JRO protocol
As per the UCLH JRO template headings
See UCL/UCLH Interventional studies protocol template (section number matches the template)
Use this document to manage version control and print or convert to a word document directly from here when ready to submit
1. Introduction
Leave for now: will be written after Background and Rationale complete
Via JRO: Overview of the study; it should be sufficient to guide the reader to the main purpose of the study, how it will be conducted, on which population(s) and its expected benefits. It should say how the results of the study would benefit in terms of clinical practice, policy or the NHS as whole. The introduction may include a study flowchart (recommended: allows users of the document to follow the participant and study pathway with ease, e.g. via a Gantt chart or timeline of activity), and should detail whether this project is being conducted in relation to an academic qualification (a student project), or is related to any previous research sponsored by UCL.
2. Background and Rationale
Antimicrobial Resistance
Antimicrobial resistance (AMR) is a global problem, with an estimated 1.9 million deaths annually attributable to AMR forecast by 2050 (GBD 2021 Antimicrobial Resistance Collaborators, 2024, 10.1016/S0140-6736(24)01867-1). In 2019, it was estimated to cause more deaths than HIV/AIDS or Malaria. Within England, the burden of AMR is increasing with London comprising the highest rates of resistant bacteraemia. AMR rates are highest in deprived areas and certain ethnic groups, with particular impact on the elderly (ESPAUR 23-24 report). Human antibiotic consumption is a major driver in resistance at a population level and at an indicidual level(Thom KA, et al. Clin Infect Dis 2019; 68: 1581–84.)
The Antibiotic Prescribing Problem
Most antibiotics are given empirically before culture results are available. Such results are available in fewer than half of patients, and take 2-3 days.
Empiric prescribing is through probabilistic clinical assessment of the likely source of infection e.g. urine, chest etc, but clinicians may also start treatment to cover multiple potential sources where there is uncertainty.
Where a particular source is suspected, clinical guidelines exist to help standardise care. Guidelines are produced by combining best available evidence on population level treatment effectiveness, with local data on bacteria prevalence and resistance patterns. Guidelines also reflect best practices with respect to antimicrobial stewardship, e.g. limiting the use of extended-spectrum agents where possible to reduce resistance.
However, clinicians vary in how they apply these guidelines in practice, in some cases showing low levels of concordant prescribing e.g. 58% concordance in CAP (BTS audit data).
Variation may be warranted, for example under conditions of diagnostic uncertainty, or where there is additional information about pre-existing treatment failure or known antimicrobial resistance.
However, variation may also be unwarranted where guideline-based care is appropriate and not followed, potentially leading to over- or under-treatment, treatment failure, and antimicrobial resistance.
Disease specific background
Urinary Tract Infections (UTIs) contribute to approximately 190,000 hospital admissions in England, resulting in 1.2 million bed days in 2023/4. They are the most frequently occurring healthcare associated infection and account for approximately 23% of all antibiotic prescriptions in primary care, second only to respiratory tract infections in the UK (Dolk et al. 2018, 10.1093/jac/dkx504).
Most UTI prescriptions are given empirically before culture results are available. An estimated 40% of bacteria that cause UTIs can be resistant to the antimicrobials used, with 18-21% of patients experiencing treatment failure. (McCowan et al 2022, https://doi.org/10.1186/s12879-022-07768-7, Moon et al 2022, https://doi.org/10.1371/journal.pone.0277713). UTI recurrence is common, with repeated infections, further compounding antimicrobial use and resistance.
Inadequately treated UTIs can lead to bacteraemia, with approximately 50% of E Coli bacteraemia originating from a urinary tract source (ESPAUR 2025). In 2024, Enterobacterales comprised 80.6% of urinary isolates, Escherichia coli (E. coli) accounting for around 70% of all episodes (ESPAUR 2025). These enterobacterales (E. coli, K. pneumoniae, and K. oxytoca) which commonly cause both UTIs and bacteraemia, accounted for 85% of resistant bacteraemia in 2024 (ESPAUR 2025).
