Health Insurance Markets Call for Papers Innovations in Health Insurance Chair: Richard Lindrooth, Ph.D. Sunday, June 6 • 9:30 a.m.-11:00 a.m. • The Incidence of the Health Care Costs of Obesity M. Kate Bundorf, Ph.D., M.B.A., M.P.H., Jay Bhattacharya, M.D., Ph.D. Presented by: M. Kate Bundorf, Ph.D., M.B.A., M.P.H., Assistant Professor, Health Research and Policy, Stanford University School of Medicine, HRP Redwood Building, Room 257, Stanford, CA 94305-5405; Tel: 650.725.0067; E-mail: bundorf@stanford.edu Research Objective: Americans are increasingly either overweight or obese. Because obese individuals tend to be sicker and to spend more on health care, this trend suggests poor future population health and rising health care expenditures. In a health economy where third parties pay for most health care expenditures, the insured population at large pays for these deleterious trends, not just the obese. In the context of employer provided health care, these increased payments will take the form of increased health insurance premiums, and may have unintended consequences ranging from employers dropping health insurance coverage to regressive distributional effects. Whether these adverse consequences arise depends critically upon who pays for rising health insurance premiums due to obesity—employers or employees. Our research objective is to examine the incidence of obesity-related health care costs among individuals with employer-sponsored health insurance. Study Design: Using data from the National Longitudinal Survey of Youth from 1989-1999, we examine differences between obese and non-obese full-time workers in the difference in wages between those with and without health insurance from their employer. We estimate multivariate models of hourly wages as a function of insurance status and indicator of obesity and their interaction. We include individual fixed effects and estimate models with a variety of control variables including year fixed effects as well as timevarying individual characteristics including marital status, indicator of rural residence, age, worker industry and occupation, and employer-size. Population Studied: Full-time workers aged 18 and over from 1989 to 1998. Principal Findings: We find a striking increase in obesity rates among full-time workers in the NLSY sample from 12% to 26% from 1989 to 1998. We find that obese workers in firms offering health insurance experience a greater wage offset for health insurance than non-obese workers in firms offering health insurance. While cash wages for insured workers are greater than those for uninsured workers for both obese and non-obese workers, the wage premium for insured workers is smaller for obese than for non-obese workers. We also find that the estimate of the incidence of health insurance premiums on cash wages has increased over time. Our findings suggest that obese workers pay between $0.09 and $0.23 per hour more than non-obese workers for health insurance coverage from an employer in the form of lower cash wages. Conclusions: Our results are consistent with the incidence of at least some of the health care costs associated with obesity falling on obese workers in the form of lower cash wages. Implications for Policy, Delivery or Practice: The results of this project are particularly notable given the difficulty other researchers have faced in identifying the wage offset and the lack of evidence of the extent to which wage offsets exists for individual characteristics. Our results provide additional support for economic models indicating that employees bear the cost of employer-sponsored health insurance in the form of lower wages. They also suggest that individuals obtaining health insurance in the employer-sponsored market are increasingly facing risk rating based on their individual health status. • Who Uses Individual Health Insurance and for How Long? An Analysis of the 1996-2000 SIPP Erika Ziller, M.S., Andrew Coburn, Ph.D., Timothy McBride, Ph.D., Courtney Andrews, M.A. Presented by: Andrew Coburn, Ph.D., Professor and Director, Institute for Health Policy, Muskie School of Public Service, University of Southern Maine, P.O. Box 9300, Portland, ME 04104-9300; Tel: 207.780.4435; Fax: 207.228.8138; E-mail: andyc@usm.maine.edu Research Objective: This study examines patterns of individual insurance coverage among non-elderly adults in the United States. Specifically we consider three questions: (1) Who uses individual insurance? (2) What characteristics predict how long a person has coverage? And, (3) What insurance status do people have before and after individual insurance spells? Study Design: This study uses the 1996-2000 panel of the Survey of Income and Program Participation (SIPP), a nationally representative, longitudinal survey by the Census Bureau. We use bivariate and multivariate analyses to determine what characteristics predict individual health insurance use, including state regulation of individual and small group markets. We also conduct multivariate survival analysis to determine what factors affect individual insurance spell lengths. Population Studied: Adults aged 18-64. Principal Findings: Compared to those with employer coverage, people with individual insurance are more likely to be part-time workers, not working, self-employed and/or working for small business. They are also more likely to be unmarried and Asian or non-Hispanic White. In addition, living in a state with a high-risk pool is associated with being individually insured. The median length for new spells of individual insurance is 8.2 months. Although 47% of spells last for less than six months, nearly one-fifth last more than two years. Multivariate survival analysis indicates that spells are at least 20% longer for the unemployed, self-employed, non-Hispanic Whites or Asians, or those with employer-based coverage the month before gaining individual coverage. Being out of the labor force, a small business employee, unmarried, and female also increase time with individual coverage, as does living in a state with guaranteed renewal laws for individual plans. The number of months with individual insurance is lower for those in fair/poor health and people who gained individual insurance after having public coverage. Living in a state with small group community rating laws is also associated with shorter spells. Around two-thirds of individual insurance spells begin when a person loses employer-based coverage, and the same number end with a person gaining employer-based coverage. Eightythree percent of those who lose coverage from an employer regain employer-based insurance. About one-sixth of people who lose individual insurance become uninsured. Conclusions: Most new individual insurance spells appear to bridge periods of employer-based coverage, and these spells are longer on average than among those who gain individual insurance after being uninsured. For many, these gaps in employer-based coverage may arise when people become unemployed, work part-time, work for a small business and/or for themselves. However, a large minority of those who lose individual insurance become uninsured. This, coupled with the fact that those in poorest health have the shortest coverage, points to potential failings of the individual insurance market and warrant further research. State policies, including high-risk pools, community rating and guaranteed renewal relate to the likelihood or length of time a person has individual insurance, although the causal direction remains unclear. Implications for Policy, Delivery or Practice: This research provides crucial information about patterns of individual health insurance use, and can assist federal and state policymakers who seek strategies for improving access to affordable individual insurance coverage. Primary Funding Source: RWJF • Prescription Drug Demand for Therapeutic Substitutes: Do Copayments and Insurer Non-Price Rationing Influence Patient Utilization? Dominick Esposito, Ph.D. Presented by: Dominick Esposito, Ph.D., Researcher, Mathematica Policy Research, Inc., P.O. Box 2393, Princeton, NJ 08543; Tel: 609.275.2358; Fax: 609.799.0005; E-mail: desposito@mathematica-mpr.com Research Objective: To determine if differential copayments and type of insurance arrangements affect the demand for therapeutically equivalent cholesterol-lowering prescription drugs in the statins class. Study Design: The statins were chosen due to their widespread use among patients with cardiovascular disease, their aggregate U.S. expenditure as a class, and the expectation of considerably greater utilization in the future. This study examines drug choice, with a multinomial logit, for a sample of 44,642 patients who each use only one statin (monotherapy). The data source is medical and pharmacy claims from the 1997 and 1998 MarketScan Commercial Claims and Encounters database, which allows for simultaneous examination of patient health care utilization, prescription drug use and patient health status. A variety of health plans, including preferred provider organizations, point of service plans, indemnity plans, and health maintenance organizations, provide healthcare to the individuals represented in this database. Average copayments vary across and within insurer type; for example, HMO copayments range from $11.48 for Lipitor to $8.96 for Lescol. Explanatory variables used to model drug choice include relative copayments, insurance type, demographic characteristics (age, gender, and region), and clinical factors (comorbidities, utilization of other lipid-modifying drugs, and evidence of or cardiovascular disease). Population Studied: Commercial health plan enrollees aged 18-94 representing 33 health plans throughout the United States. Principal Findings: An increase in the relative copayment of the patient’s drug of choice by 10% decreases that drug’s market share by as much as 2.8%. Patients’ insurance type also influences drug choice independently from differential copayments. HMOs shifted patient utilization away from Lipitor and Mevacor to Zocor, Pravachol and Lescol in this sample. Compared to utilization among indemnity plans, Lipitor’s market