Biohacking
MDCalc's new tool strips patient data from AI queries before they leave the clinic
By Life and Health Today Staff, . Life and Health Today.
A new tool from MDCalc, a medical reference website used by clinicians to look up dosing calculators and clinical decision rules, is trying to solve one of the quieter problems in hospital AI adoption: doctors are reluctant to type patient details into a chatbot they do not fully trust.
CrossCover, announced Wednesday, takes the opposite approach from most clinical AI products on the market. Rather than expanding what patient, provider, and payer data an AI system can access, it strips identifying information out of a clinician's query before it goes anywhere, according to the report by Brittany Trang in STAT News.
How it differs from the tools already in use
Most clinical AI products, including those from OpenEvidence, Abridge, and electronic health record company Epic, are moving toward incorporating more patient and provider data into their recommendations. Some require clinicians to register with national identification numbers; others manage the privacy risk through agreements that allow protected health information to enter the system.
MDCalc's CrossCover does neither. The privacy protection sits at the query level: the tool removes the sensitive information before it reaches the AI, rather than governing what happens to it afterward.
Graham Walker, head of AI at MDCalc, described the tool to STAT News as an accelerator for clinical AI adoption. "We can take away some of the challenges and risks and problems with the technology so that physicians can feel more confident in using it," he said.
What the company says it is solving
The framing MDCalc is using is that physician hesitation, not the AI itself, is the bottleneck. If a clinician will not enter a real patient scenario because they are uncertain where that data goes, the AI tool does not get used regardless of how capable it is. CrossCover is positioned as a way to remove that hesitation by making the privacy question moot at the point of entry.
The report does not include independent clinical validation of CrossCover, published trial data, or third-party assessment of how well the tool actually strips identifying information. What MDCalc has released is a product announcement, not a study.
What this does not settle
The deeper question, which the announcement does not address, is whether removing patient context from a query changes the quality of the AI's answer. Clinical AI tools that incorporate detailed patient histories argue that richer data produces more relevant recommendations. A system that works from a de-identified or stripped query may be safer from a privacy standpoint while being less useful for complex cases. That tradeoff has not been tested publicly.
For clinicians weighing which AI tools to use, the choice between privacy-first and data-rich architectures is not yet settled by evidence. Whether CrossCover changes physician behavior in practice, and whether that behavior change improves care, are questions the product launch leaves open.
Life & Health Today provides information for educational purposes only, not medical advice, diagnosis, or treatment. Always consult a qualified healthcare professional.