Urgency Prediction for Medical Laboratory Tests Through Optimal Sparse Decision Tree: Case Study With Echocardiograms

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Precision Medicine

Background:

• Laboratory tests are crucial for precision medicine but face accessibility challenges.
• Echocardiograms are vital but have high demand and scheduling complexities.
• Limited research exists on optimizing echocardiogram appointment scheduling.

Purpose of the Study:

• Develop an interpretable machine learning model to determine echocardiogram appointment urgency.
• Prioritize patient scheduling for echocardiograms efficiently.
• Identify key patient attributes influencing echocardiogram appointment prioritization.

Main Methods:

• Utilized a large-scale real-world echocardiogram appointment dataset (34,293 records).
• Employed the Optimal Sparse Decision Tree (OSDT), a state-of-the-art interpretable machine learning algorithm.
• Analyzed administrative data, referral diagnoses, and patient conditions.

Main Results:

• The OSDT model showed satisfactory performance, outperforming baseline models.
• Achieved F1-score of 36.18% (1.7% improvement) and F2-score of 28.18% (0.79% improvement).
• Extracted decision rules from the OSDT model provided medical insights for identifying urgent patients.

Conclusions:

• The interpretable OSDT model demonstrated effective predictive performance for prioritizing echocardiogram urgency.
• Decision rules derived from the model align with established medical knowledge.
• The approach can be extended to prioritize other laboratory test appointments using electronic health record data.