Reforming Medicare
Issues in Long-term Contracting
"Even with successful cost-reducing reforms, Medicare's burden on future taxpayers will grow with the size of the retired population."
Several issues about long-term contracting arise. The first arises when insurers attempt to attract or avoid customers based on their expected health care costs. The second is the expenditure risk insurers would face over the length of a lifetime contract. Expenditure risk arises from price level changes, technological advances and market adjustments to the new form of contracting. The third issue in long-term contracting is the ability of beneficiaries to move to other insurers should they become dissatisfied with their care. The fourth is how insurers would deal with end-of-life spending.
Health Risks versus Longevity Risks. Adverse selection and screening are hurdles in every insurance market. Screening by insurers is a particular problem in the Medicare market. As noted above, CMS is currently developing more precise ways to risk-adjust prospective payments to Medicare+Choice providers. These prospective payments are for a one-year contracting period. In the context of long-term contracting, beginning with initial entry into Medicare at age 65, risk adjusting takes on another dimension. Insurers must consider longevity risks in addition to the typical medical care risks over a one-year health insurance period.
When contracts run to the end of a beneficiary's life, these two risks oppose one another. Individuals who are in ill health at age 65 represent high immediate health expenditure risks, but they are also more likely to die at younger ages. Conversely, individuals in robust health at age 65 have low expected spending in the next few years, but they may have more years of receiving benefits.
"Thirty percent of annual Medicare spending is on patients in the last two years of life."
The lifetime spending patterns associated with various ages of death are depicted in Figure VI for Medicare beneficiaries born in 1912.16 The four series of annual spending are for beneficiaries dying at the ages of 70, 75, 80 and 85. Note the three points of interest about this graph. First, average annual spending rises in the year of death and in the years immediately preceding.17 Medicare spending on beneficiaries in their last two years of life accounts for upward of 30 percent of Medicare's total reimbursements every year.18 Second, spending in the last year of life for this group was fairly constant in real 1997 dollars - with spending of $14,421, $13,137, $13,973 and $15,476 for those who died at the ages of 70, 75, 80 and 85, respectively. Finally, even though the cost of death is roughly the same for the four groups of beneficiaries, those who live longer spend more Medicare dollars in the intervening years.
"Medicare spends more on those who live longer than on those who die earlier."
Figure VII illustrates the same point in a different way. The figure shows the present value of spending between the age of 65 and the age of death, using a 4 percent real discount rate, for individuals born in 1912 and 1922. The two series show that Medicare spends more on beneficiaries who survive to higher ages than those who die earlier.
"In 1977, 5 percent of beneficiaries accounted for more than 80 percent of Medicare spending among individuals born in 1912."
The Concentration of Spending: Lorenz Curves and Gini Coefficients. Another way of looking at the trade-off between health care risks and longevity risks is illustrated in Figure VIII, which graphs three Lorenz curves for beneficiaries born in 1912. A Lorenz curve is a depiction of a distribution's inequality and a convenient way of summarizing Medicare's spending distribution. It relates the cumulative share of beneficiaries to the cumulative share of Medicare spending, after spending has been sorted from lowest- to highest-cost patient. The Lorenz curve for spending at age 65 for the beneficiaries born in 1912 reflects the fact that more than 70 percent had no reimbursements in 1977. Further, the lowest 95 percent of beneficiaries accounted for less than 20 percent of aggregate spending at age 65. This implies that 5 percent of beneficiaries account for more than 80 percent of spending. If spending were uniformly distributed (if spending by all beneficiaries was the same), the Lorenz curve would fall on the 45 degree line in Figure VIII. Lorenz curves farther from the 45 degree line indicate more inequality in health care costs.
