Development of an Artificial Intelligence Powered Medication Risk Score Calculator Application

Area of Science:

• Pharmacology
• Artificial Intelligence
• Health Informatics

Background:

• Clinical pharmacists face challenges in evaluating complex medication therapies and identifying high-risk patients.
• Processing unstructured electronic health records (EHRs) for medication risk assessment is a significant hurdle.
• Existing methods may not efficiently integrate drug interactions and kidney function (e.g., eGFR) into risk scores.

Purpose of the Study:

• To develop and validate the Augmented Medication Risk Score (AUGMERIS) calculator, an AI-powered web application.
• To automate the evaluation of medication therapies using the Danish Medication Risk Score (MERIS) method.
• To create a globally accessible tool for identifying patients requiring medication reviews.

Main Methods:

• Developed a hybrid text processing model combining traditional algorithms with Generative Pre-trained Transformer (GPT) technology.
• Integrated the text processing model and automated risk calculation into a web application using Python 3 and Streamlit.
• Validated the application's performance using data from volunteer testers in eight countries.

Main Results:

• The AUGMERIS web application successfully automates medication risk assessment, integrating drug combinations and eGFR.
• The hybrid text processing model effectively handles unstructured EHR data for risk evaluation.
• 383 trial calculations were performed by volunteer testers, indicating satisfactory performance despite validation limitations.

Conclusions:

• The AUGMERIS calculator has the potential to significantly improve global pharmacotherapy and patient safety.
• Wider adoption of AUGMERIS can enhance medication reviews and optimize treatments across diverse healthcare systems.
• The tool demonstrates the feasibility of using AI for automated medication risk assessment in clinical practice.