share is nearly 30% lower in HMOs while use of Zocor is 20% higher. Results relating to demographic and health status variables also provide clinically interesting comparisons across statins. Conclusions: Results demonstrate that copayments imposed by insurers do influence drug choice, shifting market share in health plans among therapeutic alternatives. Consequently, since copayments are set insurers company rather than pharmaceutical firms, health plans have a device with which they can shift demand. Findings also indicate that insurers affect choice through non-price rationing methods. Implications for Policy, Delivery or Practice: Results have implications for bargaining between insurers and drug manufacturers as well as the implementation of the Medicare prescription drug benefit. By setting differential copayments for therapeutically equivalent drugs, insurers can alter market share by shifting patients towards preferred agents. If drug firms are willing to lower prices to obtain higher volume, insurers could use differential copayments as a negotiating tool. This strategy could also be employed to encourage utilization of cost-effective alternatives within therapeutic drug classes. Prescription drug plans (PDPs) who enroll Medicare beneficiaries could also utilize this approach. The implementation of pharmacy benefit schemes targeted to influence drug selection among beneficiaries would give PDPs substantial bargaining power to command price reductions and generate cost savings to Medicare. • The Causal Effect of Managed Care on Quality: Evidence from Cancer Screening Guideline Discontinuities Srikanth Kadiyala, B.A., Grant Miller, M.P.P. Presented by: Srikanth Kadiyala, B.A., Doctoral Student, Health Policy, Harvard University, NBER / 1050 Massachusetts Avenue, Cambridge, MA 02138; Tel: 617.588.0346; E-mail: kadiyala@fas.harvard.edu Research Objective: Although voluminous, empirical research on the impact of managed care on quality has generally not been conclusive because of difficulties in disentangling true causal effects from well-known associations. The principle methodological obstacle is the sorting of enrollees into plan types according to unobserved heterogeneity in health status and other patient characteristics. We circumvent this difficulty and provide estimates of the causal effect of various forms of managed care on cancer screening rates. Although only one dimension of quality of care, preventive screenings are critical in reducing the incidence of major cancers. Study Design: Specifically, we employ a difference-indifference design that exploits age discontinuities in the amount of cancer screening recommended by widelyrecognized national guidelines. This approach eliminates unobserved patient differences across plan types – importantly, including ones that change over time – using within-plan comparisons that exploit discrete increases in screening rates at recommended age thresholds. Furthermore, using data on offer and take-up rates, we are able to disentangle supply-side effects (i.e., being offered a screening) from demand-side effects (i.e., following a clinician’s advice to have one). Population Studied: We analyze nationally-representative samples of individuals around national screening guideline thresholds using MEDSTAT's MarketScan Research Databases (1998-2001) and the 1995-2001 National Health Interview Surveys. Specific populations studied vary with cancer screening types. For example, individuals aged 45-54 are used for colorectal cancer screening analyses, for which the threshold is age 50. Principal Findings: Preliminary findings suggest that on average, managed care plans are about 20% more likely to provide cancer screenings (mammograms, PSA tests, and fecal occult blood tests) than are traditional indemnity plans. Results specific to various tools employed by managed care as well as extensions to health outcomes are presented. Conclusions: Financial and non-financial incentives appear to have a direct causal effect on the amount of appropriate medical services provided. In particular, high-power payment contracts and more restrictive networks can result in better quality of care along some dimensions. Implications for Policy, Delivery or Practice: Qualityimprovement initiatives should be cognizant of evidence that the traditional tools of managed care can exert influence on quality. Primary Funding Source: NIA • Immigrants and Employer-Provided Health Insurance Anthony LoSasso, Ph.D., Thomas Buchmueller, Ph.D., Sarah Senesky, Ph.D. Presented by: Anthony LoSasso, Ph.D., Research Associate Professor, Institute for Policy Research, Northwestern University, 2040 Sheridan Road, Evanston, IL 60208; Tel: 847.467.3167; Fax: 847.467.4040; E-mail: alosasso@northwestern.edu Research Objective: Immigrants are nearly three times as likely as native-born Americans to lack health insurance: 31.6% of immigrants were uninsured in 2000 versus 11.8% of natives. However, while the difference in group health insurance coverage is striking (66% for natives vs. 48.7% for immigrants), the difference in Medicaid coverage is less than one percentage point (10.4% for natives, 9.9% for immigrants). Thus, in order to understand the lower rates of health insurance coverage for immigrants and to evaluate policies designed to improve this situation, it is necessary to understand the factors that contribute to the low rates of employer-provided insurance coverage of immigrants. The goal of our study is to decompose the reasons behind the striking difference in employer health insurance coverage between natives and immigrants in the US. Study Design: Our analyses use data from the 1996 panel of the Survey of Income and Program Participation (SIPP). The SIPP is a longitudinal survey in which respondents are interviewed every four months over a four-year period. A distinct advantage of the SIPP relative to other more commonly used data sets such as the Current Population Survey (CPS) is that it contains detailed measures of the reasons why respondents are uninsured. For example, unlike the CPS we know whether uninsured workers and their families were offered health insurance, whether they eligible for offered health insurance, and whether they chose not to take-up offered health insurance. In addition, the SIPP contains detailed measures of immigrant status, including year of arrival and country of origin and current citizenship status. In addition to the typical demographic variables (age, race, gender, family size, income), information is also available on work history, tenure with the firm, health status, and functional limitations. Detailed information on educational attainment, including information on degrees received and fields of study, makes it possible to improve upon the years of schooling variable used in many empirical studies. We use regression analysis to decompose the difference in employer-provided health insurance coverage between natives, naturalized citizens, and non-naturalized residents. Population Studied: We use a national random sample of immigrants and native citizens in the US from 1996 through 2000. Principal Findings: We find that employed non-naturalized residents have a baseline 22 percentage point lower level of being offered health insurance by their employer relative to natives. Interestingly, naturalized citizens do not have a statistically significant different rate of being offered health insurance by their employers. When we control for age, education, and family characteristics, the baseline difference in the employer offer rate falls to 11 percentage points, indicating that those factors explain half of the observed difference in the employer health insurance offer rate between natives and nonnaturalized residents. When we add additional variables to control for firm size, tenure with the firm, industry fixed effects, and state fixed effects, the difference is nearly halved again to roughly 6.6 percentage points. Differences in the rate of eligibility and take-up between immigrants and natives are tiny and do not account for a significant portion of the employer-provided health insurance coverage gap. Conclusions: Our results clearly indicate that the primary reason for lack of employer-provided health insurance coverage for non-naturalized residents is that they disproportionately work for firms that do not offer health insurance. The same is not true for naturalized citizens, who generally do not differ significantly from natives. Much of the baseline difference between natives and non-naturalized residents is accounted for by observable differences in characteristics and experiences of the individual as well as characteristics of the firms immigrants work for. Implications for Policy, Delivery or Practice: Our results point to the difficulties that policymakers face in crafting solutions to the employer-provided health insurance gap between natives and non-naturalized citizens. The employers in question are small, low-wage firms most likely with high turnover rates that are unlikely to be able to afford even a basic health insurance policy, thus policy options must take care to avoid causing job loss in an effort to increase insurance coverage. Primary Funding Source: RWJF Related Posters Poster Session B Tuesday, June 8 • 7:30 a.m.