Developing and maintaining clinical decision support using clinical knowledge and machine learning: the case of order...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare

Background:

• Developing and maintaining clinical order sets is complex and requires significant expertise.
• Current machine learning algorithms struggle with the knowledge-intensive nature of order set management alone.

Purpose of the Study:

• To investigate a hybrid approach combining clinical knowledge and machine learning for order set development and maintenance.
• To evaluate the effectiveness of integrating clinical expertise with machine learning in promoting best ordering practices.

Main Methods:

• Simulated the revision of an "AM Lab Order Set" using six distinct approaches.
• Utilized electronic health record (EHR) data from 2014-2015 for revision criteria and 2016-2017 for evaluation.
• Assessed revisions based on clinical appropriateness, workload impact, and temporal generalizability.

Main Results:

• Order set revisions incorporating both clinical knowledge and machine learning showed promise.
• The hybrid approach demonstrated potential for improving order set usability.
• This method facilitates content updates aligned with the latest clinical knowledge and best practices.

Conclusions:

• A combined approach of clinical knowledge and machine learning offers an effective strategy for order set revision.
• This integrated method can enhance the maintenance of clinical order sets.
• Leveraging both data-driven insights and expert knowledge optimizes order set usability and relevance.