Using Medicare Data to Measure Vertical Integration of Hospitals and Physicians

Tuesday, June 25, 2019: 2:30 PM
Wilson B - Mezzanine Level (Marriott Wardman Park Hotel)

Presenter: Sasa Tapaneeyakul

Co-Authors: Vivian Ho; Leanne Metcalfe; Lan Vu; Marah Short

Discussant: Herbert S Wong

Researchers, healthcare providers, and policy makers have become increasingly interested in the cost and quality effects of vertical integration between hospitals and physicians. Hospitals report their employment or other contractual relationships with physicians in the American Hospital Association annual survey. The SK&A database provides self-reported information from physicians on their hospital and other affiliations, and has been used in multiple recent vertical integration studies. However, SK&A is a for-profit company, so the cost of obtaining their data is higher than from nonprofit and government sources.

The Medicare Data on Provider Practice and Specialty (MD-PPAS) annual dataset may be a more cost-effective source for tracking vertical integration between hospitals and physicians. The MD-PPAS reports the National Provider Identifier, the two most common tax identification numbers (TINs), and legal names used by each physician when filing claims with the Medicare program. We matched NPIs from the MD-PPAS with annual spending for Blue Cross Blue Shield Texas PPO patients between the ages of 19 to 65 that contained the NPI for the physician responsible for the majority of primary care. We also have information on the physician group and/or hospital affiliation for each physician who signed a contract with BCBSTX in 2016. Physician group or hospital affiliation in the MD-PPAS was determined by internet searches of the provider name associated with each TIN.

We found a relatively close match between the NPI reported in the MD-PPAS and the NPI used by each physician in the BCBSTX claims. There was a small percentage of physicians in the BCBSTX who filed claims under multiple TINs, although the most common TIN tended to match the primary TIN in the MD-PPAS. When provider names were used to distinguish physician versus hospital-owned physician practices, we also found a relatively close match in ownership type between the two data sources. We estimated a regression of patient-level annual spending for BCBSTX patients on an indicator variable for physician versus hospital ownership for the primary attributed physician. Whether we used an indicator variable based on the MD-PPAS or the BCBSTX data source, patients treated by physicians in hospital-owned practices had higher spending than patients in physician-owned practices. The spending differential was precisely estimated and close in magnitude for the two data sources.

The results suggest that the MD-PPAS dataset, which costs less to obtain than SK&A data, can be used to reliably track vertical integration between hospitals and physicians.