-8:45 a.m. • Effects of Medicare HMO Penetration on In-Hospital Mortality and Post-Acute-Care Utilization of Ischemic Stroke John Bian, Ph.D., William Dow, Ph.D., Elizabeth Richardson Vigdor, Ph.D., Morris Weinberger, Ph.D., David Matchar, M.D. Presented by: John Bian, Ph.D., Assistant Professor, Department of Medicine, University of Alabama at Birmingham, MT 640, 1530 3rd Avenue S., Birmingham, AL 35294-4410; Tel: 205.934.7608; Fax: 205.934.7959; E-mail: jbian@uab.edu Research Objective: To understand how increased Medicare HMO penetration affects in-hospital mortality and subsequent use of skilled nursing facilities (SNF) and home health postacute-care (PAC) at the market level. Study Design: Using the 1993-98 Nationwide Inpatient Sample and county-level Medicare HMO enrollment data, we constructed a hospital panel data set for estimation. We focused on in-hospital mortality and, for those discharged alive, use of SNFs and home health. The key independent variable is Medicare risk HMO penetration. The unit of analysis is the individual patient. Discrete-time hazard models were used to estimate the effects of Medicare HMO penetration on in-hospital morality, and logit models were used to estimate the effects of Medicare HMO penetration on utilization of SNFs and home health. Both analyses used hospital fixed effects to control for HMO selection. Population Studied: Patients who were hospitalized for ischemic stroke at age = 65 years old and had no missing state/county codes were included for the analysis. Principal Findings: The analytical file had 327,793 patients during the 6-year period. A total of 27,152 patients (8.3%) died during hospitalization. The rate of in-hospital mortality declined from 9.6% in 1993 to 7.3% in 1998. During the same period, the use of SNF increased from 45.8% in 1993 to 55.2% in 1998, while the use of home health care decreased slightly. Medicare HMO penetration grew steadily from 7.5% in 1993 to 18.0% in 1997, but dipped to 12.2% in 1998; Medicare IPA HMO penetration was 3.0%, 10.9%, and 6.8%, respectively, in 1993, 1997, and 1998. Multivariate analysis showed that Medicare HMO penetration was not associated with inhospital mortality or SNF utilization, but was negatively associated with home health utilization (p < .001). Further analysis suggested that the effects of Medicare IPA and nonIPA HMO penetration on home health utilization differed. Alternative specifications showed the robustness of the above results. Conclusions: After accounting for the censored nature of mortality data and HMO selection, we found no evidence suggesting that increased Medicare HMO penetration affected in-hospital mortality for ischemic stroke. Although no evidence showed any impact of Medicare HMO penetration on utilization of SNF, the utilization rate of home health was negatively associated with Medicare HMO penetration. Additional evidence on post-stroke outcomes (including longterm follow-up of patients) is needed to fully assess the impact of Medicare HMO penetration on stroke care. Implications for Policy, Delivery or Practice: This study suggests that increased Medicare HMO penetration might not have adversely affected outcomes of in-hospital care. Primary Funding Source: AHRQ/NRSA fellowship • Bias in Estimates of Non-Group Health Insurance Coverage: Comparison of Surveys and Administrative Data Joel Cantor, Sc.D., Alan Monheit, Ph.D., Susan Brownlee, Ph.D. Presented by: Joel Cantor, Sc.D., Director and Professor, Center for State Health Policy, Rutgers University, 317 George Street, Suite 400, New Brunswick, NJ 08901; Tel: 732.932.3105 Ext. 228; Fax: 732.932.0069; E-mail: jcantor@cshp.rutgers.edu Research Objective: The non-group health insurance is subject to risk selection and exclusionary insurer practices. Policy responses to these practices are controversial and difficult to evaluate. The Current Population Survey (CPS), the main dataset used to evaluate the non-group market, classifies respondents as having non-group coverage if they answer that they were “covered by a plan that (they) PURCHASED DIRECTLY, that is, not related to current or past employer” in the prior year. This study examines the adequacy of the CPS for studies of the non-group market. Study Design: We compare the number and characteristics of New Jersey residents with non-group coverage in the CPS to aggregate state regulatory reports filed by insurers and to household data from the 2001 NJ Family Health Survey (NJFHS) (n=6,466; response rate=59.4%). CPS and NJFHS estimates of the composition of the NJ non-group market are also compared to a unique survey of known non-group enrollees sampled from the rosters of insurers representing over 95% of the market (n=732; response rate=52.0%). Population Studied: NJ residents with non-group health insurance. Principal Findings: CPS estimates of NJ non-group enrollment are 1.7 to 2.9 times higher than regulatory reports for 1995 to 2001. Moreover, non-group enrollment in regulatory reports declined in each of those years by -0.7 to 0.2 percentage points while the CPS shows annual changes of -2.6 to +1.0 percentage points with no clear trend. Estimates of the NJ non-group population are 91,449 (Q3 2001 NJ regulatory reports), 259,096 (2002 CPS), and 218,585 (2001 NJFHS). The composition of NJ non-group enrollees in the 2002 CPS and 2001 NJFHS is considerably different than the 2002 known sample. CPS and NJFHS non-group enrollees are younger (mean age 44.4 & 44.2 years) compared to the known sample (48.4 years) and more likely to be minority (23.5% & 27.9% versus 12.8%, respectively), non-US born (17.0% & 23.2% versus 12.3%), below 200% FPL (20.3% & 29.3% versus 15.3%), and less than college educated (67.5%& 66.0% versus 57.2%). The three surveys provide similar estimates of health status and utilization (CPS does not ask utilization). Conclusions: Population surveys appear to over-estimate non-group enrollment by two fold or more even after accounting for differences in survey reference periods. The CPS measures coverage at any time in the prior year while regulatory reports and NJFHS estimates reflect a point in time. Thus, we expect CPS to be higher than point-in-time sources. However, the point-in-time 2001 NHFHS overstates non-group enrollment by nearly as much as 2002 CPS compared to regulatory reports (2.8 times higher and 2.4 times higher, respectively). Self-reported non-group enrollees in the CPS and NJFHS are of lower socioeconomic status, younger and more likely minority and non-US born compared to known non-group enrollees. The predominance of managed care in Medicaid may lead some public enrollees to mistakenly report that they are “direct purchasers”. Also, low education and immigrant respondents may not understand survey questions or health coverage complexities. Implications for Policy, Delivery or Practice: Studies of the non-group market using the CPS and other population surveys appear to suffer significant reporting bias. Primary Funding Source: CWF, The Robert Wood Johnson Foundation • TennCare: Cost Saving or Cost Shifting? William Cecil, M.B.A., Steven Coulter, M.D. Presented by: William Cecil, M.B.A., Director, Health Policy Research, Health Care Services, BlueCross BlueShield of Tennessee, 801 Pine Street, Chattanooga, TN 37402; Tel: 423.763.3372; Fax: 423.755-5100; E-mail: bill_cecil@bcbst.com Research Objective: The TennCare program is credited with coverage expansion to the uninsured and uninsurable and health care cost savings. The objective of our study is to determine whether professional health care providers, faced with the cost cuts associated with the TennCare program implementation and coverage expansion absorbed the cost cuts or shifted costs to non-Medicaid non-Medicare payment sources. Study Design: Our analysis is based on a simple theoretical model in which non-Medicare non-Medicaid spending on physician and other professional health care services and total per capita spending depends on the number of beneficiaries in the Medicaid/TennCare program. We hypothesize that physician and other professional health care services providers who faced decreased revenues due to Medicaid/TennCare payment reductions would increase revenues from other health care payment sources, that is, shift costs to the nonMedicare non-Medicaid population. To test our hypothesis, we conducted an analysis of state health accounts health care spending for physician and other professional services and total spending by source of funds. Using this data we estimate time series regression models of non-Medicare non-Medicaid spending on physician and other professional services and total per capita spending by source of funds and service for each state and the District of Columbia, in addition to Tennessee for comparison. The independent variables are trend, TennCare/Medicaid beneficiaries, Medicare enrollment, and the population not covered by Medicare or Medicaid/TennCare. Population Studied: The population reported on in the CMS national and state health accounts reports for Medicare enrollment, Medicaid beneficiaries and the total population from 1980 through 1998 for Tennessee and each state including the District of Columbia. Principal Findings: We find that physician and other professional services and all services cost shifting did occur with TennCare and with Medicaid programs in 17 other states during the time period studied. Six states showed evidence of cost savings associated with the Medicaid program while twenty-six states and the District of Columbia showed no significant effect on spending. The largest cost shift for TennCare occurred in 1995 when 1) the number of TennCare beneficiaries increased by 56%, 2) TennCare spending for physician and other professional services increased by 11%, 3) the non-Medicare non-Medicaid population decreased by 13.5%, 4) non-Medicaid non-Medicare spending for physician and other professional services spending increased by 17%, 5) TennCare per capita spending declined by $656, and 6) allpayer per capita spending for Tennessee rose $229. Conclusions: Consistent with our expectations the TennCare program did not show savings for physician and other professional services and total spending. Instead of cost savings TennCare created cost shifting to those whose health care is largely privately funded. Implications for Policy, Delivery or Practice: The shift of costs to private funding sources represents an additional and unexpected cost burden borne by consumers and employers that would otherwise be funded through tax policy. Policymakers would play an important role by managing employer health care cost burdens at levels that would be nationally and globally competitive. Primary Funding Source: , BlueCross BlueShield of Tennessee • Health Status and the Price Elasticity of Health Plan Choice Bryan Dowd, Ph.D., Adam Atherly, Ph.D., Roger Feldman, Ph.D. Presented by: Bryan Dowd, Ph.D., Professor, Division of Health Services Research and Policy, University of Minnesota, Box 729 MMC, Minneapolis, MN 