Assessing the Impact of Health Savings Accounts on Insurance and

Assessing the Impact of Health Savings Accounts on Insurance and Coverage Costs Stephen T Parente, Ph.D. Roger Feldman, Ph.D. Jean Abraham, Ph.D. Jon B Christianson, Ph.D. University of Minnesota Abstract This analysis discovers new insights offered by the introduction of Consumer Directed Health Plans (CDHPs) into how price affects health insurance choices. Specifically, we find a positive impact of the health care spending account from which medical expenses are debited and negative responses to the new CDHP deductible developed as the difference between a traditional high deductible health plan and the consumer’s spending account. With respect to their respective elasticities, the CDHP deductible, known as the ‘donut hole,’ has a far more elastic price response than the health spending account. In terms of new policy options to reduce the uninsured through tax credits targeted at Health Savings Accounts (HSAs), we find, on average, elastic cross-price responses to increase the take-up of insurance among those without coverage. However, take-up is considerably greater among the higher-income population due to wealthier individuals having a distinct preference for CDHPs among all other health plan choices. JEL Codes: I1, D12, C93 For Presentation at the 2006 ASSA Meeting in Boston, MA December, 2005 Original NBER 2005 Summer Institute Paper Attached as Appendix 1 Do not quote or cite without permission of authors. Research originally presented at the Health Care Program of the 2005 National Bureau of Economic Research Summer Institute, Cambridge, MA, USA on July 29, 2005. Corresponding Author: Stephen T. Parente, Ph.D., Assistant Professor, Department of Finance, Carlson School of Management, University of Minnesota, 321 19th Ave., Rm. 3-149, Minneapolis, MN 55455, 612-624-1391, sparente@csom.umn.edu, www.ehealthplan.org This paper was funded jointly by the Robert Wood Johnson Foundation's initiative on Changes in Health Care Financing and Organization and DHHS Contract HHSP233200400573P: Analytic Support in Assessing the Impact of Health Savings Accounts on Insurance Coverage and Costs. 1.0 Introduction Consumer directed health plans (CDHPs) are attracting attention from consumers, employers, and policy-makers. CDHPs are high-deductible health insurance plans coupled with a tax-advantaged account that can be used to pay for eligible medical expenses. If an enrollee spends all of the dollars in the health spending account during a given year, she then spends her own money until the health insurance plan deductible requirement is met. Starting from a handful of venture capital funded initiatives in 2001, this market has evolved considerably with CDHP enrollment currently estimated at approximately three to five million covered lives (Gabel et, 2005). The 2003 Medicare Modernization Act (MMA) gave a huge boost to CDHPs by approving tax-advantaged accounts for certain high-deductible health insurance plans. Section 1201 (and subsequent guidance by the Treasury Department) approved a new form of health plan known as a ‘Health Savings Account’ or HSA. Beginning on January 1, 2004, most non-elderly individuals can purchase a health plan with an annual deductible of at least $1,000 for an individual and $2,000 for a family, coupled with a tax-advantaged account to which both the employer and the enrollee may contribute. Total annual contributions can be as large as the plan’s deductible amount (up to $5,000 for an individual and $10,000 for a family). Contributions to the HSA, as well as interest and investment earnings in the HSA, are not taxable, and unlike previous designs, the HSA is fully portable so an individual may use it without being dependent on the provisions of a particular employer. In response to MMA, mainstream insurers such as Blue Cross and Blue Shield plans and UnitedHealth Group have hurried to offer HSAs. 1 The purpose of this paper is to examine the effect of ‘price’ on health plan choice. More specifically, we will determine whether the health care spending account in the new CDHP designs provides an incentive for consumers to choose these plans over conventional health insurance plans; and we will estimate the extent to which the ‘donut hole’ (i.e. the gap between the account and point where insurance coverage begins) is a disincentive for health plan choice. These two ‘prices’ are largely absent from the design of traditional health insurance plans, which feature relatively low deductibles and do not have health spending accounts. HSAs are likely to increase the range of health insurance options available to individuals. To date, however, very little is known about whether HSAs will be popular with individuals seeking private insurance coverage in general, and more specifically, with lowincome individuals who might have very limited access to alternative sources of health insurance. We will illustrate the usefulness of our results by applying them to a simulated national population in the United States to assess the potential of Health Savings Accounts (HSAs) for increasing the number of Americans, and especially those with low incomes, with health insurance. The research questions of this examination are: • • • How does the introduction of Consumer Directed Health Plans (CDHPs) into mainstream health insurance affect plan choice? Specifically, what is the impact of the account, donut hole and deductible on the ownprice elasticity of CDHPs? What is the impact of the account, donut hole and deductible on the cross-price elasticity of insurance take-up among the uninsured? This paper builds upon the empirical findings from two earlier projects: a study of the impact of CDHPs on health plan choice supported by the Robert Wood Johnson Foundation (Parente, Feldman & Christianson, 2004); and an analysis of coverage and costs of alternative 2 policy proposals for reducing the uninsured through the use of HSA subsidies funded by the Assistant Secretary of the Department of Health and Human Services (Feldman et al, 2005; Parente et al, 2005). 2.0 Model Our model builds on the existing literature of high-deductible health plans (HDHPs) discussed by Keeler, Newhouse, and Phelps (1977), in which a high deductible introduces a ‘kink’ in the consumer’s budget constraint. In contrast, the health care account in a CDHP creates a ‘double kink’ in the budget constraint as shown in Figure #1. This differentiates it from high-deductible health plans. The budget constraint of the CDHP plan starts out flat to reflect the monetary account a consumer or their employer arranges to purchase health care with a straight debit from the account for the price paid for each unit of medical care, such as a prescription drug or a physician office visit. The first kink in the budget constraint occurs when the account is exhausted and the consumer now faces the full price of medical care. The budget constraint continues to slope downward until the second kink at the point where the highdeductible threshold is reached. Depending on the high-deductible benefit design, the consumer faces either a limited coinsurance or no-coinsurance, as depicted in Figure #1. The design depicted is consistent with the early versions of the CDHP plan design (Christianson, Parente, Taylor, 2002). An additional feature of CDHP designs is the lack of a ‘use it, or lose it’ provision.1 Unspent dollars in the health account can be saved and used in following years to pay for future health contingencies, in addition to new health account dollars deposited. Figure 1 represents only a one-year budget constraint and set of tangency points. At subsequent time periods, the Such provisions are found in “flexible spending accounts” (FSAs), also known as “Section 125 accounts” for the section of the IRS code that authorizes them. Money in a flexible spending account is forfeited if not used by the end of the year. 1 3 budget constraint could be quite different depending upon how much money was saved. For example, a healthy individual with no expenditure from her account may save enough so that the budget constraint would become flat and she would face no kinks in that year. Thus, future models need to consider a multi-period budget constraint as well as where the kinks could be located, based on contract availability, health account savings and variations in coinsurance as the market evolves. For this analysis, we restrict our attention to first-year adopters of CDHPs so that the one-period conceptual model represented in Figure 1 applies. Figure 1 Conceptual Model of CDHP Money a b c d CDHP Budget A to B: Care Account B to C: Deductible C to D: Catastrophic insurance w/no coinsurance Medical Care Spending HRA/HSA deductible We propose that the CDHP design introduces two new ‘prices’ for health insurance demand analyses to consider. First, the gap between the kinks in the budget constraint is a new form of out-of-pocket price that is neither deductible, co-payment, nor premium. Also known as the ‘donut hole,’ this amount represents a new price to which we would expect consumers to 4 have a negative demand response as the size of this potential out-of-pocket contribution increases. The second new price is the amount of the health savings account. This account can be funded by an individual, their employer, or some combination of the two. We would expect it to have a positive response in demand since it can be thought of as a form of consumer income. Typically, the unused portion of the account accumulates in the consumer’s name over time and in the HSA design represents a genuine financial asset similar to a defined-contribution personal retirement account. By framing the insurance choice of a CDHP we can consider the ‘donut hole’ and savings account as new ‘prices’ in the demand for health insurance with expected negative and positive responses, respectively. This will enable us to estimate a model to examine the price sensitivity to different benefit options (i.e., premium, account, and donut hole) that could significantly affect take-up of CDHPs. To examine the choice of a CDHP, we recognize the need for a model than can accommodate existing mainstream health insurance designs, specifically preferred provider organizations (PPOs) and Health Maintenance Organizations (HMOs). We begin by considering a model where the utility of the jth plan to the ith consumer is specified as: (1) U ij = α j + β ′Z j + θ ′Z j X i + eij The utility function in (1) depends on plan benefit attributes (Zj) including: • • • • Annual tax-adjusted employee premium ($1000s dollars) Savings/reimbursement account size ($1000s dollars) Donut hole: difference between annual deductible and account size ($1000s dollars) Coinsurance rate (i.e., .10 = 10% coinsurance) In addition, we allow interactions between plan and consumer attributes (Xi), such as age, sex, wage income, and family contract as well as plan-specific constants (αj). The random error term 5 eij represents unobserved, consumer-specific aspects of utility from alternative j. This model provides an approach to estimate the price elasticity for the consumer’s portion of their health insurance premium and coinsurance as well as the new CDHP arguments for the response from the savings account and donut hole. 3.0 Estimation Using the utility framework from (1), we assembled a set of data sources to complete an econometric analysis of plan choice. In this section, we first describe the data sources from employers and their contracted health plans. Second, we discuss our estimation approach using a conditional logistic regression to generate estimates of plan choice intercepts, health plan design attribute effects, and interaction effects. Third, own-price elasticity responses are presented from the model. Finally, the conditional logistic regression coefficients are used in a micro-simulation of plan choices, including an HSA choice, to develop a set of cross-price elasticities for the takeup of any health insurance. 