An unsupervised learning approach to identify immunoglobulin utilization patterns using electronic health records.

Area of Science:

• Health Services Research
• Pharmacoeconomics
• Data Science in Healthcare

Background:

• Managing immunoglobulin (Ig) product allocation in Canada faces challenges due to rising demand, costs, and global shortages.
• Identifying patient groups with high Ig utilization is key for effective resource planning and supply chain management.
• Electronic Health Records (EHRs) offer a valuable data source for analyzing Ig utilization patterns.

Purpose of the Study:

• To identify distinct subgroups of immunoglobulin recipients based on their utilization patterns.
• To leverage EHR data for data-driven segmentation of Ig users.
• To inform resource allocation strategies and demand forecasting for Ig products.

Main Methods:

• Analysis of EHR data from all intravenous or subcutaneous immunoglobulin recipients in Calgary (2014-2020).
• Application of K-means clustering to patient characteristics and laboratory data to derive patient clusters.
• Descriptive analyses and visualization techniques to interpret identified patient clusters.

Main Results:

• Six distinct patient clusters were identified among 4112 recipients.
• Two clusters (9.9% and 30.9% of patients) accounted for 62.2% and 27.1% of total Ig utilization, respectively.
• Specific clusters were associated with high inpatient use (86.4%), pediatric populations (median age 4 years), emergency department infusions (77.3%), and subcutaneous Ig treatments.

Conclusions:

• The study successfully segmented patients based on Ig utilization using EHR data.
• Identified patient clusters highlight high-utilization groups and those at risk for short-term inpatient use.
• These findings provide a data-driven foundation for improving Ig product demand forecasting and resource allocation during shortages.