Redesigning Kidney Disease Care to Improve Value Delivery

Area of Science:

• Health Informatics
• Nephrology
• Machine Learning Applications in Healthcare

Background:

• Chronic kidney disease (CKD) management is complex due to frequent comorbidities.
• Existing electronic medical record (EMR) systems may not optimally identify or manage CKD patients.
• A need exists for proactive, structured interventions to improve CKD patient outcomes.

Purpose of the Study:

• To develop and deploy a machine learning (ML) model-driven solution for enhancing CKD care within a health system.
• To integrate ML-driven alerts with structured clinical checklists for primary care physicians (PCPs).
• To improve the management of comorbidities, medication, and diagnostic testing for CKD patients.

Main Methods:

• An ML model was developed to identify CKD patients within the EMR.
• The ML model triggered alerts for both CKD patients and their PCPs.
• PCPs utilized structured checklists to manage comorbidities, adjust medications based on CKD stage, and order relevant tests before nephrology referral.

Main Results:

• The ML model successfully identified CKD patients and facilitated timely primary care engagement.
• Structured checklists enabled PCPs to effectively address comorbid conditions and optimize CKD-specific treatments.
• Operational results exceeded expectations, demonstrating the program's high value and effectiveness.

Conclusions:

• ML-driven solutions integrated with structured checklists can significantly improve the value of care for CKD patients.
• This approach provides a facile method for understanding and managing the comorbid burden in CKD.
• The ML-driven, checklist-enabled care paradigm is adaptable across different EMR platforms for broad application in health systems.