Modeling Physician Supply, and the Impacts of Health Reform

Monday, June 13, 2016: 10:55 AM
G55 (Huntsman Hall)

Author(s): Hao Yu; Scott Ashwood; William Vogt; Jose J. Escarce; Emmett Keeler

Discussant: David Auerbach

The ACA has succeeded in substantially increasing the number of individuals enrolled in health insurance coverage. But, researchers and policy makers have expressed concerns regarding both the capacity of the health care system to treat an influx of newly insured and the geographic mismatches between areas with adequate supplies of physicians and areas with large populations gaining insurance coverage. The ACA is affecting physician supply and geographic distribution through several explicit provisions, including funding for new residency slots in primary care, and a temporary increase in Medicaid fees for primary care. The ACA is also indirectly affecting physician supply by expanding coverage and altering payer mix.

In this paper, we use historical data to examine the factors affecting the supply of physician services in the U.S. and their geographic distribution, and project the impacts of health reform on physician supply. We use a two-level model, in which we first project the national aggregate supply of different types of physicians (i.e., primary care providers versus specialists), and then project the geographic distribution of the aggregate supply. To estimate the aggregate national supply of physicians, we use historical data to measure how population growth, insurance rate, and Medicaid fees affect overall physician supply and supply by physician specialty. The historical data sources include the Area Health Resources File, the U.S. Census Bureau’s American Community Survey, the public-use version of the Community Tracking Study Physician Survey, and the National Ambulatory Medical Care Survey (NAMCS).

To estimate the geographic distribution of physicians under alternative policy scenarios, we first use the SK&A data to create a panel data set with primary care services areas (PCSAs) as the unit of analysis. We then perform a Poisson regression with PCSA-level random effects to examine physician distribution across the country, including several types of predictors: sociodemographic characteristics of the local population (e.g. share of the population over age 65, share in poverty, share eligible for Medicaid), medical education activity (e.g. having a residency training program in the PCSA, having a medical school in the PCSA, distances to the nearest residency training program, and  distance to the nearest medical school), factors affecting the financial viability of physician practices (e.g. medical malpractice insurance premiums, Medicaid physician fees, and eligibility for loan repayment programs), and local amenities (e.g. prevalence of extreme weather, crime rates, entertainment venues).

We then combine projections based on the aggregate national model with the PCSA-level estimates and compare projections of physician supply under current law (with the ACA) and an alternative scenario without the coverage expansions and other provisions of the ACA. Based on preliminary analyses, we find that population insurance rate and Medicaid physician fees have significant impact on physician numbers and locations, and the impact differs by physician specialty. The finding suggests that over time,  the ACA will not only impact on the aggregate supply of physician services, but also lead to a redistribution of physicians to areas with large gains in coverage, and/or increases in Medicaid physician fees.