New Research Finds Medicare Soft Spending Cap Lowers Costs without Harming Patient Health, But Introduces Inequities

A new working paper from Assistant Professor Maggie Shi and UCLA’s Ashvin Gandhi sheds light on how “soft” spending limits – rules that impose expenditure caps but allow exceptions – can curb healthcare costs without harming patient health. But, despite this encouraging finding, it also highlights a troubling side effect: providers’ ability to navigate the paperwork required to approve exceptions leads to disparities that leave vulnerable patients behind.

The study, "Screening Through Soft Spending Limits: Evidence from the Medicare Therapy Cap," examines a Medicare policy that capped annual per-patient spending on physical therapy but allowed patients to exceed the cap if their provider submitted documentation showing medical necessity.

Shi and Gandhi find that the cap lowered overall Medicare spending on physical therapy by about 8 percent, without pushing patients toward opioids, surgery, or other potentially risky substitutes. These results suggest that soft caps can reduce unnecessary care while preserving access to necessary treatment, and without adverse health effects.

The gains – the fact that lower-need patients were more likely to stop care at the cap – stemmed both from patients choosing to stop just short of the cap without attempting to receive an exception (what the authors call the “deterrence channel”), and from those who attempted but were denied (what the authors called the “denial channel”).

“Soft spending caps in physical therapy allow policymakers to cut waste, but the exception cases make it possible for people who need to continue with care able to do so,” said Shi. “What we show is that this works: the exceptions are going to the high-need patients. The trade-off is it introduces an odd distortion, because in order to do the exceptions, you have to do paperwork, which is a skill that not every provider is equally good at.”

Denials were most common at small provider practices, which often lacked the administrative expertise to successfully submit documentation. Lower-income and minority patients – who are more likely to seek care from smaller practices – were disproportionately denied further care. The authors call this a form of horizontal inequity: people with similar needs receiving different treatment, based not on health, but on providers’ paperwork skills.

The research shows, however, that providers are learning. Large providers improve faster simply because they have more patients and more chances to learn from mistakes. When a provider’s first denial gets reversed, they are much more likely to succeed with documentation in future cases—a sign that experience leads to better outcomes. Still, small providers are learning the paperwork as well, albeit more slowly.

“It’s important to note that American healthcare operates in this system where it's not just one insurer who's introducing a rule: Medicare, say, may introduce a rule, and then two years later, Blue Cross Blue Shield introduces a different rule with slightly different requirements,” Shi added. “While providers are indeed learning, this fragmented system means that the learning is never-ending. While our paper focuses on one policy journey with one rule, there are many different payers and different rules across the health insurance policy ecosystem.”

The study ultimately points to a core tension in policy design: soft spending limits can improve efficiency and target resources more effectively, without harming health outcomes, but they can also introduce inequities if success depends on administrative savvy. As Shi and Gandhi write, these policies “screen on agents’ administrative ability,” making it easier for large, experienced providers to serve their patients while others struggle to comply, at least at the beginning of a new policy regime.

Ultimately, soft caps do help Medicare rein in costs without harming health, though paperwork and administrative ability may leave some patients without the care they need.