The Abbott coronavirus test hailed by President Trump and used by the White House failed to detect infected samples in a large number of cases that were caught by a rival firm, a preliminary study has found.

The speedy Abbott test, which is supposed to determine in five to 13 minutes whether a person has the virus, missed a third of the positive samples found by the diagnostic company Cepheid when both tests used nasopharyngeal swabs, according to the study done by a group from New York University. It missed more than 48 percent of cases when dry nasal swabs were run on Abbott machines, compared to nasopharyngeal swabs run on the Cepheid testing platform. Nasopharyngeal swabs penetrate deeply into the nasal passages, while dry nasal swabs are less invasive.

The study, while preliminary and not yet peer-reviewed, raised questions about a test that has been praised by Trump, who displayed it at a news conference in the Rose Garden on April 2 and said it created “a whole new ballgame.” As the pandemic was creating a sense of urgency about testing, the Abbott test triggered a scramble among governors and other state officials because bottlenecks were causing waits of as long as a week or longer for test results.

Abbott denied Wednesday there were major flaws in its test. “It’s unclear if the samples were tested correctly in this study. Abbott has distributed about 1.8 million ID NOW tests and the reported rate of false negatives to Abbott is at 0.02 percent,” Abbott spokesman John Koval said. “This rate has been previously shared with the FDA. In communications with the users of the test, it is performing as expected.”

The White House was shaken last week when two staff members tested positive for the virus. Four officials who have been screened said the White House was using only the Abbott test.

Abbott’s customers include small physician groups and big retailers such as CVS, which is deploying the ID NOW test at its five drive-through testing sites in large public parking lots, including the Georgia Institute of Technology lot in Atlanta and the Twin River Casino lot in Lincoln, R.I.

The New York University group initiated its study because it was searching for a diagnostic tool that would be faster than what it currently has. The NYU laboratory uses two platforms — the Roche Cobas SARS-CoV2 and the Cepheid Xpert Xpress SARS-CoV-2. The run times for each test are 3.5 hours and 45 minutes, respectively.

The Abbott ID NOW test was a logical alternative. It processes results through a device about the size of a toaster in as few as five minutes and as many as 13 minutes. Other companies rely on devices about the size of a washing machine; the samples take longer to analyze, although the devices can process larger numbers of samples at a time.

Even before the NYU study, there were concerns that the Abbott test might miss more cases than traditional laboratory tests.

Before a Senate committee last week, Francis Collins, director of the National Institutes of Health, testified the rapid Abbott test had a 15 percent false-negative rate, an apparent reference to unpublished data from the Cleveland Clinic.

“If you’re in a circumstance where you really don’t want to miss a diagnosis of somebody who’s already carrying the virus, you’d like to have something that has a higher sensitivity than that,” Collins said.

The Abbott rapid test missed about 25 percent of positives compared with a slower Abbott test, according to a study published April 23 in the Journal of Clinical Microbiology.

The company has pushed back on these studies for different reasons. The Cleveland Clinic study, not yet published, involved putting the swabs in a liquid that is no longer recommended, Abbott’s Koval said. He questioned whether the samples in the NYU study were run correctly. “We are working with the author to better understand the analysis as part of the scientific process,” Koval said of the Journal of Clinical Microbiology study.

At Loyola University Medical Center, where researchers found 1 in 4 positive cases were missed, health-care providers are using the Abbott rapid test to screen patients — but performing a laboratory test to double-check any negatives.

“We are going to do some additional testing to investigate our initial results, and we will continue to monitor test performance going forward,” said Amanda Harrington, medical director of Loyola’s Clinical Microbiology Laboratory and an author on the Journal of Clinical Microbiology study.

Brian D. Stein, a pulmonary and critical care physician and acting chief quality officer at Rush University Medical Center in Chicago, said the center didn’t have exact numbers, but Abbott’s test was “not quite as sensitive as other tests that take longer to run.”

“The question is, ‘Do we have something better that’s that fast?’” he said. “No.”

Some medical centers using the tests early on found the rate of false negatives — people who test negative but are actually positive — were too great for the tests to be useful. Abbott worked with the Food and Drug Administration in April to amend instructions on how to avoid diluting samples.

But rapid tests can be useful in quickly sifting out sick people, even if they misidentify some cases. The rapid influenza test may miss up to half of all cases. And people who are infected could test negative for a variety of reasons — if they are early in the illness, for example, and do not yet have much virus in the back of their nose.

“The guidance we are giving is that if an ID NOW test is negative and the clinical suspicion is high, to run another test to confirm,” said Bob Kocher, a physician, partner at a venture capital firm and member of California’s testing task force, noting the odds of two false negative ID NOW tests are low.

Josh Dawsey contributed to this report.

Correction: A previous version of this story incorrectly described one of the comparisons by NYU researchers. They compared dry nasal swabs run on Abbott machines to nasopharyngeal swabs run on the Cepheid testing platform.