AI could cut medical device recalls by nearly a third, study finds

by Institute for Operations Research and the Management Sciences

edited by Robert Egan

Robert Egan

Senior Editor

Meet our editorial team
Behind our editorial process
Editors' notes

This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

peer-reviewed publication

trusted source

proofread

The GIST Add as preferred source

Most medical devices do not enter the market through rigorous safety testing but by being deemed similar enough to something already approved—the "predicate" device—through a mechanism known as the FDA's 510(k) pathway. But similar is not a synonym for safe. While the assumption may be that substantial equivalence is a reasonable stand-in for a safety review, some devices cleared with this method are still recalled, and the FDA spends valuable time reviewing devices the same way, regardless of how risky they appear on paper.

A new study in Management Science finds that a human-plus-algorithm approach could help the FDA focus its limited review resources on medical devices whose safety is harder to assess while potentially reducing future recalls. The study found that the FDA could catch more unsafe devices and spend less time on those that do not need scrutiny, with a 40.5% reduction in review workload, a 32.9% improvement in the recall rate—the FDA's current recall rate is 10.3%—and up to an estimated $1.7 billion in annual health care savings from fewer device replacements.

Researchers from Indiana University, Harvard Kennedy School and Emerging Health Consulting built a tool that flags which submissions are safe bets, which are risky enough to reject after a brief review and which need more careful consideration by human experts. The machine-learning system estimates recall risk and, coupled with a data-driven policy, recommends which medical devices could be cleared or rejected algorithmically with minimal human oversight and which cases should be flagged for in-depth human review.

"AI is most useful in this setting when it works alongside human judgment, not when it tries to replace it," said Soroush Saghafian, co-author. "An algorithm can systematically identify patterns in thousands of devices and their predicate histories, but there will always be cases where the evidence is ambiguous, or the stakes are too high for an automated decision. By reserving those cases for expert review, the FDA can use its human expertise where it matters most while making the overall process more efficient."

Triaging risk by predicate history

The system works with predicate devices, or devices that may be similar enough—or "substantially equivalent"—to an existing approved device on the market that regulators can be reasonably confident they are safe and effective. Researchers sought to make the comparison system more informative by systematically examining the history of predicate devices, particularly their age and recall status, to estimate the likelihood that a new device would eventually be recalled.

For the devices, an optimization model set thresholds in an algorithm for three categories: devices that could be cleared, devices that may carry a high enough risk to be rejected, and devices in between that should be reviewed by human experts. The model balanced the goal of identifying potentially unsafe devices with the FDA's limited review capacity.

"Right now, the FDA has to devote substantial resources to reviewing medical devices even when the available evidence suggests they pose relatively little risk," said Mohammad Zhalechian, lead author. "Our results suggest that a data-driven approach could help distinguish the straightforward cases from the ones that deserve a closer look. The goal isn't to replace FDA reviewers—it's to give them better information about where their time and expertise can have the greatest impact."

The study and its findings coincide with an existing regulatory conversation. In 2023, the FDA issued draft guidance on predicate selection best practices, proposing to evaluate predicate devices based on their safety track records rather than just availability. The study's model speaks directly to that proposal by tracking predicate recall history and age, a safety track record that the FDA's own guidance supports. If tools like this can inform the direction the FDA takes, the result could be greater safety, more efficient use of resources and more money saved.

Publication details

Mohammad Zhalechian et al, Harmonizing Safety and Speed: A Human-Algorithm Approach to Enhance the FDA's Medical Device Clearance Policy, Management Science (2026). DOI: 10.1287/mnsc.2024.06477

Journal information: Management Science

Provided by Institute for Operations Research and the Management Sciences Who's behind this story?

Robert Egan

Bachelor's in mathematical biology, Master's in creative writing. Well-traveled with unique perspectives on science and language. Full profile →

Citation: AI could cut medical device recalls by nearly a third, study finds (2026, September 22) retrieved 22 September 2026 from https://medicalxpress.com/news/2026-09-ai-medical-device-recalls.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.