Machine learning models may be able to predict who will develop Covid-19, FDA authorizes an at-home test to detect the coronavirus, flu, and respiratory syncytial virus (RSV), and more in this week's roundup of Covid-19 news.

• Vaccinated patients who had breakthrough infections had a lower risk of several diseases, including long Covid, hypertension, diabetes, and heart disease than Covid-19 patients who were unvaccinated, according to a new study published in Open Forum Infectious Diseases. For the study, researchers analyzed retrospective data from TriNetX on 1,578,719 adults who tested positive and sought care between September 2020 and December 2021. Of the patients, 25,225 (1.6%) were vaccinated. Among vaccinated patients, the average age was approximately 55, 60% were women, and 68% were white. At baseline, 47% of vaccinated patients had hypertension, 23% had diabetes, and 13% had chronic kidney disease. Among unvaccinated patients, the average age was 43, 56% were women, and 62% were white. At baseline, 28% had hypertension, 14% had diabetes, and 6% had chronic kidney disease. At 90 days after initial diagnosis, the researchers found that patients in the vaccinated cohort had a lower risk of new or persistent outcomes compared with patients in the unvaccinated cohort. Incidences per 1,000 in the vaccinated cohort compared with the unvaccinated cohort, respectively, were 7.19 and 20.26 for heart disease, 6.45 and 25.53 for mental disorders, 6.42 and 19.59 for hypertension, and 2.69 and 9.69 for diabetes. In addition, patients who were vaccinated had lower risk of new respiratory symptoms, headache, fatigue, body ache, and diarrhea or constipation at 90 days. "We hypothesize that [vaccination's] effect on reducing the inflammatory responses during the acute phase does also explain the lower rates of all [post-acute sequelae of SARS-CoV-2] outcomes observed in our study among the vaccinated group," the researchers wrote. (Walker, MedPage Today, 5/9)

• Pfizer and BioNTech on Monday announced that three doses of their vaccines were effective at generating a strong immune response in children 6 months to under 5 years of age. According to a preliminary analysis, a three-dose regimen of the pediatric vaccine, which is one-tenth of the adult dose, has an efficacy rate of 80.3% and meets "all immunobridging criteria required for Emergency Use Authorization [EUA]," the companies said. Overall, 1,678 children under five received a third dose two months after their second dose as part of a Phase 2/3 clinical trial. "The study suggests that a low 3-[microgram] dose of our vaccine … provides young children with a high level of protection against the recent Covid-19 strains," said Ugur Sahin, CEO and co-founder of BioNTech. Both companies said they plan to complete their EUA submission of vaccine in children under five this week. FDA plans to convene its Vaccines and Related Biological Products Advisory Committee (VRBPAC) on June 15 to discuss Covid-19 vaccination for children as young as six months. At the meeting, VRBPAC will consider applications from both Pfizer-BioNTech and Moderna. (Chappell, NPR, 5/23; Reed, Axios, 5/23; Walker, MedPage Today, 5/23; Branswell, STAT News, 5/23)

• FDA last week authorized a non-prescription Covid-19 test that can also be used to detect flu and RSV. To use the test, which is made by Labcorp, individuals collect a nasal swab sample and send it to the company for testing. Then, they can access their results online, with a health care provider following up about any positive or invalid results. "While the FDA has now authorized many COVID-19 tests without a prescription, this is the first test authorized for flu and RSV, along with COVID-19, where an individual can self-identify their need for a test, order it, collect their sample and send it to the lab for testing, without consulting a health care professional," said Jeff Shuren, director of the FDA’s Center for Devices and Radiological Health. "The rapid advances being made in consumer access to diagnostic tests, including the ability to collect your sample at home for flu and RSV without a prescription, brings us one step closer to tests for these viruses that could be performed entirely at home." (Shapero, Axios, 5/16)

• Viral antigen was found in Covid-19 patients' cerebrospinal (CSF) fluid and associated with immune activation in the central nervous system (CNS), according to a small study in Sweden. For the study, researchers analyzed CSF samples from adult patients admitted to Sahlgrenska University Hospital between March 2020 and June 2021. In total, there were 44 Covid-19 patients, 10 health controls, and 41 Covid-negative controls. Among the Covid-19 patients, 89% had detectable nucleocapsid antigen in their CSF samples; however, there was no detectable coronavirus RNA in any of the samples. In addition, the Covid-19 patients had significantly higher CSF neopterin, β2-microglobulin, interleukin (IL)-2, IL-6, IL-10, and tumor necrosis factor α than the controls. According to the researchers, these findings suggest that viral components can affect CNS immune response even without direct viral invasion. "The presence of viral antigen has not been known previously and provides additional information regarding what may cause immune responses within the central nervous system in patients with COVID-19," said Arvid Edén, one of the study's authors. (George, MedPage Today, 5/24)

• Machine learning models developed by NIH can accurately identify patients who are likely to have long Covid, according to a study published in The Lancet Digital Health. For the study, a National Covid Cohort Collaborative team analyzed HER data from 97,995 adults who were diagnosed with Covid-19 at least 90 days earlier and 597 adults currently undergoing treatment at a long Covid clinic. Three machine learning models were developed using this data and designed to detect patterns of symptoms, health care use, demographic, and prescriptions to identify Covid-19 patients who were likely to have lingering symptoms. Tests of the models estimate the area-under-the-curve scores, which indicate sensitivity and specificity, were 0.92 in all Covid-19 patients, 0.90 in hospitalized patients, and 0.85 in outpatients. So far, more than 200,000 long Covid patients have been identified by the models. According to Josh Fessel, senior clinical advisor at the NIH's National Center for Advancing Translational Sciences, the models can be used to identify patients who specialized care."Once you're able to determine who has long COVID in a large database of people, you can begin to ask questions about those people," Fessel said. "Was there something different about those people before they developed long COVID? … Was there something about how they were treated during acute COVID that might have increased or decreased their risk for long COVID?" (CIDRAP News, 5/17)