Privacy-preserving Sequential Pattern Mining in distributed EHRs for Predicting Cardiovascular Disease.

Area of Science:

• Health Informatics
• Data Mining
• Privacy-Preserving Technologies

Background:

• Electronic Health Records (EHRs) contain valuable data for healthcare research, including risk prediction.
• Mining distributed EHR data presents challenges due to data silos and privacy regulations.
• Existing methods struggle to balance data utility with patient privacy in distributed environments.

Purpose of the Study:

• To develop a privacy-preserving framework for sequential pattern mining in distributed EHR data.
• To enable effective risk prediction by discovering discriminative patterns across multiple data sources.
• To address the challenges of data fragmentation and privacy concerns in EHR analysis.

Main Methods:

• A novel framework utilizing sequential pattern mining on distributed data sources.
• Pattern extraction at each data source followed by secure sharing of patterns.
• Integration of differential privacy mechanisms to guarantee patient confidentiality.
• Case study on predicting Cardiovascular Disease in type 2 diabetes patients.

Main Results:

• Demonstrated the framework's effectiveness in discovering representative and discriminative patterns from distributed EHRs.
• Successfully applied the framework to a real-world risk prediction task (Cardiovascular Disease).
• Showcased the ability to maintain patient privacy while achieving valuable insights.

Conclusions:

• The proposed framework offers a viable solution for privacy-preserving data mining in distributed EHRs.
• It facilitates accurate risk prediction by leveraging data from multiple sources securely.
• This approach enhances the utility of EHR data for research while upholding stringent privacy standards.