Impact of hospital pharmacy interventions on length of stay in hospitals: a protocol for a systematic review and meta-analysis

Objectives

To our knowledge, no systematic review or meta-analysis has been conducted on this topic. Therefore, we aim to perform a systematic review and a meta-analysis to evaluate the existing evidence of the impact of hospital pharmacy interventions on LOS.

This systematic review will be an integral component of the Pharmacist in the Emergency Department (PharmED) trial¹⁵, which explores the integration of clinical pharmacists into Norwegian emergency departments. By synthesizing available data, we hope to contribute to the broader understanding of the role of the pharmacists in optimizing patient outcomes.

Protocol development

While developing this protocol, we have adhered to guidelines outlined in the PRISMA-P check list. This will ensure a structured and transparent approach in preparing and reporting our systematic review protocol.

Eligibility criteria

Participants. Studies with a primary focus on adult patients aged 18 years and older are eligible for inclusion. Any disease specific studies, e.g. studies including solely patients with Parkinson’s disease, will be excluded. Similarly, studies focusing on one group of medication, such as only anticoagulation therapy or antimicrobial therapy, will be excluded. Studies where the care unit is intensive care, psychiatric or maternity we will exclude, as well studies undertaken in nursing homes or residential care settings. Those studies might involve a disease/medication/care-specific approach from the pharmacists, potentially introducing bias into our meta-analysis.

Intervention. Studies addressing any intervention involving a hospital pharmacist or other hospital pharmacy staff such as pharmacy technicians or students, will be eligible. The intervention has to be aimed on general medical or surgical patient population, e.g. medication reconciliation and review performed in general internal medicine, surgical or emergency department. Studies describing interventions performed by pharmacists employed outside the hospital, for example in community pharmacy, will be excluded. Studies which describe hospital pharmacy services delivered by hospital pharmacy personnel in other settings, for example a pre-admission service delivered in the community, are eligible for inclusion.

Outcome. Studies reporting LOS as a primary or secondary outcome

Study design. Peer-reviewed trials with a control group using a sample size greater than 100 participants will be considered for inclusion if they have conducted an analysis and presented quantitative data on LOS. A larger sample size enhances statistical power, increasing the likelihood of detecting meaningful differences or effects, and lead to narrower confidence intervals, improving the reliability of findings.

Information sources

Five databases will be searched: MEDLINE (Ovid), EMBASE (Ovid), Cochrane Trials, Cochrane Reviews and the Cumulative Index to Nursing and Allied Health Literature (CINAHL, via EBSCO). We might need to request some additional data from the study authors, such as providing mean values for length of stay when only medians are published, or supplying standard deviations or confidence intervals, to facilitate meta-analyses.

Search strategy

The searches for all databases will be drafted by an experienced librarian (MS). All searches will include a combination of free text terms and controlled vocabulary terms related to the core elements of the research question (pharmacists/pharmacy service, hospital, length of stay). The full search strings are available in Table 1 and Table 2. Once developed, the search will be validated by ensuring that targeted pre-identified papers are retrieved by the search.

Selection of studies

The search results from each database will be uploaded to our review project in Covidence (https://www.covidence.org/) and automatically deduplicated. Two reviewers will then independently screen the titles and abstracts of the retrieved records for eligibility within Covidence and acquire the full texts of the potentially eligible studies. Only studies published in English since January 1^st 2014, to reflect the most recent evidence, will be eligible. Inclusion or exclusion of an article from further review require consensus between two reviewers. Any discrepancies between two reviewers will be resolved by a third reviewer. Covidence will monitor the screening progress and generate a study selection flow chart for the review.

Data extraction process

A data extraction template will be used to extract all necessary data to complete the review, structuring data about the study design and LOS definition; participant, intervention, comparator and outcome characteristics, in addition to the statistical methods employed in each study to examine the relationship between the intervention and LOS, and the study findings. The LOS in the hospital will be given as the primary outcome of interest for this review, and reported as measured by the authors, for example, as the mean, median, percentage reduction or other.

Data analysis

The LOS data for each study group, along with the statistical methods used to compare the intervention and comparator groups, will be documented as reported by the authors in each included study. Where possible, continuous variables will be presented as Standardized Mean Difference, while dichotomous variables will be expressed as risk ratios.

Continuous data, not presented as a mean or standard deviation, will be examined for potential for mathematical conversion. This may involve request for additional data from the study authors, and/ or confirmation that the data were normally distributed.

For data reported as median and interquartile range (IQR), bar charts will be used to express the difference in median days between intervention and comparator groups.

A narrative synthesis will be used. If possible, results of comparable trials will be pooled using the fixed effect or random effect model and 95% confidence intervals, as appropriate relative to the degree of statistical heterogeneity.

We will use the I² statistic and the Cochran’s Q test to assess the heterogeneity among the studies, and explore the potential sources of heterogeneity using subgroup analyses and meta-regression, if feasible. We will assess the publication bias using funnel plots and Egger’s test, if there are at least 10 studies in the meta-analysis. We will use Covidence to export the data into Excel or RevMan formats, to support data synthesis and analysis.

Data presentation

The extracted data will be systematically organized, described and presented to illustrate key patterns. The aggregation and analysis of the data will be conducted with the aim of ensuring comprehensive synthesis of the information. Specific findings related to LOS will be detailed in tabular and graphical format.

Risk of bias and quality assessment

We will assess the risk of bias of the included studies in Covidence using the appropriate tools according to the design of the study. Randomised trials will be assessed using the Cochrane Risk of Bias Tool (ROB 2)¹⁶, which evaluates the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of bias. Non-randomised trials will be assessed using the Risk Of Bias In Non-randomized Studies - of Exposure (ROBINS-E) tool¹⁷. We will assign a judgment of low, high, or unclear risk of bias for each domain, and provide a rationale and a supporting quote for each judgment. One reviewer will perform the risk of bias assessment, and this will be independently validated by a second reviewer for a proportion of the included studies. The risk of bias results will be reported in a summary table and a graph.

For the randomised studies, the overall level certainty of evidence for the benefit of the intervention on reducing LOS will be scored as high, moderate or low certainty using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) with GRADEpro GDT¹⁸. The outcome will be assessed for ‘not serious’, ‘serious’ or ‘very serious’: bias risk, inconsistency, indirectness, imprecision and publication bias. For each ‘serious’ score, the level of certainty will be downgraded by one unit.

Research ethics committee approval

Given the nature of systematic reviews, which involve the analysis of previously published data, our study does not require approval from a research ethics committee.

Study status

The literature search phase of our systematic review has been completed successfully. We have meticulously screened the titles, abstracts, and eventually full texts of the articles identified. Currently, we are engaged in extracting data from the articles that have undergone rigorous selection criteria and constitute the final collection for inclusion in our review.