Alternative Payment Models Save, But Drop-Out Rate Still a Concern

Alternative payment models (APM) primarily run by Medicare have generally produced modest savings and quality improvements. However, popular APMs including accountable care organization (ACO) and bundled payment models have high drop-out rates, which may indicate a problem with downside risk, according to a new review article in Health Affairs.

The article reviewed 40 APMs operated by the CMS Innovation Center, including the Medicare Shared Savings Program (MSSP), Bundled Payments for Care Improvement (BPCI), Comprehensive Care for Joint Replacement (CJR) model, and the Comprehensive Primary Care Plus (CPC+) initiative.

The review found that population- and episode-based payment models – which largely cover ACO and bundled payment programs – generated some savings on average, with ACO models making the biggest impact overall. Bundled payment savings largely depended on the model.

The APMs also realized some quality improvements compared to fee-for-service providers, including higher overall care ratings from patients. This was especially seen in ACO models, which not only attained high quality scores from the start, but also improved over time with limited evidence of upcoding.

“APMs offer providers the ability to capture the financial benefits of some of the efficiencies that must be generated if the health care system is to become fiscally sustainable,” wrote Michael E. Chernew, the Leonard D. Schaeffer Professor of Health Care Policy and director of the Healthcare Markets and Regulation Lab in the Department of Health Care Policy at Harvard Medical School.

“Yet we are only at the beginning of the APM journey, and success has been modest at best. Even the most promising models are achieving savings of only a few percentage points, and some of those savings are returned to providers through shared savings bonuses,” he continued in the article.

Co-authors of the review article included Patrick H. Conway, former president and CEO of Blue Cross and Blue Shield of North Carolina, and Austin B. Frakt, director of the Partnered Evidence-Based Policy Resource Center at the Veterans Affairs Boston Healthcare System.

The three industry experts found that drop-out rates were high for both population- and episode-based APMs.

For example, of the 624 ACOs that existed at some point during the MSSP, 187 have exited the program overall, and incoming ACOs only offset the losses until 2019 when exiting ACOs outnumbered incoming ACOs for the first time.

Higher drop-out rates were also common for voluntary episode-based APMS, such as the BPCI. But exits have been offset for most models.

Coupled with modest cost and quality improvements, the industry experts agreed that the findings show that downside financial risk may be impeding progress.

“Downside risk could discourage the participation of organizations that intend to make an effort to save but are not confident that they will succeed and are not ready to take on financial risk,” they explained. “As a result, downside risk should be imposed gradually and in a limited fashion, particularly when it is applied to smaller practices.”

CMS and its Innovation Center have largely used a gradual transition to downside risk in its APMs – most notably in the MSSP. However, the agency has recently decided to ramp up risk-based arrangements, particularly in the MSSP, which it renamed Pathways to Success in 2018.

In Pathways to Success, ACOs have significantly less time to assume downside risk although the risk arrangements come with more flexibilities, including a beneficiary incentive program and more telehealth benefits.

Program changes received severe criticism, with industry stakeholders arguing that new downside risk requirements would harm program participation. But CMS stands behind its decision to overhaul the MSSP to encourage more meaningful savings and quality improvements.

But to produce greater savings, Chernew et al. also recommended that CMS and other payers separate future spending targets from past performance, which proved to be the downfall for the Next Generation ACO model. They also stated that limited the number of conditions part of episode-based APMs and the number of APMs, in general, would also reduce complexity and refine models.

“We need improved APMs, not more APMs,” they stressed.