3.1 Data Our primary source of data for this analysis was employee health plan choices from three large employers participating in a Robert Wood Johnson Foundation (RWJF) funded study of Consumer Directed Health Plans (CDHPs). In addition, we used the 2001 Medical Expenditure Panel Survey (MEPS) developed and supported by the Agency for Healthcare Research and Quality (AHRQ) and premium data for individual health insurance policies, including HSAs, from the eHealthinsurance.com web site.2 The data from three large employers represented approximately 80,000 covered lives of information (including dependents). Two of the three employers were national firms with substantial populations of employees; one was a large employer located in Minnesota. Each of 6 these employers offers a CDHP along with traditional managed care plans. For the CDHP plans, each employer has received a take-up rate ranging between 4% and 15% in their first year offered. In our analysis we consider two models of CDHPs, the HSA or Health Savings Account and the HRA or Health Reimbursement Account arrangement. The HRA is usually offered by self-insured employers with 1,000 or more covered lives. The health account in an HRA is not owned by the employee as an asset and is not transferable if the employee were to leave the firm. On the other hand, the HSA represents a benefit design where the health account is a financial asset owned by the individual. HSAs are being offered by employers as well, but their largest market segment is the individual and small group risk-rated health insurance market. In our plan choice analysis, we principally focus on HRAs offered by three employers. Table 1 Plan Choice – Descriptive Statistics Total Observations Employee Premium in $1,000 Health Account in $1,000 Donut Hole in $1,000 Coinsurance Female, 1=female, 0 else Age (scaled in 100 years) Income in $1,000 Family contract=1, else 0 Highest value = * We use a Low HRA in our simulations for approximating an HSA benefit design. 28,737 0.627 0.169 0.336 0.084 0.514 0.419 47.81 0.528 Low HRA* High HRA 821 0.591 0.859 1.507 0.074 0.474 0.426 55.91 0.481 2,734 0.489 1.513 0.983 0.021 0.413 0.415 83.28 0.681 HMO 9,254 0.164 --0.033 0.540 0.419 40.98 0.435 Low PPO Med PPO High PPO 1,685 0.973 -0.797 0.200 0.500 0.426 36.21 0.595 8,276 0.499 -0.326 0.108 0.532 0.417 45.35 0.541 5,967 1.491 -0.284 0.128 0.504 0.407 47.73 0.571 7 Table 1 provides the descriptive statistics of the pooled employer sample. Across the three employers there were six possible health plan choices. A given employee faced a minimum of four choices and maximum of five choices. The total possible choices faced by an employee include: • • • • • • Low-option HRA (low premium, larger deductible, some coinsurance) High-option HRA (high premium, smaller deductible, no coinsurance) Low-option PPO (low premium, smaller physician panel) Medium-option PPO (median premium, coinsurance) High-option PPO (high premium, less copayments than other PPOs, larger provider panel) HMO (e.g., group model, lower premium than PPOs) Amongst the six choices, the high-option PPO had the highest premium and the HMO had the lowest premium. The HRA premium was most similar to the medium PPO. The HRA had the highest deductible and the low-option PPO had the greatest coinsurance. The highoption HRA had the least coinsurance. HMOs and medium PPOs had the greatest share of female enrollees and there was no statistically significant difference in the average age of employees among plans. Those who chose the high-option HRA had the highest income and those who chose the low-option PPO had the lowest income. The HRA had the highest proportion of family contracts and the HMO had the least proportion of family contracts. 3.2 Plan Choice Estimation Using our pooled health plan choice data from the three employers, we specified a conditional logistic regression model similar to our earlier work (Parente, Feldman and Christianson, 2004) to estimate the utility function, based on the observed health plan choices. This method is motivated by a random utility function because there are errors in maximization due to imperfect perception and optimization, as well as errors due to unobserved relevant variables. We assume the random terms in equation (1), which represent unobserved, consumer- 8 specific aspects of utility from alternative j, are independently and identically distributed with an extreme value distribution. Given this assumption, the probability that consumer i chooses alternative j from among the set of K possible plan choices can be represented as: K (2) Pij = exp (α j + β ′Z j + θ ′Z j X i ) / ∑ exp (α k + β ′Z k + θ ′Z k X i ) k =1 A very important constraint in our modeling was that any plan attribute used in the model from the employer data also had to be available in the MEPS data to permit a prediction of uninsured take-up as part of our policy analysis. As a result, the key variables used in the plan choice model were: • • • • • • • • After-tax premium paid by the employee3 The amount of money in the employee’s health reimbursement account (HRA), if any. The difference between the employee’s plan deductible and the HRA. Coinsurance rate Employee’s age Employee’s gender (1=female, 0=male) Employee has a 2-person or family contract=1, else =0 Employee’s annual wage income. Also included in the model were alternative-specific constants (intercepts) for each of the possible health plan choices. These intercepts are used to capture plan-specific features not represented by other identifiers of plan design. They are also included as interaction terms with age, gender, family status and income. The plan choice estimation results are presented as Table 2. The reference category for the regression was the high-option PPO. The signs of the coefficients for premium and coinsurance are negative as expected. The sign for the donut hole, representing the deductible The price of each health plan (one of the Zj variables in the model) was measured by its “tax adjusted” out-ofpocket premium, because the three employers in our study let employees pay their out-of-pocket premiums with taxfree dollars (Dowd, et al., 2001). 3 9 less the amount in the health account, is negative as well. The sign for the health account variable is positive, showing that an increase in the starting account balance would lead to higher likelihood of a consumer selecting a CDHP. All four of these price responses are highly significant. With respect to the attributes of the consumer most likely to be associated with a particular health plan choice, we find that older workers are less likely to choose either a low or high-option HRA. This result is different than our descriptive statistics and suggests that once price and other demographic variables are accounted for, HRAs may be prone to more favorable selection compared with a high-option PPO. Similar to earlier findings by Parente et al (2004), higher-income employees are more likely to choose an HRA. In addition, no other health plan choice appears to be as positively linked to income as an HRA. Female employees are significantly less likely to prefer an HRA than a PPO. Also, employees selecting family coverage policies for benefits for spouses and dependents are less likely to choose a high-option PPO compared with an HRA. This result is also somewhat different than our bivariate descriptive results in Table 1 where the HRAs were associated with the greatest share of family contracts. 10 Table 2 Conditional Logistic Regression Adjusted R-square: ~.363 Variable Tax adjusted Employee Premium in $1,000 Employee's Health Account in $1,000 ∆ Between Deductible and Health Account in $1,000 Coinsurance (e.g., 15% = .15) PPO Medium Plan Intercept=1, else=0 PPO Low Plan Intercept=1, else=0 High HRA Plan Intercept=1, else=0 Low HRA* Plan Intercept=1, else=0 HMO Plan Intercept=1, else=0 Premium & Family Contract (0/1) Interaction PPO Medium & Age (in 100 years) Interacttion PPO Low & Age (in 100 years) Interaction High HRA & Age (in 100 years) Interaction Low HRA & Age (in 100 years) Interaction HMO & Age (in 100 years) Interaction PPO Medium & Income (in $1,000) Interaction PPO Low & Income (in $1,000) Interaction High HRA & Income (in $1,000) Interaction Low HRA & Income (in $1,000) Interaction HMO & Income (in $1,000) Interaction PPO Medium & Female (0/1) Interaction PPOLow & & Female (0/1) Interaction High HRA & Female (0/1) Interaction Low HRA & Female (0/1) Interaction HMO & Female (0/1) Interaction PPO Medium & Family Contract (0/1) Interaction PPO Low & Family Contract (0/1) Interaction High HRA & Family Contract (0/1) Interaction Low HRA & Family Contract (0/1) Interaction HMO & Family Contract (0/1) Interaction Standard Coefficient Error T-Statistic P-value -1.7299 0.6179 -0.8502 -7.5675 0.1153 -1.0210 -1.9816 -1.7031 1.9338 1.1067 -1.5547 0.5609 -2.0759 -2.5626 -3.9740 -0.0005 0.0038 0.0100 0.0102 -0.0017 0.0852 -0.1566 -0.1665 -0.0278 -0.1892 0.0218 0.3809 -0.2979 0.0067 -0.3651 0.0665 0.0974 0.0272 0.5300 0.0860 0.1311 0.1646 0.1252 0.1048 0.0721 0.1673 0.2917 0.3506 0.2410 0.1946 0.0006 0.0014 0.0009 0.0006 0.0007 0.0369 0.0617 0.0776 0.0514 0.0435 0.0690 0.0773 0.1140 0.1144 0.0909 -25.9996 6.3407 -31.2867 -14.2794 1.3404 -7.7887 -12.0383 -13.6042 18.4506 15.3418 -9.2926 1.9228 -5.9217 -10.6318 -20.4199 -0.8201 2.6689 11.6295 17.5883 -2.3806 2.3100 -2.5373 -2.1456 -0.5399 -4.3506 0.3154 4.9275 -2.6132 0.0586 -4.0175 <.0001 <.0001 <.0001 <.0001 0.1801 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 0.0545 <.0001 <.0001 <.0001 0.4122 0.0076 <.0001 <.0001 0.0173 0.0209 0.0112 0.0319 0.5893 <.0001 0.7525 <.0001 0.0090 0.9533 <.0001 N=28,737 * We used Low HRA to create HSA 'predicting' coefficients for an individual version where the employee pays all (S) and the employer offered version where the premium and the account is heavily subsidized (E). 