Design and Implementation of Individualized Drug Dosage Regimen Based on Bayesian Hierarchical Model — A Case Study of Vancomycin

This study proposes a vancomycin individualized dosing regimen based on a Bayesian hierarchical model, integrating Hamiltonian Monte Carlo (HMC) sampling and dynamic adjustment of prior strength. This method addresses the challenges of sparse data and inter-patient variability by dynamically adjusting prior strength, enhancing the model’s precision and reliability in predicting drug concentrations. The results demonstrate a significant reduction in prediction errors after dynamic adjustment of prior strength, with improvements in both root mean square error (RMSE) and mean prediction error (MPE). This approach provides a robust framework for real-time personalized vancomycin therapy, improving patient safety and treatment efficacy. Future work will focus on large-scale clinical validation and integrating more advanced computational techniques.