ÂÒÂ×Ðã

Laura Shallcross has been successful in her NIHR Clinician Scientist application on antimicrobial resistance using hospital EHRs and host transcriptomics. 

Scientific summary of research

Background: Reducing inappropriate antibiotic prescribing to tackle antimicrobial resistance (AMR) is a global health priority. This is widely acknowledged, yet physicians continue to prescribe antibiotics for patients in whom bacterial infection is unlikely. Antimicrobial stewardship is the only approach with potential to immediately impact on AMR by preserving the lifespan of existing antibiotics. 

The major burden of drug-resistant infection is in hospital, but antibiotic prescribing in this setting is poorly understood due to lack of individual-level prescribing data. Without this information we cannot estimate the number of patients who could be safely treated without antibiotics, or devise evidence based stewardship initiatives to improve prescribing quality. 

Aims: I propose a novel, precision medicine approach to antibiotic prescribing in hospital, using high resolution electronic health records (EHRs) and host transcriptomic bio-signatures to guide prescribing decisions. I propose to test the safety and acceptability of this approach in a proof of principle study. I focus on urinary tract infection (UTI) syndromes as an exemplar of antibiotic overuse in hospital. 

Plan of investigation: I will analyse electronic health records (EHRs) from more than 9000 patients at University Hospital Birmingham to describe the clinical course, management and outcomes of patients with UTI in order to estimate antibiotic overuse. Applying conventional statistical and machine learning approaches in parallel to these EHRs, I will develop a series of risk models to predict each patient’s need for antibiotic treatment. For individuals who require antibiotics, I will predict the choice and duration of antibiotic treatment based on the patient's clinical syndrome, their risk of drug-resistance and the severity of their illness. In order to determine whether inclusion of omic data can improve predictive accuracy, I will collect RNA samples from peripheral blood from 300 patients with UTI syndromes.  I will investigate whether previously defined host transcriptomic bio-signatures can discriminate between patients with different UTI syndromes which merit different types and duration of antibiotic treatment. 

Provided my models can adequately discriminate patients who are unlikely to require antibiotic treatment for UTI, I will use them to identify a target population of patients for inclusion in a proof of principle study.  This study will be designed in collaboration with patients, members of the public and healthcare professionals to test the safety and acceptability of an intervention (such as antibiotic discontinuation at 24 hours following hospital admission) to reduce unnecessary antibiotic prescribing. If this study suggests the intervention is feasible, this work will lay the foundation for a future randomised controlled trial of risk-assessed antibiotic prescribing beyond this fellowship. 

Benefits to patients and the NHS: This work will inform the development of novel diagnostic and risk prediction tools to reduce inappropriate antibiotic prescribing in hospital.  It will benefit patients by reducing the number of people who get side effects from unnecessary antibiotics and will benefit society by reducing antibiotic resistance. In addition, we will develop a unique resource of clinical samples linked to high-quality electronic health records as the basis for a future program of research on infection and AMR in hospital.