11 3.3 Own-Price Elasticity Response In Table 3 we present the price elasticities for the four prices in our health plan choice model, including employee premium, health account, donut hole and coinsurance. These results are calculated based on an expected probability of a given health plan choice of 15%.4 The largest of the elasticities is for the tax-adjusted employee premium with a value of -0.9213.5 The least elastic response is for the employee health account. An interesting finding is the greater elasticity of coinsurance (-.5405) compared with the donut hole (-.2430), or the difference between the deductible and the health account. This result suggests that consumers have greater sensitivity to variations in coinsurance than to changes in the donut hole structure. This finding may challenge some detractors of CDHP plans who suggest consumers will not embrace a plan design with obvious increased cost sharing in the form of a large potential deductible, compared with coinsurance, a much more conventional method of cost-sharing. Table 3 PriceVariable Tax adjusted Employee Premium in $1,000 Employee's Health Account in $1,000 ∆ Between Deductible and Health Account in $1,000 Coinsurance (e.g., 15% = .15) Elasticity -0.9213 0.0885 -0.2430 -0.5405 A health plan choice of 15% was chosen because two of the three employers had CDHP take-up rates that were close to that share. Also, 15% provides an approximately equal probability of plan choice among the six options analyzed in our conditional logistic regression. The results reflected here do not include confidence intervals for the premium elasticity measures. However, given the large T-statistics associated with the price-variable coefficients, we doubt that any of the elasticity estimates would be insignificant. 5 4 12 These results can be very helpful to future insurance benefit designers. For example, if an employer wants to increase enrollment in a CDHP, it may get the best result by favoring an increase in deductibles over an increase in coinsurance, as well as increasing the size of the account, in order to hit the target enrollment with a projected premium constraint. 3.4 Cross-Price Elasticity Response on Take-up by the Uninsured Our plan choice estimates can be used for policy analysis. Specifically, we can use them to estimate what proposed tax credits to purchase HSAs could yield in terms of new insurance take-up by the uninsured. To examine this policy impact, we estimate cross-price elasticity responses with respect to subsidies applied to the HSA premium, health account, and the donut hole. To estimate the cross-price take-up elasticities, we used the plan choice regression results to predict plan choice probabilities for each MEPS sample respondent.6 We developed two sets of plan choice predictions for the simulation: one set of data for workers with insurance offers and a second set for individuals who do not have employer offers of coverage.7 This second set includes both uninsured individuals, as well as those who take up non-group policies One group of individuals that we exclude from the simulation are non-offered individuals who reported having employer group coverage through another household member. The analytic steps taken to develop this database are described in Parente et al (2005) and Feldman et al (2005). The Medical Expenditure Panel Survey is an annual survey of the non-institutionalized, civilian population in the U.S. For this analysis, we use the 2001 MEPS Household Component (HC), which is a public-use file containing detailed demographic, health status, employment, insurance, medical care utilization and expenditure information on individuals. We restricted our attention to individuals who are 19-64 years of age, not enrolled in public insurance programs, and not full-time students. Our full sample has 16,282 individuals. When weighted to produce population estimates, this corresponds to 147,955,033 non-elderly adults in the United States. We converted HMO co-pays to actuarially-equivalent coinsurance rates for predicting the HMO enrollment probability. 7 6 13 In order to predict plan choice probabilities, we needed to develop some specific assumptions about benefit plan design and premiums for individual plans. To get premium estimates, we used MEPS linked insurance data to develop a hedonic price model to predict premiums for individual plans. Once we estimated the premiums we inflated them from 2001 prices to 2005 prices based on medical insurance price inflation during the period. In addition, we also tax-adjust the out-of-pocket premiums used in the national simulation.8 One significant issue with our simulation is that we were not able to predict whether an individual would take up insurance in the employer-offered market or purchase insurance in the individual market. We faced this limitation because the CDHP employer data only include information on offered workers who held coverage. To address this issue, we calibrated our model to accurately reflect both the actual percentage of people who turn down employer offers and the actual percentage of people in the individual market who are uninsured. Our cross-price elasticity estimates were based on a series of steps designed to predict a nationally representative sample of health plan choices. We first estimated the probability of being uninsured assuming the presence of HRAs in the group market and HSAs in the individual market. This first involved generating predicted estimates of plan choice take-up using the MEPS data and our conditional logistic regression coefficients. Since there were no uninsured in our employer databases, we created intercept variables for the uninsured and turn-down populations. For our simulations, we also assumed the high-option HRA represents the overall HRA market and the low-option HRA represents the HSA market. Our rationale for assigning the HSA to the low-option HRA is that the actuarial benefit of low-option HRA is more compatible as an HSA. These intercepts allowed us to calibrate the predictions to represent We are only able to adjust at the federal income tax level since the MEPS does not provide state-specific locations of the survey respondents. 8 14 national estimates of the uninsured. We then added subsidies to premiums, health accounts, and deductibles in separate simulations. Since we are generating person-specific probabilities of plan choice, we are also able to tally by different subgroups of the US population including by income quartile and family insurance contract type. From the status quo and other simulations we used the expression (3) to generate cross-price elasticities. (3) cross-price elasticity = - (pr uninsured|status quo - pr uninsured|∆ HSA premium) * (HSA premium @status quo/pr uninsured|status quo) Table 4 presents cross-price elasticity estimates with respect to a $1,000 increase in the employee’s out-of-pocket premium.9 At lower incomes we find far less elastic response to an increase in premium than at higher income. This response is consistent in both single and family insurance contracts. Subsidy responsiveness is higher for family contracts than single contracts. This finding suggests plan choice decision-makers for family contracts have greater interest in taking up insurance than purchasers of single insurance contracts because more lives are at risk of facing random catastrophic illness or injury. The overall response is similar to other work by Marquis and Long (1995) and Marquis et al. (2004), in the individual market as well as Chernew et al., 1997; Shiels et al., 1999; Blumberg et al., 2001; and Gruber and Washington, 2005. One interesting aspect of our findings is the differences in elasticity with respect to income. Initially, we expected that individuals with lower incomes would respond more positively to a subsidy than those with high incomes. The result can be explained by our earlier observation that the HRA plan choices were positively associated with income. Given the construction of cross-price elasticity estimates and the preference of higher-income individuals for CDHPs, it would be expected that CDHP designs would have a less-elastic price response 9 The $1,000 increase is an addition to the tax-adjusted employee out-of-pocket premium. 15 among lower-income individuals. The opposite effect would be found if we were to subsidize the premiums for HMO plans since wealthier individuals have less desire for an HMO compared with persons with less income. Table 4 Cross price-elasticity* of insurance take-up with respect to HSA premium subsidy Single Adults 0.080 0.138 0.498 0.754 0.205 Adults with Dependents 0.039 0.107 0.250 0.378 0.107 Income Quartile 0 to 25 25 to 50 50 to 75 75 to 100 All *Calculated as the MEPS survey-weighted average of each person’s: -(pr uninsured|status quo - pr uninsured|∆ HSA premium) * (HSA premium @status quo/pr uninsured|status quo) Table 5 presents the cross-price elasticities of insurance take-up with respect to a decrease in the donut hole as well as an increase in account size. We find a similar result to Table 4 with respect to income where take-up response is greater among the wealthier citizens. Consistent with our own-price elasticity response for the donut hole and the account, we find greater cross-price elasticity for a decrease in the donut hole than an increase in the account size. Interestingly, a reduction in the donut hole will yield roughly the same magnitude of take-up responses as a reduction in the premium paid from a subsidy. 