Hello and welcome to Healthbeat's weekly report on stories shaping public health in the United States.

I am Dr. Jay K. Varma, a physician, epidemiologist, and public health expert currently serving as chief medical officer at Fedcap, a global nonprofit focused on economic mobility and well-being for vulnerable communities. Views expressed here are my own.

During the peak of the Covid pandemic, the phrase "follow the science" became a popular refrain, used first by those advocating stronger infection control policies and later by those opposed to them.

In this newsletter, I am discussing how, in the past few weeks, scientific evidence has been selectively chosen to meet a specific political or economic end, and how AI is making it harder than ever to know which scientific studies we can trust.

Politics of Covid vaccines and origins continue to cause trouble

The Washington Post broke a story last week that acting Centers for Disease Control and Prevention Director Jay Bhattacharya has been preventing CDC career scientists from publishing a study about the benefit of Covid vaccines during the recent winter respiratory virus season.

The report, which had cleared the agency's full scientific review process and was scheduled for publication in the CDC's flagship journal, the Morbidity and Mortality Weekly Report, on March 19, found that, between September and December 2025, healthy adults who received a Covid vaccine reduced their likelihood of emergency department and urgent care visits by 50% and the likelihood of Covid-associated hospitalizations by 55%. 

According to the story, Bhattacharya, a health economist, thinks that the methodology used by CDC scientists is flawed. He is arguing that the CDC should only be publishing randomized control trials, a position completely at odds with the overwhelming consensus of vaccine scientists and epidemiologists who work in academia, government, and industry. 

Randomized controlled trials are used to answer the question: "Does this vaccine prevent an infectious disease when used as intended in a selected population?" In contrast, the methods used in the suppressed Covid study are used to answer a different question: "Does this vaccine prevent an infectious disease when used in a real-world population under real-world conditions?" The Food and Drug Administration and industry consider this "real world evidence," and the CDC has been using the exact same method for decades to estimate the effectiveness of the seasonal flu vaccine after each flu season has ended. 

A report about flu vaccine effectiveness this past winter, using the exact same methods, was published in the MMWR just a week before Bhattacharya blocked the Covid study.  (I covered the flu study last week.) Why did his concerns about scientific methods apply to the Covid vaccine study and not the flu vaccine study?

I fear that the real reason Bhattacharya has squashed this study is that he does not want the CDC publishing any data about the benefit of Covid vaccines. Despite his repeated podcasting and writing about the importance of scientific debate, he rose to prominence as a Covid contrarian and is using his positions as National Institutes of Health and CDC director to promote his views and squash opposing scientific views. 

On March 20, in his role as NIH director, Bhattacharya hosted the inaugural lecture of a series he is calling "NIH Scientific Freedom." That talk featured a science writer from the UK named Matt Ridley about the "search for the origin of Covid-19." Ridley argues that Covid was engineered in a lab and released accidentally. His talk completely disregarded the extensive evidence that points to an opposing hypothesis: The virus spilled over from animals into humans and amplified into an outbreak at a live animal market in Wuhan, China. 

Thankfully, a diverse team of experts recently pushed back. Twenty-three of the 27 original members of the World Health Organization's Scientific Advisory Group for the Origins of Novel Pathogens recently published a careful, detailed review concluding that most peer-reviewed scientific evidence supports that SARS-CoV-2 spilled over from an animal into humans.

They found no conclusive evidence for a lab leak, and noted that government and intelligence reports favoring a lab leak rely on political, rather than scientific, arguments. Virologist Angela Rasmussen has written a considerably more caustic assessment of the Ridley lecture that is also worth reading. 

In studies about the origins of a virus, we can rarely say anything is 100% certain. While a lab leak is certainly possible, the evidence strongly and consistently points elsewhere.

Politics of raw milk products cause trouble at FDA

Science remains a point of contention in an ongoing outbreak of E. coli O157 infections that CDC and FDA investigators have linked to cheddar cheese sold by a company known as RAW FARM.

Identifying the source of a foodborne disease outbreak is straightforward when people are exposed to the contaminated food at a single event — what's known as a "point source" outbreak. When the exposure occurs over a wide geographic area and a long period — such as a food product sold around the country over weeks — the investigation is far more challenging. 

