Transporting difference-in-differences estimates to assess health equity impacts of payment and delivery models.

OBJECTIVE

To demonstrate the use of transportability methods to extend findings from payment model evaluations to groups of historically underserved beneficiaries.

STUDY SETTING AND DESIGN

We used a simulation study to transport the effects of the Comprehensive Primary Care Plus (CPC+) model to a target population of Black fee-for-service (FFS) Medicare beneficiaries living outside the original 18 CPC+ regions. Our main outcome variable was total Medicare spending per beneficiary per year (pbpy).

DATA SOURCES AND ANALYTIC SAMPLE

We simulated practice-level spending in 18 CPC+ regions and 32 non-CPC+ regions (1200 practices per region). We calibrated the simulation parameters to values from the literature and then varied four key parameters to create 16 realistic simulation scenarios. These scenarios varied the representativeness of practices in CPC+ regions that joined CPC+ (i.e., the sample) relative to the target population by changing the distribution of Black beneficiaries across practices and the distribution of practices across regions. Practices were characterized by their experience with the Medicare Shared Savings Program (SSP) and system/hospital ownership because these are known to modify the effect of CPC+ on spending.

PRINCIPAL FINDINGS

Across the 16 simulation scenarios, transporting the treatment effect of CPC+ to Black FFS beneficiaries in non-CPC+ regions yielded median treatment effects that ranged from $15.5 pbpy smaller to $10 pbpy larger than in the sample. These differences are roughly the same magnitude as the estimated overall effect of $13 pbpy.

CONCLUSIONS

The Center for Medicare and Medicaid Innovation has pledged to put equity at the center of its demonstration models. However, offering models in limited geographic areas with voluntary provider participation may result in unrepresentative samples. Naively generalizing CPC+ effects from geographically limited, voluntary samples to all Black FFS beneficiaries could be misleading. Under some circumstances, transportability methods can be used to estimate effects in this target population.