16 Table 5 Cross price-elasticities of uninsured take-up with respect to “donut hole” and account Income Quartile 0 to 25 25 to 50 50 to 75 75 to 100 All with respect to a decrease in donut 0.070 0.147 0.408 0.587 0.175 with respect to a increase in account size 0.039 0.076 0.179 0.233 0.082 4.0 Implications This analysis offers new insights on price effects from the introduction of CDHPs into mainstream health insurance choices. Specifically, we find a positive impact of the account and negative responses to the donut hole. The donut hole has a far more elastic price response than the health account. In terms of new policy options to reduce the uninsured through tax credits targeted at HSAs, we find, on average, elastic cross-price responses to increase the take-up of insurance among those without coverage. However, we also find that take-up is considerably greater among the higher-income population due to wealthier individuals having a distinct preference for CDHPs among all other health plan choices. We discuss the policy implications of this research as well as the next steps for continued investigation as the CDHP market continues to evolve. 17 4.1 Policy Implications There are three policy implications from this analysis. The first is that plan design matters if the goal of targeting subsidies or other economic incentives, through the use of a tax exemption, is to reduce the uninsured by an increase in take-up in the individual market. For example, we find greater take-up from a reduction in the donut hole than an increase in the account size. A second implication is that HSA subsidies might not be the most effective economic incentive to increase take-up among the lower-income uninsured population. This population is likely to face a greater financial burden to purchase health insurance than wealthier individuals. One way to design better incentives is adjust the size of the subsidy based on income. In fact, the 2004 Republican presidential campaign proposed such a means-tested policy prescription targeting HSAs as the plan choice to increase take-up among the uninsured. The approach used in this paper can be extended and applied to approximate an optimal sliding scale for means testing. It can also be used to compare the relative efficiency of HSAs as the target of tax credits to increase take-up as opposed to HMO or PPO benefit designs. A third implication is the relative complexity of the use of economic incentives coupled with HSAs to increase take-up. Critical questions as to when the tax credit or subsidy would be available would need to be addressed. In this analysis, we assume that the consumer will realize an immediate benefit from the tax credit. However, the tax credit may take 12 to 16 months to materialize if it is made available as an after-purchase credit. If the credit is available at time of purchase, how will eligibility be verified? At the very least, policy makers should try to avoid 18 the unnecessary administrative eligibility barriers like those experienced in the Trade Adjustment Assistance Reform Act of 2002.10 One advantage of using the HSAs is that the supply of the plans is far more abundant across the nation. Almost all major commercial insurers, including staff-model HMOs such as Kaiser Permanente, offer one or more variant of the plans. A positive macro-economic externality of favorable selection into HSAs by the young and the healthy would yield an increase in personal savings to stem the rate of erosion of national savings. Furthermore, most plan designs have also included no-cost preventive care for physicals and well-child visits and some even offer free prescription drugs for chronic illnesses such as asthma or diabetes that require regular pharmaceutical use for preventive care. 4.2 Research Implications This research provides an economic model and an empirical test for CDHP benefit and plan choice. Specifically, we respond to the early call from Rosenthal (2004) for a starting template to model and evaluate ‘donut hole economics’. Although Rosenthal’s intent was to highlight how the use of donut hole is simply re-invented cost-sharing to give consumers a greater incentive to engage in health care investment as well as temporary hedge against health care premium increases rising faster than wage rate increases, it does introduce a second kink in the traditional high-deductible budget constraint. Our analysis provides an approach to model CDHPs in future plan choice analyses found commonly in the health economics literature. Additional complexities will need to modeled, such as the role of free preventive services running concurrently with other services on the budget constraint. To date, we have found the expenditure associated with these services to be fairly minor. However, the recent inclusion of 10 Lack of knowledge by employers and employees were the principal barriers prohibiting greater success of the Trade Adjustment Assistance Reform Act of 2002. Future initiatives targeted at specific populations may require aggressive public education campaigns. 19 prescription drugs could lead to greater impact on the overall budget constraint that would require further consideration. Another further development will be the changes in the starting point of the account due to roll-over of unused resources. We anticipate that the roll-over contribution may very well be another price that needs to be considered or, at the very least will show the effect of roll-over dollars in the difference between the deductible and the account plus roll-over.11 A future research topic will be the welfare implications of CDHP plans. The use of HSAs opens the possibility of a consumer changing her allocations between long and short term savings. A consumer may decide to contribute more to her HSA if she decides her health needs are more critical than retirement savings and she finds herself healthier, but less wealthy, in old age. Alternatively, a consumer may decide to invest more in retirement and under-invest in her health only to find that she has not been able to live a long healthy life to enjoy retirement due to a premature death. These questions can be addressed using a combination of health plan choice and retirement savings information available from employers. Pursuing this question will help address the open policy question of whether the young and healthier might invest more in health and, by default, increase their retirement savings. This work would contribute to the joint health and retirement decision literature summarized by Gruber & Madrian (1995). 5.0 Conclusions Using an adaptation of the high-deductible health plan economic model, we developed a consumer utility model of health plan choice that considers the new developments in the CDHP market. We found a positive impact of the account and a negative response to the donut hole. With respect to sensitivity of the price effects, the donut hole has a far more elastic response than 11 In our empirical analysis, rollover was not an issue since we used the first year of CDHP offer for all three employers. 20 the health account. In terms of new policy options to reduce the uninsured through tax credits targeted at HSAs, we find, on average, elastic cross-price responses to increase the take-up of insurance among those without coverage. However, we also find that take-up is considerably greater among the higher-income population due to wealthier individuals having a distinct preference for CDHPs among all other health plan choices. 21 References Christianson J, Parente ST, Taylor R. Defined contribution health insurance products: development and prospects. Health Affairs 2002; 21(1):49-64. Dowd B, Feldman R, Maciejewski M,. Pauly MV. The effect of tax-exempt out-of-pocket premiums on health plan choice. National Tax Journal 2001; 44(4): 741-756. Feldman, R. Parente, S., Abraham, J., Christianson, J. “Assessing the Impact of Health Savings Accounts on Insurance and Coverage Costs.” Forthcoming in Health Affairs in November, 2005. Gruber, J., and B. Madrian. 1995; Health Insurance Availability and the Retirement Decision: American Economic Review. 85: 938–948. Keeler, Emmett B & Newhouse, Joseph P & Phelps, C E, 1977. "Deductibles and the Demand for Medical Care Services: The Theory of a Consumer Facing a Variable Price Schedule under Uncertainty," Econometrica, vol. 45(3), pages 641-55. Parente, S.T., Feldman, R., Abraham, J., Christianson, J. “Final Technical Report: Assessing the Impact of Health Savings Accounts on Insurance and Coverage Costs.” Department of Health and Human Services, Washington, DC, July, 2005. Parente ST, Feldman R, and Christianson JB. “Employee Choice of Consumer Driven Health Insurance in a Multiplan, Multiproduct Setting,” Health Services Research Vol. 39, No. 4, Part II (August 2004), pp. 1091-1111. Rosenthal, M. “Doughnut-Hole Economics”, Health Affairs.2004; 23: 129-135. U.S. Congress (108th, 1st, 2003). “Medicare Prescription Drug, Improvement, and Modernization Act of 2003” conference report to accompany H.R.1. (Washington, DC, U.S. G.P.O., 2003). U.S. Department of the Treasury. General Explanations of the Administration’s Fiscal Year 2005 Revenue Proposals. February, 2004. 22 Appendix A National Bureau of Economic Research Summer Institute Working Paper The following working paper describes the full policy simulations mentioned to in the preceding text and are provided as additional background material. 23 Assessing the Impact of Health Savings Accounts on Insurance and Coverage Costs Stephen T Parente, Ph.D. Roger Feldman, Ph.D. Jean Abraham, Ph.D. Jon B Christianson, Ph.D. University of Minnesota JEL Codes: I1, D12, C93 July, 2005 Do not quote or cite without permission of authors. Paper presented at the Health Care Program of the 2005 National Bureau of Economic Research Summer Institute, Cambridge, MA, USA on July 29, 2005. Corresponding Author: Stephen T. Parente, Ph.D., Assistant Professor, Department of Finance, Carlson School of Management, University of Minnesota, 321 19th Ave., Rm. 3-149, Minneapolis, MN 55455, 612-624-1391, sparente@csom.umn.edu, www.ehealthplan.org This paper was funded from joint support from the Robert Wood Johnson Foundation's initiative on Changes in Health Care Financing and Organization and DHHS Contract HHSP233200400573P: Analytic Support in Assessing the Impact of Health Savings Accounts on Insurance Coverage and Costs. 