Dilemma of empirical prescribing
Given this risk of resistance, prescribing the correct antibiotic for the correct pathogen is crucial to reduce immediate harm to the patient. However, a dilemma exists in prescribing broader-spectrum or unnecessary antibiotics will drive resistance, causing future harm to both the patient and society.Therefore, optimal antibiotic prescribing should be guided by confirmation of infection, and microbiology susceptibility testing of isolates, whilst taking into account the future harm of resistance. Several challenged hinder optimal prescribing in practice.
Microbiology results take days, meaning initial treatment decisions are made empiricially before susceptibility data is available. Prior antibiotic usage and resistance history is therefore crucial to guide therapy, yet this is often fragmented and not readily accessible due to patients accessing multiple providers across different non-interoperable systems. Even when information is available, the volume and complexity of clinical and microbiological data can make accurate, timely decision-making difficult, particularly in emergency settings where clinicians are under pressure. This challenge is further amplified by the fact that most prescribing decisions are not made by infection specialists, meaning that the nuanced judgement required to balance immediate patient need against the future risk of driving resistance is not always available at the bedside.
As a result, prescribers often rely on heuristics and local guidelines, rather than patient specific data. Whilst an infection specialist input can be sought, the volume of cases would make individual review impractical. Consequently, empirical therapies are frequently initiated without full consideration of patient-specific risk factors (9).
The EHR Opportunity for Antibiotics
The NHS 10 year plan calls for all NHS trusts to adopt an Electronic Health Record Systems (EHRS),presenting a significant opportunity to address the challenges of antimicrobial prescribing. Increasing adoption of comprehensive EHRS, encompassing prescribing, patient records and laboratory results including microbiology data creates an opportunity to creates the foundation of learning health system that can improve antibiotic use.
This may be achieved through modification of EPR architecture to embed better prescribing practices. This can be through real-time intervention via clinical decision support systems (CDSS) at the point of prescribing. Although conceptually straightforward, successful CDS alerts must deliver accurate information, in clinical context and at the point of care, and must be well-integrated into the clinical workflow. Otherwise they suffer from poor usability, alert fatigue which can itself be a patient safety risk.
The potential of this approach is supported by studies such as the INSPIRE trial which demonstrated that prompts at the time of prescribing significantly reduced the use of extended spectrum antibiotics.
The NHS 10-year plan calls for all NHS trusts to have adopted a digital electronic patient record (EPR) by March 2026.
With increasing adoption of comprehensive EPRs, including electronic prescribing and linkage across other electronic health records, there is the potential to create learning health systems for improving antibiotic use.
This may be through:
Modification of the EPR architecture to improve antibiotic prescribing practices.
[REF]
Interceding with antibiotic prescriptions in real-time via computerised clinical decision support systems (CDSS).
The INSPIRE cluster randomised trial demonstrated that computerised provider order entry (CPOE) prompts reduced use of extended spectrum antibiotics.(Gohil et al., 2024, https://doi.org/10.1001/jama.2024.6248)
Linkage across EHRs to deliver additional information not normally visible to clinicians at the point of antibiotic prescribing.
[REF]
Research Question
In patients attending secondary care, diagnosed with an acute infection, can an electronic nudge embedded into clinical workflows improve guideline-concordant antibiotic prescribing safely and effectively?
In adult patients presenting to secondary care with a diagnosis of urinary tract infection or pyelonephritis, does the availability of in situ recent and clinically relevant microbiological culture and sensitivity data at the point of antibiotic prescribing reduce the rate of antimicrobial pathogen mismatch, compared with standard empirical prescribing without prior microbiology?
Research Question 2
From UKHSA , ESAPUR report
3. Objectives and outcome measures
Primary objective
Evaluate the effectiveness of a clinically-integrated digital prescribing nudge for improving administration of guideline-concordant antibiotic treatment.
Evaluate the effectiveness of a clinically integrated digital prescribing nudge for improving administration of antibiotics which reduce pathogen drug mismatch.
Secondary objectives
Evaluate the safety of the digital prescribing nudge
Evaluate the acceptability and utility of the nudge to clinicians
Outcome measures/endpoints
Effectiveness:
Proportion of guideline-concordant antibiotic prescriptions in the nudge vs. standard care study arms at 24 hours after initial treatment initiation.
Proportion of antibiotic switches from initial prescription at 24, 48 and 72 hours.
