AI Provides Same-Day Prediction of Bloodstream Infection and Antimicrobial Resistance in ICU Patients

Infections in the bloodstream can become resistant to antibiotics, leading to the potentially life-threatening condition known as sepsis. Once an infection escalates to sepsis, there is a high likelihood that patients will quickly develop organ failure, shock, and even death. Current methods for assessing patients in intensive care units (ICUs) are time-consuming and involve lengthy laboratory tests that require culturing bacteria, a process that can take up to five days. This delay can severely impact patient care outcomes, particularly for ICU patients who are often critically ill. Access to this information sooner would allow clinicians to make faster and more informed decisions regarding treatment, including the use of antibiotics. The appropriate use of antibiotics is closely linked to improved patient outcomes. Researchers are now leveraging the power of artificial intelligence (AI) to evaluate antimicrobial resistance in ICU patients and identify bloodstream infections that cause sepsis.

Patients with drug-resistant infections often arrive in the ICU in critical condition, and they may not survive long enough for traditional diagnostic methods to determine their infections. Factors such as prior exposure to antibiotics, genetic predispositions, and dietary influences can contribute to varying levels of antimicrobial resistance among patients, affecting their microbiomes. Consequently, clinicians face a challenging scenario in which they must administer broad-spectrum antibiotics in a "blinded fashion" to save the patient’s life, despite the risk of harming beneficial microbes in the microbiome and potentially exacerbating the pathogen's resistance to treatment.

A collaborative team from King's College London’s Faculty of Life Sciences & Medicine (London, UK) and clinicians at Guy’s and St Thomas’ NHS Foundation Trust (London, UK) has undertaken an interdisciplinary study aimed at improving outcomes for critically ill patients. This research utilized data from 1,142 patients at Guy’s and St Thomas’ NHS Foundation Trust, laying the groundwork for ongoing investigations involving datasets of over 20,000 individuals. The team has made notable advancements in demonstrating how AI and machine learning can facilitate same-day triaging for ICU patients, particularly in settings with limited resources. This technology proves to be significantly more cost-effective than traditional manual testing. The researchers hope that a more sophisticated version of this study, particularly within a multi-hospital framework using Federated Machine Learning technology, could meet regulatory requirements for actual deployment of this AI approach.

“Our study provides further evidence on the benefits of AI in healthcare, this time relating to the crucial issues of antimicrobial resistance and bloodstream infections,” said first author Davide Ferrari, King’s College London. “Our use of machine learning provides a new way of tackling the important clinical issue of antimicrobial resistance. We hope that the AI will provide a useful tool for clinicians in making important decisions, particularly in relation to ICU.”

“The findings of this study are incredibly promising as using AI to speed up the diagnostics of infection to allow for prescription of the correct antibiotic could not only have a huge impact on the patient’s survival and their care outcomes; but could help to preserve the antibiotics we already have developed and prevent the development of further antibiotic resistance," added Dr Lindsey Edwards, expert in microbiology at King’s College London.