The percentage of patients being taken into hospitals for “observation” is increasing and so is the length of their time under such observation, a new study from Brown University found.

The ratio of observation stays to inpatient admissions grew 34 percent between 2007 and 2009, from an average of 86.9 such stays per 1,000 inpatient admissions per month in 2007 to 116.6 in 2009, study authors Zhanlian Feng, Brad Wright, and Vincent Mor said.

Two to three out of every 1,000 Medicare patients were put under observation in any given month from 2007 to 2009, rising from an average of 2.3 per 1,000 patients per month in 2007 to 2.9 per 1,000 in 2009, the study found.

But Medicare patients were increasingly put under “observation” and then discharged for outpatient treatment, “which can expose them to greater out-of-pocket expenses if they are eventually admitted to skilled nursing facilities,” the authors found in the report, released June 5.

Overall, there were 814,692 hospital observation stays involving 742,888 unique fee-for-service patients in 2007. By 2009, those figures had grown to 1.02 million observation stays for 918,180 patients, increases of 25 and 24 percent, respectively, the study found.

Even as the numbers and percentages of observation stays were rising, so, too, were the length of observation stays, the study found. In 2007, patients were kept under observation for 26.2 hours on average. By 2009, the length of observation stays had grown to an average of 28.2 hours, an increase of 7 percent, the study found.

Only a handful of Medicare beneficiaries—less than 3 percent per month—were held for more than 72 hours over the years in the study. But “their absolute number more than doubled, from an average of 1,025 each month in 2007 to 2,258 each month in 2009,” Feng, Wright, and Mor said.

The study’s authors relied exclusively on Medicare claims data, they said. “Our counts of observation stays should be taken as conservative estimates because we followed official instructions regarding the coding and reporting of hospital observation services,” the authors wrote. “To our knowledge, the accuracy of Medicare claims data … is less than perfect.”

The spike in observation may be “an unintended consequence” of several Medicare policy changes, including the Centers for Medicare & Medicaid Services’ 2004 decision to authorize the so-called Condition Code 44, the authors said.

Under that code, “hospitals may retroactively change a patient’s status from inpatient to outpatient with observation services if the utilization review committee determines, and the attending physician concurs, that an inpatient admission was not medically necessary,” the authors wrote.

Another change stemmed from the Medicare Recovery Audit Contractor program, which began as a pilot program under the 2003 Medicare Modernization Act and was formally nationalized in 2006.

“Presumably, both policy changes—especially the latter—may have motivated hospitals and physicians to reduce inpatient admissions, especially of patients whose prognosis might be thought to require only short-term treatment,” Feng, Wright, and Mor said. “Facing more stringent criteria for inpatient admissions and uncertainties over the prospects of retroactive payment denial, physicians may choose to place their patients under observation more often than they would otherwise.”