Multicenter Validation of a Risk Prediction Model for MDRGNB Infection

Adult ICU patients were prospectively and consecutively enrolled across 11 research centers. MDRGNB infection risk predictions were generated within 24 hours of ICU admission using routinely available clinical data. Without interfering with clinical decision-making, the study assessed the discrimination, calibration, and clinical net benefit of the previously developed MDRGNB infection risk prediction model in real-world practice, applying the predefined threshold of 0.318.

Inclusion criteria

• Age ≥18 years;
• ICU stay ≥48 hours;
• At least one clinical microbiological specimen collected and submitted for testing within 48 hours of ICU admission, with the first submission time designated as the index time;
• Core predictive variables of the model extractable from electronic medical records or laboratory information systems.

Exclusion criteria

• For patients with multiple ICU admissions, only the first ICU stay was retained;
• Missing or indeterminate primary outcome;
• Missing key predictive variables that could not be handled according to prespecified rules.