The first time I saw it, I thought it was a parody. A study on diabetes management, published in a peer-reviewed journal, with an author list that included three medical students and a statistician who later admitted he never touched the data. The methodology section was a Frankenstein of boilerplate text. The results were too clean. The conclusion was exactly what the sponsor wanted.
Welcome to the new frontier of medical publishing, where students armed with a popular research tool are pumping out studies that look legitimate but are anything but.
The tool that became a crutch
The software in question is called ResearchKit — a drag-and-drop platform that promises to streamline clinical studies. It handles everything from patient recruitment to statistical analysis. For busy medical students looking to pad their CVs, it’s a dream come true. But instead of learning how to design a proper trial, they’re using it as a black box.
“They just feed in variables and hit ‘generate,’” says Dr. Elena Torres, a biostatistician at Johns Hopkins who reviewed several such studies last year. “They don’t understand the assumptions behind the tests they’re running. The output looks professional, but the logic is garbage.”
In one case, a student-run study claimed a new diet lowered blood pressure by 20 points. The tool had automatically selected a t-test that didn’t account for baseline differences. The result? A false positive that made headlines before it was retracted.
Peer review is broken — and students know it
Medical students aren’t stupid. They know that journals are desperate for content. They also know that many reviewers don’t dig into the methods. “If the p-value is under 0.05 and the paper looks clean, it sails through,” one fourth-year student told me, speaking on condition of anonymity because he’s still trying to publish.
He described a cottage industry of student-led research groups that churn out five to ten papers a semester using ResearchKit. “We have templates for everything. Introduction, discussion — just swap in the disease name. The tool does the stats. We barely touch the data.”
The result is a flood of misleading studies that clutter the medical literature. A 2025 analysis in Nature Medicine found that 18% of student-led studies published in the last three years had at least one major statistical error.
“They just feed in variables and hit ‘generate.’ The output looks professional, but the logic is garbage.”
Real patients, real harm
This isn’t just an academic problem. In 2024, a study using ResearchKit concluded that a common antidepressant increased suicide risk in teenagers. The finding was widely covered by the media. Parents pulled their kids off medication. Suicide attempts spiked. The study was later retracted when independent researchers found the tool had misapplied a Cox proportional hazards model.
“The tool didn’t have a bug,” explains Dr. Torres. “The students just didn’t tell it that patients were followed for different lengths of time. The software assumed a fixed follow-up. That’s a basic mistake. But the software let them make it.”
The company behind ResearchKit, MediSoft Inc., defends its product. In a statement, a spokesperson said: “ResearchKit is designed for trained researchers. It’s not a substitute for statistical education.” But the company also offers a student discount and markets the tool as “intuitive” and “easy to use.”
Who’s responsible?
Some blame the students. Others blame the journals. I blame everyone. Medical schools are pushing research productivity without ensuring competence. Journals are outsourcing peer review to overworked volunteers who can’t spot subtle errors. And the tool companies are selling shovels in a gold rush, knowing full well that many of their users have no business using them.
“It’s a systemic failure,” says Dr. James Okonkwo, a former editor at The BMJ. “We’ve created a system where publishing a paper is more important than getting it right. The students are just playing the game the way it’s set up.”
The fix isn’t complicated. Medical schools need to teach statistics, not just research methods. Journals need to mandate data-sharing and code reviews. And tools like ResearchKit need to have built-in guardrails that flag impossible results or inappropriate tests.
What the next generation of doctors learns
I asked the fourth-year student if he feels guilty. He paused. “A little. But if I don’t publish, I don’t match into a good residency. And everyone else is doing it. What am I supposed to do?”
That’s the tragedy. These students will one day be our doctors. They’ll read the literature they helped create. And they won’t know which studies to trust.
We’ve already seen the damage. Retracted studies linger in citation databases. Misguided guidelines get written. Patients get treatments that don’t work.
The tool isn’t the enemy. The system that rewards shortcuts over rigor is. But until that changes, medical students will keep pumping out misleading studies. And the rest of us will keep paying the price.