1 Introduction Consumer directed health plans (CDHPs) are attracting attention from consumers, employers, and policy-makers. CDHPs are high-deductible health insurance plans coupled with a taxadvantaged account, which can be used to pay for eligible medical expenses. If an enrollee spends all of the dollars in the health spending account during a given year, she then spends her own money until the health insurance plan deductible requirement is met. The 2003 Medicare Modernization Act (MMA) gave a huge boost to CDHPs by approving taxadvantaged health savings accounts (HSAs) for certain high-deductible health insurance plans. Section 1201 (and subsequent guidance by the Treasury Department) approved a new form of health plan known as a “Health Savings Account” or HSA. Beginning on January 1, 2004, most non-elderly individuals can purchase a health plan with an annual deductible of at least $1,000 for an individual and $2,000 for a family, coupled with a tax-advantaged account to which both the employer and the enrollee may contribute. Total annual contributions can be as large as the plan’s deductible amount (up to $5,000 for an individual and $10,000 for a family). Contributions to the HSA, as well as interest and investment earnings in the HSA, are not taxable, and unlike previous designs, the HSA is fully portable so an individual may use it without being dependent on the provisions of a particular employer. In response to MMA, mainstream insurers such as Blue Cross and Blue Shield plans and UnitedHealth Group have hurried to offer HSAs. Research Questions The purpose of this project is to examine the potential impact of Health Savings Accounts (HSAs) on increasing the number of Americans, and especially those with low incomes, with health insurance. The research questions of this examination are: • • What is the expected take-up rate of HSAs in the individual market from the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA)? What is the impact of the Administration’s proposed HSA subsidies? o Take-up rate of HSAs o Impact on the uninsured o Cost of the subsidy What is the impact of other possible subsidy designs? • Background Congress recently enacted and the President signed into law the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA)1. The MMA contains a number of significant changes to the Medicare program, including the introduction of a prescription drug 1 P.L. 108-173 2 benefit and the use of competitive bidding to determine which plans (and at what prices) will be allowed to participate in the Medicare program. The MMA also establishes Health Savings Accounts (HSAs), which are tax-advantaged savings vehicles that can be used to pay for medical expenses incurred by individuals and their dependents.2 Prior to MMA, employers were able to provide their employees either Flexible Spending Accounts (FSAs) or Health Reimbursement Accounts (HRAs), which paid for qualified medical expenses out of pre-tax wage income. In addition, employees in small firms, the self-employed, and others purchasing insurance in the non-group market could establish a tax-advantaged Medical Savings Account (MSA), which could be used in connection with a high-deductible insurance plan. In general, HSAs are similar to MSAs in that both accounts are tax-advantaged and must be used in combination with a high-deductible insurance policy. HSAs, however, are more expansive than MSAs, in that eligibility is not limited to employees of small firms and the self-employed. Larger employers can now offer an HSA, which was not the case previously with MSAs. HSAs are likely to increase the range of health insurance options available to individuals. To date, however, very little is known about whether HSAs will be popular with individuals seeking private insurance coverage in general, and more specifically, with low-income individuals who might have very limited access to alternative sources of health insurance. This project builds upon an existing effort supported by The Robert Wood Johnson Foundation. The objective of this analysis is to produce estimates of coverage and costs of alternative policy proposals for reducing the uninsured through the use of HSA subsidies. This is being accomplished by: (1) developing an analytic database that uses information from the 2001 Medical Expenditure Panel Survey (MEPS) as well as the Contractor’s existing employer-based data files; (2) estimating a health plan choice model; and (3) using these results to perform a micro-simulation of HSA take-up. Data & Analytic Approach Three data sources were used to complete this analysis. These data sources and the steps taken to prepare the database are described in Figure 1. The data sources include: 1. The 2001 Medical Expenditure Panel Survey (MEPS) developed and supported by the Agency for Healthcare Research and Quality (AHRQ). 2. Health plan choice data from three large employers participating in a Robert Wood Johnson Foundation (RWJF)-funded study on Consumer Directed Health Plans (CDHPs). 3. Premium data for individual health insurance policies from the eHealthinsurance.com web site. 2 See Section 1201 of the MMA. 3 These data sources were used for three major analysis tasks as outlined in Figure 1: Model estimation; Choice Set Assignment/Prediction; and Policy Simulation. As illustrated in this figure, often more than one database was required to complete the task. Integral to this analysis was the use of consumer directed health plan data from three large employers working with the study investigators. Below, we provide greater detail on database attributes, use of the databases, and the analytic methods used. Figure 1 Analysis Design Data Sources MEPS Estimate plan offerings using linked data Estimate hedonic premium regression Assign plan choices to full MEPS sample Define HSA plan design & premium CDHPs Merge employer data Estimate plan choice regression eHealthinsurance Model Estimation Choice set Assignment/ Use parameter estimates Prediction to predict plan choice probabilities for MEPS Re-scale take-up rates Simulate impact of proposed policies Policy Simulation Database Descriptions Medical Expenditure Panel Survey (2001): The Medical Expenditure Panel Survey is an annual survey of the non-institutionalized, civilian population in the U.S. For this project, we use the 2001 MEPS Household Component (HC), which is a public-use file containing detailed demographic, health status, employment, insurance, medical care utilization and expenditure information on individuals. We restrict our attention to individuals who are 19-64 years of age, not enrolled in public insurance programs, and not full-time students. Our full sample has 16,282 individuals. When weighted to produce population estimates, this corresponds to 147,955,033 non-elderly adults in the United States. A breakdown of the 19-64 population for 2001 is provided in Figure 2. 4 Figure 2 Consumer Directed Health Plan data (2001-2003): The project investigators had access to de-identified data on the selection of health plans by employees, as well as their demographics. For this analysis, data from three large employers representing approximately 80,000 covered lives of information (including dependents) were available. Two of the three employers were national firms with substantial populations of employees; one was a large employer located in Minnesota. Each of these employers has offered a CDHP along with other traditional managed care plnas. For the CDHP plans, each employer has received a take-up rate ranging between 4% and 15% in their first year offered. eHealthInsurance Data (2005 HSA premiums): We used the eHealthinsurance.com web site to sample premiums for Health Savings Accounts. The web site provides a monthly estimated premium cost based on county of location, age, family size, and health history. 3For individual contracts, we assumed a single, non-smoking male age 40. For family contracts, we assumed the policyholder was a nonsmoking, 40 year old, married man with a spouse and two children under the age of 10. As a robustness check, we randomly sampled fifty other large metropolitan statistical areas around the U.S., and found that only two did not offer an HSA policy and moreover, that most policies had comparable premiums. Model Estimation The model estimation had several steps. As a first step, we pooled the data from the three employers offering CDHPs to estimate a conditional logistic plan choice model similar to our For our first round of simulations we used Santa Clara County, California, as the geographic location. For our reported results we use premiums based on a sample taken from 100 metro areas. 3 5 earlier work (Parente, Feldman and Christianson, 2004). Conceptually, we used a choice model based on utility maximization, where utility is considered to be a function of personal attributes such as age, gender, income, and family status; health plan attributes such as the tax-adjusted, out-of-pocket premium and the deductible amount; and the interaction of personal and plan attributes. Personal characteristic variables were entered into the model as interactions of plan attribute variables. The coefficient estimates produced by this model represent the utility of each plan attribute to an employee. In the second step we used the estimated choice-model coefficients to predict health plan choices for individuals in the MEPS-HC. In order to complete this step, it was necessary first to assign the number and types of health insurance choices that are available to each respondent in the MEPS-HC. For this purpose we turned to the smaller, but more-detailed MEPS Household Component-Insurance Component linked file, which contained the needed information. The steps taken to estimate this predictive model are highlighted in Figure 1. More detail of how these steps were executed is described below. Estimate plan offerings using the MEPS linked data: The MEPS “linked” Household Component-Insurance Component data file is a random sample of individuals who reported being employed and offered health insurance in Round 1 of the Household Component survey. These individuals were asked to provide contact information regarding their place of employment. Employers of these individuals were surveyed to provide detailed information about the number and types of plans that they offered to eligible workers. For each offered plan (up to four plans for private establishments and all plans for government organizations), an employer was asked to include the total premium, employee and employer shares of the total premium, and plan characteristics including hospital and physician coinsurance, hospital and physician