Models for Heart Failure Admissions and Admission Rates, 2016 through 2018.

Area of Science:

• Cardiovascular Medicine
• Health Economics
• Data Science in Healthcare

Background:

• Heart failure affects 6.5-6.9 million Americans, costing $43.6 billion in 2020.
• Rising incidence and costs necessitate improved predictive modeling and resource allocation strategies.

Purpose of the Study:

• To develop geographical incidence and cost models for heart failure in the U.S.
• To identify key variables influencing hospital heart failure diagnoses and expenditures.
• To inform healthcare leaders with data-driven insights for strategic planning.

Main Methods:

• Utilized regression and ensemble techniques (random forests, gradient boosting) for diagnosis forecasting.
• Employed spatial and non-spatial regression models for admission rate analysis.
• Developed descriptive maps and interactive online tools for visualizing geographical patterns.

Main Results:

• Heart failure incidence shows increasing trends, particularly in Eastern and Central U.S. states.
• Extremely randomized trees model best predicted diagnoses (R²=0.86), with workload and hospital type as key factors.
• Gradient-boosted models identified facility workload and hospital financial metrics as significant cost predictors.

Conclusions:

• Heart failure and associated healthcare expenditures are escalating, with significant cost increases noted in specific DRGs.
• Projected increases in demand and limited supply of specialists suggest future cost escalations.
• The developed models are crucial for healthcare leaders to manage increasing heart failure burden and costs.