Palma AI cuts ICU sepsis admissions by four points

At Hospital Universitario Son Llàtzer in Palma, the share of sepsis patients admitted to intensive care has fallen from 34.4% to 30.4%. An artificial intelligence system now flags a patient’s sepsis risk up to 24 hours ahead. The study tracked 8,039 people treated at the hospital between January 2011 and June 2024.

Researchers compared results before and after BIAlert Sepsis was introduced. Before the AI tool, the hospital used a conventional warning system based on clinical rules. Average hospital stays also fell by almost one day per patient.

BIAlert Sepsis analyses more than 70 pieces of clinical information from electronic patient records in real time. They include laboratory results, vital signs, demographic information and other relevant data. It refreshes the assessment every 30 minutes and estimates the risk of sepsis over the next 24 hours.

When the risk is high, the system sends an alert through the tools already used by medical staff. It also shows which patient data triggered the warning, helping doctors assess the patient and decide on tests, monitoring or treatment.

Marcio Borges, who leads Son Llàtzer’s multidisciplinary sepsis unit and headed the study, said earlier warning lets staff assess patients and act before their condition becomes more serious. The findings were published in PLOS Global Public Health.

The AI does not make medical decisions. A clinician still examines each patient and decides what happens next. Elisa Martín of the Knowledge Engineering Institute said the challenge is to fit such systems safely into hospital routines and provide clear, useful information when professionals need to act.

Topics: Court cases · Health

Source: Europa Press / Mallorca.com