ADAPT-AST (Adaptive Antimicrobial Susceptibility Testing)

Overview

The goal of this study is to improve the way urinary tract infections (UTIs) are tested for antibiotic resistance. The main questions it aims to answer are: * Can the investigators use a method called Bayesian causal inference to create or check clinical prediction models that help predict if certain antibiotics will work for a urinary infection, using patient information from the National Health Service (NHS)? * Can this new ADAPT-AST method, which uses data and a smarter approach, do a better job of testing for urinary infection than the old methods? Will it help doctors make quicker decisions and save resources by being more efficient? Participants in this study will not be receiving treatments. The study will involve: Using statistical methods to predict UTI test results based on patient data. Evaluating whether this new approach can provide doctors with more timely and useful information for treating UTIs. Assessing whether it can help save money and resources in the lab and pharmacy.

The aim of this study is to develop and evaluate an adaptive informatics approach for laboratory antimicrobial susceptibility testing (AST) for urinary tract infection (UTI) pathogens compared with current practice to improve patient outcomes, reduce AMR risks and reduce waste of laboratory resources. UTI is a leading cause of community and hospital acquired infection and a major driver of antimicrobial prescribing in primary and secondary care. The continued proliferation of AMR also increasingly limits treatment choices for many UTIs. Despite the importance of UTI, antimicrobial susceptibility testing (AST) of urine specimens is based on inflexible 'one-size-fits' all standard operating procedures (SOPs). Either a very large unfocused panel of antimicrobials is immediately tested (leading to wasted resources), or more commonly, and particularly in low or middle income (LMIC) settings, a selected subset of antimicrobials is tested at day one prior to a second or even third panel of antimicrobials. Such an approach does not adapt to prior information such as previous resistance patterns, antimicrobial prescribing, or demographic information, despite these factors being powerful (strong) predictors of resistance. This results in imprecise, inefficient, and inequitable provision of antimicrobial susceptibility information, which provides suboptimal support of decisions for treatment of UTI. This project will use statistical techniques based on Bayesian causal inference to predict urine AST results and prioritise testing using patient demographics, prescribing, admission, and microbiology laboratory care data. The clinical utility of resulting algorithms will be evaluated in terms of their ability to increase the number, timeliness and appropriateness of usable AST results available to clinicians, and their ability to reduce laboratory resource costs through better test prioritisation. The anticipated benefits of a successfully developed, evaluated, and implemented system are faster and more precise treatments of UTI in patients with drug-resistant organisms and more efficient resource management, particularly in laboratory and pharmacy workflows.