Machine Learning-based Characterization of Longitudinal Health Care Utilization Among Patients With Inflammatory...

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Clinical Operations

Background:

• Inflammatory bowel disease (IBD) is linked to high healthcare utilization.
• Predicting high resource users can optimize resource allocation.
• Developing models to forecast patient needs is crucial for efficient healthcare management.

Purpose of the Study:

• To develop machine learning models for clustering IBD patients by clinical utilization patterns.
• To predict longitudinal healthcare utilization based on baseline clinical data.
• To enhance resource allocation and patient risk assessment in IBD care.

Main Methods:

• Retrospective study of 1174 adult IBD patients (2015-2021) across two academic centers.
• Utilized machine learning for patient clustering and prediction of healthcare utilization.
• Compared model performance against ordinal regression and random choice methods.

Main Results:

• Clustering identified distinct low, medium, and high resource utilization groups.
• Machine learning models predicted longitudinal healthcare utilization with 81%-85% accuracy (AUC 0.84-0.90).
• Models demonstrated superior predictive performance compared to traditional methods.

Conclusions:

• Machine learning effectively clusters IBD patients by resource utilization.
• Accurate prediction of longitudinal utilization is achievable using baseline clinical factors.
• Integration into EHRs can support risk assessment, care coordination, and resource allocation.