Effectiveness of interactive dashboards to optimise prescribing in primary care: a protocol for a systematic review

Version Changes

Revised. Amendments from Version 1

In response to reviewer feedback, some revisions have been made to the manuscript. The introduction section has been updated to strengthen the rationale for this systematic review, additional references have been added. Key terms such as implicit and explicit medication appropriateness tools and effectiveness have been elaborated on. Clarity on the population of interest has also been provided, with minor changes made to table 1 to reflect this. In the study selection section, the fact that full text articles will be screened independently has been clarified. The analysis section has been updated to clarify under which criteria a meta-analysis will be performed. Information around sensitivity analyses has been added.

Abstract

Introduction

Advances in therapeutics and healthcare have led to a growing population of individuals living with multimorbidity and polypharmacy making prescribing more challenging. Most prescribing occurs in primary care and General Practitioners (GPs) have expressed interest in comparative feedback on their prescribing performance. Clinical decision support systems (CDSS) and audit and feedback interventions have shown some impact, but changes are often short-lived. Interactive dashboards, a novel approach integrating CDSS and audit and feedback elements, offer longitudinal updated data outside clinical encounters. This systematic review aims to explore the effectiveness of interactive dashboards on prescribing-related outcomes in primary care and examine the characteristics of these dashboards.

Methods

This protocol was prospectively registered on PROSPERO (CRD42023481475) and reported in line with PRISMA-P guidelines. Searches of PubMed, EMBASE, Medline, PsychINFO, CINAHL, Scopus, the Cochrane Library, and grey literature, including trial registries were performed to identify interventional studies (randomised and non-randomised) that assess the effectiveness of interactive dashboards on prescribing related outcomes. The search will be supplemented by searching references of retrieved articles with the use of an automated citation chaser. Identified records will be screened independently by two reviewers and data from eligible studies extracted using a purposely developed data extraction tool. We will narratively summarise the intervention types and those associated with improvements in prescribing outcomes. A quantitative synthesis will be carried out if a sufficient number of homogenous studies are identified. Methodological quality will be assessed by two reviewers using the Cochrane Effective Practice and Organisation of Care risk assessment tool.

Discussion

This systematic review will explore the effect of interactive dashboards on prescribing related outcome measures in primary care and describe the characteristics of interactive dashboards. This research may inform future intervention development and shape policymaking particularly in the context of ongoing and planned developments in e-prescribing infrastructure.

Introduction

Reducing medication-related harm has been identified within the World Health Organisation’s (WHO) third global patient safety challenge in 2017, ‘Medication without Harm’ ¹ . In the United States alone it is estimated that adverse drug reactions represent the third leading cause of death ² . While English data shows 1.5% of hospital admissions are as a result of an Adverse Drug Reaction (ADR) ³ . In Ireland ADRs are the third most common type of reported adverse event in the health care system ⁴ . Most prescribing occurs in primary care ⁵ and qualitative data from general practitioners (GPs) indicates prescribing has become more challenging, particularly for patients with multimorbidity and polypharmacy ⁶ . Irish GPs in a nationwide cluster randomised controlled trial (RCT) evaluating a deprescribing intervention, viewed participation as an opportunity to review their prescribing practices and were interested in getting performance feedback ⁷ . There is evidence that audit and feedback is effective in improving professional behaviour and that ongoing or repeated feedback is more effective ^{8– 10} .

Primary care prescribers receive feedback on their prescribing through various means, such as clinical decision support systems (CDSS) and audit and feedback. CDSS are real-time, electronic tools that provide prescribers with knowledge and person-specific information at the point of care, which supplement decision-making processes ¹¹ . CDSS are embedded in clinical software and typically appear as “alerts” for the prescriber. However problems such as interrupting work flow and too many alerts can cause “alert fatigue” resulting in the user ignoring recommendations ¹² . Evidence suggests CDSS probably have a small effect on practitioner performance but the effect on patient reported and clinical outcomes is less clear ^{13, 14} . Audit and feedback involves retrospectively reviewing clinical performance or practices, enabling peer comparison and social norm feedback and it has been identified as an effective strategy for improving prescribing ^{15, 16} . However, audit and feedback data typically provide a snapshot at one time point and therefore improvements may be temporary ¹⁷ . Interactive dashboards combine elements of both CDSS and audit and feedback; the data is longitudinal and updated on an ongoing basis but is outside the clinical encounter. The prescriber can visualise their data graphically and the data can be manipulated and interacted with through various interactive elements, identifying both time trends and comparisons with peers ¹⁸ .

