Improving Medicare's Hospital Compare Mortality Model

Area of Science:

• Health Services Research
• Biostatistics
• Health Informatics

Background:

• Medicare's Hospital Compare (HC) aims to inform public hospital choice.
• Current HC predictions may not fully account for hospital-specific factors.
• Improving prediction accuracy is crucial for patient decision-making.

Purpose of the Study:

• To enhance Medicare's Hospital Compare predictions.
• To enable more informed hospital selection by the public.
• To investigate the impact of hospital characteristics on prediction accuracy.

Main Methods:

• Bayesian cohort analysis of Medicare fee-for-service claims (2009-2011) for Acute Myocardial Infarction patients.
• Comparison of current HC model assumptions with an expanded model incorporating hospital attributes (volume, capabilities, staffing).
• Direct standardization for comparing hospital predictions.

Main Results:

• The expanded model, including hospital characteristics, yielded significantly different predictions compared to the current HC model.
• Hospitals with lower volume and poorer characteristics showed higher predicted mortality rates.
• Example: Chicago analysis suggested avoiding smaller hospitals with suboptimal technology and staffing.

Conclusions:

• Medicare's Hospital Compare model should be updated to incorporate hospital attributes like volume, capabilities, and staffing.
• Systematic variation in predictions based on these attributes will aid patient hospital selection.
• Enhanced predictions will lead to better-informed healthcare decisions.