copayments, and deductibles for individual and family coverage. Since the linked sample only represents a subset of all offered workers in the Household Component, we checked the representativeness of the linked sample using a binary logistic regression and found: • Individuals in professional services and public administration were more likely to link than those in agriculture, mining, entertainment/recreation, personal services, and active military. Midwesterners were more likely to link relative to westerners. Whites were less likely to link relative to persons of “other” race. Government workers had a higher response rate than private-sector workers. • • • The link process was a function of the following variables: age, sex, race, marital status, dependents, geographic region, metropolitan (MSA) location, government employment, 6 establishment size, industry category, wage income, and chronic illness (defined as a binary variable). The linked data have 3,127 individuals and 7,802 plan-person observations. We do not have good information on response rates because we do not know what fraction of offered workers in MEPS was considered for the linked survey. In absolute terms, it appears that approximately 36% of offered workers linked. Approximately 40% of linked workers have one plan offered to them, 19.7% have two plans offered, 11.8% have three plans, and the remaining 29.5% have four or more plans from which to choose. These percentages are not representative of the national proportions of workers who have one, two, three, and four or more plans offered to them because of the over-representation of government workers, who commonly have more offered plans than private-sector workers. To predict the number and type of plans offered, we followed two steps: 1. Used the MEPS linked insurance file to estimate a model for the number and types of health plans offered to eligible workers (age 19-64, non public enrollees, non full-time students). More specifically, we estimated an ordered probit model with the dependent variable taking the values of 1, 2, 3, or 4+ plans. The model included the following explanatory variables: age, male, white, black, marry, total number dependents, wage income, union member, works for government, establishment size, whether the establishment has more than one location, northeast, midwest, south, and MSA. The total number of observations was 2,891 and the R2 was .12. 2. Apply the model estimates to the MEPS-HC full sample to predict the number of plans for all respondents who were offered insurance by an employer. Using the model estimates, for each individual who reported being offered employer group coverage in the MEPS-HC, we predicted the probability of each outcome (1, 2, 3, 4+ plans offered). We then identified the category that had the maximum probability among the four options. We used a specific decision rule to assign the number of plans to each individual It included using both the category with the highest predicted probability as well as the individual’s direct response to a question asked in the MEP-HC about whether he/she had a choice of plans. If he/she was reported not having a choice of plans, then the individual was assigned one plan. If he/she reported having plan choice, then the assigned number of plans reflected the outcome with the highest predicted probability among the 2, 3, and 4 plan options. The types of plans were based on the distribution of plan offerings from the linked sample, conditional on the total number of plans offered. For example, individuals who 7 had one plan offered to them were most likely to be offered a Preferred Provider Organization (PPO) plan.4 So, we assigned a PPO to those with one offered plan. The other assignments were as follows: 2 plans: PPO and HMO 3 plans: 2 PPOs and 1 HMO 4+ plans: 3 PPOs and 1 HMO Estimate Hedonic Premium Regression: One challenge we faced was how to designate specific plan attributes (e.g., coinsurance rate, deductible, etc.) for the assigned plan choices. We used summary statistics from the MEPS linked insurance file to identify the median characteristics of plans by type (PPO versus HMO) as well as coverage type (single versus family). To predict the premium that would be associated with a particular bundle of attributes, we estimated “hedonic” premium models. The specific equation used was: Total premium = f(hospital coinsurance, physician coinsurance, and deductible). The estimates for HMOs used patient co-payments (dollar payments per unit of service) rather than the physician coinsurance rate. These equations were estimated separately by coverage type, plan type, and establishment size (e.g., single-coverage PPO offered by establishments with <50 workers). The model estimates were then used with the summary statistics to predict premiums for each plan, coverage type, and establishment size category (< 50; 50-200; >200) combination. Finally, to obtain the employee’s out-of-pocket premium cost, we multiplied predicted total premiums by the average proportion paid by employees for single and family coverage. We did not feel that the sample sizes were large enough reliably to perform this multiplication separately by coverage type and establishment size. Estimate Plan Choice Regression: We pooled plan choice data from the three employers offering CDHPs to specify a conditional logistic regression model similar to our earlier work (Parente, Feldman and Christianson, 2004). Conceptually, we use a choice model based on utility maximization, where utility is considered to be a function of personal attributes such as health status, health plan attributes such as the out-of-pocket premium, and the interaction of premium and health status, formally stated as: Uij = f(Zj,Yi,Xij) MEPS follows the unconventional notation of “Mixed” provider organization for PPO, “Exclusive” provider organization for HMO, and “Any” provider organization for conventional open access fee-for-service plan. Relatively few of the latter plans were represented in the data; therefore we did not assign a conventional plan to any worker with an employment-based offer. 4 8 Where i is the decision-making employee choosing among: • j = health plan choices, • Yi = employee personal attributes, • Zj = health plan attributes and • Xij = interactions between alternative-specific constants and personal attributes. A very important constraint in our modeling was that any plan attribute used in the model from the employer data also had to be available in the MEPS data to permit a simulation. As a result, the key variables used in the plan choice model were: • • • • • • • • SCALEDPREM After tax premium paid by the employee CLB The amount of money in the employee’s health reimbursement account (HRA), if any. CUB The difference between the employee’s plan deductible and the HRA. COIN Coinsurance rate AGE Employee’s age (years) FEM Employee’s gender (1=female, 0=male) FAM Employee has a 2-person or family contract=1, else =0 INC Employee’s annual wage income. Also included in the regression were alternative-specific constants (intercepts) for each of the possible health plan choices. These intercepts are used to capture plan-specific features not represented by other identifiers of plan design. They are also included as interaction terms with age, gender, family status and income. The intercept terms include: • • • • • • • PPO_L PPO Low (e.g., restrictive network, high co-pay, 15% coinsurance) PPO_M PPO Medium (e.g., better network, lower co-pay and coinsurance) PPO_H PPO High (e.g., open network, lowest co-pay, no coinsurance) HRA Health Reimbursement Account CDHP HSA_E Employer-sponsored HSA, modeled on higher premium cost HRA HSA_S Employee-paid HSA, no employer contribution, modeled on lower premium cost HRA HMO Health Maintenance Organization The final plan choice regression results are presented as Table 1. The reference category for the regression was PPO_M based on its market dominance. 9 Table 1 Conditional Logistic Regression Adjusted R-square: ~.363 Variable SCALEDPREM CLB CUB COIN PPO_M PPO_H HRA HSA_S/E* HMO SCALEDPREM * FAM PPO_M * AGE PPO_H * AGE HRA * AGE HSA_S/E * AGE HMO * AGE PPO_M * INC PPO_H * INC HRA * INC HSA_S/E * INC HMO * INC PPO_M * FEM PPO_H * FEM HRA * FEM HSA_S/E * FEM HMO * FEM PPO_M * FAM PPO_H * FAM HRA * FAM HSA_S/E * FAM HMO * FAM Standard Coefficient Error T-Statistic P-value -1.7299 0.6179 -0.8502 -7.5675 0.1153 -1.0210 -1.9816 -1.7031 1.9338 1.1067 -1.5547 0.5609 -2.0759 -2.5626 -3.9740 -0.0005 0.0038 0.0100 0.0102 -0.0017 0.0852 -0.1566 -0.1665 -0.0278 -0.1892 0.0218 0.3809 -0.2979 0.0067 -0.3651 0.0665 0.0974 0.0272 0.5300 0.0860 0.1311 0.1646 0.1252 0.1048 0.0721 0.1673 0.2917 0.3506 0.2410 0.1946 0.0006 0.0014 0.0009 0.0006 0.0007 0.0369 0.0617 0.0776 0.0514 0.0435 0.0690 0.0773 0.1140 0.1144 0.0909 -25.9996 6.3407 -31.2867 -14.2794 1.3404 -7.7887 -12.0383 -13.6042 18.4506 15.3418 -9.2926 1.9228 -5.9217 -10.6318 -20.4199 -0.8201 2.6689 11.6295 17.5883 -2.3806 2.3100 -2.5373 -2.1456 -0.5399 -4.3506 0.3154 4.9275 -2.6132 0.0586 -4.0175 <.0001 <.0001 <.0001 <.0001 0.1801 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 0.0545 <.0001 <.0001 <.0001 0.4122 0.0076 <.0001 <.0001 0.0173 0.0209 0.0112 0.0319 0.5893 <.0001 0.7525 <.0001 0.0090 0.9533 <.0001 N=28,737 * HSA_S/E we used the HSA coefficients for a individual version where the employee pays all (S) and the employer offered version where the premium and account is heavily subsidized. Choice Set Assignment and Prediction Assign Plan Choices to Full MEPS Sample: We used the three data sources to develop two sets of plan choice predictions for the simulation: one set of data for workers with insurance offers and a second set for individuals who do not have employer offers of coverage. This second set includes both uninsured individuals, as well as those who take up non-group policies One group of individuals that we exclude from the simulation are non-offered individuals who reported having employer group coverage through 10 another household member. Below we outline the analytic steps taken to develop the individuals’ choice sets for the simulations. 