Quantifying the physician contribution to managed care pharmacy expenses: a random effects approach

Area of Science:

• Health Services Research
• Pharmacoeconomics
• Biostatistics

Background:

• Ordinary least squares (OLS) regression is commonly used for physician profiling, but its effectiveness in capturing prescribing variability is debated.
• Limited research exists on interphysician variability in managed care pharmacy expenses and the impact of different statistical models.

Purpose of the Study:

• To quantify interphysician variability in managed care pharmacy expenses.
• To compare statistical approaches for profiling physician prescribing patterns and their impact on cost estimations.

Main Methods:

• Compared three statistical models: OLS, fixed effects, and random effects (hierarchical) regression.
• Analyzed data from two distinct managed care populations with varying pharmacy expenditures.
• Calculated intraclass correlation coefficients (ICCs) and projected expense ranges per physician.

Main Results:

• Intraclass correlation coefficients (ICCs) for pharmacy expenditures were low (≤0.04), indicating minimal physician-driven cost variation.
• Ordinary least squares (OLS) identified significant expense ranges between the highest and lowest-cost physicians.
• Random effects models yielded a 37% smaller range in projected physician performance compared to OLS.

Conclusions:

• Systematic prescribing differences among physicians account for a small proportion of overall pharmacy expenses.
• Factors beyond physician prescribing style are likely the primary drivers of pharmacy costs.
• Random effects models provide more conservative estimates of individual physician cost contributions than OLS or fixed effects models.