Health Practitioner Bonuses and Their Impact on the Availability and Utilization of Primary Care Services. C. Econometric Framework


We used the provider-level data to address the key research questions related to the Medicare PCIP. We estimated several regression-based models using the DID method to obtain consistent estimates of the PCIP policy effect.39 In order to implement the DID model, we:

  1. Form a group of providers (treatment group) who are likely to be affected by the PCIP program, including the following specialties: family practice, internal medicine, pediatrics and geriatrics;

  2. Identify a natural relevant comparison/control group (other non-primary care practitioners) who are not likely to be affected by the PCIP;

  3. Assess the changes in our key outcomes of interest for the treatment group relative to the control group before and after the PCIP policy is in effect.

We estimate the following econometric model to identify the effect of Medicare PCIP program on the magnitude of primary care E&M services:

We let be the volume of eligible E&M services provided by provider i in period t; is an indicator for treatment group that takes the value 1 if the provider is a PCP, and zero otherwise; the variable is an indicator variable taking the value of 1 in the years when the PCIP policy is likely to impact the behavior of the providers and 0 otherwise; the vector includes provider characteristics, such as age and gender. The term represents the net influence of random unobserved factors affecting E&M services. Importantly, the estimated coefficient (the coefficient of the interaction term between the treatment group indicator and the post-PCIP policy period indicator) in equation (1) represents an unbiased estimate of the effect of the Medicare primary care incentive payment on the volume of primary care E&M services. The derivation of this estimated effect is presented in Appendix C (C.1). A similar methodology can be applied to estimate the effect of the incentive payment on other outcomes of interest.

The choice of the control group for the DID model was based on several criteria. Ideally, we would have included providers with non-eligible specialties who also supplied E&M services. In order to determine some preferred comparison group providers we:

  1. Evaluated the frequency of PCIP eligible and non-eligible providers by (primary) specialty available in the Provider360 dataset after merging that data set with the list of NPIs of 2011 PCIP recipients.

  2. Examined the volume of claims related to evaluation and management (E&M) services by specialty, based on 2009 Medicare claims by service and specialty.

  3. Chose, based on the volume of claims and the number of providers, 8 PCIP non-eligible specialties. These PCIP non-eligible specialties are: (1) Psychiatry & Neurology; (2) Obstetrics & Gynecology; (3) Urology; (4) Ophthalmology; (5) Pathology; (6) Psychologist; (7) Podiatrist; and (8) Optometrist. Total number of PCIP eligible claims submitted by these providers with PCIP non-eligible specialties is presented in Appendix B Exhibit B14.

39 An alternative estimation method is a pre-post analysis, but if there are unobserved factors that affect the magnitude of E&M services supplied by any provider (primary or non-primary care) and the effect of these unobserved factors varies over time, a pre-post model would not yield an unbiased incentive effect.

View full report


"FinalReport_HealthPractitionerBonuses_8_13_15 (v2).pdf" (pdf, 2.1Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®