55455; Tel: 612.6245468; Fax: 612.624.2196; E-mail: dowdx001@tc.umn.edu Research Objective: The purpose of this paper is to (a) interpret the theory of price elasticities that vary with the consumer’s health status, (b) present empirical evidence on that variation, and (c) explore the policy implications of healthrelated variations in plan-choice price elasticities. Study Design: We develop a model of information and preferences that predicts lower price-elasticities for chronically ill individuals. Data on health plan choices are observed for samples of employed individuals and Medicare beneficiaries. Nested logit health plan choice models are estimated and the out-of-pocket premium elasticity of choice is compared for individuals who are healthy versus those who are chronically ill. Population Studied: Twin Cities employees and a national sample of Medicare beneficiaries drawn from the Medicare Current Beneficiary Survey data. Both datasets contain individuals who are healthy and who have self-reported chronic illnesses. Principal Findings: We find a statistically significant, but numerically small reduction in out-of-pocket premium price elasiticities for chronically ill employees relative to their healthy counterparts. The effect is limited to plans offering a broad choice of providers, rather than prepaid group practice HMOs. We also find that price elasticities decline with the length of time the employee has been in the plan. We do not find an effect among Medicare beneficiaries. Conclusions: Although we do not find a strong effect of chronic illness on the price elasticity of health plan choice, other authors have found evidence of an effect, and thus the topic should remain open for further research. Specifically, it would be helpful to have data on charateristics of health plans that are difficult for consumers to learn about without enrolling in the plan. Implications for Policy, Delivery or Practice: If a subset of consumers with low price elasticity could be identified and isolated, health plans might be able to enroll, acclimate, and then charge monopoly premiums to high-risk consumers or consumers with longer enrollment periods. If health plans cannot identify and isolate high risk / low elasticity consumers who prefer their plans, then certain types of benefits or types of health plans might be subject to "death spirals." • Realistic Consumerism: Assessing the Prospects for Consumer Empowerment with Respect to Health Plan Experiences Mark Schlesinger, Ph.D., Brian Elbel, M.P.H. Presented by: Brian Elbel, M.P.H., Doctoral Student, Health Policy and Administration, Yale University, P.O. Box 208034, New Haven, CT 06520; Tel: 203.809.0517; Fax: 203.785.6287; E-mail: brian.elbel@yale.edu Research Objective: American health policy has increasingly relied on consumer empowerment to improve the performance of health plans. Consumers who experience problems are expected to respond by either switching to a new plan (exit) or filing a grievance with the plan (formal voice). In this study, we assess how consumers actually respond to problems of various sorts. We estimate the potential for enhancing empowerment by examining the behavior of consumers who are already well-educated and well-informed about their options in the health care system. Study Design: We assess the prevalence of consumer responses, focusing on problems that have the most severe consequences for consumers and on those consumers who attribute their problem, at least in part, to the performance of their health plan. To simulate the potential for enhanced consumerism, we identify various subsets of “empowered” consumers, using various plausible definitions of empowerment to test the reliability of our findings. Population Studied: Nationally representative telephone sample of 5000 individuals with health insurance. Approximately half of all individuals had a problem with their health care in the last year, and 67% attribute this problem at least in part to their health plan. Principal Findings: Exit and formal voice were utilized by only 18% and 6%, respectively, of individuals who had a problem with their health care that they blamed, at least in part, on their health plan. Active consumerism increased for problems with more severe consequences, but was still reported by less than half of all respondents. Among individuals who had to pay at least $1000 in out of pocket expenses, aggrieved patients filed a grievance only 35% of the time and left their plan in only 11% of all cases. Among enrollees who reported having experienced a serious health decline because of the problem in question, 38% formally complained to the plan and 17% switched health plans. Our sub-sample of empowered consumers did not respond appreciably more often, though there is some variation across types of problems. We continue to explore alternative groupings to characterize “empowerment”. Although exit and formal voice appear to be difficult to establish as reliable forms of consumerism, health plan enrollees do more consistently respond to problems in other ways, including informal contacts with plans and communicating with their physicians. Conclusions: Market-oriented policies that rely on empowered consumers to police the behavior of health plans have evident limitations. Although the subset of enrollees who do respond through exit or voice may create an adequate signal to encourage plans to improve their average performance, these consumer responses are too inconsistent to provide reliable protection for individual patients. Implications for Policy, Delivery or Practice: At all levels of empowerment, health plan enrollees respond to problems primarily through informal mechanisms, which are unlikely to be identified by existing state monitoring programs or conveyed to the public on report cards measuring plan performance. There appears to be only limited potential to increase empowerment by providing enrollees with more information or educating them about the appropriate ways in which to respond to problems related to their health care. • The Effects of State Managed Care Patient Protections on Physicians' Attitudes Frank Sloan, Ph.D., Mark Hall, J.D., John Rattliff, Ph.D., Mark Hall, J.D. Presented by: Mark Hall, J.D., Professor of Law and Public Health, Department of Public Health Sciences, Wake Forest University Medical School, 2000 W. 1st Street, WinstonSalem, NC 27157-1063; Tel: 336.716.9807; Fax: 336.716.7554; Email: mhall@wfubmc.edu Research Objective: During the 1990s, as part of the backlash against managed care organizations (MCOs), 47 states adopted laws protecting against some of the perceived abuses of MCOs. These laws included: external review of coverage denials; subjecting MCOs to tort liability; requiring direct patient access to specialists without obtaining a gatekeeper’s approval; requiring payment for emergency room visits under a prudent layperson’s standard of necessity; and restricting use of financial incentives. In most states, this legislation was spearheaded by medical societies and other groups representing providers. This study assesses whether these laws improved physicians’ views about the conditions of their medical practice. Study Design: Data came from 3 waves of the Community Tracking Study physician surveys, conducted between 19962001, with about 15,000 respondents nationally. Most physicians participated in more than 1 wave. The attitude measures included physician reports of career satisfaction, clinical freedom, the quality of care the physician is able to provide, ease of referrals, and the degree to which profiling affects compensation. Each physician’s change in these views between survey rounds was modeled as dependent on whether the state had enacted a set of managed care patient protection laws (or a “patients’ bill of rights”), either prior to or during the respective time intervals between survey rounds. We estimated separate regressions for primary care providers (PCPs) and specialists. Principal Findings: Between 1996-7 and 1998-9, specialists, but not PCPs, perceived improvements in states with a set of patient protection laws. It did not matter whether the laws had been implemented prior to or during this time interval. However, in the analysis of 1998-9 to 2000-1 changes, neither PCPs nor specialists perceived improvements in states that enacted or retained patient protection laws in those years. Conclusions: Overall, this analysis suggests that these laws improved specialist physicians’ views of their medical practices when relatively few states had enacted the laws. However, after many states had done so, this effect disappeared, both for states with newly enacted and continuing laws. Qualitative interviews reported elsewhere from a different part of this study suggest that there was a tipping point after which the MCOs changed their practices in all states in which they operated rather than differentiate according to the particular regulatory environments. Thus, although there were improvements overall in physicians’ attitudes between 1998-9 and 2000-1, these improvements were not more likely in states with patient protection laws. Also, these laws appear more relevant to specialists than PCPs. Implications for Policy, Delivery or Practice: This analysis does not identify the precise mechanism through which physicians’ attitudes may have been affected by these laws. Changes in attitudes may reflect concrete underlying changes in the behavior of health plans, or physicians’ perceptions may be improved by knowing that they achieved a desired legislative victory. The fact that improved attitudes were shortlived in the first group of enacting states suggests the latter possibility. However, the facts that improvements were seen only among specialists, and that physicians’ attitudes continued to improve without regard to these laws, suggest these laws may have been one of the social forces that succeeded in changing managed care practices across the industry. Primary Funding Source: RWJF • The Effect of HMO Coverage on the Choice of Outpatient or Inpatient Surgery Hsou Mei Hu, Ph.D., Niccie McKay, Ph.D. Presented by: Hsou Mei