Medicines optimisation interventions that target a heterogonous population often use prescribing-related outcome measures, as clinical outcome’s such as ADRs or unplanned hospital admissions may take time to manifest or reach measurable levels ¹⁹ . Various tools have been developed to assess the quality of prescribing and broadly speaking these can be categorised into two groups: explicit tools and implicit tools. A systematic review published in 2014 identified 46 different explicit and implicit tools that have been developed to assess medication appropriateness. Explicit tools are focused on drugs, measuring how they fit pre-defined criteria whereas implicit tools are based on clinical guidelines and clinical evaluation criteria ²⁰ . Examples of explicit measures of medication appropriateness include the United States (US) Beers criteria ²¹ and the European Screening Tool for Older People’s potentially inappropriate Prescriptions (STOPP) criteria ²² . Multiple observational studies have demonstrated an association between potentially inappropriate prescribing, measured using these indicators, and clinical outcomes such as increased emergency admissions, ADRs and reduced health related quality of life ^{23– 25} . In addition, specific research groups have identified high-risk and low-value prescribing criteria and evaluated the effectiveness of interventions utilising these criteria ^{26– 28} . For example the Data-driven Quality Improvement in Primary Care (DQIP) intervention included an informatics tool that provided weekly updates of selected high risk prescribing indicators to clinicians, and facilitated medication review by graphically displaying relevant drug history data ²⁹ . The pharmacist-led information technology intervention (PINCER) was effective at reducing hazardous prescribing, however the effect may have been temporary as the original intervention provided a snapshot of data from the electronic health record ²⁶ . More recently an interactive dashboard utilizing the PINCER criteria has been developed whereby the user can track their performance across different criteria compared to other practices and over time ³⁰ , and this intervention resulted in a reduction of potentially hazardous prescribing by 27.9% (95% CI 20.3% to 36.8%, p < 0.001) ³¹ .

In line with national and international campaigns to reduce medication-related harm and recent developments in e-prescribing infrastructure, this systematic review aims to explore the effectiveness of interactive dashboards in improving prescribing-related outcomes in primary care. These outcomes include potentially inappropriate prescribing (PIP) and drug utilisation rates (e.g., reducing prescribing volumes where lower rates are preferable or optimising prescribing patterns in line with guidelines). Additionally, it aims to describe the characteristics of these interventions to inform future intervention development and e-prescribing infrastructure.

Methods

This systematic review was prospectively registered on PROSPERO (CRD42023481475), it will be conducted in line with guidance set out in the Cochrane Handbook for Systematic Reviews of Interventions ³² , and reported in adherence to PRISMA-P reporting guidelines ³³ . At the time of writing, the search strategy has been finalised, title and abstract screening has been completed, and full text review is currently in process.

Search strategy

An information specialist in the host institution’s library with extensive experience in supporting systematic reviews was involved in developing search strategies. A systematic literature search was conducted and included the following databases; PubMed, EMBASE, MEDLINE (OVID), PsycINFO (EBSCOhost), CINAHL (EBSCOhost), Scopus and the Cochrane Library (OVID).

A search of grey literature was conducted by running keyword searches in OpenGrey, CADTH Grey Matters and web-based clinical trial registries. The search was supplemented by searching references of retrieved articles with the use of an automated citation chaser ³⁴ . No restrictions were placed on language or year of publication. Search terms included “interactive dashboard” and the medical subject heading (MeSH) “clinical audit”, “medical audit”, “benchmarking” and “feedback” and keywords to capture concepts related to providing prescribers with feedback, such as “electronic health record” and “alerts”. See supplementary file 1 for electronic search reports, including the full search terms.

Study selection

Identified records were uploaded to Covidence systematic review software and de-duplicated. Reviewers were blinded to minimise potential bias and ensure impartial evaluation of the included studies. Two reviewers independently read the titles/abstracts of identified records and eliminated studies not meeting inclusion criteria. The full text of the remaining studies will be reviewed again independently by two reviewers who will assess their suitability for inclusion. Disagreement will be resolved through discussion with the wider study group. Eligibility criteria are described in Table 1. All outcome measures detailed in Table 1 will be considered, we do anticipate however, based on scoping searches, that explicit criteria and utilisation patterns will feature more prominently. All interventional designs will be included including randomised controlled trials (RCTs) (e.g. cluster RCTs, step wedged RCTs and individually randomised RCTs) and non-randomised interventional studies (e.g. interrupted time series design and controlled before and after studies) ³⁵ .

