Evaluating hospital performance.

Area of Science:

• Health Services Research
• Health Economics
• Medical Statistics

Background:

• Accurate hospital performance evaluation is crucial for quality improvement.
• Traditional methods often rely on observable patient characteristics for risk adjustment.
• Patient selection bias can confound hospital quality assessments.

Purpose of the Study:

• To compare hospital performance estimates using instrumental variable (IV) methods versus traditional risk adjustment.
• To assess the impact of patient observables on hospital quality indicators.
• To evaluate hospital performance regarding in-hospital mortality, 30-day mortality, and 30-day readmission.

Main Methods:

• Utilized quasi-random patient assignment based on hospital rotation schedules in Linz, Austria.
• Employed an instrumental variable (IV) framework with high-quality administrative data.
• Contrasted IV estimates with traditional risk adjustment models using patient observables and socio-economic covariates.

Main Results:

• Hospital performance assessment is sensitive to the inclusion of patient observables.
• Increased socio-economic covariates did not consistently align risk-adjusted estimates with IV estimates.
• Divergence was most significant for readmissions, with risk adjustment showing large differences and IV estimates showing smaller, non-significant ones.

Conclusions:

• Standard risk adjustment models may not adequately control for patient differences between hospitals.
• Hospital quality indicators derived from administrative data require cautious interpretation.
• Personalized medicine trends may facilitate better quality comparisons through enhanced clinical data collection.