Given the heterogeneity of available instruments, we will first group studies according to the type of tool used (judgement-based such as MAI,³⁰ criterion-based such as STOPP/START,³¹ or other validated measures). Where studies use the same tool, we will conduct analytic synthesis using effect measures appropriate to the scale. If pooling across different tools is not methodologically appropriate, we will present results in a structured narrative synthesis, highlighting consistencies or divergences across instrument types. This approach ensures both comparability and transparency in handling outcome variability.⁴⁵

The rationale is summarised in table 2.

Table 2. Rationale for outcome measure selection.

Justification	Explanation
Direct relevance with review objective	The outcome directly addresses whether educational interventions for general practitioners improve prescribing quality in older adults with multimorbidity and polypharmacy.
Clinical complexity in target population	Older adults with multiple chronic conditions and polypharmacy face increased risks of both overuse and underuse of medications. Assessing prescribing appropriateness in this context requires tools sensitive to patient-specific nuances.
Use of validated tools	Tools such as the Medication Appropriateness Index (MAI) and Screening Tool of Older Person’s Prescriptions/Screening Tool to Alert to Right Treatment (STOPP/START) are widely used as they reflect clinical complexity. MAI supports contextualised clinical judgement, while STOPP/START offers structured, evidence-based criteria.
Flexibility in tool inclusion	Additional tools may be included if they systematically and meaningfully assess prescribing appropriateness in older adults.
Handling of multiple instruments	To ensure comparability, studies will be grouped by type of tool (judgement-based, criterion-based or other validated measures). Where ≥4 studies use the same tool, analytic synthesis will be performed using mean difference, standardised mean difference or risk ratio if dichotomised. Where pooling is not feasible, structured narrative synthesis will be undertaken following Synthesis Without Meta-analysis guidance52.

Secondary outcomes of interest will be organised into three priority levels.

Patient-level outcomes (highest priority): hospital admissions, including all-cause admissions and unplanned readmissions; health-related quality of life, as measured by validated instruments (eg, EQ-5D⁴⁶); overall well-being measured by validated patient-reported outcome measures (eg, WHO-5 Well-Being Index^{47 48}); and medication-related problems (eg, adverse drug reactions).

Physician-level outcomes (second priority): improvements in physicians’ skills related to medication appropriateness, assessed either by objective measures (eg, performance in case vignettes) or validated self-assessment questionnaires where available. Changes in prescribing behaviours (eg, number and type of PIPs prescribed, adherence to STOPP/START criteria) will also be included.

System-level outcomes (third priority): healthcare utilisation (eg, emergency visits, primary care visits) and cost-effectiveness data from economic evaluations.

Setting

Only studies conducted in a primary care or general practice setting will be included.

Language

No language restrictions will be applied. For articles published in languages other than English, translation tools will be used, and any issues related to translation will be transparently reported in a dedicated appendix.

Information sources

A comprehensive literature search will be conducted via the EBSCO interface (MEDLINE, CINAHL, EMBASE and CENTRAL), with no language or time restrictions. Reference lists of included articles and relevant reviews will be hand-searched, and citation tracking will be used to identify additional studies.

Search strategy development and peer-review calibration

The search was primarily conducted in MEDLINE using both controlled vocabulary (eg, MeSH) and free-text terms related to GP/family medicine and medication appropriateness or inappropriate prescribing. To enhance sensitivity and capture studies where educational components may be implicit (eg, ‘medication review’, ‘polypharmacy management’), terms explicitly related to educational interventions were excluded following iterative pilot testing in MEDLINE. The search strategy was developed by GC (systematic review expert) and VF and peer-reviewed by TZ to ensure an optimal balance between comprehensiveness and specificity. A calibration exercise will be undertaken by VF and TZ to refine screening criteria and ensure alignment with the review objectives. The MEDLINE search strategy is provided in online supplemental appendix 1. Additionally, reference lists of included studies will be screened for relevant articles.

Study record and selection process

Search results from MEDLINE, CINAHL, EMBASE and CENTRAL will be imported into Covidence for deduplication and screening. Two independent reviewers (VF and TZ) will screen titles and abstracts against inclusion criteria, full texts of all titles that appear to meet the inclusion criteria or where there is any uncertainty will be obtained; disagreements will be resolved by discussion or by a third reviewer (GC). The record review and selection process will be illustrated using a PRISMA flow chart.⁴¹ Companion studies will be identified and appropriately merged or excluded to prevent duplication. Missing data will be addressed by contacting corresponding authors (up to three attempts). Data will be stored in Covidence and backed up in Excel by VF.