Table 1. Study eligibility criteria.

Criteria	Inclusion	Exclusion
Population	Primary care prescribers (e.g. General Practitioners, non-medical prescribers based in primary care such as pharmacists and advanced nurse practitioners). Patients within primary care settings (No restrictions on patient characteristics)	Primary care prescribers working in a secondary care setting. Dentists Patients in other settings not considered a primary care setting.
Intervention	An interactive dashboard designed to provide feedback on prescribing data to prescribers and including the following characteristics: Visual display of data: Data is presented in the form of graphs or tables. Interactivity: Allows direct manipulation with visual analytical tools or provides multiple parameters from the dataset, accessible online or via email. Real-time data: Offers real-time or relatively contemporaneous data, no older than one year. Frequent data feedback: Provides data feedback more than once. Comparative analysis: Compares data to peers or set standards.	Simple CDSS interventions Audit and feedback interventions that do not give longitudinal and ongoing feedback
Comparator	Usual care
Outcomes	Primary Outcome: Prescribing related outcome measures such as implicit/explicit criteria, high-risk or low-value criteria or where relevant prescribing rates (e.g. where a higher rate may reflect lower quality such as benzodiazepine or opioid use).	No prescribing outcomes measured
Setting	Primary care (Family practice, general practice)	Studies focused on specialist clinics, nursing homes, hospital based, dental surgeries.
Study design	Interventional studies both randomised and non-randomised designs (e.g. Randomised controlled trials, non-randomised controlled trials, controlled before and after studies and interrupted time series)	Systematic reviews, descriptive study designs (e.g., case reports, uncontrolled before and after studies). Letters, commentaries, editorials.
Publication language	No language restriction
Dates of publication	No year limitation

Data extraction and management

Two review authors will independently extract data using a purposely developed data extraction tool in Covidence, developed with use of the Template for Intervention Description and Replication (TIDieR) checklist ³⁶ . Extracted data will include study details (e.g. setting, design), population (e.g. GPs), intervention details, comparison group, outcome measures and results. Table 2 outlines example data that may be extracted to describe the intervention using the TIDieR checklist. We will attempt to contact the lead authors of primary studies to locate missing data. Discrepancies will be resolved through discussion and consensus between two reviewers with consultation with a third reviewer if necessary.

Table 2. Intervention data extraction using adapted TIDieR checklist.

Description	Detail	Example Intervention
1. Brief name	Provide a concise name for the intervention.	"Prescribing Dashboard"
2. Why	Rationale or goal of the intervention.	To improve prescribing quality by providing GPs with real-time data on high-risk prescribing for comparative benchmarking against peers.
3. What (materials and procedures)	Materials used and procedures for the intervention.	Web-based or practice software-embedded dashboard displaying prescribing quality metrics. GPs access their prescribing metrics on the dashboard at their convenience for self-assessment and benchmarking.
4. Who provided	Description of the intervention providers.	Developed by healthcare IT specialists in collaboration with clinicians.
5. How and where	Modes and location of delivery.	Delivered via a secure web-based platform or embedded within practice management software, accessible in primary care clinics and offices on various devices.
6. When and how much	Frequency and duration.	Dashboards updated monthly with new prescribing data to reflect recent prescribing practices.
7. Tailoring	Personalization of the intervention.	Dashboard data tailored to individual prescriber level or practice level, depending on the setting and objectives of the feedback.
8. Modifications	Changes made during the study.	N/A or describe any modifications based on user feedback or technological updates.
9. How well (planned and actual)	Assessment of adherence or fidelity and actual engagement.	Usage analytics to monitor frequency of dashboard interactions, time spent on the dashboard, and engagement with specific metrics.

Methodological quality assessment. Methodological Quality Assessment will be performed by two reviewers using the Cochrane Effective Practice and Organisation of Care (EPOC) risk of bias tool ³⁵ . Discrepancies will be resolved through discussion and consensus between two reviewers with consultation with a third reviewer if necessary.