1. Workers With Offers We started with the original four choices predicted earlier, including three PPOs and an HMO. Since a worker was assigned between one and four plans, we needed to make some assumptions for each. • 4 choices: Low PPO, Medium PPO, High PPO, HMO • 3 choices: Low PPO, High PPO, HMO • 2 choices: Medium PPO, HMO • 1 choice: Medium PPO Here, low, medium, and high refer to the cost and quality of the plans (e.g., low implies low cost and lower quality). To these choices we added four additional options: • Self-financed (full cost) HSA – Additional choice for all workers • Turned down health coverage – Additional choice for all workers • Employer sponsored HSA – Available to all workers in establishments with >500 employees, not available to other workers • Employer sponsored HRA – Available to all workers in establishments with >500 employees, not available to other workers 2. Individuals Without an Insurance Offer Individuals who did not have health insurance offered to them at work or who were not employed, faced five health plan choices regardless of income, age or gender: • High PPO • Medium PPO • Low PPO • Self-financed HSA • Uninsured Use Parameter Estimates to Predict Plan Choice Probabilities: With a total set of possible choices for workers with insurance offers and individuals without insurance offers, we used the plan choice regression results to predict plan choice probabilities for each MEPS-HC sample respondent.5 However, before we could predict the probabilities, we needed to develop some specific assumptions about benefit plan design and premiums for individual plans. To get premium estimates, we used MEPS linked insurance data to develop a hedonic price model to predict We converted HMO co-pays to actuarially-equivalent coinsurance rates for predicting the HMO enrollment probability. 5 11 premiums for individual plans. We worked with the same hedonic plan regressions described above, except that for individuals without offers of coverage, we used the premium model for the smallest establishment size category, based on the assumption that this most closely represents an individual policy in terms of the loading charge for plan administrative costs. Once we estimated the premiums we inflated them by from 2001 prices to 2005 prices based medical insurance price inflation during the period. The plan characteristics that we used to define the three PPOs (low, median, and high) came from the 2002 HIAA/eHealthinsurance.com survey of plans purchased in the individual market. Roughly speaking, we used the 25th, 50th, and 75th percentiles of coinsurance and deductibles for assigning the plan characteristics. We also recognized that premiums in the individual market vary a lot by a person’s age. This survey included a table of average premiums by age cohort. We created an index using the information on this table. The index was set equal to 1.0 for the age group corresponding to the median age of adults in our sample (35-39). Older individuals, who had higher premiums, had index values that were greater than 1.0. Younger individuals, who had lower premiums, had index values less than 1.0. The index values ranged from .59 to 2.18 for single coverage policies and .453 to 1.65 for family coverage policies. Finally, we adjusted all premiums to 2005 dollars. Rescale Take-up Rates One significant issue with our simulation is that we were not able to predict whether or not an individual would take-up insurance in the employer-offered market or be uninsured in the individual market. We faced this limitation because the CDHP employer data only includes information on offered workers who held coverage. To address this issue, we needed to calibrate our model to accurately reflect both the actual percentage of people who turn down employer offers and the actual percentage of people in the individual market who are uninsured. To obtain more accurate estimates, we completed these calibrations by four quartiles of income and then compared our results to national, non-take-up and uninsurance rates. We also applied the national population weights to the calibrated model to represent the entire adult population, excluding full-time students, those with public insurance, and individuals with employer-based coverage through another household member. This fairly tedious process was performed for each re-estimation and/or modification of the conditional logistic regression. Policy Simulation To complete the simulations, two final steps remained. The first was to generate 2005 HSA premiums and benefit designs. The second was to specify the various simulation proposals. Define HSA Plan Design and Premium: Starting in 2004, we assumed that all individuals in the non-group (“individual”) market would have access to an HSA. We relied on the eHealthinsurance.com website 12 (www.eHealthinsurance.com) for current information on HSA premiums and plan characteristics. We collected information on two HSA policies offered in the largest two cities across every state. Next, we estimated a hedonic premium equation that allowed us to predict the premium for different HSA designs. For all of the simulations, except one (described below), we used an HSA with a $1,000 spending account and a $3,500 deductible for single coverage and $2,000/$7,000 for families. The average monthly premium for our prototype HSA for a 40-year old non-smoking single male was $102.78 per month; for a 40-year old married male (also a non-smoker) with a spouse and two children under the age of ten, the monthly premium was $226.97. The HSA premiums used in our simulations are the sum of the catastrophic policy price plus a $1,000 account. For example, a $6,500 HSA premium in our simulation for a family policy would be based on a $5,500 premium for a catastrophic insurance policy and a $1,000 HSA. Benefit differences in HSAs can be large. For example, below we list two different HSA options, a high a low deductible HSA plan in Santa Clara County, CA, that we found on eHealthinsurance.com: HSA Option #1 Single Coverage: • $1,000 HSA Account • $3,500 Deductible • $2,500 ‘Donut Hole’ (DH starts at $1,001 of expenditure - ends at $3,500) • 0% Coinsurance • Premium includes catastrophic and $1,000 HSA Account. • Thus, 100% catastrophic coverage starts at $3,501 Family Coverage: • $1,000 HSA Account • $7,000 Deductible • $6,000 ‘Donut Hole’ (DH starts at $1,001 of expenditure - ends at $7,000) • 0% Coinsurance • Premium includes catastrophic and $1,000 HSA Account. • Thus, 100% catastrophic coverage starts at $7,001 HSA Option #2 Single Coverage: • $1,000 HSA Account • $2,600 Deductible • $1,600 ‘Donut Hole’ (DH starts at $1,001 of expenditure - ends at $2,600) • 0% Coinsurance • Premium includes catastrophic and $1,000 HSA Account. 13 Family Coverage: • $1,000 HSA Account • $2,600 Deductible • $1,600 ‘Donut Hole’ (DH starts at $1,001 of expenditure - ends at $2,600) • 0% Coinsurance • Premium includes catastrophic and $1,000 HSA Account. HSA premiums were age-adjusted using the same method described above to rescale individual PPO plan coverage. Note, the premiums used in the predictions included an annual payment of $1,000 into an HSA for both the single and family policies. We chose $1,000 because it was the lowest amount for a family coverage personal care account in our analysis of employer HRAs and a low to moderate amount for a single coverage personal care account. Finally, it is important to note that for the Offered-turned down population, we have not explicitly taken account of whether these individuals have employer group coverage through another source (e.g., a working spouse). From the MEPS data, we do know that approximately 25% of those who turn down an offer of employer coverage are uninsured. Also, in our take-up estimates, we have excluded all non-offered individuals who reported having employer group coverage from their partner through the offered-group market. This group represents approximately 29 million insured individuals. National Simulation Overview The total possible health plan choices available to individuals for simulations are described in Figures 3 and 4. Several policy parameters can be examined. For example: • • • We can add different tax subsidies for purchase of individual HSA plans We can vary the characteristics of the HSA (e.g. make the ‘donut hole’ larger or smaller) We could remove the tax subsidy for employee or employer-paid premiums in the employer-offered market We proceeded with a ‘baseline’ simulation, which simply reflects the current state of HSAs defined under the 2003 MMA legislation. Following that, we simulated the impact of several policy proposals to subsidize HSA premiums and obtained national estimates of the change in estimated plan take-up rates in both the employer group and individual markets, the reduction in the number of uninsured in the individual market, and the associated cost to the federal government in 2005 dollars for each. 14 Figure 3 Possible Health Plan Choices from Simulation OFFERED 1. HMO 2. PPO_L: Low option (low premium, higher coinsurance) 3. PPO_M: Medium option 4. PPO_H: High option 5. HRA (Health Reimbursement Account) 6. HSA_E: Employer-sponsored HSA (employer shares premium cost) 7. HSA_S: Individual HSA (employee pays premium cost) 8. TURND: Turn down offered insurance NOT-OFFERED – Self-paid premiums 1. PPO_L: Low option (low premium, higher coinsurance) 2. PPO_M: Medium option 3. PPO_H: High option 4. HSA_S: Individual HSA (employee pays premium cost) 5. UNINSURED: No coverage Figure 4 Possible Health Plan Choices from Simulation - Detail OFFERED CHOICE SETS Large Employers > 500 Five Choices: HRA, HSA_E, HSA_S, PPO_M, TURND Six Choices: HRA, HSA_E, HSA_S, PPO_M, HMO, TURND Seven Choices: HRA, HSA_E, HSA_S, PPO_L, PPO_H, HMO, TURND Eight Choices: HRA, HSA_E, HSA_S, PPO_L, PPO_M, PPO_H, HMO, TURND Small Employers <500 Three Choices: HSA_S, PPO_M, TURND Four Choices: HSA_S, PPO_M, HMO, TURND Five Choices: HSA_S, PPO_L, PPO_H, HMO, TURND Six Choices: HSA_S, PPO_L, PPO_M, PPO_H, HMO, TURND NOT-OFFERED – Self-paid premiums ALWAYS THE SAME Five Choices: HSA_S, PPO_L, PPO_M, PPO_H, UNINSURED 15 Results The following are the results of the simulation model. We focus most of our discussion on changes in the individual market, since employers’ adoption of HSAs is still small and because the majority of the uninsured population does not have an offer of employer coverage. We begin in Figure 5 with a national simulation of the baseline effect from MMA for calendar year 2005. Four additional simulations are described. Baseline: In the Baseline simulation we see a non-trivial take-up of HSAs by 2005 without any additional change in health policy. Specifically, we estimate that take-up for HSA in the individual market should be 3.2 million people. We attribute this impact to the relatively lower premium of HSAs compared with a PPO and the high price elasticity associated with coinsurance. Figure 5 Baseline Impact of MMA 2003 Plan Choice HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Uninsured Baseline Baseline (unsubsidized) (unsubsidized) Population % Project Pop. 9% 13% 1% 4% 74% 3,155,982 4,651,023 310,041 1,426,040 27,273,018 INDIVIDUAL MARKET EMPLOYER INSURANCE OFFERED MARKET HMO HRA HSA-Shared Prem HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Turned Down 31% 2% 1% 0% 7% 2% 41% 16% 26,295,237 1,811,281 530,882 332,249 5,930,246 1,571,384 34,949,793 13,298,512 NOTE: Population is 19-64, non public insurance Administration’s proposal: In 2004, to encourage the purchase of HSAs, the Administration proposed to change the tax law providing significant tax credits to lower income workers. 