Data collection process and data extraction

Two independent reviewers (VF and TZ) will extract data in duplicate using Covidence. The extraction form will be piloted on three to five studies and refined until inter-rater agreement of κ≥0.70 is achieved, with this calibration exercise ensuring consistency. Adjusted estimates will be prioritised when both adjusted and unadjusted values are reported. Full coding guidance and operational categories are provided in online supplemental appendix 2. Implementation fidelity will be captured systematically across four dimensions: adherence (extent to which intervention components were delivered as intended, including content, frequency, duration and coverage), exposure (amount and frequency received), quality of delivery (skill and accuracy of intervention delivery) and participant responsiveness (degree of engagement and completion by participants).⁴⁹

Risk of bias

The quality of the studies included will be assessed using appropriate tools depending on the study design. For RCTs, the Cochrane Risk of Bias tool will be employed,⁵⁰ and for non-randomised studies, the Risk Of Bias In Non-randomized Studies of Interventions tool.⁵¹ Assessments will be conducted independently by two reviewers, with disagreements resolved by consensus or, if needed, by consultation with a third reviewer. Risk of bias assessments will directly inform the synthesis. Studies judged at high risk of bias will not be excluded a priori but will be addressed through sensitivity analyses, in which pooled estimates will be recalculated after excluding these studies to test the robustness of findings. Where a body of evidence for a given outcome consists mainly of high-risk studies, results will be reported descriptively with explicit caution, rather than pooled analytically.

Data analysis

Data synthesis

We will follow PRISMA 2020 guidelines for data synthesis.⁴¹ This framework will guide us in presenting the selection flow and the criteria for excluding studies that do not meet our inclusion criteria. Any variability observed in the studies—such as differences in study samples, interventions or other factors—will be assessed and discussed in the results and limitations sections of the review. A meta-analysis will be conducted if at least four sufficiently comparable studies are available. Studies will be considered comparable if they are similar in terms of participants, interventions and outcomes. Candidate outcomes for pooling are prespecified as follows: medication appropriateness (analysed separately by type of tool, eg, judgement-based such as MAI or criterion-based such as STOPP), hospital admissions, adverse drug reactions and health-related quality of life. These outcomes were selected for their clinical relevance and consistent use in previous research on polypharmacy and deprescribing in older adults.^{19 23 24 35} For continuous outcomes, mean differences or standardised mean differences will be calculated; for dichotomous outcomes, risk ratios or ORs, each with 95% CIs. For cluster RCTs, we will adjust for clustering (eg, using effective sample size/design effects) if not already reported. Where scales differ in directionality (eg, MAI indicates better appropriateness), we will harmonise them so that higher scores consistently indicate improvement (or decline). Heterogeneity will be assessed using the I² statistic and Q-test (p<0.10) and interpreted according to Cochrane guidance: 0–40% might not be important, 30–60% may represent moderate heterogeneity, 50–90% may represent substantial heterogeneity and 75–100% may represent considerable heterogeneity.⁴⁵ Studies with substantial or considerable heterogeneity will be explored in subgroup or sensitivity analyses or excluded from meta-analysis if pooling is not appropriate. Potential publication bias will be assessed with Egger’s and Begg’s tests. Subgroup analyses will explore differences by intervention type and duration, GP characteristics (eg, years of experience, solo vs group practice) and patient-related variables (eg, age, comorbidity, baseline polypharmacy). Where sufficient data are available (≥10 studies), meta-regression will be used to examine the influence of intervention and contextual characteristics.⁴⁵ Candidate covariates, pre-specified for their theoretical and clinical relevance, include educational theoretical model, intervention intensity, delivery mode, follow-up, GP experience, baseline patient polypharmacy and chronic conditions; definitions and operationalisation are detailed in Appendix 2. When multiple time points are reported, the longest follow-up will be prioritised; earlier time points will be explored in sensitivity analyses if comparable. If a meta-analysis is not feasible due to clinical or methodological heterogeneity, narrative synthesis will be conducted following Synthesis Without Meta-analysis reporting guidelines⁵² structured by key research questions, intervention type and outcomes.

Quality of evidence

The quality of evidence for primary and secondary outcomes will be assessed using the Grading of Recommendations Assessment, Development, and Evaluation approach,⁵³ which considers risk of bias, inconsistency, imprecision, indirectness and publication bias. Outcomes will be rated as high, moderate, low or very low quality, reflecting the level of confidence in the effect estimates.

Discussion

This review will explore educational interventions as key components of effective strategies to improve medication appropriateness and promote deprescribing in older adults within GP. By examining how these interventions are designed, implemented and reported, the review aims to clarify their role in changing prescribing behaviours among GPs. Understanding these mechanisms may help identify which elements contribute most to successful outcomes and guide the development of more targeted, scalable and context-sensitive interventions. The findings will support future implementation efforts and inform educational frameworks, research and policy initiatives aimed at enhancing strategies to reduce low-value care practices, such as inappropriate prescribing.

Data availability statement

No data are available.