Estimating Variance of Log Standardized Incidence Ratios Assessing Health Care Providers' Performance: Comparative...

Area of Science:

• Health Services Research
• Biostatistics
• Health Equity

Background:

• Standardized incidence ratios (SIRs) are vital for assessing healthcare provider performance by comparing observed vs. expected event rates.
• Accurate estimation of SIR variance is critical for reliable performance evaluation and decision-making.
• Limited data exists on how different variance estimation methods impact these crucial healthcare decisions.

Purpose of the Study:

• To compare three statistical methods—the delta method, bootstrapping, and a Bayesian approach—for estimating the variance of the logarithm of the SIR.
• To evaluate the impact of these variance estimation methods on healthcare provider performance assessment and health equity measurement.

Main Methods:

• Utilized patient-level data from the Australia and New Zealand Dialysis and Transplant Registry (2012-2023).
• Employed a random effects model to predict treatment at home.
• Compared delta, bootstrapping (over 5000 iterations), and Bayesian (Markov chain Monte Carlo sampling) methods using bias, variance, and mean squared error (MSE).
• Applied funnel plots to assess hospital performance in treating Indigenous and non-Indigenous patients, measuring equity.

Main Results:

• Bayesian method showed the lowest bias (0.01922) and variance (0.00005), indicating superior stability and accuracy.
• Bootstrap method exhibited the highest variance (0.00027) and MSE (0.00094), signifying the least reliability.
• All methods demonstrated potential for measuring equity in patient-centered outcomes across service providers.

Conclusions:

• The choice of variance estimation method significantly influences the interpretation of healthcare provider performance.
• The Bayesian Markov chain Monte Carlo method is recommended for its superior accuracy and stability in SIR variance estimation.
• These methods can effectively assess health equity for patient-centered outcomes within and between healthcare providers.