The impact of a weekly antimicrobial stewardship team intervention on antibiotic prescriptions in Dutch nursing homes: a protocol for a randomised controlled trial Authors: Kelly C. Paap, Lisa M. Kolodziej, Laura W. van Buul, Menno D. de Jong, Maarten F. Schim van der Loeff, Sacha D. Kuil Published: 2025 ABSTRACT BackgroundInappropriate prescribing of antibiotics for suspected urinary tract infections (UTI), respiratory tract infections (RTI) and skin and soft tissue infections (SSTI) in nursing homes (NHs) is common. Antimicrobial stewardship (AMS) has proven effectiveness in reducing inappropriate antibiotic prescriptions in hospitals. AMS has not been widely adopted in NHs due to limited evidence of its effectiveness in that setting. Therefore, the aim of this study is to evaluate the effect of a weekly AMS team on appropriate antibiotic prescribing in NHs. An additional aim is to determine internal and external validity and identify barriers and facilitators of implementing the intervention.Methods and analysisIn this multicentre, two-arm, randomised, controlled, non-blinded trial in Dutch NHs, four medical microbiology laboratories with their clinical microbiologists and eight NH organisations participate. NH clinicians will be randomised to the intervention or control arm (usual care) in a 1:1 ratio. The intervention consists of a weekly AMS intervention per NH organisation where audit and feedback is provided by a clinical microbiologist during 36 weeks. The primary outcome is the incidence rate (IR) of inappropriate antibiotic prescriptions per clinician per week. A sample size of 54 clinicians will be sufficient to demonstrate a reduction of 35% in the IR of inappropriate prescriptions, with a power of 90% at alpha = 0.05. Secondary outcomes are the IR of total antibiotic prescriptions, and the IR of inappropriate prescriptions with 95%CI will be estimated per arm and compared using univariable Poisson regression.DiscussionThis RCT will evaluate whether an AMS intervention in Dutch NHs results in a reduction in the incidence rate of inappropriate antibiotic prescriptions for the most common infections in NHs; a parallel process evaluation will inform possible future implementation. If proven effective, AMS should be considered for NHs.Ethics and disseminationThe Medical Ethics Committee of the Amsterdam University Medical Center waived ethical approval for this study. Written informed consent is obtained from all participants. Outcomes will be published in a peer-reviewed journal.Trial registrationClinicalTrials.gov NCT05998226. Registered on August 18, 2023.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13063-025-09303-w. Introduction Background and rationale Antibiotics are frequently prescribed in nursing homes (NHs), mainly to treat suspected infections of the urinary tract (UTI), respiratory tract (RTI) and skin and soft tissue (SSTI) [1]. A substantial proportion of the prescribed antibiotics in NHs is inappropriate [2–4]. A systematic review found that the decision to start antibiotics was appropriate in only 49% to 62% of the cases [4]. The main reason for inappropriate antibiotic use appears to be the prescription of antibiotics in case of a suspected, rather than a confirmed infection [5–7]. Diagnosing a bacterial infection in NH residents is challenging for several reasons. Many residents are unable to communicate symptoms, e.g. due to cognitive impairments, often have an atypical clinical presentation and may not always exhibit classical signs and symptoms of infection. In addition, pressure from nursing staff and NH residents and their family may influence prescribing decisions, as physicians may feel inclined to comply with family wishes, especially in uncertain clinical situations or during end-of-life care [8–10]. As a consequence, antibiotics are frequently prescribed out of risk avoidance (“better be safe than sorry”) rather than guideline-based [11]. Finally, guidelines adherence for antibiotic prescribing is limited [12]. In addition to the use of antibiotics when not indicated, incorrect prescribing, e.g. incorrect choice of antibiotic agent, dosage or duration, is also common [4, 13–16]. Inappropriate antibiotic use is a major contributor to antimicrobial resistance (AMR) development and associated with adverse outcomes, such as drug side effects, multi-drug interactions and Clostridioides difficile infection, especially in vulnerable older adults [17–22]. Therefore, optimising antibiotic use in NHs is important. Antimicrobial stewardship programs Antimicrobial stewardship (AMS) aims to optimise the appropriate use of antibiotics, ensuring the best therapeutic outcomes, while minimising the development of antibiotic resistance [23]. AMS is considered fundamental in combating the global rise of AMR and has therefore been implemented in hospitals worldwide to increase compliance to antibiotic guidelines [24–26]. Evidence for effectiveness of AMS in NHs varies and seemed to be most promising for UTI and interventions using audit and feedback, but high-quality clinical trials are required [27–29]. In a previously conducted pilot study from our group (a single-centre, before-after trial), the proportion of appropriate antimicrobial prescriptions increased significantly [30]. Therefore, the aim of this study is to assess the impact of a weekly AMS intervention on reducing inappropriate and overall antibiotic prescriptions for UTI, RTI or SSTI among NH residents. Methods Design and study setting This is a multicentre, two-arm, randomised, controlled, non-blinded trial to evaluate a weekly AMS team intervention compared to usual care in eight Dutch NH organisations with four allied medical microbiology laboratories (MML). A complete