1. Community Uninsurance Rates and Expenditures on Emergency Room Care: Is there Evidence of a Spillover? Community Uninsurance Rates and Expenditures on Emergency Room Care: Is there Evidence of a Spillover? James Kirby AHRQ

2. What is “Spillover”? The number of uninsured people in a community may affect medical care for everyone, even those with health insurance The number of uninsured people in a community may affect medical care for everyone, even those with health insurance

3. Why Might Spillover Occur? A large number of uninsured residents may: A large number of uninsured residents may: – provide a less stable revenue base for providers – result in high levels of uncompensated care

4. Literature Previous literature focuses on access Previous literature focuses on access Spillover to cost is not examined Spillover to cost is not examined Spillover in the context of emergency rooms is not examined Spillover in the context of emergency rooms is not examined

5. Research Questions Do insured individuals living in areas with many uninsured people pay more for emergency room care? Do insured individuals living in areas with many uninsured people pay more for emergency room care? How does this differ by insurance type? How does this differ by insurance type?

6. Data Sources Individual-level data: Medical Expenditure Panel Survey, 2009 Individual-level data: Medical Expenditure Panel Survey, 2009 – Adults with at least one emergency room visit who live in a county with 65,000 residents or more (N=3,773) County-level data: American Community Survey, 2009 County-level data: American Community Survey, 2009

7. Main Variables Main Outcome variable Main Outcome variable – Average expenditure per emergency room visit Main Independent variables Main Independent variables – County-level: Number of uninsured individuals per emergency room – Individual-level: Insurance status

8. Control Variables Individual-level Race/ethnicity Race/ethnicity Sex Sex Age Age Subjective health Subjective health Serious chronic conditions Serious chronic conditions Poverty status Poverty status County-level Poverty rate Poverty rate Unemployment rate Unemployment rate Number of Federally Qualified Health Centers (FQHC) per capita Number of Federally Qualified Health Centers (FQHC) per capita MSA vs non-MSA MSA vs non-MSA

9. Methods Generalized Linear Model Generalized Linear Model – Family: Gamma – Link: Log Marginal predictions in dollars Marginal predictions in dollars

10. Means of Main Variables TotalPrivately Insured Publically Insured Uninsured Mean per-visit ED Expenditure$1,118$1,268$836$977 Percent Uninsured in County14% 15%16% Number of Uninsured per Emergency Room in County 19,50018,76019,54021,280

11. Marginal Predictions for Per-Visit Emergency Room Expenditure, All Adults

13. Marginal Predictions for Per-visit Emergency Room Expenditure

14. Summary & Conclusion The number of uninsured people per ED in a county is positively associated with average expenditures per ED visit The number of uninsured people per ED in a county is positively associated with average expenditures per ED visit – This association exists only among insured individuals – The association is strongest among those with public insurance Reducing the number of uninsured people in communities may lower ED expenditures for the insured, but especially for those with public insurance Reducing the number of uninsured people in communities may lower ED expenditures for the insured, but especially for those with public insurance

15. Limitations Unobserved differences in intensity of use? Unobserved differences in intensity of use? Unobserved differences in county characteristics? Unobserved differences in county characteristics?