More than 1,100 people would have to be offered HIV tests in the emergency room to find just one new infection with the potentially deadly virus, researchers from France said.

Their conclusion -- that universal HIV screening in the ER is not a practical option -- contradicts current recommendations from the U.S. Centers for Disease Control and Prevention.

"Unexpectedly, nontargeted screening identified only a few new diagnoses, often already at late stages, and most newly diagnosed patients belonged to a high-risk group and had been tested previously," Dr. Kayigan Wilson d'Almeida and colleagues from the Emergency Department HIV-Screening Group write in the Archives of Internal Medicine.

They assigned 29 Paris emergency departments to offer HIV blood tests over six weeks to patients ages 18 to 64. Nearly 21,000 people were offered the test, and more than 12,700 took it.

Only 18 new HIV cases, or 14 per 10,000 tests, surfaced over the study. And seven of those were gay men, who are already considered high-risk.

The U.S. Preventive Services Task Force, a federally funded expert panel, makes no recommendations about HIV screening of the general public, but recommends testing high-risk groups such as prostitutes, drug addicts and homosexual men.

According to an editorial published along with the new findings, there are likely about 240,000 undiagnosed cases of HIV in the U.S.

But to Dr. Jason Haukoos at the Denver Health Medical Center who wrote the editorial, screening all patients at ERs is not the best way to find those cases.

In a study from last year, he found only a quarter of more than 28,000 patients at his hospital accepted the test, resulting in just 10 new HIV diagnoses.

"These new studies basically say you need to test thousands to identify a handful of patients," Haukoos told Reuters Health. "The question is, is there a way to use scarce resources to target patients at a higher risk?"

According to Haukoos, the test used in the new study costs about $10, and limiting it to patients at risk would be the best idea. According to unpublished data from his group, he said, it even appears that doing so would identify more HIV cases.