Ventilator-associated pneumonia (VAP) is the most frequent infection in the intensive care setting. For VAP there is currently no reliable diagnostic criteria. We aimed with the present study, using data from the mechanical ventilator to identify early this infection using artificial intelligence methods .

Ventilator-associated pneumonia (VAP) is defined as a hospital-acquired pneumonia occurring in patients submitted to invasive mechanical ventilation (MV) for at least 48 hours. VAP represents the most prevalent nosocomial infection in the intensive care setting. VAP is burdened by prolonged duration of MV and hospital length of stay and consequently increases hospital costs. Moreover, mortality and antibiotic use are also significantly affected. Unfortunately, there is currently no valid, accurate diagnostic criteria of VAP because even the most widely used ones are neither sensitive nor specific.. The insufficient sensitivity of these criteria to rule out VAP carries the risk of antibiotic overuse with the consequently emerging of antibiotic resistance and superinfections. On the other hand, the insufficient specificity to rule in VAP carries the risk of delayed administration of antimicrobial therapy leading to increased mortality. Ventilator-associated event surveillance failed to accurately identify VAP, too . The purpose of the present study is to develop different AI-algorithms using data continuously recorded form the mechanical ventilator in supporting clinicians for the early detection of VAP. An accurate AI-algorithm for early VAP identification has the potential to reduce morbidity, mortality, exposure to broad-spectrum and/or unnecessary antibiotics and finally to reduce costs.

Inclusion Criteria:

• adult patients admitted to our ICU requiring invasive respiratory support for at least 48 hours

Exclusion Criteria:

• previuos pneumonia

All consecutive adult patients admitted to our ICU requiring invasive respiratory support for at least 48 hours will be enrolled