Meaningful variation in performance: what does variation in quality tell us about improving quality?

Area of Science:

• Healthcare Quality Improvement
• Health Services Research
• Biostatistics

Background:

• Variance reduction is a common goal in clinical quality improvement.
• Intraclass correlation (ICC) measures variance at physician or facility levels, with low ICCs often suggesting limited improvement potential.
• Absolute variance and changes in variance components during quality improvement are less studied.

Purpose of the Study:

• To analyze changes in variance components at primary care physician and medical facility levels.
• To assess these changes across four quality indicators: systolic blood pressure, low-density lipoprotein-cholesterol, patient-reported care experience, and mammography screening rates.
• To evaluate how performance improvements impact variance.

Main Methods:

• Utilized data from 62,596-410,976 adult members of Kaiser Permanente in Northern California (2001-2006).
• Employed multilevel linear and logistic regression models.
• Adjusted for case-mix to examine interphysician and interfacility variances over six years.

Main Results:

• Intraclass correlations (ICCs) were low for all four indicators (0.0021-0.086).
• Statistically and clinically significant variances existed at both physician and facility levels.
• Systolic blood pressure and care experience showed decreased variance with improved quality, particularly at the facility level.
• Low-density lipoprotein-cholesterol variability increased with quality improvement; mammography screening showed non-significant declines.

Conclusions:

• Low ICCs do not preclude significant quality improvement potential.
• Quality improvement initiatives targeting facilities effectively enhanced quality across all four indicators, despite low facility-level ICCs.
• Facility-level interventions are crucial for driving quality improvements in primary care.