Clinical Impact of a Machine Learning Decision Support System for Empirical Antibiotic Therapy

The goal of this quasi-experimental study is to analyze if a Machine Learning Clinical Decision Support System can improve the empirical antibiotic treatment in patients with pneumonia, urinary tract infection and / or sepsis. The main questions it aims to answer are: * Primary outcome: clinical success defined as clinical cure (resolution of all signs and symptoms related to infection); no complications until day 30 (recurrence, or development of adverse events- AEs-); no new acquisition of MDROs; and survival at day 30. * Secondary outcomes: a subgroup analysis of the primary outcome according to the department participants, infectious syndrome, severity of the infection assessed by the SOFA score, and in microbiological confirmed infections. In microbiological confirmed infections, desirability of Outcome Ranking (DOOR) for the Management of Antimicrobial Therapy (MAT) according to the beta-lactam classification Researchers will compare a pre-intervention group with a post-intervention to see if improve in the DOOR MAT score Participants in the post-intervention group will: • Received empirical antibiotic therapy prescribed by their treating physicians according to the machine-learning recommendations

Inclusion Criteria: * Adult patients (aged ≥18 years) * Admitted to the Nephrology, Oncology or ICU wards * Diagnosis of sepsis, pneumonia and/or UTI * Empirical antibiotics prescribed Exclusion Criteria: * informed consent obtained \> 48 hours since the infection onset * beta-lactam allergy * infection syndrome other than sepsis, pneumonia or UTI * confirmed no-bacterial infection * death within the first 48 hours of inclusion or imminent risk of death at time of the inclusion * pregnancy and/or breastfeeding * inclusion in a clinical trial of antimicrobial treatment