list of participating study sites is available upon request. This trial hypothesises that a clinician receiving a weekly AMS intervention will reduce inappropriate and overall antibiotic prescriptions for UTI, RTI and SSTI among NH residents. Table 1 provides a schematic overview of clinician enrolment, intervention and data collection (i.e. SPIRIT schedule) [31]. Table 1Clinician enrolment, intervention and data collection in the Trial among Nursing Home clinicians in the Netherlands, 2023-2024, SPIRIT scheduleAbbreviations: IR Incidence Rate Table 1Clinician enrolment, intervention and data collection in the Trial among Nursing Home clinicians in the Netherlands, 2023-2024, SPIRIT scheduleAbbreviations: IR Incidence Rate Clinician enrolment, intervention and data collection in the Trial among Nursing Home clinicians in the Netherlands, 2023-2024, SPIRIT schedule Abbreviations: IR Incidence Rate In the Netherlands, high-level care for vulnerable older adults in NHs is provided by various clinicians: (1) elderly care physicians (i.e. a separate medical specialisation focused on care of frail older adults with complex conditions in long-term care settings), (2) elderly care physicians in training or other physicians and (3) nurse practitioners or physician assistants. Microbiological testing and a consultative service is provided by clinical microbiologists of associated medical microbiology laboratories (MML). Inclusion criteria NH organisations are eligible to participate when their medication prescription system enables weekly antibiotic prescription data extraction. Clinicians are eligible to participate if (1) they are authorised to prescribe independently and (2) they expect to be employed at the NH for at least 4 months from date of randomisation. Intervention The intervention consists of a weekly AMS meeting with a clinical microbiologist and all clinicians of the same NH organisation randomised to the intervention arm for 36 weeks. In this AMS meeting, audit and feedback is provided by the clinical microbiologist on all antibiotic prescriptions during the prior week by the attending clinicians, based on a standardised protocol according to the current national guidelines [32–34]. Advice during the weekly AMS meeting will be updated if treatment guidelines change. No formal checks of clinicians’ adherence to the protocol are performed. No restrictions are imposed on participating clinicians, and participants cannot experience any form of harm during the trial. No provisions for ancillary or post-trial care are made. Data on the antibiotic prescriptions are extracted weekly from the medication prescription system by an NH employee or pharmacist. Control Clinicians in the control group do not participate in the AMS meetings and provide care as usual. Outcomes Primary outcome measure The primary outcome is in the domain of antibiotic prescribing, with the measure being inappropriate antibiotic prescriptions. The metric is the incidence rate of inappropriate prescriptions per clinician per week with 95% confidence intervals. The aggregation method calculates the average incidence rate per group (intervention vs. control) with 95% confidence intervals. The timepoint for assessment is weekly throughout the intervention period. Appropriateness will be assessed based on an algorithm of the current clinical infection treatment guidelines (see further) [32–34]. Secondary outcome measures The secondary outcomes are in the domain of antibiotic prescribing, with the measure being overall antibiotic prescriptions. The metric is the incidence rate (IR) of overall antibiotic prescriptions per clinician per week. The aggregation method involves calculating the average incidence rate per group (intervention vs. control) with 95% confidence intervals. The timepoint for assessment is weekly throughout the intervention period. Sample size We expect a full-time clinician to be responsible for an average of 80 NH residents and assume clinicians work for 0.6 full-time equivalent (Fte) on average. This leads to an expected average of 50 residents (0.6*80 = 48≈50) per clinician. We expect an average rate of 0.12 inappropriate prescriptions per clinician per week, conservatively based on an estimate derived from van Ewijk et al. [30]. We expect that the intervention will lead to a reduction of 35%, which means an average of 0.078 inappropriate antibiotic prescriptions per clinician per week. We will use a Poisson model with μ = 1.2 and SD = 1.5 to compare the incidence rate of inappropriate prescriptions between arms. To achieve 90% power for detecting the assumed effect, at a significance level of α = 0.05, 42 clinicians, with in total 1512 observation weeks are required. Allowing for 20% of loss of observation time due to drop out, we would need: 1.25*42 = 52.5, rounded to 54 clinicians (27 per arm). Recruitment NHs collaborating with the medical microbiology laboratories were identified and invited to participate by the laboratories. Following NH consent, lists of active prescribers were obtained. Clinicians were individually contacted, informed of participation requirements and assessed for eligibility by the researchers. Allocation All active prescribers were approached and assessed for eligibility. Those who provided participants informed consent will be randomly assigned to either control or intervention group with a 1:1 allocation as per a computer-generated randomisation schedule (Stata version 17, College Station, TX, USA). Because we expected that prescription behaviour might differ by clinician type, randomisation is stratified by clinician type in a 1:1 ratio. Randomisation is conducted by an independent statistician outside the study team, who was blinded to