Effective April 2016, hundreds of hospitals across the country are at financial risk for the results of total joint replacement surgery, not simply for what happens during the hospital stay but for what happens for up to 90 days after the patient is discharged. This mandated program builds on the voluntary pilots and demonstration projects that Medicare has implemented since the mid-1990s. It should also build on what the private sector has piloted for more than two decades, but it does not. Instead, Medicare has designed the financial risk arrangements with what appears to be a simple and intuitive formula but is instead a rigged game. As a result, the decks are being stacked against many providers. They do not have to be.
Medicare’s financial risk formula calls for creating a target price (think of it as a virtual budget) based on a combination of the market average and the provider average, with the price moving increasingly to the market average over time. The rationale is that this will avoid excessively punishing today’s good performers while creating immediate but bearable pressure on the poorer performers. Furthermore, the average will not be adjusted for the underlying severity of the patients. After a defined period, the providers are at full risk, meaning that they gain or lose 100% of the difference between the actual total price paid for all claims incurred for each episode and the target price. While the losses and gains are limited by a hard ceiling and floor (±5% to 20% of the target price), this double-sided equal share risk contract is very similar to what private-sector payers are doing, and it is inherently flawed.
The equal share assumes that risk is distributed evenly and that there is as much potential for gain as there is for loss. That assumption is completely false, because there is a shallow floor to the gains and a high ceiling for losses, as illustrated in figures 1 and 2, representing the distribution of costs for total knee replacements for Medicare beneficiaries (figure 1) and commercial plan members (figure 2).
Figure 1 shows that a reasonable floor for the Medicare price of a total knee replacement is approximately $20,000. Of note, the inpatient costs are not compressible, since they trigger a diagnosis-related group (DRG) payment. As such, even if all other costs of the procedure were driven close to zero, which is not realistic, the absolute floor would be the DRG payment. The figure also shows that individual case costs can be as high as $100,000. If the target price is about $30,000 per episode, then the potential for gain on any episode is $10,000 and the potential for loss is $70,000. Even if the gain or loss is capped at a specified percentage of the average, the probability of losses reaching the limit is much higher than the probability of gains reaching the limit.
Figure 2 shows a similar distribution for commercially insured patients, but with a higher floor, since inpatient prices in the private sector are significantly higher than Medicare prices.
Figure 3 illustrates a similar distribution for patients with diabetes, but for this particular episode of care, the floor is about $1,000, the ceiling can be as high as $40,000, and the average is $2,500.
In all cases, it is important to note that most of the costs above the average are driven by potentially avoidable complications—negative sequelae of treatment that result from poor patient management, medical errors, or other factors that can be controllable by providers. But they can also result from lack of patient adherence to treatment regimens, social determinants of health, and various social and economic factors. All of these are potentially but not absolutely avoidable, and some of the excess costs should be adjusted based on the severity of the patient’s condition and the patient’s general health status.
In order to determine the extent to which the asymmetrical distribution of risk could affect providers under various risk arrangements, HCI3 developed a series of simulations to model hypothetical provider-specific financial gains and losses. We varied the parameters of the risk contracts along two dimensions.
- First, we established contracted budgets for each provider by using either the average episode costs across all patients in the sample (the “average market” price) or risk-adjusted, episode-specific budgets. The risk-adjusted budgets were calculated from a regression model that predicted episode costs based on an individual’s demographics, preexisting comorbidities, and episode severity. Each provider’s contracted total annual budget under both approaches was equal to the sum of the budgeted episode costs for all of the provider’s patients.
- Second, we created four different hypothetical risk contracts with varying levels of upside and downside risk exposure: 100% upside risk only, 100% upside and 60% downside risk, 50% upside and downside risk, and 100% upside and downside risk. Upside risk refers to the gains that a provider would receive from the payer if the provider’s actual costs were less than its allotted budget, whereas downside risk refers to the losses that a provider would incur if its actual costs exceeded its budget.
For each combination of budgeting mechanism and risk exposure pair, HCI3 ran four simulations of 1,000 hypothetical providers that were randomly assigned 200 patients under varying levels of patient selection risk. To do this, we separated episodes into “typical” and “high” risk, with high risk defined as episodes in the top 10% of risk-adjusted episode costs. Our base simulation was a complete random assignment of episodes to providers. This represents an approximation of the range of potential gains and losses for providers, whose mix of typical- and high-risk patients was similar to that observed in the overall sample. For the remaining simulations, we applied higher levels of insurance risk by incrementally increasing the proportion of providers’ high-risk patients in their patient panels. In the second simulation, the mix was 20/80 (20% high risk and 80% typical risk); in the third and fourth, the ratios were increased to 50/50 and 80/20, respectively. Our analysis compared the magnitude and range of providers’ financial gains and losses for each simulation under each of the eight possible risk contracts (two budgeting mechanisms times four risk-sharing scenarios).
Figures 4 and 5 represent the results for a risk contract to manage patients with diabetes. Figures 6 and 7 represent the results for a risk contract to manage patients with an elective total knee replacement.
The implications are clear. First, using the market average and not adjusting for patient severity can result in systematic losses with even a minor adverse selection of patients. In fact, providers with a slight overweighting of high-risk patients have almost no potential for gains when managing patients with diabetes and only a small potential for gains when managing patients with an elective joint replacement.
Second, even when adjusting for severity, the construct of the financial arrangement has an implication on the odds of gains and losses. Due to the asymmetrical distribution of risk, an “equal” share of gains and losses is, in fact, unequal. A mechanism to mitigate for that asymmetry is to implement a matching asymmetrical risk arrangement in which the share of gains is systematically higher than the share of losses. That type of arrangement is illustrated in HCI3’s 100% upside, 60% downside model.
There are also other ways of mitigating for the asymmetrical distribution of risk, and that includes instituting individual-case stop loss arrangements and global stop loss arrangements. But while a stop loss would have the effect of shifting the box plots up in each graph, it would not eliminate the potential for losses when the prices are not adjusted for the severity of patients.
As private- and public-sector payers continue to implement alternative payment models at an ever-increasing pace, it is essential to understand that, in most cases, whether for total cost-of-care payments or for episode-of-care payments, the potentials for gains and losses are not symmetrical and, ex ante, the decks may be stacked against the providers. The playing field can, however, be leveled by adjusting for patient severity, implementing asymmetrical gain/loss sharing, and instituting stop losses. These mechanisms are simple to put in place and should be by every responsible payer.
Unfortunately, the largest single payer in the country, Medicare, has so far failed to understand how much it is stacking the odds against providers.
Andrew Wilson, MPH, MA, contributed to this article.