The Effects of Physician-Hospital Integration on Health Care Spending and Use

Background: Integration between hospitals and physicians has been increasing as hospitals have purchased or entered into close contractual relationships with physicians. Hospital-physician integration could have positive or negative effects on health care delivery and costs. Integration may affect prices paid to hospitals or physicians by private health plans. Integrated physicians may use more services either if integration makes it easier for them to admit patients or to use other hospital-based services, or if integration generates incentives for physicians to increase their use of services. Vertical integration, however, also has the potential to reduce utilization through improved coordination and management of care.

Research Question: In this paper we analyze whether physician practices vertically integrated with hospitals produce health care for Medicare patients at higher cost than non-integrated practices and whether there are differences in care utilization across these settings.

Data: We use Medicare claims data from 2000-2014. Our data contain information about all Medicare-billed services provided to a 20% random sample of traditional (fee-for-service) Medicare beneficiaries. Our study sample includes only geographic movers - Medicare beneficiaries whose residence changes from one hospital referral region (HRR) to another over the sample period.

Methods: We begin by using a difference-in-difference model to investigate the effect of vertical integration on expenditures, among those who move, by comparing beneficiaries who do and do not switch practice type, where practice type refers to whether the practice is vertically integrated with a hospital or not. While we expect that using geographic movers will mitigate the selection problem associated with moving to a new practice type, it may not eliminate it entirely. If changes in health that occur around the time of the move are correlated with choice of practice type in the new location, our estimates could be biased. To address this possible bias, we combine our difference-in-difference framework with an instrumental variables approach. Our instruments are based on the observation that a geographic mover’s “destination HRR” may have a different distribution of practice types (integrated versus non-integrated practices) than the origin HRR. We expect that the prevalence of different types of practices in the destination HRR will affect patients’ choice of practice type but should have no direct effect on an individual patient’s expenditures or utilization.

Measures: We identify the integration status of physicians based on the place of service designated on claims for office visits and consultations. If the place of service is a physician office, we designate it a non-integrated practice and if the place of service is a hospital outpatient department we label it an integrated practice. We assign patients to practices based on their patterns of claims for primary care services.

Outcomes and Results: Our models relate measures of utilization and spending to indicators of physician practice type. Our models also control for demographics, chronic conditions, year, and HRR of residence. Preliminary results indicate that there is small negative effect of integration on provider charges; however, we expect that any large effects are likely to appear in expenditures for hospitalizations.