By 2017, the Department of Health and Human Services wants 85 percent of traditional Medicare reimbursements to be based on quality or value, with 30 percent of payments reimbursed through alternative payment models like Accountable Care Organizations (ACOs). Toward that end, the Centers for Medicare and Medicaid Services (CMS) has sponsored several voluntary initiatives, including the Shared Savings and Pioneer ACO models that are thought to be critical to Medicare's ability to bend the cost curve.
Dartmouth-Hitchcock Health System (DH) is the preeminent academic health system in Northern New England. DH began developing a population health based strategy in 1983 and entered Medicare's Physician Group Practice (PGP) demonstration project in 2005.
In three of the five PGP demonstration's years, DH earned incentive payments based on estimated savings in Medicare expenditures. Since then, DH expanded ACO-like models to non-Medicare populations; by 2015, DH had a total of approximately 65 percent of its primary care service population enrolled in, and about 30 percent of its total system net patient revenues generated from, an ACO-like model.
The Triple Whammy
A leader in the development of and participation in ACO models, DH applied for the Pioneer ACO Demonstration and earned a $1 million payment in shared savings during the program's first year; however, financial performance for 2013 and 2014 resulted in $1.4 million and $3.6 million shared loss repayments, respectively.
Despite already having the sixth lowest per capita expenditures in the Pioneer program (among the 19 ACOS remaining in the program), DH continued to save CMS substantial money in 2014: for approximately 41,000 attributed patients, DH reduced annual per-capita expenditures by 3.9 percent, generating a cost savings of about $16.0 million while maintaining high quality care. While DH provided care at a lower cost than that of the reference population, care costs were higher than the target set by CMS, hence the penalty.
In September 2015, DH made the difficult decision to withdraw from the Pioneer program. DH was not alone in exiting the program: while 32 programs joined the Pioneer ACO program in 2012, by the program's third year, only 19 organizations remained.
Below, we describe the three specific—and solvable—issues that led DH to withdraw from the Pioneer program, all of which arose by virtue of DH having operated in a very low-cost environment for a long time — a characteristic that seemingly makes it very difficult for Pioneer ACOs to be successful. Medicare should address these deficiencies in future ACO project iterations.
A Flawed Risk-Adjustment Methodology
During the first three years of the Pioneer ACO program, Medicare used a 'matched cohort' methodology that compared performance of a national sample of Medicare beneficiaries (with at least one evaluation and management visit during the baseline period, adjusted for age and gender distribution — the 'reference population') to that of DH's actual service population. DH operates in a low-utilization region, and areas with higher use of health care services have been shown to use more diagnostic codes than areas with lower use of health care services.
Consequently, the matching process did not accurately adjust for illness burden — because of the coding disparities, high utilization areas were expected to have higher utilization and low utilization areas were expected to have lower utilization. This caused low-utilizing areas to experience higher than might be anticipated cost growth.
While the fourth year of the program will use a different process, Medicare should incorporate measures of visit intensity or poverty measures into illness burden estimates so that baseline performance of low-utilizing regions will not be biased and better adjustment for illness burdens and expected future service utilization across high- and low-utilizing areas can be made. High-utilizing areas should not be rewarded a second time for generating higher risk scores.
A Moving And Flawed Target
The Pioneer ACO model generated a target benchmark against which financial performance was assessed. After defining the reference population, a target benchmark was calculated by first adding to baseline actual expenditure 50 percent of the absolute dollar difference between ACO-specific baseline expenditures on the reference population and the ACO-specific actual performance year expenditures. To that was added 50 percent of the product of the ACO-specific baseline expenditure and the percentage change between the ACO-specific baseline reference expenditure and the ACO-specific performance year expenditure.
There were at least two challenges with this somewhat convoluted method of assessing financial performance. First, the final target measure was not generated until the performance year was over, making it impossible for managers to assess their performance in real time. Second, changes in health care utilization in other geographic areas, far outside of one's own, inherently impacted the target. The relative slowdown of cost growth in the US during this time period allowed organizations that had a history of providing excess care to generate a higher percentage of cost savings while concurrently generating unachievable targets for ACOs that operate in historically low-cost environments.
In future ACO models, CMS should allow organizations to benchmark their performance against their own prior performance and use those data to set prospective, achievable benchmarks at the beginning of the evaluation period, like those demonstrated to be effective at achieving quality improvement goals.
Identical Incentives Regardless Of Baseline Performance
While the target varied by ACO, the challenge of operating in a low-cost environment was magnified because CMS applied the same financial incentives to all participating organizations, regardless of their baseline performance. Although the average Medicare cost per beneficiary per year for all participating Pioneer ACOs was $11,138, DH's was $9,297, or 17 percent lower than the average and one of the lowest utilization areas in the nation.
Just as performance ceilings limit improvement gains, cost floors—spawned by medicine's high costs of labor, technology, and compliance—make cost reductions difficult when they are already low. To encourage both high- and low-cost ACOs to participate in such programs in the future, CMS should use local benchmarks to evaluate improvement, but national cost benchmarks to determine shared savings distribution levels: ACOs with higher baseline costs might retain less of their cost savings, while those with lower baseline costs might keep more or at least not be penalized.
Crediting Pioneer participants for developing and operating the infrastructure necessary to participate in the program—DH spent about $2 million per year to do so—might help retain ACOs that operate in low-utilization areas. Furthermore, CMS could use an improvement or attainment model so that consistent low utilization is rewarded.
Looking Ahead
DH did not make the decision to leave the Pioneer program lightly. A leader in the adoption and application of ACO principles, DH is committed to pursuing high value population-based care through continuous quality improvement and a relentless focus on cutting waste from health care delivery by engaging providers and payers, waste that is perhaps more difficult to identify and reduce from within a very low-cost environment.
To meet its ambitious goals, CMS must provide fair incentives that are large enough to encourage more health care systems to enter new reimbursement models and attract and retain providers in both high- and low-cost settings; otherwise, unrestrained fee-for-service cost growth will continue and the full promise of "accountable care" will remain unrealized.
CMS must rectify the flaws in the Pioneer model so that high-cost ACOs are rewarded for meaningful improvement, and low-cost ACOs—that have already benefited CMS and Medicare beneficiaries by applying ACO principals over the long term and whose ability to generate cost savings is likely to be modest—are rewarded and not punished while demonstrating improvement.
from Health Affairs Blog http://ift.tt/1S0RhHz
No comments:
Post a Comment