Insurance expansions: Do they hurt those they are designed to help?

Health Affairs

Volumn 26, Issue 5, 2007, Pages 1345-1357

  Richman, Barak D ^a  

^a DUKE UNIVERSITY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

ARTICLE; ECONOMICS; FEE; HEALTH CARE; HEALTH CARE DELIVERY; HEALTH CARE QUALITY; HEALTH INSURANCE; HEALTH SERVICES RESEARCH; HUMAN; MANAGEMENT; ORGANIZATION AND MANAGEMENT; PATIENT ATTITUDE; POLICY; STATISTICS; UNITED STATES; UNIVERSITY; UTILIZATION REVIEW; VULNERABLE POPULATION;

COMPREHENSIVE HEALTH CARE; FEES AND CHARGES; HEALTH BENEFIT PLANS, EMPLOYEE; HEALTH SERVICES ACCESSIBILITY; HEALTH SERVICES RESEARCH; HEALTHCARE DISPARITIES; HUMANS; NORTH CAROLINA; ORGANIZATIONAL POLICY; PATIENT ACCEPTANCE OF HEALTH CARE; POLICY MAKING; PROGRAM EVALUATION; UNIVERSITIES; VULNERABLE POPULATIONS;

EID: 35148857101     PISSN: 02782715     EISSN: 02782715     Source Type: Journal    
DOI: 10.1377/hlthaff.26.5.1345     Document Type: Article

References (23)

