Estimating the indirect costs of teaching

Area of Science:

• Health Economics
• Econometrics
• Healthcare Management

Background:

• Medicare's Prospective Payment System (PPS) utilizes a log-log cost function to reimburse teaching hospitals.
• Previous research indicated the regression coefficient is sensitive to covariate selection.
• A common specification for teaching intensity (intern and resident-to-bed ratio) introduces bias.

Purpose of the Study:

• To re-specify Medicare's teaching hospital payment equation to minimize bias.
• To assess the sensitivity of the regression coefficient to covariate inclusion after bias correction.

Main Methods:

• Log-log regression analysis was applied to Medicare cost data.
• A corrected specification for teaching intensity was implemented.
• The impact of covariate inclusion on the teaching intensity coefficient was evaluated.

Main Results:

• The standard specification of teaching intensity (log(1 + ratio)) introduces substantial bias in the regression coefficient.
• The re-specified equation significantly reduces this bias.
• In the corrected model, the coefficient shows less sensitivity to the inclusion of other covariates.

Conclusions:

• The bias introduced by the common teaching intensity specification is significant.
• A corrected model improves the reliability of the Medicare payment coefficient.
• Further research is needed before policy implications can be determined.