Hu, Ph.D., Postdoctoral Fellow, Institute for Health, Health Care Policy and Aging Research, Rutgers the State University of New Jersey, 30 College Avenue, New Brunswick, NJ 08901; Tel: 732.932.6942; Fax: (732) 9326872; E-mail: hhu@ihhcpar.rutgers.edu Research Objective: To examine the effect of health maintenance organization (HMO) coverage on the choice of surgery setting for a surgery that is feasible in either the inpatient or outpatient setting. Study Design: We constructed a dataset pooling 1997, 1998, 1999, and 2000 Medical Expenditure Panel Survey (MEPS) data. Cases included in the final dataset are for privately insured individuals under age 65 with an ICD-9 procedure code that was reported in both the 1996 National Hospital Discharge Survey and the 1996 National Survey of Ambulatory Surgery. Surgeries that were done in only one of the settings (inpatient or outpatient) are excluded. The final dataset contains 1,044 cases (702 outpatient and 342 inpatient). The dependent variable is whether the surgery was performed outpatient or inpatient. The independent variable (type of health plan) has four possible values: HMO, fee-for-service (FFS), non-HMO with gatekeeper, or multiple plans. The percentage of cases by type of health plan (HMO, non-HMO with gatekeeper, FFS, multiple plans) is 46.7%, 7.2%, 35.2%, and 10.9%, respectively. A logistic regression model was specified to analyze the effect of HMO coverage on having an outpatient surgery (vs. inpatient surgery), controlling for severity, patient characteristics, and payer characteristics. Population Studied: Privately insured patients under age 65 who had a surgery that is feasible in either inpatient or outpatient setting Principal Findings: The percentage of cases with HMO coverage having an outpatient surgery (69.1%) was similar to that for those with FFS coverage (70.5%), while 77.3% of cases covered by a non-HMO with gatekeeper and 65.4% of those with multiple plan coverage were outpatient. The regression results showed that HMO coverage did not increase the likelihood of having an outpatient surgery, holding other factors constant. However, the interaction between HMO status and facility payment had a significant negative effect on the likelihood of having an outpatient surgery. Specifically, when facility payment increased, the likelihood of having an outpatient surgery dropped more for those with HMO coverage than for those with FFS coverage. Non-HMO with gatekeeper and multiple plan coverage did not significantly affect the likelihood of having an outpatient surgery. Conclusions: This study found that HMO coverage did not increase the likelihood of having outpatient surgery relative to those with FFS coverage. However, HMOs did pay less for both types of surgery. These results are inconsistent with the general belief that HMOs control costs by directly controlling the use of services, at least for inpatient vs. outpatient surgery. Rather, HMOs seem to have focused more on controlling payments to providers than on controlling utilization. Implications for Policy, Delivery or Practice: When examining how health plans control expenditures, it is important to distinguish between payment-control approaches and methods that focus on controlling utilization. • Decomposing the Changes in the Number of Uninsured in USA Since 1994 Mahmud Khan, Ph.D. Presented by: Mahmud Khan, Ph.D., Professor, Health Systems Management, Tulane University, 1440 Canal Street, #1900, New Orleans, LA 70112; Tel: 504.584.1979; Fax: 504.584.3783; E-mail: khan@tulane.edu Research Objective: To decompose the uninsured population into a number of components or determinants like poverty, age, gender, race and source of insurance Study Design: Difference equations were used to decompose the uninsurance rates to identify the proportion of change affected by the selected social and economic determinants. Population Studied: Macro uninsurance data for a number of years since 1994 have been used. Principal Findings: The decomposition exercise indicates that if there were no changes in the insurance coverage rates of different races from the base year levels, uninsured population would have increased by 2.3 million over the years 1994-98 and 0.6 million over 1998-99. Employer sponsored insurance (ESI) was the most important factor explaining the reduction in the number of uninsured. Even during 1994-98, employer sponsored insurance would have reduced the size of uninsured if other insurance programs did not add so many to the uninsured pool. Prevalence of poverty was not a very strong factor in explaining the uninsurance rate. The more important factor was the improvements in the income-classspecific coverage rates. Conclusions: Changes related to economic growth were crucial in determining the size of the uninsured population in the USA since 1994. In order to reduce the variability of insurance coverage, policy makers need to identify alternative methods of providing health insurance coverage to the population so that access to insurance becomes less sensitive to macroeconomic changes. Implications for Policy, Delivery or Practice: Uninsurance rate in a market-oriented health care system will remain at a relatively high level even with strong performance of the macro economy. Public sector intervention appears to be the only feasible way of reducing uninsurance rates significantly in the longer run. • Impact of Three-Tier Pharmacy Benefit Design on Demand for Prescription Medications: Results from Three Prescription Benefit Plans Winnie Yi, Pharm.D., BCPS, Pamela Landsman, M.P.H., Dr.PH, XiaoFeng Liu, Ph.D., Steven Teutsch, M.D., Marc Berger, M.D. Presented by: Pamela Landsman, M.P.H., Dr.PH, Sr. Manager, Outcomes Research & Management, Merck & Co., Inc., PO Box 4, WP39-166, West Point, PA 19486-0004; Tel: 215.652.7492; Fax: 215.652.0860; E-mail: pamela_landsman@merck.com Research Objective: This study estimates the price elasticity of demand for specific therapeutic classes among three managed care commercial populations undergoing different two-tier to three-tier pharmacy benefit plan changes for specific therapeutic classes. Study Design: Retrospective pharmacy claims database analysis for consumers enrolled in one of three prescription benefit plans filling prescriptions for an NSAID, SSRI, ACE-I, CCB, ARB, HMG-CoA, or triptan. Population Studied: During the period 1999-2001, three prescription drug benefit plans managed by a large pharmacy benefit management organization moved enrollees from a two- to a three-tier formulary structure. Copayments for enrollees of Plan 1 increased from $5 for tier 1 and $10 for tier 2 drugs ($5/$10) to $5 for tier 1, $15 for tier 2, and $25 for tier 3 drugs ($5/$15/$25). Plan 2 changed from $10/$20 to $10/$20/$40 and Plan 3 from $5/$10 to $5/$20/$35. Plans were geographically dispersed; Plan 1 enrollees resided in the Northeast, Plan 2 enrollees in the South, and Plan 3 enrollees in the Southeast. Patients were continuously enrolled for 24 months with 12 months experience pre- and post- benefit change and have at least 2 prescriptions filled for a drug of interest 3-months prior to the benefit change. Elasticity of demand was calculated for each drug class within the individual plans. Principal Findings: Elasticity of demand varied by therapeutic classes and by plans. In general, drugs that are used for chronic or asymptomatic conditions (e.g. ACEIs, ARBs, statins, SSRIs) tend to be less elastic than drugs used for acute symptomatic conditions (e.g. triptans, Cox-2 and NSAIDs). Elasticity of demand varied also from plan to plan without consistent patterns For example, Plan 2 enrollees were more price sensitive (elasticity -0.24) for ARBs compared to the other two plans (-0.16 and -0.13). Elasticity for Cox-2s (0.50) in Plan 2 was also nearly twice that of Plans 1 and 3 (0.26 for both). Plan 1 enrollees demonstrated a greater elasticity of demand for SSRIs in contrast to those in Plans 2 and 3 (-0.47 versus -0.20 and -0.31). Neither age nor gender, appear to explain elasticity differences seen among the plans. Conclusions: The demand of pharmaceuticals in these populations was relatively inelastic for medications treating chronic asymptomatic conditions and somewhat higher for medications treating acute symptomatic conditions. While demand decreased with an increase in copayment, the magnitude of the decrease varied with both drug class and copayment. Utilization changes across plans may be explained by other factors not examined such as overall and baseline drug utilization, total member copayment burden and other member education and disease management initiatives in place. More detailed analyses examining the tiers for specific products may also explain the inconsistent patterns of elasticity seen across plans. Implications for Policy, Delivery or Practice: Patients are sensitive to changes in copayments for prescription medications. Their sensitivity varies with medical condition and class of medication. Other factors along with cost may explain the variation in sensitivity across plans. Primary Funding Source: Merck & Co., Inc. • The Health Care Safety Net and Crowd-Out of Private Health Insurance Anthony LoSasso, Ph.D., Bruce Meyer, Ph.D. Presented by: Anthony LoSasso, Ph.D., Research Associate Professor, Institute for Policy Research, Northwestern University, 2040 Sheridan Road, Evanston, IL 60208; Tel: 847.467.3167; Fax: 847.467.4040; E-mail: alosasso@northwestern.edu Research Objective: To examine the impact of the health care safety net on the health insurance coverage of children. Study Design: We conduct an individual-level analysis of children's health insurance coverage using data from the Current Population Survey for the years 1990-2000 combined with a long panel of state level data on hospital uncompensated care and free and reduced price care offered by Federally Qualified Health Centers (FQHCs). Because measures of the safety net are likely to be confounded with insurance coverage, we use an instrumental variables approach in estimating the impact of the safety net on health insurance coverage. The instruments for our measures of the safety net