JMIR Publications

ABSTRACT

Background:

Traditional therapeutic drug monitoring (TDM) for vancomycin has primarily focused on post-administration phases, but it is crucial to determine the patient-specific initial drug dosage.

Objective:

We aim to develop and evaluate an algorithm that utilizes machine learning (ML) techniques to optimize the initial vancomycin dosage.

Methods:

We established an internal cohort of patients who were administered intravenous vancomycin and underwent a pharmacokinetic TDM consultation for vancomycin (n=415). The internal cohort was randomly divided into training and testing datasets in a 7:3 ratio, and several ML techniques were employed to develop an algorithm for determining the dosage for initial vancomycin administration. And the optimal algorithm was selected referred to as OPTIVAN algorithm, and its performance was validated using an external cohort (n=268). We implemented a web-based vancomycin TDM application.

Results:

The Support Vector Machine algorithm demonstrated the best predictive performance, with an AU-ROC of 0.817 for the training dataset and 0.722 for the external validation dataset. The algorithm included seven covariates: age, body mass index (BMI), glucose, blood urea nitrogen, estimated glomerular filtration rate, hematocrit, and daily dose per weight. Subgroup analyses showed consistent performance across different patient categories, such as renal function, gender, and BMI. A web-based TDM analysis tool was developed using the OPTIVAN algorithm.

Conclusions:

The OPTIVAN algorithm represents a significant advancement in personalized initial vancomycin dosing, overcoming the limitations of current TDM practices. This algorithm could minimize the need for dosage adjustments after the initial administration. The algorithm's web-based application is user-friendly, making it a practical tool for clinicians. The study highlights the potential of ML to enhance the effectiveness of vancomycin treatment.