Predicting Beta-Lactam Target Non-Attainment in ICU Patients at Treatment Initiation: Development and External...

Area of Science:

• Pharmacokinetics and Pharmacodynamics
• Clinical Pharmacy
• Infectious Diseases

Background:

• High infection-related mortality in ICUs underscores the need for effective antibiotic therapy.
• Beta-lactam target non-attainment affects up to 45% of ICU patients, reducing treatment success.
• Optimizing antibiotic dosing is critical for improving clinical outcomes in critically ill patients.

Purpose of the Study:

• To develop and validate prediction models for beta-lactam target non-attainment in ICU patients.
• To identify key patient variables associated with suboptimal antibiotic exposure.
• To provide tools for optimizing antibiotic treatment strategies in the ICU.

Main Methods:

• Development and validation of prediction models using random forest (RF), logistic regression (LR), and naïve Bayes (NB) algorithms.
• Utilized data from two multicenter studies including 376 patients for model development.
• External validation performed on 150 ICU patients to assess model performance (discrimination, calibration, net benefit).

Main Results:

• Age, sex, serum creatinine, and beta-lactam antibiotic type were identified as predictors of non-attainment.
• External validation showed good discrimination for RF (AUC 0.79), LR (AUC 0.80), and NB (AUC 0.75) models.
• RF and LR models demonstrated significant net benefit, indicating clinical utility.

Conclusions:

• Successfully developed and validated prediction models for beta-lactam target non-attainment within 12-36 hours of antibiotic initiation.
• Online-accessible models utilize readily available patient data to aid in antibiotic therapy optimization.
• The RF and LR models offer promising performance for improving antibiotic treatment precision in ICU settings.