New research is raising more questions about tests used around the U.S. to diagnose COVID-19 patients, with some of the tests producing a surprisingly high rate of false negatives that incorrectly show a person isn’t infected.

One study released Wednesday examining an Abbott Laboratories test that’s used at the White House to get rapid diagnoses indicated it may miss as many as half of positive cases. A second peer-reviewed study released hours later suggested that results for another type of widely used diagnostic test are particularly unreliable early on in an infection.

The first study found that Abbott’s ID NOW machine missed at least one-third of positive cases detected with a rival test, and as many as 48% when using the currently recommended dry nasal swabs, according to the report on BioRxiv, a server where researchers post early work before it has been reviewed by other scientists.

Abbott immediately disputed the results, which came from researchers at New York University, and said that they didn’t use the test as intended. The company said its data showed a false-negative rate of 0.02%.

The second analysis, published in the Annals of Internal Medicine and authored by Johns Hopkins researchers, analyzed seven previously published studies evaluating 1,330 patient samples and focused on a group of widely used COVID-19 diagnostic tests known as “polymerase chain reaction” tests. It suggested that false negatives are especially likely before the onset of symptoms, such as when a patient is tested soon after exposure to the virus.

When patients were tested immediately after infection, typically before symptoms occur, the false-negative rate was 100%. On the first day of symptoms, the false-negative rate was 38%. After three days of symptoms, false-negatives dropped to 20%. The rate began getting worse after five days, suggesting a narrow window for the most accurate results.

Accurate diagnosis of coronavirus infections is considered a crucial tool to help identify and control infections. But the technology used by the tests can depend on getting an adequate sample from the patient, often with a swab that reaches deep into the nasal passages. If the sample collected doesn’t have enough viral material on it, either because of how it was collected or because of how far along the infection is, it can affect the result. Low accuracy from the tests could diminish their usefulness as a catch-all screening tool for public health workers and medical personnel treating patients.

The White House is frequently testing staff, governors, lawmakers, reporters and others who come in close contact with the president using Abbott’s ID Now. The testing machine, which is roughly the size of a toaster, is portable and doesn’t require deep nasal swabs, about which President Donald Trump once said there is “nothing pleasant.”

Trump touted the Abbott machine with great fanfare in the White House Rose Garden after the Food and Drug Administration approved it for emergency use, saying it could be a godsend for hospitals and workplaces because it “delivers lightning-fast results in as little as five minutes.”

“That’s a whole new ballgame,” the president said at the March 30 event.

The White House declined to comment on the study, referring questions to the Food and Drug Administration. The agency is “reviewing the information in this non-peer-reviewed study,” a spokeswoman said Wednesday.

Shares of Abbott fell 1.8% to $92.16 in New York on Wednesday. They are up 6.1% this year.

The results of the NYU study of Abbott’s machine, if confirmed, indicate the test may still suffer from high false-negative rates despite changes recommended by the company to avoid such a problem. In April, after other academic researchers raised concerns about the issue, Abbott told clinicians to avoid storing samples in so-called transport media, a liquid used to hold the sample, as that might dilute the test and lead to false negatives.

But when the NYU authors attempted to account for changes and avoid false negatives by using dry swabs, the performance was even worse. It missed the virus in 48% of the samples that were positive using Cepheid GeneXpert, a standard reference test sold by Danaher Corp. The NYU study hasn’t yet been reviewed by outside experts or published in a scientific journal.

Abbott said its scientists are still reviewing the 12-page paper. It has distributed about 1.8 million ID NOW tests and said it hasn’t found the problem identified by the NYU report.

“It’s unclear if the samples were tested correctly, and we’re further evaluating the results,” said Darcy Ross, an Abbott spokeswoman. “The test is performing as expected when it’s used correctly.”

The researchers, who tracked infections among 101 patients treated in the emergency department at NYU Langone Tisch Hospital, said they stand by their findings and the conclusion that the technology is unacceptable for use in a clinical setting.

They also acknowledged the limitations of the study in a statement, including its small size and the testing of the samples in a laboratory rather than at the point of care. They declined interview requests.

Abbott’s ID NOW is a point-of-care test that can be done quickly in a health care setting. Other tests require a sample to be shipped to a lab, with results available after several days. The rapid approach has been widely viewed as a key to ramping up hamstrung U.S. testing efforts, giving answers more efficiently and directing resources to those who are truly sick.

Concerns about false negatives, though, call into question whether the Abbott test will be able to alleviate the strain on U.S. testing efforts as intended.

Scientists have developed a variety of tests to detect the novel coronavirus, and some haven’t been vetted with the usual rigor by federal health regulators. It is also still unclear at what point in the incubation period of the virus any test is most likely to produce an accurate result. Enough information about the virus is simply not yet known.

Several tests have sparked concern about false negatives. Doctors have reported patients with obvious signs of the disease testing negative. While the scale of the problem hasn’t yet been determined, experts agree that it is prevalent.

False negatives can hinder diagnosis of disease, obscure the full scope of an outbreak and raise the potential that sick people could unknowingly spread a pathogen.

Doctors at hospitals around the country have reported patients testing negative sometimes multiple times before eventually testing positive. Some doctors have instead at times turned to other methods of diagnosis, such as chest X-rays or CT scans.