# Model-Informed Precision Dosing: Conceptual Framework for Therapeutic Drug Monitoring Integrating Machine Learning and Artificial Intelligence Within Population Health Informatics

## TL;DR

This paper explores how combining machine learning and electronic health records can improve drug dosing precision for better patient outcomes and safety.

## Contribution

The paper introduces a framework integrating model-informed precision dosing with AI/ML in population health informatics to enhance drug dosing for vulnerable groups.

## Key findings

- MIPD using Bayesian methods and AI/ML can enable real-time, precise drug dosing adjustments.
- Integration of MIPD with EHRs can improve patient safety and reduce healthcare costs for vulnerable populations.
- Successful implementation requires collaboration among clinicians and secure data management solutions.

## Abstract

Background/Objective: Traditional therapeutic drug monitoring is limited by manual interpretation and specific constraints like sampling at steady-state and requiring a minimum of two drug concentrations. The integration of model-informed precision dosing (MIPD) into population health informatics represents a promising approach to address drug safety and efficacy. This article explored the integration of MIPD within population health informatics and evaluated its potential to enhance precision dosing using artificial intelligence (AI), machine learning (ML), and electronic health records (EHRs). Methods: PubMed and Embase searches were conducted, and all relevant peer-reviewed studies in English published between 1958 and December 2024 were included if they pertained to MIPD and population-level health, with the use of AI/ML algorithms to predict individualized drug dosing requirements. Emphasis was placed on vulnerable populations such as critically-ill, geriatric, and pediatric groups. Results: MIPD with the Bayesian method represents a scalable innovation in precision medicine, with significant implications for population health informatics. By combining AI/ML with comprehensive electronic health records (EHRs), MIPD can offer real-time, precise dosing adjustments. This integration has the potential to improve patient safety, optimize therapeutic outcomes, and reduce healthcare costs, especially for vulnerable populations where evidence is limited. Successful implementation requires collaboration among clinicians, pharmacists, and health informatics professionals, alongside secure data management and interoperability solutions. Conclusions: Further research is needed to define, implement, and evaluate practical applications of AI/ML. This insight may help develop standards and identify drugs for MIPD to advance personalized medicine within population health informatics.