include measures of tax appropriations and state and local support received by hospitals and FQHCs, state Disproportionate Share payments, state uncompensated care pool dollars, and the amount of state budget surplus or deficit. We also control for Medicaid eligibility of sample members with standard techniques. By additionally including state fixed effects we can avoid the bias that can occur in previous cross-sectional designs because of unobserved differences across states that result in, for example, high levels of both uninsurance and safety net dollars. Including state fixed effects lets us restrict the analysis to the within-state variation in health insurance coverage and the health care safety net measures. Population Studied: A national sample of children aged 14 and under for the years 1990-2000. Principal Findings: When we omit state fixed effects, we observe that high levels of hospital uncompensated care are associated with statistically significantly higher levels of uninsurance and lower levels of both public insurance and private insurance. Our results imply that a one standard deviation increase in hospital uncompensated care per population (a 46% increase) would be associated with a 3 percentage point increase in the uninsurance rate among children. However this result, which generally conforms to prior findings, does not control for unobserved state heterogeneity. When we include state fixed effects we find no statistically significant impact of hospital or FQHC uncompensated care on health insurance coverage. Conclusions: While there are clear cross-state differences in insurance coverage and safety net support, it does not appear that differences in safety net support are causally related to insurance coverage for children. Implications for Policy, Delivery or Practice: Our results suggest that, contrary to previous work, unintended health insurance coverage consequences associated with the health care safety net are likely to be minimal. Primary Funding Source: RWJF • The Impact of Benefit Buy-Down on Health Services Utilization Raymond Phillippi, Ph.D., Phil Johnson, FSA, MAAA, Allen Naidoo, Ph.D., Natalie Ramey Presented by: Raymond Phillippi, Ph.D., Sr. Biostatistical Research Scientist, Health Services Research, BlueCross BlueShield of TN, 801 Pine Street, 3E, Chattanooga, TN 37402; Tel: 423.752.8243; Fax: 423.755.5100; E-mail: raymond_phillippi@bcbst.com Research Objective: With the rise in health care costs, many employers have been changing the benefit structure that they offer their employees in order to reduce their premiums, called benefit buy-down in the industry. While this clearly shifts costs to the employee, little is know about the impact of buydown on health care utilization. This research was meant to answer two questions; 1. Does benefit buy-down affect the utilization of services by members, and 2. If buy-down does affect utilization, are there threshold levels of buy-down that must be reached in order to have an effect? Study Design: Data were collected on small groups that renewed their health insurance with BlueCross BlueShield of Tennessee between July 2001 and June 2002. Utilization data, consisting of admissions, inpatient days, hospital outpatient visits, emergency room visits, physician office visits, and overall utilization, for these groups were collected for the twelve months prior and twelve months after the groups renewed. Groups who changed their benefit structure by increasing coinsurance, deductibles, out of pocket maximums, emergency room co-pays, outpatient surgery coinsurance, or office visit co-pay or who changed to a more restrictive network were classified as changed groups. All others were classified as no change groups. In addition, a second independent variable consisting of the sixteen patterns of change among the first four changes was created. All utilization measures were examined pre and post renewal to determine if the benefit buy down had an effect with the no change groups serving as controls. Population Studied: Members of commercial small groups, under 150 employees, insured by BlueCross BlueShield of Tennessee. Principal Findings: Among groups that changed benefits, there was a statistically significant decrease in inpatient days, ER visits, and office visits while the groups that did not change benefits either had no change in utilization or showed an increase in utilization. The pattern analysis revealed that few patterns reduced utilization across the board and almost all patterns reduced office visits. Further analysis revealed that seventy-seven percent of those utilization measures that showed significant reductions were related to a pattern that involved an increase in office visit co-pay. The threshold analysis showed that regardless of how much office visit copay increased, only co-pays of twenty dollars or more reduced utilization and the greater the increase over twenty dollars, the greater the reduction. Conclusions: Changing benefit structure in general has little impact on utilization, however increasing physician office visit co-pay not only reduces office visits, but also reduces other utilization parameters. This makes sense if the physician drives all other health care utilization. Implications for Policy, Delivery or Practice: From an insurer’s perspective, reducing unnecessary utilization, which will help to control health care costs and insurance premiums, can be accomplished by increasing the physician office visit co-pay. • Health Benefits Offer Rates: Are We Reaching the Wrong Conclusions Because of Nonresponse? Jeremy Pickreign, M.S., Jon Gabel, M.S. Presented by: Jeremy Pickreign, M.S., Statistician, Health System Studies, HRET, One Empire Drive, Rensselaer, NY 12144; Tel: 518.431.7827; Fax: 518.431.7915; E-mail: jpickreign@aha.org Research Objective: To determine if a bias due to nonresponse exists when estimating the offer rate for health benefits in firms with fewer than 50 workers Study Design: We conducted a follow-up survey to the 2003 Employer Health Benefits Survey sponsored by the Kaiser Family Foundation and Health Research and Educational Trust. This follow-up survey collected health benefits offering data from firms with fewer than 50 workers that did not fully participate in the main survey. We used McNemar’s Test to verify that the follow-up survey provided comparable results to the main survey, and t-tests were used to identify bias in offer rates between responders and nonresponders. We calculated and applied a simple weighting adjustment to the main survey data to control for the observed bias in health benefits offer rates and compared the results. Population Studied: Private and public firms with at least three workers and as many as 49 workers throughout the United States. Principal Findings: Firms with 50 or fewer workers not responding to the main survey were significantly less likely to offer health benefits. Combining the responses from the two surveys, we find that the unweighted offer rate is being overstated by 4.6 percentage points, a rate that exceeds a common rule of thumb for determining nonresponse bias. After adjusting the weights to control for this bias, we find that the weighted offer rate for firms with fewer than 50 workers drops from 66.7 percent to 63.4 percent. Conclusions: A positive bias exists in the offer rates of firms with 50 or fewer workers, most likely due to nonresponse by firms that do not offer health benefits. An adjustment to the sampling weights to reflect this bias resulted in a reduction of the offer rate by over three percentage points in firms with fewer than 50 workers. Implications for Policy, Delivery or Practice: These findings send a proceed with caution warning to analysts studying offer rates. Because the offer rate appears inversely correlated with the survey response rate, analysts need to account for differences in survey response rates when analyzing and reporting on health benefits offer rates. • Quality and Utilization in Managed Care: Is More Care Better? Sarah Scholle, M.P.H., Dr.PH, Russell Mardon, Ph.D., Gregory Pawlson, M.D., M.P.H., FACP Presented by: Sarah Scholle, M.P.H., Dr.PH, Assistant Vice President, Research & Analysis, National Committee for Quality Assurance, 2000 L Street, N.W., Suite 500, Washington, DC 20036; Tel: 202.955.1726; Fax: 202.955.3599; E-mail: scholle@ncqa.org Research Objective: Previous studies of Medicare data show that large regional and local differences in health care spending are unrelated to access to care, quality of care, health outcomes and satisfaction. As health care spending rises, it is unclear whether the purchasers and consumers of care are getting value for their health care dollar. The purpose of this study was to examine correlations of commercial health plan performance on HEDIS effectiveness of care with access to care, outpatient use, and hospital use, as a proxy for cost. Study Design: Health care utilization and quality measures reported by 316 commercial managed care plans in the 2003 HEDIS data were analyzed. Plans report on a standardized set of performance measures using detailed specifications. To be considered in the HEDIS data set, an independent audit of data collection procedures is required. This study used data from all reporting plans (including unaccredited plans that do not allow public reporting of their data). Utilization measures include Access to care (the proportion of adults with at least one primary care or preventive visit) and the rate of outpatient visits. As a proxy for potentially avoidable hospitalization, the rate of hospitalization (excluding surgical and obstetric care) examined. Quality measures include all the measures available in the HEDIS Effectiveness of Care domain, such as the diabetes measurement set, anti-depressant medication management, follow-up after hospitalization for mental illness, use appropriate medication for people with asthma, breast cancer screening, cervical cancer screening, cholesterol management after acute cardiovascular event, beta blocker treatment after a heart attack, controlling high blood pressure, Chlamydia screening, childhood immunizations