Epidemiologists need to interview every patient about any foods they may have eaten before they became sick, relying on human memory and electronic records, neither of which are completely reliable. Microbiologists need to obtain specimens from patients and link them together using genomic analyses. Ideally, they are also able to obtain specimens from food, but this is often difficult because the contamination may have been intermittent or limited to a batch that has already been sold and discarded.

Because it is so difficult to find a "smoking gun," public health officials need to use a preponderance of scientific evidence to decide what to do, similar to the debates about how much evidence we need to decide Covid originated in nature or that vaccines work in the real world.

In this outbreak, the epidemiological case against RAW FARM is substantial. As of early April, nine people had been infected with the outbreak strain of E. coli across three states. Eight of these nine people were interviewed, and all reported eating raw dairy products. Of the seven who remembered the brand, all named RAW FARM. Genomic sequencing confirmed all the infections were genetically identical, meaning everyone probably got sick from the same source.

And yet RAW FARM twice refused FDA requests to voluntarily recall its cheese, posting on social media that the outbreak was not connected to its products and resharing testimonials from customers who continued to buy and feed the cheese to their children. When the company finally issued a recall on April 2, it did so "under protest," continuing to deny the investigation's findings. 

Unfortunately, this is not just a problem of one defiant cheese maker. The FDA took weeks longer than it would have previously to compel action, and the company's defiance was likely emboldened by the current political moment, in which the head of the federal department overseeing food safety has celebrated raw dairy as a health food and framed food safety regulations as an assault on consumer freedom.

When leading government officials signal that food safety standards have changed, some producers are going to test that proposition and sicken us in the process.

AI is causing trouble with the scientific literature

An article published last week in Nature illustrates a different and perhaps more insidious threat: how AI can corrupt scientific communication and literature. 

A medical researcher at the University of Gothenburg named Almira Osmanovic Thunström invented a fake skin condition called "bixonimania," a name she chose specifically because it sounded absurd, a psychiatric term attached to an eye condition.

She uploaded two obviously fake preprints about the condition to an academic website in early 2024, complete with a fictional author, a made-up university, and acknowledgements thanking "Professor Sideshow Bob" and "the Galactic Triad" for funding support. One of the papers stated explicitly that it was "made up." Her goal was to test whether AI systems would pick up and repeat fabricated medical information.

The experiment worked faster and more thoroughly than she had anticipated. Within weeks, major AI chatbots were telling users that bixonimania was a real condition, describing its prevalence, and advising them to see an ophthalmologist.

Even more troubling, the fake papers were then cited in actual peer-reviewed medical literature, suggesting that some researchers were using AI-generated references without reading the underlying papers. One such paper, published in a Springer Nature journal, was retracted only after Nature's reporters contacted the journal to ask about it.

As recently as March, some versions of major AI chatbots were still describing bixonimania as an "emerging condition" or a "proposed new subtype" of a real disorder.

Millions of people now use AI chatbots as a first stop for health information. If the scientific literature that feeds those systems can be poisoned by fake papers, bad actors could potentially manufacture medical "evidence" to sell products, cast doubt on real treatments, and distort public understanding of disease.

Who controls what counts as evidence?

Who controls what counts as evidence, and what happens when that control is abused? In the past week, we saw examples of this problem from government agencies and the private sector.

Scientific reasoning requires understanding what a particular study can and can’t tell you, what the weight of evidence across multiple studies suggests, and where genuine uncertainty remains. That kind of reasoning is hard to do, easy to distort, and, as this week makes clear, under assault from multiple directions at once.

ICYMI

Here’s a recap of the latest reporting from Healthbeat:

• Measles: Where measles is spreading in the U.S.: Outbreaks fuel infections in states coast to coast

• Health policy: Spurred by Harlem Legionnaires’ disease outbreak, new rules for NYC cooling tower testing to take effect

• HIV and malaria aid: The next U.S. aid disruption: HIV, malaria drugs for Africa and Asia

  • 🌎 Sign up here for the Global Health Checkup, a weekly report that explains how global health matters to all.

Until next week,

Jay