Machine Learning to Improve Decision Support for Preventing Adverse Drug Events

Area of Science:

• Pharmacovigilance
• Health Informatics
• Artificial Intelligence in Medicine

Background:

• Adverse drug events (ADEs), including drug interactions, pose significant risks in healthcare.
• Clinical decision support systems (CDSS) are implemented to mitigate ADEs.
• Current rule-based CDSS in Swedish healthcare require evaluation for predictive accuracy.

Purpose of the Study:

• To assess the accuracy of existing rule-based CDSS for ADE prediction in Swedish healthcare.
• To explore the potential of machine learning (ML) to enhance ADE prediction accuracy.
• To compare the performance of ML models against current rule-based CDSS.

Main Methods:

• Analysis of real-world healthcare data from a Swedish region.
• Utilizing a 10-year data span for comprehensive analysis.
• Developing and evaluating machine learning models for ADE prediction.

Main Results:

• Machine learning models demonstrate potential for improved ADE prediction.
• ML-based predictions may offer higher accuracy compared to current rule-based systems.
• The study highlights the efficacy of ML in a real-world healthcare setting.

Conclusions:

• Machine learning presents a viable approach to enhance the prediction of adverse drug events.
• ML has the potential to significantly improve upon the accuracy of existing rule-based CDSS.
• Further integration of ML into healthcare systems could lead to safer medication management.