and adolescent immunizations. Population Studied: Three hundred sixteen plans representing 63% of MCOs in the US and representing 87% of individuals enrolled in commercial MCOs. Average number of enrollees per plan is 211,635. The commercial enrollee population is 52% female and less than 2% are ages 65 years or older. Plan product types include HMOs (36%), HMO/POS combined (61%), and POS (3%). Principal Findings: Rates of medical hospitalizations and outpatient visits vary by more than six-fold between the highest and lowest utilization plans. For example, the average annual number of medical Hospitalizations among adults age 45-64 ranged from 11.5 to 76.5 per 1000 members, with a mean of 48.4. Access to care (defined as at least one primary or preventive care visit) was not as variable, with a mean of 94.2 and range of 67.7 to 98.4 for this age group. Higher rates of access to primary and preventive visits were associated with higher performance on HEDIS measures. For example, access to care for adults age 45-64 was correlated 0.31 with the plan’s average performance on diabetes control, 0.31 with acute phase depression management, 0.39 with appropriate asthma medication use, 0.45 with breast cancer screening, and 0.30 with childhood immunizations. In contrast, higher rates of hospital care were associated with lower quality performance with performance on HEDIS quality of care measures: -0.27 for Diabetes control, -0.32 for acute phase anti-depressant medication management, -0.11 for appropriate asthma medication use, -0.22 for breast cancer screening, and -0.11 with childhood immunizations. Quality was only slightly, if any amount, correlated with outpatient visit utilization rates:0.00 for diabetes HbA1c control, 0.09 for acute phase depression management, 0.03 for asthma medication, 0.04 for breast cancer screening, and 0.00 with childhood immunizations. Conclusions: High quality health plans may reduce hospitalization care, while increasing access to primary care services. Some plans may achieve high quality more efficiently than others, given the lack of correlation between quality and utilization. Further research should examine whether potential differences in patient populations, health plan characteristics and provider supply factors account for these correlations. Implications for Policy, Delivery or Practice: With health care costs rising and placing greater burdens on purchasers and consumers, there is a critical need to learn more about ways to identify and reward efficient provision of high quality care. Initiatives to identify health plans that achieve higher performance while minimizing costs are needed. • Favorable HMO Selection among Medicare Enrolled Veterans Min-Woong Sohn, Ph.D., Denise Hynes, Ph.D., R.N., Kristin Koelling, M.P.H., Linda Kok, M.S., Noreen Arnold, M.S. Presented by: Min-Woong Sohn, Ph.D., Associate Director, VA Information Resource Center, Edward Hines Jr. VA Hospital, 5th Avenue and Roosevelt Road, Building 1, Room C303, Hines, IL 60141; Tel: 708.202.2413; Fax: 708. 202.2415; E-mail: sohn@research.hines.med.va.gov Research Objective: To investigate whether there has been favorable HMO selection among Medicare enrolled veterans aged 65 years old or older. Favorable HMO selection refers to the phenomenon that healthier individuals selectively join managed care programs more than those with poorer health. Study Design: We used the VA-Medicare linked data to identify veterans who newly enrolled in Medicare+Choice in 2000. For risk adjustment, we used the Hierarchical Condition Categories (HCC) risk adjustment model, which will be used by the Centers for Medicare and Medicaid Services (CMS) to risk adjust Medicare reimbursements in 2004. We combined the VA administrative databases and the Medicare claims files for 1999 to compute the “concurrent” HCC risk scores for patients. Hierarchies in HCC are designed to allow for coding variations due to different financial incentives under the VA and Medicare systems. Four risk groups were defined based on the quartile values on the risk scores. Multiple logistic regression was used to model the odds of a veteran joining Medicare+Choice in 2000, controlling for individual, access, and market characteristics. Population Studied: Our sample consisted of veterans who were 65 years old or older on January 1, 1999 and were Medicare enrolled but did not participate in Medicare+Choice in any month in 1999. Principal Findings: There were 1.65 million Medicare enrolled veterans who did not participate in Medicare+Choice in 1999; about 16,000 (0.1%) newly enrolled in this program in 2000. Bivariate analysis shows that only 0.78% in the highest risk group joined the program, compared with 1.27% in the lowest risk group (p<0.0001). Adjusted for individual, access, and market characteristics, veterans in the highest risk group were about 35% less likely to join the program than those in the lowest risk group (Odds Ratio = 0.65, p < 0.001). Additionally, we found that blacks (OR = 1.29, p < 0.001) and veterans with low incomes (OR = 1.27, p < 0.001) were more likely to join, while older veterans (OR = 0.54, p <0.001 for veterans 85 and older, compared to veterans 65-74) were less likely to join Medicare+Choice. Conclusions: Like the general elderly population in the United States, healthier veterans joined Medicare+Chioice more frequently than veterans with poor health, indicating favorable HMO selection among veterans. Unlike previous studies that relied on self-reported health status, we used the HCC risk scores as an objective measure of health status for veterans. Implications for Policy, Delivery or Practice: The proposed VA+Choice Medicare plan for Medicare eligible veterans needs to carefully consider how favorable HMO selection may affect this program. Primary Funding Source: VA percutaneous transluminal coronary angioplasty (PTCA) in a managed care environment. Both competition and volume are known to affect quality of care independently, but their combined effect on mortality has not been studied in detail. Study Design: In-hospital mortality of patients who received coronary angioplasty were used as the main outcome measure. Patient, hospital and environmental characteristics were included as covariates in multivariate logistic regression models. The APR-DRG severity index was used to adjust for patient severity. Population Studied: A cross-sectional sample of 37,206 patients who received coronary angioplasty at 116 hospitals in California in 1995. Principal Findings: Competition was negatively and significantly associated with adjusted mortality in hospitals with 350 or less procedures a year. A 10% increase in competition was associated with a 12% decrease in mortality at a volume of 50 (Odds Ratio [OR] = 0.881; 95% Confidence Interval [CI], 0.850 – 0.978; P < 0.000) and with a 5.7% decrease at a volume of 350 (OR = 0.943; 95% CD, 0.896 – 0.993; P = 0.026). Competition was associated with increased mortality at 650 or more procedures. Hospitals with higher volume had lower mortality when under low to moderate competition. An increase in volume by 10 procedures was associated with a decrease in mortality of 1.8% for hospitals under no competition (OR = 0.982; 95% CI, 0.972 – 0.992; P < 0.000) and of 0.7% for hospitals under moderate competition (OR = 0.993; 95% CI, 0.988 – 0.998; P = 0.004). For hospitals under intense competition, an increase in volume did not affect mortality. Conclusions: The effect of volume was not constant across different levels of competition. For hospitals that operated under intense competition, increasing volume did not improve quality of care for patients who received coronary angioplasty. This suggests that price competition among high volume hospitals to obtain contracts may have resulted in increased pressures to reduce cost and the resulting cost constraint might have increased the likelihood among hospitals under intense competition to engage in qualityskimping behavior. Implications for Policy, Delivery or Practice: Regionalization of specialized clinical services may not necessarily improve outcomes if it involves concentrating volume in hospitals under intense competition. Recent proposal to regionalize services in metropolitan areas may need to be reconsidered, because hospitals in metropolitan areas are likely to have higher volume and be under stronger competition than those in the fringes or in rural areas. Primary Funding Source: VA • The Association between Competition, Volume and Mortality among Patients Receiving Coronary Angioplasty Min-Woong Sohn, Ph.D., Paul Rathouz, Ph.D., Larry Manheim, Ph.D., Neeraj Jolly, M.D. • Utilization and Outcomes of Chiropractic Care for WorkRelated LBP: Impact of Workers' Compensation Reimbursement Policies Radoslaw Wasiak, Ph.D., Eileen McNeely, Ph.D., Sandra Magnetti, Ph.D. Presented by: Min-Woong Sohn, Ph.D., Associate Director, VA Information Resource Center, Edward Hines Jr. VA Hospital, 5th Avenue and Roosevelt Road, Building 1, Room C303, Hines, IL 60141; Tel: 708.202.2413; Fax: 708.202.2415; Email: sohn@research.hines.med.va.gov Research Objective: To examine the association between competition, volume, and mortality for patients who received Presented by: Radoslaw Wasiak, Ph.D., Researcher, Center for Disability Research, Liberty Mutual, 71 Frankland Road, Hopkinton, MA 01748; Tel: 508.497.0242; Fax: 508.435.8136; Email: radoslaw.wasiak@libertymutual.com Research Objective: In workers’ compensation (WC), utilization and pricing of chiropractic care for work-related low back pain (LBP) is subject to several cost containment measures (reimbursement limits, provider choice/change, utilization/bill review, visit limits). We examined the impact of state reimbursement policies (fee schedules and payment policy) on the utilization and cost of chiropractic care for occupational non-specific LBP to answer the following questions: 1. Do WC reimbursement policies effectively control costs associated with chiropractic care for occupational LBP? 2. Is high/low reimbursement for chiropractic care associated with high/low utilization? Study Design: A cross-sectional analysis of