[Methodological requirements for center comparisons (profiling) in healthcare]

Area of Science:

• Health Services Research
• Biostatistics
• Quality Improvement in Healthcare

Background:

• Quantitative comparisons of health institutions are common in Germany for quality assurance.
• Current evaluation methods often rely on raw center means, which are insufficient for comprehensive assessment.
• Raw means do not account for random effects or baseline differences, leading to inaccurate comparisons.

Purpose of the Study:

• To highlight the limitations of using raw means for evaluating health institutions.
• To advocate for the use of multilevel statistical models for more accurate center effect estimation.
• To promote alternative presentation methods like profile plots and control charts over rankings.

Main Methods:

• Application of multilevel statistical models with institutions as random effects.
• Inclusion of carefully selected potential confounders as fixed effects.
• Comparison of multilevel model outputs with traditional raw mean calculations.

Main Results:

• Raw mean comparisons inflate perceived differences among institutions.
• Multilevel models provide adjusted center effects, allowing for fairer comparisons.
• Rankings and league tables based on raw means do not reflect true institutional performance or development.

Conclusions:

• Multilevel statistical models are superior to raw means for evaluating health institution quality.
• Adjusting for random effects and baseline differences is crucial for accurate healthcare comparisons.
• Profile plots and control charts offer more informative visualizations than traditional rankings.