Using Computational Approaches to Improve Risk-Stratified Patient Management: Rationale and Methods

Area of Science:

• Health Services Research
• Biomedical Informatics
• Predictive Analytics

Background:

• Chronic diseases account for 52% of American illnesses and 86% of healthcare costs, with a small patient subset driving most resource utilization.
• Current risk stratification models for chronic disease management are limited by poor prediction accuracy, lack of explainability, and suboptimal risk thresholds.
• Existing models often exclude physician characteristics, which significantly impact patient outcomes and healthcare costs.

Purpose of the Study:

• To enhance risk-stratified patient management for improved healthcare delivery and patient outcomes.
• To develop more accurate predictive models for patient health outcomes and costs by integrating diverse data sources.

Main Methods:

• Integrating patient, physician, and environmental variables to improve prediction accuracy for individual patient health outcomes and costs.
• Developing novel algorithms for explaining prediction results and suggesting tailored interventions.
• Creating algorithms to compute optimal thresholds for risk strata and conducting simulations to evaluate management strategies.

Main Results:

• Data extraction from an integrated healthcare system's data warehouse is underway.
• The study is projected to be completed within approximately five years.

Conclusions:

• The developed methods will significantly improve risk-stratified patient management.
• Expected outcomes include better clinical results, increased patient satisfaction, enhanced quality of life, reduced healthcare utilization, and lower overall costs.