Fewer patients are returning to the hospital within 30 days of discharge, and it’s not because hospitals are holding patients in observation units instead of admitting them as a means of avoiding penalties, according to new federal data.

Readmission rates dropped significantly for more than 3,300 U.S. hospitals between 2007 and 2015. A small increase in the number of Medicare observation claims was also seen at that time. But researchers say the changes in observation stays can’t account for the drops in readmission rates.

The trends, rather, imply changes in the organization of care that are consistent with hospitals’ responding to the penalties imposed under the Affordable Care Act (ACA), according to findings published Wednesday in the New England Journal of Medicine.

Hospitals began facing financial penalties for having high numbers of 30-day readmissions in 2010, when the federal Hospital Readmission Reduction Program took effect. Health policy analysts worried, however, that the pressure to avoid those fines could result in patients being held longer than necessary in observation units, areas meant for short-term care which are coded as an outpatient admission.

In a post in Health Affairs in October, AARP policy analysts said at least some hospitals are substituting observation status for inpatient readmissions, both for Medicare and privately insured patients.

The AARP analysts said the trends raise a number of questions about whether patients get the same quality of care, whether the drops in readmission rates truly mean hospitals are providing better quality, and if treatment under observation could mean denial of coverage for a patient’s necessary follow-up care.

But new study’s authors, including two policy analysts who work for HHS’ Office of the Assistant Secretary for Planning and Evaluation, the upward trend in observation claims may be driven by factors that are largely unrelated to the 30-day readmission program, “such as confusion over whether an inpatient stay would be deemed inappropriate by Medicare recovery audit contractors.”

The ASPE researchers, along with colleagues at Brigham and Women’s Hospital and Harvard Medical School, examined changes in readmission rates and stays in observation units from October 2007 through May 2015, before and after the 30-day readmission reduction program was implemented.

They looked at targeted therapies, such as heart attack, pneumonia and heart failure, and general conditions to assess whether hospitals that had greater increases in observation claims also had greater reductions in readmissions.

The percentage of patients held in observation increased from 2.6% in 2007 to 4.7% in 2015 for targeted therapies, and 2.5% to 4.2% for general therapies. At the same time, readmission rates for targeted conditions declined from 21.5% in 2007 to 17.8% in 2015, and went from 15.3% to 13.1% for non-targeted therapies.

While readmission rates may be dropping, still only a quarter of more than 3,400 hospitals avoided penalties in the latest round. As one health economist put it when the data was released in August, “Everyone sucks.” With fines ranging from 0.01% to 3% of base operating DRG payments, the CMS estimated Medicare would save $420 million in 2016.

But health policy researchers and industry groups representing U.S. hospitals continue to argue that many factors affecting whether a patient needs to be readmitted are beyond a hospital’s control. In particular, facilities in poor communities may be unfairly penalized, some of the program’s critics say.

Understanding which factors are within a provider’s control and how to adjust for these factors in value-based payments remains fuzzy, concluded a January reportfrom National Academies of Sciences, Engineering and Medicine.

The NEJM study notes that the most rapid declines in readmission rates happened before October 2012, two years after the readmissions reduction program took effect, and that reductions continued after that, though at a slower pace. “Presumably, hospitals made substantial changes during the implementation period,” the authors wrote, “but could not sustain such a high rate of reductions in the long term.”