the study assignments. Blinding This is an unblinded study. Data collection and assessment of appropriateness Participant data are collected via standardised forms in Castor EDC (Castor EDC, Amsterdam, Netherlands) (data collection) and assigned a unique participant ID to protect personal identification (data coding). Data are checked for completeness and consistency through automatic validation rules and manual quality checks (data entry and validation) and securely stored in Castor EDC, compliant with relevant security and privacy regulations (data storage). Access is limited to three study researchers, with personal identification information stored separately and pseudonymisation applied (data handling). Data will be retained for 15 years and then securely destroyed in accordance with Dutch laws and regulations (data retention and destruction). All antibiotics prescribed by clinicians in the intervention and control arm during the 36-week intervention period will be extracted from the electronic health records (EHRs) including type of antibiotic, dosage, duration and interval by an NH employee or pharmacist. The appropriateness of prescribed antibiotics will be assessed based on retrospective chart reviewing. Information on demographics, clinical symptoms and diagnostic test results will be extracted from the EHR by two independent medically trained staff research staff members to assess appropriateness. Discrepancies in the assessments will be discussed with and decided by a third researcher. Assessment will be based on the current national guidelines and previously published criteria for classification of appropriateness in five categories (Table 2). The date of enrolment of the first participant is July 2023. Table 2Appropriateness of antibiotic prescriptions: 5 mutually exclusive categories. AMS study, the Netherlands, 2023–2024CategoryCategory nameDescription of categoryAppropriate/inappropriate1AppropriateAn infection was present that required antibiotic treatment, and the appropriate antibiotic was given (according to guideline and formulary) with correct dosage, interval and durationAppropriate2Incorrect use of an antibioticAn infection was present that required antibiotic treatment, and the correct antibiotic was prescribed but with incorrect duration, dosage or intervalInappropriate3Incorrect choice of an antibiotic*An infection was present that required antibiotic treatment, but an incorrect antibiotic was prescribed (i.e. prescription not according to guideline or formulary)Inappropriate4No indication for an antibioticAntibiotic treatment was started without sufficient indications of an infection needing antibiotic treatmentInappropriate5Not assessableThe appropriateness of the antibiotic treatment could not be assessed because of lack of information recorded in the EHRNot assessableCategorisation is based on van Ewijk et al. [30]*Additional outcome options: 3a. Deviated with good reason; 3b. Deviated without good reason Table 2Appropriateness of antibiotic prescriptions: 5 mutually exclusive categories. AMS study, the Netherlands, 2023–2024CategoryCategory nameDescription of categoryAppropriate/inappropriate1AppropriateAn infection was present that required antibiotic treatment, and the appropriate antibiotic was given (according to guideline and formulary) with correct dosage, interval and durationAppropriate2Incorrect use of an antibioticAn infection was present that required antibiotic treatment, and the correct antibiotic was prescribed but with incorrect duration, dosage or intervalInappropriate3Incorrect choice of an antibiotic*An infection was present that required antibiotic treatment, but an incorrect antibiotic was prescribed (i.e. prescription not according to guideline or formulary)Inappropriate4No indication for an antibioticAntibiotic treatment was started without sufficient indications of an infection needing antibiotic treatmentInappropriate5Not assessableThe appropriateness of the antibiotic treatment could not be assessed because of lack of information recorded in the EHRNot assessableCategorisation is based on van Ewijk et al. [30]*Additional outcome options: 3a. Deviated with good reason; 3b. Deviated without good reason Appropriateness of antibiotic prescriptions: 5 mutually exclusive categories. AMS study, the Netherlands, 2023–2024 Categorisation is based on van Ewijk et al. [30] *Additional outcome options: 3a. Deviated with good reason; 3b. Deviated without good reason When multiple reasons for inappropriateness apply, the highest category will be assigned. For example, when both dosage and choice of antibiotic are incorrect, the antibiotic prescription is assessed as 3 “incorrect choice of antibiotic”, and not as 2 “incorrect use of antibiotic”. For the analysis of the primary endpoint, the categorisation is dichotomised into appropriate and inappropriate; antibiotic prescriptions that are not assessable will be described separately. A process evaluation will be conducted in parallel, as shown in Fig. 1, where the structure of the study and the interplay between the randomised controlled trial (RCT) and process evaluation are visualised.Fig. 1Schematic overview of the structure and the interplay between the RCT and process evaluation. Legend: The RCT consists of a baseline and follow-up period for data collection, with an intervention period or usual care in between. A process evaluation follows parallel at the RCT. Quantitative data via questionnaires are obtained at baseline, 16 weeks and 32 weeks and will be supplemented after 32 weeks with qualitative data obtained via semi-structured interviews Schematic overview of the structure and the interplay between the RCT and process evaluation. Legend: The RCT consists