1. Duke has employees living in 97 of North Carolina's 100 counties, but 95 percent live in the six counties surrounding the Raleigh-Durham area. The region is home to many urban, suburban, and rural residential areas.
2. To protect employees' privacy, and to ensure that the data remained deidentified, individual salaries were not released. However, the Duke Human Resources Department categorizes each position by job code, each with a fairly precise salary range and required levels of education, which permitted imputing education and annual income for each person. Income was determined by the midpoint of the income range for each job code, coded in units of $10,000 in 2004 dollars. For job codes where wages are hourly, the hourly rate was multiplied by the individual's full-time equivalent. Job code salary ranges were not available for 2001, so 2001 incomes were imputed for each job code from the salary ranges in 2002-2004. Finally, faculty members' salaries and the salaries of certain administrators are not determined by job code; thus, people in those positions were not included in the sample. Anecdotal evidence suggests that results would be even stronger if these high-income workers remained in the data. Also omitted from the analyses were 784 people with missing data on race.
3. The data set is skewed by sex since women are heavily represented in health care occupations. Approximately 65 percent of the people in the data set were female. The median income for females in the sample was nearly identical to the median income for males, which is just above the median for males in Durham County.
4. It should be noted that the benefits staffs of most large employers should have access to similarly useful data, but very few share their data with researchers. Medicare claims data exhibit some of these advantages, since they follow heterogeneous people with known insurance benefits, but those data do not cover the working population.
5. Logit estimations were also used to estimate the probabilities of use, and the same variableswere found to be statistically significant. Ordinary least squares (OLS) regression was used instead because of the ease of interpreting OLS coefficients. Some of the claims in the data were for $0. Probability estimations were made both for claims greater than $0 and for claims of any amount, including $0. Results were consistent and robust. Estimates shown are for claims of any amount.
6. i is the residuals from that regression. This is the same two-stage smearing estimation method used in W.G. Manning et al., "Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment," American Economic Review 77, no. 3 (1987): 251-277.
7. The Latinos in the data set appear to be misrepresentative of other Latinos in Durham and North Carolina. Median incomes for Latinos in the sample held steady at approximately $34,000 throughout the sample, just slightly below the overall median, and Latinos' median education was at least one year higher than the sample's overall median. Many Latino low-wage earners working at Duke are employees of subcontractors and not Duke employees, which might explain this skewed sample. Since few generalizeable conclusions can be drawn from studying the Latinos in the sample, results for that group are omitted.
8. Income and education have a correlation of 0.62 (p < 0.0001), and income and exemption status have a correlation of 0.66 (p < 0.0001).
9. Duke also provides employees with short-term counseling, or Personal Assistance Services (PAS), free of charge. Use of PAS was not captured in the claims data, but since this constitutes another form of employer-based mental health care, a complete understanding of employees' use of mental health care services requires taking PAS into account. Data on PAS usage were not at a level of detail that would allow a replication of the analyses executed on the claims data. Overview statistics of PAS use were available, and they suggested that the findings on race and income would not measurably change if PAS use were included in the larger sample. For example, PAS data revealed that white employees visited PAS in greater proportions than African American or Asian employees (no income data were available for PAS clients).
10. To ensure that patterns in mental health care use were not driving the results on pharmaceutical use, the effect of race and income were tested separately on the use of psychotropic and nonpsychotropic drugs. Results remained robust.
11. The RAND Health Insurance Experiment made important observations of how out-of-pocket payments affect use of health services. See J.P. Newhouse and the Insurance Experiment Group, Free for All?: Lessons from the RAND Health Insurance Experiment (Cambridge, Mass.: Harvard University Press, 1996). But there has been little research in this area since the RAND experiment, and the experiment focused on how cost sharing affects families with different incomes without substantial investigation into consumption patterns across race or ethnicity. Understanding how cost sharing affects use across different demographic populations is an important area for future research.
12. We tested but found no statistical significance for race-income interaction effects for either mental health care or pharmaceutical use.
13. B.D. Smedley, A.Y. Stith, and A.R. Nelson, eds., Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care (Washington: National Academies Press, 2003).
14. See also K.A. Schulman et al., "The Effect of Race and Sex on Physicians' Recommendations for Cardiac Catheterization," New England Journal of Medicine 340, no. 8 (1999): 618-626. If providers' discriminatory referral practices alone explain disparities in mental health and pharmaceutical use, one might expect to find no disparities across race for visits to PCPs, only disparities for the referred services. However, regressions on outpatient care find similarly regressive disparities by race and income. Of course, patients might anticipate discriminatory referrals and decide to forgo seeing a PCP in the first place.
15. United States Surgeon General, Mental Health: Culture, Race, and Ethnicity (Washington: U.S. Government Printing Office, 2001).
16. Sullivan Commission, Missing Persons: Minorities in the Health Professions, September 2004, http://minorityhealth.pitt.edu/archive/ 00000040/01/Sullivan_Final_Report_000.pdf (accessed 21 June 2007).
17. There is evidence, for example, that social support systems and religious participation provide some protection to the African American population from the causes of mental health disorders. See D.R. Williams and H.W. Neighbors, "Social Perspectives on Mood Disorders," chap. 9 in Textbook of Mood Disorders, ed. D.J. Stein, D.J. Kupfer, and A.F. Schatzberg (Arlington, Va.: American Psychiatric Publishing Inc., 2006), 145-158.
18. Regressions on use of outpatient care indicated that whites and higher-income people in the sample used more outpatient care than nonwhites and those with lower incomes. Results from analyzing use of inpatient care, however, did not follow the same trend. This suggests that some people might obtain more preventive care to substitute for inpatient care. Mental health care and pharmaceuticals might similarly be more effective substitutes for other forms of expensive and disruptive inpatient care.
19. It deserves emphasis that major mental health disparities have been documented and deserve attention from policymakers. African Americans, for example, are overrepresented among people with very severe major depressive disorders and with persistent and debilitating mental illnesses. See D.R. Williams et al., "Prevalence and Distribution of Major Depressive Disorder in African Americans, Caribbean Blacks, and Non-Hispanic Whites," Archives of General Psychiatry 64, no. 3 (2007): 305-315.
20. See C.C. Havighurst and B.D. Richman, "Distributive Injustice(s) in American Health Care," Law and Contemporary Problems 69, no. 4 (2006): 7-82.
21. See, for example, J. Gruber, "Health Insurance and the Labor Market," Working Paper no. 6762 (Cambridge, Mass.: National Bureau of Economic Research, 1998).
22. See M.V. Pauly, "The Tax Subsidy to Employment-Based Health Insurance and the Distribution of Well-Being," Law and Contemporary Problems 69, no. 4 (2006): 83-101.
23. This accounting method was presumed by the White House when it contemplated how to implement the president's proposal to cap the tax exclusion on health insurance. L. Burman et al., "The President's Health Insurance Proposal: A First Look," Tax Policy Center Brief (Washington: Urban Institute and Brookings Institution, 23 January 2007).