Analysis. We will narratively summarise the intervention types and those associated with improvements in prescribing outcomes. Additionally, we will narratively describe the prescribing related outcomes used by included studies. A quantitative synthesis (i.e. meta-analysis) will be considered if a sufficient number of homogenous studies are identified which examine the same outcome. We defined homogeneity based on several key factors, (i) study design (ii) outcome definition (iii) level of measurement.

If a meta-analysis is conducted, a random-effects model will likely be appropriate given the review question. We will not combine results from different study designs and interventions in an overall meta-analysis. Results will be presented in separate subgroups in the same forest plot (with no summary effect estimate) Heterogeneity will be assessed through a visual assessment and a logic-based assessment of study differences. We will conduct a standard Q-test statistic for heterogeneity and evaluate the heterogeneity via the I² statistic, which can be interpreted as the proportion of variability in the meta-analysis due to between study heterogeneity. Funnel plots will explore publication bias if more than ten studies are identified. These plots will help assess the relationship between effect size and study precision. As we believe that heterogeneity exists regardless of whether we happen to detect it using a statistical test, we will focus less on significance tests and instead further investigate the sources and impact of the heterogeneity (e.g Risk of Bias) through sensitivity analysis. In the event of substantial clinical or methodological heterogeneity, we will not report study results as pooled effect estimates and will synthesise study findings using the approach suggested in the Synthesis Without Meta-analysis (SWIM) guidance ³⁷ .”

All missing outcome data for included studies will be recorded on the data extraction form and reported in the risk of bias table. If there is insufficient information on the primary outcomes (due to inability to contact authors, unavailable data), these studies will be reported separately. Reasons for exclusion will be described and included in a supplementary table. As this systematic review is assessing an intervention targeted at primary care prescribers, included studies may have aggregate level patient data, thus it may not be possible to conduct population level subgroup analysis.

Discussion

With a growing population of individuals living with multimorbidity and polypharmacy, prescribing has become more challenging with a greater propensity for adverse outcomes ^{6, 19} . Preventable drug related morbidity (PDRM) has significant economic and social consequences at both the individual patient-level and for the wider healthcare system ³⁸ , it is thus vital to develop interventions to support safe and effective prescribing.

Interactive dashboards have become increasingly prevalent in healthcare settings, offering a versatile tool for visualising clinical data across various levels ranging from organisational, physician to patient-focused applications. They have the potential to enhance patient care and safety by providing contemporaneous feedback on potentially suboptimal treatment or care when integrated into clinical record systems ³⁹ .

Interactive dashboards have demonstrated varied effects on prescribing-related outcomes, such as antibiotic prescribing rates and appropriate statin use ^{40, 41} . Current evidence suggests they are most effective when combined with additional strategies which include education and/or behavioural components ⁴¹ . Given the limited number of eligible studies identified in previous reviews, the present systematic review will not restrict its focus to specific medication classes.

Strengths and limitations

This research will be conducted in line with Cochrane guidance. To increase transparency and reduce the risk of selective reporting this systematic review has been prospectively registered on PROSPERO, and will involve a search of the grey literature and trial registries to reduce the risk of publication bias. Title and abstract screening, full-text review, and methodological quality assessment will be performed by two reviewers working independently and blinded to each other's assessments, thereby minimising the potential for bias and errors. Excluding studies in progress but not yet published may lead to publication bias.

Potential implications for future research, policy and clinical practice

This research may identify gaps in the current literature and inform future intervention development with respect to how prescribing data may be fed back to prescribers. The findings from this review may inform policies aimed at enhancing or expanding the infrastructure necessary for effective e-prescribing, particularly those focused on optimising prescribing behaviours. In addition, this review will provide prescribers with a synthesised understanding of how interactive dashboards have been used, highlighting their potential benefits and limitations. This may lead to more informed decisions in regards adopting or optimising use of such tools in clinical practice, with the ultimate aim of improving patient safety and reducing medication related harm.

Funding Statement

PM is funded by an ICGP Post CSCST Fellowship award. CMC is funded by a HRB Clinician Scientist Fellowship Award (CSF- 2023-012).

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

Underlying data

No underlying data are associated with this article.

Extended data

This project contains the following extended data:

figshare: Supplementary file 1: Electronic search reports https://doi.org/10.6084/m9.figshare.25859506.v1 ⁴²

fighare: Supplementary file 2: PRISMA-P 2015 checklist https://doi.org/10.6084/m9.figshare.25887193.v1 ⁴³

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0) ( https://creativecommons.org/licenses/by/4.0/)