of a baseline and follow-up period for data collection, with an intervention period or usual care in between. A process evaluation follows parallel at the RCT. Quantitative data via questionnaires are obtained at baseline, 16 weeks and 32 weeks and will be supplemented after 32 weeks with qualitative data obtained via semi-structured interviews The RCT data will be used to map the process of recruitment, randomisation and reach (sampling quality). Questionnaires at 16 and 32 weeks after start of the intervention will be sent to all clinicians in the intervention arm and one clinical microbiologist per MML. The questionnaires will evaluate participants’ perceptions on relevance, feasibility and the extent to which the intervention was performed as planned (fidelity) [35]. Semi-structured interviews will be conducted after 32 weeks with clinicians and clinical microbiologists who are purposively selected. Statistical methods A descriptive analysis of the enrolled clinicians by study arm will be provided. The incidence rate of inappropriate antibiotic prescriptions per clinician per week and the 95% confidence interval will be estimated per arm. The primary analysis will be on an intention-to-treat basis, including all clinicians who actually participated in the AMS meetings; we will also perform per-protocol analyses, including only clinicians who were randomised to the intervention arm, regardless of whether they actually participated in the AMS meetings. IR of inappropriate antibiotic prescriptions per study arm over the course of the study period will be plotted graphically per 4-week period. Rates will be compared between the two study arms using univariable Poisson regression. If there are substantial differences in the characteristics of the participants in the two arms, a multivariable Poisson regression will be performed to adjust for these confounders. An interaction term between time and study arm will be used to investigate whether there is a different time trend in the incidence of inappropriate prescriptions between the two study arms. There are no interim analyses or stopping guidelines. The incidence rate of all antibiotic prescriptions (regardless of appropriateness) will be estimated as secondary outcome measure and will be compared between study arms using univariable and, if indicated, multivariable Poisson regression. All analyses will be repeated for each of the three infection groups separately (UTIs, RTIs, SSTIs), but the trial is not powered to detect significant differences between arms in these sub-analyses. The level of statistical significance will be α = 0.05. Analyses will be done with Stata (Stata version 17, College Station, TX, USA). Discussion This protocol describes an RCT to evaluate whether an AMS intervention in Dutch NHs results in a reduction in the incidence rate of inappropriate antibiotic prescriptions for UTI, RTI and SSTI. Given the high proportion of inappropriate antibiotic prescriptions among NH residents, the search for effective interventions to promote appropriate antibiotic prescribing should be a priority in this vulnerable population. Although AMS has been widely implemented in hospitals, its application and effectiveness in NHs have not been well-studied. The current study aims to fill this gap. Prospective audit and feedback as used in this study is a promising way to improve the quality of antibiotic prescriptions in NHs [35, 36]. Moreover, the personalised feedback has also been shown to improve knowledge and clinical performance and to motivate clinicians to demonstrate good practice through competition [37]. This study will be the first RCT in multiple NH organisations and MMLs on AMS targeted at individual clinicians and including three of the most common bacterial infections. Furthermore, the primary outcome includes all aspects of inappropriate antibiotic use, i.e. incorrect indication, incorrect choice, incorrect dosage and incorrect duration of an antibiotic. We decided to randomise individual clinicians instead of institutions in this RCT since the prescribing behaviour of individual clinicians within the same institution varies strongly [38]. A limitation is the risk of knowledge transmission from the clinicians in the intervention group to those in the control group (contamination). Another potential limitation is that researchers and clinicians cannot be blinded due to the nature of the intervention. Through this weekly audit and feedback AMS intervention with clinical microbiologists, we intend to overcome the existing barriers to audit and feedback, i.e. resource limitations (trained personal) and difficulties in identification of prescriptions and indications [36, 39, 40]. A process evaluation will be conducted in parallel with the trial, using the framework of Leontjevas et al [41]. to assess the internal and external validity of the study [42]. The embedded process evaluation will offer important additional insights such as identifying the implementation barriers and facilitators, as NHs are a complex setting for conducting and implementing research [43]. If the intervention proves to be effective, the process evaluation will provide tools for broader application, and also provide directions and indications for further optimisation of the process. On the other hand, if the intervention proves ineffective, the process evaluation will help to interpret the results. In conclusion, this study will provide further insight into the efficacy and applicability of a weekly AMS intervention in NHs. Ultimately, robust evidence of efficacy will contribute to increased support for implementation of an AMS intervention in NHs. Supplementary Information Supplementary Material 1. Supplementary Material 1.