AI Models Helping Combat Antibiotic Resistance in ICU – Personalized Decision Support for Empirical Antibiotic Therapy

Objectives: To develop and rigorously evaluate an AI model that strategically integrates the heterogenous data to support personalized empirical antibiotics therapy decisions in ICU to advance antibiotics stewardship and combat antibiotics resistance. Hypothesis to be tested: an advanced AI model can accurately predict the pathogen and antibiotic susceptibility for critically ill patients and achieve improved empirical antibiotic treatment strategy than the experience-based practice. Design and subjects: This is a retrospective study of patients who have been admitted to ICUs at Tuen Mun Hospital and Pok Oi Hospital since 2008. Instruments: Medical history, clinical and laboratory parameters, and clinical notes available in the Clinical Information System and electronic Patient Records. Interventions: No additional interventions are required for this study. Main outcome measures: The accuracy of predicting the pathogen, antibiotic susceptibility, and the toxic antibiotic concentration for ICU patients. Data analysis: All parameters will be used for the AI model. Quantitative analysis will be carried out to evaluate the performance. Expected results: An advanced AI model could accurately predict the pathogen and antibiotic susceptibility, thereby outperforming the existing practice in terms of recommending more effective yet less broad-spectrum antibiotics.