As the period over which spending occurs is lengthened, the degree of inequality diminishes and the Lorenz curve moves closer to the 45 degree line. The Lorenz curve for cumulative spending to age 75 is to the left of the curve for spending at age 65. The curve for cumulative spending to age 85 shows a further reduction in inequality.19
The Lorenz curve can be summarized by a single statistic known as the Gini coefficient. The coefficient is equal to the area between the Lorenz curve and the 45 degree line, divided by the total area of the triangle to the southeast of the 45 degree line. The coefficient thus ranges between 0 and 1, with numbers close to 0 reflecting rather equally distributed reimbursements and coefficients close to 1 reflecting unequally distributed reimbursements.
"The longer the time horizon, the more equal is Medicare spending across all beneficiaries who enter the program in a given year."
Figure IX depicts the Gini coefficients for four birth years for successively longer computation periods. As was expected from the Lorenz curves in Figure VIII, the Gini coefficients decline as the years over which spending is aggregated increase. The coefficient for the 1912 birth year is 0.94 for spending at the age of 65, 0.61 for spending to the age of 75, and 0.49 for spending to the age of 85. The coefficient series for each birth year indicate that inequality declines at a declining rate. The separate series also show that the degree of within-birth-year inequality has declined at age 65, but as more years are added the degree of inequality is similar across birth years.
"Regardless of the birth year, longer time horizons lead to less concentration of spending among a few high-cost beneficiaries."
The previous graphs showed that as the period lengthened, the degree of inequality in the reimbursement distribution declined. Thus, insurers of long-term contracts would face a dramatically different distribution than they would face with one-year contracts. However, this evidence, combined with the evidence on spending by age of death, does not necessarily indicate that longer-term contracts pose a simpler insurance problem. Spending is less concentrated among high-cost users, but it is not uniformly distributed.20
Spending Dynamics: Movement among Quintiles. Yet another way of approaching this issue is to ask how well past spending predicts remaining lifetime spending among Medicare beneficiaries. Table I shows how well spending between ages 65 and 69 predicts remaining lifetime spending for beneficiaries born in 1912.21 Spending for the two computation periods is sorted from lowest to highest and divided into five equal-size groups of beneficiaries, or quintiles. Moves between quintiles are then tabulated. The first row in the table identifies the movements of beneficiaries who were in the bottom quintile for spending between 65 and 69 years of age. One-third of these beneficiaries were in the same quintile in the remaining lifetime spending distribution. Another 21 percent migrated to the second, 18 percent to the middle, 15 percent to the fourth, and 13 percent to the top quintile. Overall:
- One-third of those who were the lowest-cost beneficiaries from ages 65 to 69 continued to be the lowest-cost beneficiaries for the rest of their lives.
- The other two-thirds rose to a higher-cost quintile.
- Of those who were the highest-cost beneficiaries at ages 65 to 69, only 27 percent continued to be the highest-cost beneficiaries for the rest of their lives.
- The remaining 73 percent fell to a lower-cost quintile.
If spending between 65 and 69 years of age were a perfect predictor of remaining lifetime spending, all members of each group (100 percent) would fall in the same quintile in both periods. If the initial location were completely unrelated to subsequent spending, 20 percent of each original group would fall into each of the remaining lifetime spending quintiles.
The second, middle and fourth quintiles show great mobility in both directions. For example, beneficiaries originally in the second quintile were almost uniformly distributed in the remaining lifetime spending quintiles. The pattern suggests that remaining lifetime spending for each of the three middle quintiles is almost random, based on spending during the first five years.
"Only one-third of those who were the lowest-cost beneficiaries from ages 65 to 69 have the lowest cost for the rest of their lives."
Expenditure Risks. The second concern with long-term contracting is the risk insurers would face as spending on medical care rises relative to the increase in national income. Continuing with our previous example, people born in 1912 who died at the age of 75 spent $31,915 before they died, on the average. By contrast, people born in 1922 who died at the age of 75 spent $48,865 before they died. That is an increase of 53 percent, even though the age difference between the two groups is only 10 years. How can insurers cope with this risk? Rising expenditures can be addressed by indexing Medicare's annual per capita premium payments to health plans to reflect spending growth among all beneficiaries or spending growth in the private market.
