Background

Infection diagnosis in Intensive Care Units (ICUs) is a challenge given the spectrum of conditions that present with systemic inflammation, the illness severity and the delay and
imprecision of existing diagnostic methods. We hence sought to analyze the prevalence and predictors of confirmed infection after empirical antimicrobials during ICU stay.

Methods

Retrospective cohort of prospectively collected ICU data in an academic tertiary hospital in Sao Paulo, Brazil. We included all adult patients given a new empirical antimicrobial during their ICU stay. We excluded patients using prophylactic or microbiologically guided antimicrobials. Primary outcome was infection status, defined as confirmed, probable, possible, or discarded. In a multivariable analysis, we explored variables associated with confirmed infection.

Results

Out of After screening 1721 patients admitted to the ICU from November 2017 to November 2022, we identified 398 new antimicrobial prescriptions in 341 patients. After exclusions, 243 antimicrobial prescriptions for 206 patients were included. Infection was classified as confirmed in 61 (25.1%) prescriptions, probable in 39 (16.0%), possible in 103 (42.4%), and discarded in 40 (16.5%). The only factor associated with infection was deltaSOFA (OR = 1.18, 95% CI 1.02 to 1.36, p = 0.022).

Conclusions

Suspected infection in the ICU is frequently not confirmed. Clinicians should be aware of the need to avoid premature closure and revise diagnosis after microbiological results.
Development and implementation of new tools for faster infection diagnosis and guiding of antimicrobial prescription should be a research priority