Producing personalized statin treatment plans to optimize clinical outcomes using big data and machine learning

Area of Science:

• Cardiovascular Medicine
• Health Informatics
• Artificial Intelligence in Healthcare

Background:

• Statin therapy is crucial for managing cholesterol and preventing cardiovascular disease, but high discontinuation rates due to side effects are a major public health issue.
• Current statin prescribing relies on limited data and reactive management of statin-associated symptoms (SAS), leading to suboptimal patient outcomes.
• Identifying optimal statin regimens proactively is essential to minimize SAS and improve treatment adherence.

Purpose of the Study:

• To develop and evaluate a machine-learning personalized statin treatment plan (PSTP) platform.
• To proactively identify optimal statin agent and dosage to minimize SAS and discontinuation risks.
• To demonstrate a method for individualized, patient-centered statin therapy.

Main Methods:

• Leveraged de-identified administrative insurance claims data from OptumLabs® Data Warehouse (over 130 million enrollees).
• Developed a machine-learning PSTP platform to assess various statin treatment plans.
• Utilized an artificial neural network approach for enhanced predictive performance.

Main Results:

• The PSTP platform demonstrated significantly lower risks of SAS and statin discontinuation compared to standard practice.
• Machine learning, particularly artificial neural networks, improved the performance of proactive statin prescription strategies.
• A method for incorporating optimization constraints for personalized medicine and shared decision-making was successfully demonstrated.

Conclusions:

• Machine learning and big data approaches are feasible for creating personalized healthcare treatment plans.
• The PSTP platform shows promise in improving statin adherence and patient outcomes.
• Further research is needed to explore the clinical utility of this personalized approach.