Machine learning-enabled early risk stratification of β-Iactam-induced electrolyte imbalances.

Area of Science:

• Pharmacology
• Clinical Medicine
• Medical Informatics

Background:

• Beta-lactam/beta-lactamase inhibitor (BLBLI) antibiotics are widely used.
• Electrolyte imbalances are a potential adverse effect of BLBLI therapy.
• Early identification of patients at risk for electrolyte disturbances is crucial for patient safety.

Purpose of the Study:

• To determine the incidence of electrolyte imbalances associated with BLBLI use.
• To develop and validate a predictive model for early identification of patients at risk for BLBLI-associated electrolyte imbalances.

Main Methods:

• Retrospective analysis of 240 hospitalized adult patients treated with piperacillin-tazobactam or ampicillin-sulbactam.
• Electrolyte imbalance defined by post-treatment abnormalities in sodium, potassium, chloride, calcium, or phosphate.
• Logistic regression and machine learning were used for predictor selection and model development.

Main Results:

• Electrolyte imbalances occurred in 29.6% of patients.
• Respiratory disease was a significant risk factor (AOR 2.616).
• A six-predictor model (including age, cardiovascular disease, respiratory disease) showed good performance (AUROC 0.68) with high negative predictive value (~0.80).

Conclusions:

• Electrolyte imbalances are a frequent complication of BLBLI therapy.
• An interpretable prediction model utilizing routine clinical variables can aid in risk stratification.
• External validation of the developed model is recommended to confirm its generalizability.