A study led by UF researchers examining the role machine learning plays in evaluating early and cumulative beta-lactam antibiotics exposure for inpatient pneumonia was recently featured on Contagion Live — a news source for practitioners studying and treating infectious diseases.
Mohammad Alshaer, Pharm.D., Ph.D., a research assistant professor of pharmacotherapy and translational research in the University of Florida College of Pharmacy, was interviewed about the study at the 24th Annual Making a Difference in Infectious Diseases Meeting, or MAD-ID, The Antimicrobial Stewardship Meeting, which took place in Orlando on May 18-21.
The UF study used machine learning to retrospectively evaluate outcomes for more than 700 intensive care unit patients who were infected with hospital-acquired and ventilator-associated pneumonia. These infections are common in hospital ICUs and contribute to longer hospital stays and mechanical ventilation, as well as death.
The UF researchers examined beta-lactam exposure across different time periods and used machine learning algorithm to rank the top predictors. They found drug exposure during the first 24 hours to be the top predictor of clinical cure in ICU patients and recommended beta-lactam exposure should be optimized early and maintained through the duration of therapy.
“Machine learning is growing in all fields, including infectious diseases,” Alshaer told Contagion Live during the interview. “It’s promising in this area [beta-lactams’ exposure and precision dosing] that once you have rich data including antibiotics exposure and different co-variates related to your patient this may increase the accuracy of the predictions of these outcomes in the future.”
The study, “Using machine learning to define the impact of beta-lactam early and cumulative PK/PD target attainment on outcomes in ICU patients with hospital-acquired and ventilator-associated pneumonia,” was presented as a poster at the MAD-ID meeting and was published in the journal Antimicrobial Agents and Chemotherapy.