16 Using the U.S. Department of the Treasury Blue Book published in February, 2004, we changed the simulation model to effectively ‘spline’ or segment the premiums depending on income. We used $1,000 and $556 tax credits with incomes at $15,000 and $20,000 respectively. No tax credit applied once income was at $30,000. These parameters were used to develop ratios to permit a sliding scale of tax credits with two kinks at $15,001 and $20,001. We also modeled the tax credit applying to dependents (starting at $500) at higher income breaks associated with families.6 The results of the simulation are shown in Figure 6. The total, annual subsidy cost is approximately $8.1 billion with a 10.7% reduction in the uninsured (to 24.3 million). Interestingly, the subsidy also yields HSA take-up in the offered market, representing $1.2 billion of the subsidy cost. Figure 6 Sim#1: Administration’s* Proposal Plan Choice INDIVIDUAL HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Uninsured OFFERED HMO HRA HSA-Shared Prem HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Turned Down Unsubsidized Population % 9% 13% 1% 4% 74% Simulation Population % 19% 11% 1% 3% 66% Unsubsidized Project Pop. 3,155,982 4,651,023 310,041 1,426,040 27,273,018 Simulation Project Pop. 6,971,694 4,017,191 263,278 1,215,872 24,348,069 % Change 120.9% -13.6% -15.1% -14.7% -10.7% $ $ $ $ $ Subsidy Cost 6,900,791,439 - 31% 2% 1% 0% 7% 2% 41% 16% 31% 2% 1% 1% 7% 2% 41% 16% 26,295,237 1,811,281 530,882 332,249 5,930,246 1,571,384 34,949,793 13,298,512 26,232,550 1,803,079 528,590 861,387 5,921,970 1,569,135 34,627,195 13,175,679 -0.2% -0.5% -0.4% 159.3% -0.1% -0.1% -0.9% -0.9% $ $ $ $ $ $ $ $ 1,174,289,915 - NOTE: Population is 19-64, non public insurance. *Proposal as interpreted from February, 2004 U.S. Treasury Blue Book. Low income buy-in subsidy: Given that one of the policy objectives of the subsidy proposal is to reduce the proportion of uninsured in the U.S., we simulated an even more generous policy that effectively subsidizes the We have based subsidy eligibility on individuals’ wage income rather than household income, since we do not observe household income in the CDHP employer database. As a result, this will lead to a lower HSA premium for a larger number of individuals than would otherwise be the case and potentially over-estimate the magnitude of the HAS take-up response. 6 17 total or partial cost of the premium (including the $1,000 HSA contribution) for lower income individuals. Specifically, we defined the HSA premium as: • • • $0 for individuals with annual wage income of $15,000 or less 50% of the premium for individuals having $25,000 to $45,000 in wage income. 25% of the premium for individuals having $40,000 to $60,000 in wage income. The results of this simulation are presented in Figure 7. This proposal results in a greater share of the previously-uninsured population having coverage. However, the cost is significantly higher at $12.2 billion annually, of which $10.8 billion is for the individual market population. Once again, the model predicts that many in the offered population find the incentive attractive and are willing to take up an HSA as well. Figure 7 Sim #2: Low-income Buy-in Subsidy Plan Choice INDIVIDUAL HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Uninsured OFFERED HMO HRA HSA-Shared Prem HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Turned Down Unsubsidized Population % 9% 13% 1% 4% 74% Simulation Population % 24% 10% 1% 3% 62% Unsubsidized Project Pop. 3,155,982 4,651,023 310,041 1,426,040 27,273,018 Simulation Project Pop. 8,814,552 3,840,600 240,992 1,142,829 22,777,131 % Change 179.3% -17.4% -22.3% -19.9% -16.5% Subsidy Cost $ 10,832,553,072 $ $ $ $ - 31% 2% 1% 0% 7% 2% 41% 16% 31% 2% 1% 1% 7% 2% 41% 16% 26,295,237 1,811,281 530,882 332,249 5,930,246 1,571,384 34,949,793 13,298,512 26,210,822 1,798,914 528,097 1,036,512 5,920,098 1,568,187 34,506,258 13,150,696 -0.3% -0.7% -0.5% 212.0% -0.2% -0.2% -1.3% -1.1% $ $ $ $ $ $ $ $ 1,387,115,890 - Income < 15K, free; 25K to 45K, 50% off; 40K to 60K, 25% off NOTE: Population is 19-64, non public insurance Full subsidy for HSA premium for entire adult population: In the third simulation, we increased the level of subsidy and simply have the price of an HSA be zero. In effect, this proposal is a complete subsidy for all HSA designs. As seen in Figure 8, this proposal achieves a 47% reduction in the uninsured. However, the cost to do so is $69.2 billion annually. 18 Given that we are simulating a complete subsidy for HSA insurance and also offering $1,000 ‘for free’ as part of the premium to start the individual’s HSA account, it is surprising that take-up is not higher. However, when the same analysis was proposed using HSA Option #2 which has a smaller ‘donut hole’ but larger premium, the take-up rate is much greater, with only 3.8 million uninsured remaining. However, the cost of the full subsidy with HSA Option #2 is approximately $211 billion dollars. Full subsidy for HSA premium for the non-working, non-public insurance population: As a final simulation targeted at workers without jobs, we created a simulation where anyone who was not employed received a full subsidy for the HSA, regardless of income. The result, shown in Figure 9, is lower take-up than the Administration’s proposal. Also, in terms of the per person reduction in uninsured per subsidy dollar, this policy approach it is less efficient than Sim #1 or Sim #2. Figure 8 Sim #3: Full Subsidy for HSAs Plan Choice INDIVIDUAL HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Uninsured OFFERED HMO HRA HSA-Shared Prem HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Turned Down Unsubsidized Population % 9% 13% 1% 4% 74% Simulation Population % 53% 6% 0% 2% 39% Unsubsidized Project Pop. 3,155,982 4,651,023 310,041 1,426,040 27,273,018 Simulation Project Pop. 19,464,040 2,165,844 119,097 613,960 14,453,162 % Change 516.7% -53.4% -61.6% -56.9% -47.0% Subsidy Cost $ 52,302,405,014 $ $ $ $ - 31% 2% 1% 0% 7% 2% 41% 16% 30% 2% 1% 7% 7% 2% 37% 15% 26,295,237 1,811,281 530,882 332,249 5,930,246 1,571,384 34,949,793 13,298,512 25,480,910 -3.1% 1,697,603 -6.3% 505,483 -4.8% 5,694,864 1614.0% 5,804,413 -2.1% 1,534,310 -2.4% 31,684,081 -9.3% 12,317,920 -7.4% $ $ $ $ 16,911,914,862 $ $ $ $ - NOTE: Population is 19-64, non public insurance 19 Figure 9 Sim #4: Full Subsidy for Non-working Plan Choice INDIVIDUAL HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Uninsured OFFERED HMO HRA HSA-Shared Prem HSA-Full Price PPO_High $$ PPO_Low $$ PPO_Medium $$ Turned Down Unsubsidized Population % 9% 13% 1% 4% 74% Simulation Population % 19% 12% 1% 3% 66% Unsubsidized Project Pop. 3,155,982 4,651,023 310,041 1,426,040 27,273,018 Simulation Project Pop. 6,858,372 4,266,913 279,884 1,281,403 24,129,531 % Change 117.3% -8.3% -9.7% -10.1% -11.5% Subsidy Cost $ 11,234,374,712 $ $ $ $ - 31% 2% 1% 0% 7% 2% 41% 16% 31% 2% 1% 0% 7% 2% 41% 16% 26,295,237 1,811,281 530,882 332,249 5,930,246 1,571,384 34,949,793 13,298,512 26,295,237 1,811,281 530,882 332,249 5,930,246 1,571,384 34,949,793 13,298,512 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% $ $ $ $ $ $ $ $ - NOTE: Population is 19-64, non public insurance 20 The results of the four simulations can be compared in terms of their ‘efficiency’ to reduce the number of uninsured using subsidies targeted at HSA purchases. Figure 10 shows the comparative results of the four simulations and the diminishing returns to subsidy investment. This graph will be useful to benchmark future simulations to see if they can be more efficient (i.e. above the current diminishing-returns curve). Figure 10 Diminishing Subsidy Returns 14,000,000 12,000,000 12,819,857 Previously Uninsured Sim #3 10,000,000 8,000,000 6,000,000 4,000,000 2,000,000 0 $0 Sim #2 4,495,887 2,924,949 Sim #1 0 $10 $20 Sim #4 $30 $40 $50 $60 $70 $80 Billions Subsidy Cost Conclusions Using a combination of public and private data sources focused on the impact of consumer directed health plans generally and HSAs more specifically, we find that the national adoption of these plans might be significant. Untouched, the impact of the 2003 MMA could lead to approximately 3.2 million HSA covered lives among the U.S. population between the ages of 19-64 who are not students and not enrolled in public health insurance programs. The Administration’s February 2004 Blue Book subsidy 21 plan will double HSA take-up and reduce the uninsured by 2.9 million at a tax cost of $8.1 billion, an average cost of $2,761 per person newly insured. A full subsidy of HSA premiums yields the best-case reduction of uninsured by 47%, (about a 12.8 million person reduction) at a cost of $69.2 billion annually, an average cost of $5,399 per person. Finally, offering a free HSA to the non-working, non-public population reduces the uninsured, but less efficiently than income-targeted subsidies. The next step for analysis will be to simulate the impact of additional policy proposals. In addition, we will continue to refine and test the model and further map the diminishing return point of subsidy to find opportunities for greater impact per dollar spent. 22 References Parente ST, Feldman R, and Christianson JB. “Employee Choice of Consumer Driven Health Insurance in a Multiplan, Multiproduct Setting,” Health Services Research Vol. 39, No. 4, Part II (August 2004), pp. 1091-1111. U.S. Congress (108th, 1st, 2003). “Medicare Prescription Drug, Improvement, and Modernization Act of 2003” conference report to accompany H.R.1. (Washington, DC, U.S. G.P.O., 2003). U.S. Department of the Treasury. General Explanations of the Administration’s Fiscal Year 2005 Revenue Proposals. February, 2004. 23