CDS-Compare: A Web Application for Machine Learning Assisted Curation of Clinical Order Sets

Area of Science:

• Health Informatics
• Clinical Decision Support
• Natural Language Processing

Background:

• Disease-specific order sets enhance clinician efficiency and patient safety.
• Curating order sets for consistency across multiple healthcare sites is complex and time-consuming.

Purpose of the Study:

• To introduce CDS-Compare, a novel software tool designed to assist expert reviewers in efficiently curating large databases of clinical decision support order sets.
• To evaluate the effectiveness of CDS-Compare in validating clustered order sets.

Main Methods:

• Developed CDS-Compare software utilizing a clustering-based approach.
• Processed a database of order sets extracted from the Veterans Affairs Electronic Health Record using Natural Language Processing (NLP).
• Subject-matter experts reviewed the web application interface of CDS-Compare to assess clustering validity.

Main Results:

• The study successfully applied clustering-based software to a large database of NLP-processed order sets.
• Expert reviewers assessed the web application version of CDS-Compare for clustering validity.
• The software aims to reduce the burden on expert reviewers during the order set curation process.

Conclusions:

• CDS-Compare offers a potential solution for the efficient and effective curation of clinical decision support order sets.
• The software facilitates consistency in order set management across multiple sites.
• Further validation by subject-matter experts is crucial for refining the tool.