claims from a large WC insurer (10% of US market) for service years 19992001 in seven states. For most frequently utilized CPT codes, fee schedule and actual reimbursement levels were benchmarked against the Medicare fee schedules for chiropractic treatments. WC and actual reimbursement indices were developed using weighted averages to proxy state payment policies and classify states as high or low reimbursers. Utilization and outcomes measures included: chiropractic visits and services per individual, services per chiropractic visit, cost of chiropractic care per visit and per individual. Population Studied: Individuals with a work-related nonspecific LBP reported to the insurer in one of seven states (FL, ID, IL, MD, NH, NY, PA) and at least one visit to a chiropractor during the 1999-2001 period (N-99=4732, N00=4898, N-01=5524). Principal Findings: States vary in their relative reimbursement levels for chiropractic services provided to individuals with work-related LBP (low reimbursers: FL, MD, NY; high: ID, IL, NH, PA; range 0.52-2.36 with the value of 1.00 denoting average weighted reimbursement at the Medicare fee schedule level). Similarly, utilization and outcomes measures show substantial variation across the states. Median visits per individual ranged from 5 to 14, services per individual from 11 to 25, services per visit from 1 to 3, chiropractic costs per visit from $28 to $69, and chiropractic costs per individual from $290 to $655. None of the utilization measures correlated with the relative level of reimbursement for chiropractic services. Low reimbursement states had lower costs per visit (F-value 15.72, p=0.01) but such relationship was not observed for costs per individual. Conclusions: Payment policies seem to effectively control chiropractic costs per visit. Such association does not carry through to chiropractic costs per individual, as chiropractors adjust their utilization in low reimbursement states. The utilization adjustment varies by state; chiropractors increase either the number of visits per individual (NY) or the number of services per visit (FL and MD). Given other state WC policies, which are targeted more directly at utilization, high or low reimbursement does not correlate with any of utilization measures. Implications for Policy, Delivery or Practice: Restrictive reimbursement policy aimed at curbing chiropractic costs and utilization is not able to reduce both costs and utilization. Providers adjust to introduced price limitations to preserve the overall level of revenue per individual. Other WC measures may be better suited to target utilization levels in conjunction with payment policies and fee schedules. More research is required to understand the joint impact of various cost containment measures on utilization and outcomes of chiropractic care in WC. • Index of Treatment Variation: A Qualitative Approach to Measuring Practice Patterns in the Treatment of Asthma William Westerfield, M.A., Allen Naidoo, Ph.D. CHES, Ken Patric, M.D., Judy Slagle, R.N., M.P.A. Presented by: William Westerfield, M.A., Senior Research Analyst, Health Services Research, BlueCross BlueShield of Tennessee, 801 Pine Street, 3E, Chattanooga, TN 37402; Tel: 423.755.6260; E-mail: William_Westerfield@BCBST.com Research Objective: The primary objective of this study was to measure the relative variation in the courses of treatment selected by a provider or specialty to treat asthma without comorbidity. The secondary objective of this study was to determine the most common course of treatment used by a provider or specialty to treat asthma without comorbidity. Study Design: Treatment variation was measured using the Index of Qualitative Variation. Called the Index of Treatment Variation (ITV) for the purposes of this study, the ITV is an appropriate measure for determining the relative number of differences in a given set of items. ITV scores range from 0.00 to 1.00. An ITV score of 0.00 indicates no treatment variation, while an ITV score of 1.00 indicates maximum treatment variation. To measure the variation in distinct courses of treatment, data based on an “episode of care” model and extracted from a commercial Preferred Provider Organization were used. The data covered episodes of care from June 2002 – May 2003. Episodes from the Episode Treatment Group 389 “Asthma without Comorbity” were chosen to test the variation in asthma treatments. Berenson & Eggers Type of Service Codes were used to identify medical treatments. Therapeutic Class Codes were used to identify drug therapies. Population Studied: 2,923 episodes of asthma without comorbidity were studied. The episodes were treated by a variety of specialties including Internal Medicine, Family Practice, Pediatric Medicine, Nurse Practitioner, Pulmonary Disease, Emergency Medicine, and Allergy & Immunology. Principal Findings: ITV scores for individual providers ranged from 0.84 to 1.00. The most common ITV score was 1.00. The provider with the lowest treatment variation was an Allergy & Immunologist. This particular provider treated 5 episodes of asthma and used 3 different courses of treatment. The provider’s most common course of treatment included an office visit, a respiratory flow volume loop, an adrenergic agent, a sympathomimetic agent, and other unclassified drug therapies. The specialty with the lowest treatment variation was also Allergy & Immunology. This specialty treated 144 episodes of asthma and used 106 different courses of treatment. This specialties’ most common course of treatment included an office visit, a respiratory volume flow loop, an adrenergic agent, and a sympathomimetic agent. Conclusions: The variation in treatment courses used by the providers and specialists to treat asthma appears clinically significant. The most common approach to the treatment of asthma was a different course of treatment for each episode. The Index of Treatment Variation is a useful measure for comparing the amount of variation in courses of treatment selected by providers and/or specialties to treat a particular illness. Implications for Policy, Delivery or Practice: The Index of Treatment Variation is a valuable and flexible measure. Combined with an episode of care model, the Index of Treatment Variation could be used in practice pattern analysis, disease profiling, evidence-based medicine, and quality measures Primary Funding Source: BlueCross and BlueShield of Tennessee • Strategic Lessons for the Advantageous Use of Provider Incentives in Health Insurance Markets Learned from the National Evaluation of Seven Rewarding Results Local Demonstration Projects Bert White, M.B.A., D.Min., Gary Young, Ph.D., J.D., Matthew Guldin, M.P.H. Presented by: Bert White, M.B.A., D.Min., Project Director, Health Services Department, Boston University School of Public Health, Talbot Building, 715 Albany Street, Boston, MA 02118-2526; Tel: 617.232.9500 Ext. 4380; Fax: 617.232.6140; Email: bertw@bu.edu Research Objective: Programs involving monetary or nonmonetary rewards to clinicians who achieve certain quality goals are becoming increasingly popular as a mechanism for improving health care and controlling costs. This research evaluates seven demonstration sites that are measuring the results of offering financial incentives to providers. The objective of this research is to make a strategic contribution to those who need to calculate the impact and value of incentives on providers. Study Design: Multiple data collection approaches are being used including telephone interviews, written surveys, and site visits in accordance with a conceptual frame work that includes seven domains: awareness, scope of control, financial salience, scientific and clinical credibility, fairness, readiness for change and unintended consequences. Practice executives with responsibility for physician contracts in the health insurance market have been identified in each of the seven local demonstration projects. These practice executives provide demographic information and access to physicians exposed to a demonstration’s incentive plan. There will be standard data collection across all demonstrations measuring how incentives influence provider behavior. During the fouryear project, practice executives will be interviewed three times and physicians will be surveyed twice. Population Studied: A scientific sample from contracting entities in each demonstration has been used to identify and study practice executives and physicians. Principal Findings: Findings to date include major differences among the demonstration sites’ structure and implementation of incentives including: magnitude; distribution; fee-for-service versus capitation payment methodologies; and the communication features of incentives. Preliminary data collection indicates that physicians are very sensitive to any perceived association between the practice of medicine and the pursuit of incentives. In addition, the gateway role of the practice executive may be central to understanding the total value of any incentives for quality targets offered to providers. Conclusions: This research will identify and report the overarching lessons learned about incentives from across the seven local demonstration projects. Evidence based insights will document the strategic domains and requirements for incentives as a mechanism for improving health care and controlling costs. Implications for Policy, Delivery or Practice: The study will provide comparative business-case information to CMS and health insurance markets required for a more robust understanding of incentives and their application. These findings should be strategic for additional research and demonstration projects on incentives in the health insurance markets. Primary Funding Source: AHRQ, The Robert Wood Johnson Foundation Invited Papers Trends & Innovations in Insurance Design: Implications for Cost, Access & Quality Chair: Jon Christianson, Ph.D. Monday, June 7 • 4:00 p.m.-5:30 p.m. • Panelists: Richard Kronick, University of California, San Diego; Stephen Parente, University of Minnesota; Greg Scandlen, Galen Institute; Humphrey Taylor, Harris Interactive (no abstracts provided)