Is the Deal Any Good?
Published Online: December 12, 2013
François De Brantes, MS, MBA
ACOs are entering into risk-based deals. Research suggests that many of these deals will have uncertain outcomes, which a new tool might help mitigate.
The implementation of the Affordable Care Act introduced new payment models that create opportunities for providers to increase their margins if they manage resources more effectively when caring for patients. By far, the focus of Medicare’s Innovation Center has been in promoting various flavors of “global capitation” in which providers can form an accountable care organization (ACO) to manage a specified number of Medicare beneficiaries. There are 2 different types of ACO programs offered and the more popular is the one in which the ACO is held accountable for increases in total patient costs of care, and which can share in savings if the target trend rate is beat by a specified margin. It’s what’s referred to as a “one-sided” deal because the ACO will share savings if there are any but won’t be penalized if there aren’t. There also is a “two-sided” deal in which the ACO is at financial risk if costs exceed a specified rate.
Since Medicare’s introduction of these total costs of care payment programs, private sector payers have mimicked the model and most of the national health plans have a version of this payment scheme in effect with network providers. Today there are far more than 500 ACOs in the United States and the number is growing, with the majority in “one-sided” deals as opposed to “two-sided.” Are these deals any good? In other words, how confident is the payer—whether Medicare or private insurer—or the provider that retrospective calculations on actual costs of care will reflect true gains or losses?
A body of literature has examined the underlying variability in total costs of care and the potential for increases in that variability as the sample of patients decreases. This research is the foundation of the concepts of insurance in which the experience of an individual at any point in time is binary, while for cohorts of individuals the probability of a negative event is distributed along a curve. However, the specific cost of that negative event can be trivial or consequential. Overall, as the cohort size increases the variation decreases, but the opposite is also true.
In early 2012 researchers from Rutgers University published a paper that examined the potential for misclassifying ACOs as “winners” or “losers” given the underlying variability of total costs of care in a patient population. The research was based on Medicare beneficiaries but has implications for the private sector as well. Similar to the RAND study on the potential for misclassification of physicians as being efficient or inefficient, the answer lies in the size of the cohort of patients included in the ontractual arrangement between the ACO and the payer. The upshot of these studies is that there is a risk that the provider— the ACO—will be misclassified as a winner or a loser, simply due to the shift in the distribution of the probability curve of incurring a egative event for the population managed. In other words, the ACO might get lucky, or not. And unfortunately, as the size of the cohort goes down, severity adjustment or case mix adjustment models are simply not powerful enough to compensate for the random luck effect of a distribution of probabilities. as a result, depending on the specific nature of the financial deal between the payer and the ACO, as well as the size of the cohort of patients under management, the potential for misclassification can be significant, and the payer could be rewarding the ACO for savings not actually achieved or, conversely, the ACO could fail to receive a reward for savings actually realized.
That explains why a third of the “Pioneer ACOs,” who had signed up for a two-sided deal, recently dropped from the program. They realized that the deal wasn’t any good for them. It’s likely that similar realizations will hit payers and providers in the years to come. It would be regrettable to infer from that realization that new payment programs like the Medicare Shared Savings program cannot work. Instead, we should work to better understand the uncertainty embedded in the program and mitigate it as much as possible.
Building on DeLia and colleagues’ work, we’ve collaborated with IMPAQ International to create an online calculator that should help answer this basic but fundamental question of whether or not the deal offered is any good. There are several variables that influence the answer:
The size of the population to which the deal applies— Any payment model that is centered on total costs of care requires a large patient cohort to avoid the statistical uncertainties involved in the underlying variability of the population. DeLia points out that a reasonable sample size for Medicare patients can be as high as 50,000 plan members depending on the other variables in the deal. While that may be high for commercially insured plan members, the analytic work pretty clearly shows that the lower the number of plan members included in the deal, the greater the underlying uncertainty about the results. So plan member count matters a lot.
The target trend rate and the minimum savings rate— These 2 combine to create the conditions under which the ACO will succeed or fail. The target trend rate is typically the expected rate of increase in medical costs during the performance year, and the minimum savings rate is the rate below which shared savings kick in. So, for example, the expected rate could be 4% and the rate under which shared savings would start could be 3%. In other words, if the ACO were to achieve a rate of increase of medical costs of less than 3%, it would share in the savings with the payer. Depending on the spread between the target and the minimum rates, the likelihood of a “false positive” increases.
Intuitively, it’s easy to understand how these variables can impact the likelihood that the ACO (or the payer) will win or lose, but what’s not as easily understood is the extent to which these variables combine to create uncertainty in the actual results, and that’s what the calculator helps better define. With it, a user can modify these key variables and better understand the potential uncertainty in gains and losses, even if the actual results are better than what they appear to be.
PDF is available on the last page.