Movement among Health Plans. Beneficiaries may find that they would like to move to another insurer because they relocate or because they have become dissatisfied with their current insurer. In the event of a move, the receiving insurer would voluntarily accept only patients who bring with them a premium that covers the present value of their expected future Medicare reimbursements. For example, the last column in Table I reports the present value, at age 70, of remaining lifetime spending for each quintile. These amounts approximate the remaining lifetime spending expected for a beneficiary at age 70 in each of the quintiles. If the premium does not cover this amount, a severance payment from the original insurer could make up the deficiency. As John Cochrane has shown, severance payments solve many of the problems associated with long-term contracting.22
The system might work like this. Assume that each year Medicare pays an insurer a premium on behalf of each beneficiary. The premium payments continue for a prespecified number of years so that they equal the present value of expected spending for each beneficiary when the beneficiary entered the plan. Payment adjustments can take into account unexpected changes in medical care expenditures. At any point in time, the Medicare premium payment thus may be less than, greater than, or equal to the expected annual costs for the beneficiary at that time.
Now consider that a beneficiary chooses to move from Plan X to Plan Y. If Medicare makes a payment exactly equal to the beneficiary's expected health care costs, no severance payment would be required. If Medicare's payment to Plan X exceeds the expected cost of care because the beneficiary's expected costs have declined, Plan X would suffer an economic loss from the move. To induce Plan X to agree to the transfer, Plan Y would receive a reduced payment from Plan X equal to the beneficiary's lower expected cost, but Plan X would continue to receive the original stream of annual payments.
How should the amount of compensation be determined? Ideally, by the marketplace. Plan X and Plan Y will agree to the exchange if the loser is compensated. The amount of compensation would be determined by voluntary agreement.
Notice that Medicare's payments under this scenario are administratively determined, but the marketplace would determine the compensation plans receive from each other. Insurers can set premiums based on a beneficiary's risk - and beneficiaries can move between plans.
"People born in 1912 have about the same costs of dying, regardless of the year in which they die."
Plans likely will develop a comparative advantage in certain specialties. Under these circumstances, one plan will have lower costs than another to achieve the same outcome. This would create opportunities for both the health plans and beneficiaries to gain.23
End-of-Life Spending. We saw in Figure VI that people born in the same year have about the same costs of dying, regardless of the year in which they die. This may appear surprising. One might have anticipated that the cost of dying at, say, age 70 would be lower than at age 85, given rising health care costs. In order for the cost of dying to be roughly the same over the 15-year period, the health care system must be systematically spending less on 85-year-olds at the time of death than it spends on 70-year-olds at the time of death, in any given year.
"Medicare spends less on older patients than it spends on younger patients who die in the same year."
Figure X shows that this is the case. The figure depicts the cost of care in the year of death for people who died in 1997. As the figure shows, Medicare spent less on older patients than on younger patients who died that year. For example, average spending in the last year of life on those who died at age 70 was $22,649. This is 45 percent more than the $15,576, spent on those who died at age 85.24
Why does the heath care system spend less on older patients near the end of life? Perhaps older patients are less able to benefit from expensive procedures, especially invasive procedures. The value of renal dialysis and kidney transplants, for example, is thought to decline with age. Possibly the health care system makes implicit cost-benefit decisions. Since younger patients have more potential remaining years of life, they promise a larger payoff than do older patients, for the same procedure.
Regardless of the actual explanation for the pattern in Figure X, a reformed Medicare system should not give health plans an incentive to systematically underprovide care to the sick or those near the end of life.
Two forces will help minimize the problem and improve incentives over the current system. First, allowing patients to move from plan to plan should lead to better monitoring by the patients themselves. Second, requiring plans to compensate each other for their losses when beneficiaries switch plans will create a "market for sick people."