Proportion of empirical escalation of antibiotic treatment (change in antibiotic treatment independent of new microbiological evidence).
Antibiotic defined daily dose per adult per admission.
Safety:
All cause in-hospital and 90-day mortality.
Length of hospital stay.
Escalation to level 2 or 3 care.
Readmission to hospital within 30 days of discharge.
Time to intravenous to oral antibiotic switch.
Rates of antibiotic co-prescription.
Presence of infection or colonisation by antimicrobial resistant pathogens
Incidence of specific infections e.g. C. difficile
4. Trial design
A single centre, randomised, parallel assignment digitally integrated service evaluation.
Adult patients attending hospital who receive an antibiotic order for an infection, who are prescribed an antibiotic will be included. There are no exclusion criteria.
Recruitment will be automatic. Patients who fulfil the above eligibility criteria will be enrolled in the study and randomised between intervention and standard care arms. All study data will be extracted from the electronic patient record. There will be no additional testing or follow up requirements.
Description of the proposed study workflow:
Patient diagnosed with infection. The patient enters the study at the point of antibiotic order placement by the treating clinician. As such, they may be located in the emergency or outpatient departments or may be an inpatient. In all cases, they will be receiving their first antibiotic order for the current encounter.
Clinician opts for antibiotic treatment and selects desired antibiotic from orders tab search box (e.g. Co-Amoxiclav).
Clinician enters antibiotic indication data within order window (mandated step) for body system and specific indication (e.g. Lung/CVS, Community Acquired Pneumonia, Mild). The dose, timing, frequency and duration are entered. The order is accepted and appears in the orders sidebar for signature.
There are then two options for delivering the guideline-adherence nudge:
Option 1 - Order Validation Prompt
Closing the antibiotic order (or signing the order, tbc) triggers silent evaluation of logic rules determining order appropriateness (compliance against existing guidelines) and second stage eligibility criteria:
Is this the first antibiotic order associated with the current patient encounter (Y/N)
Is there discordance between the selected antibiotic, highlighted indication data and clinical guideline (Y/N)
Discordance is determined by applying the following logical criteria:
1) Does the patient have a documented allergy to penicillin or a penicillin containing antibiotic preparation?
2) Does the selected antibiotic for the stated indication align with the guideline recommendation?
Has the patient already been randomised within the study during the current patient encounter (Y/N)
If the patient is receiving the first antibiotic order for the encounter, and the order is discordant with the recommended treatment in the clinical guideline, and the patient has not been previously randomised within the study within the current encounter, then a fourth screening step will occur:
Does the patient have a documented allergy to the guideline-concordant antibiotic recommendation? (Y/N)
If the patient is allergic to the guideline concordant antibiotic then the alert is suppressed and the patient is not randomised.
Option 2 - Our Practice Advisory
Exactly the same workflow but fires after signing process. Applies same logical criteria. Under discussion whether 1 > 2 in terms of delivering randomisation and supporting above logic.
If the eligibility criteria are satisfied then the patient will be randomised to alert or standard care arms (no alert).
If randomised to the alert arm, the intervention will display to the clinician during the antibiotic order process: "You are selecting [abx X1] for [indication Y]. UCLH guidelines suggest use of [abx X2] for [indication Y]. Click here to change the current antibiotic order to [abx X2], or select a clinical justiification for deviating from the clinical guideline: 1) Requirement for extended spectrum agent use given diagnostic uncertainty 2) Requirement for extended spectrum agent use given additional clinical risk 3) Approved extended spectrum agent use from microbiology 4) Other (Comment)"
5. Sampling methods
Inclusion criteria
Exclusion criteria
Recruitment
Consent
6. Intervention
7. Trial procedures
8. Finance and supply of equipment
9. Data management
10. Statistical considerations
11. Assessment and management of risk
12. Recording and reporting adverse events
13. Oversight committees
14. Regulatory review and patient and public involvement
Regulatory review
Peer review
Patient and public involvement
15. Monitoring and auditing
16. Training
17. Insurance and indemnity
18. Record keeping and archiving
19. Intellectual property
20. Publication and dissemination
21. References
22. Appendices
Last updated