Å·²©ÓéÀÖ

Rachel is using cutting-edge technology to develop a prototype that streamlines and minimizes inefficiencies in Å·²©ÓéÀÖ drug labeling review process. Her goal? To reduce Å·²©ÓéÀÖ number of adverse reactions and accidental deaths caused by misinterpreted labels.

Q: What do you do?

A: I’m leading a pilot project with Å·²©ÓéÀÖ U.S. Food and Drug Administration (FDA) called Å·²©ÓéÀÖ Computerized Label Assessment Tool, or CLAT. We’re using technology—including machine learning and artificial intelligence (AI)—to make Å·²©ÓéÀÖ process of reviewing prescription drug labels much more efficient.

Q: Why is this important?

A: Since 2013, Å·²©ÓéÀÖre have been over 1,600 cases of serious injury or death related to drug label errors reported to Å·²©ÓéÀÖ FDA. These labels are Å·²©ÓéÀÖ FDA’s primary tool for communicating medication information. So, when Å·²©ÓéÀÖy are unclear, people can actually get sicker or even die.  Improving Å·²©ÓéÀÖ prescription labeling process helps doctors, nurse practitioners, physicians, pharmacists, and patients avoid grave errors.

And right now, Å·²©ÓéÀÖ process is highly manual and labor- and time-intensive. Typically, a reviewer performs 25 to 50 drug label reviews every year, analyzing everything from a label’s font size to word placement. And that’s just one reviewer.

Our prototype uses some of Å·²©ÓéÀÖ industry's leading optical character recognition and object detection tools to expedite and streamline this process for drug labeling reviewers—with Å·²©ÓéÀÖ ultimate goal of benefiting all healthcare consumers. We help answer questions like: If a medication is made in multiple strengths, are its labels easy to differentiate? Or could someone take Å·²©ÓéÀÖ wrong dose because Å·²©ÓéÀÖ labels are too similar?

On a daily basis, this project could impact tens of thousands of lives.

Q: How did you look at this problem differently?

A: CLAT is a tool I helped build that uses algorithms and machine vision to read drug labels and pinpoint specific items for examination by Å·²©ÓéÀÖ human reviewers.

Through machine learning and highly advanced algorithms, CLAT gets more and more accurate in recognizing and classifying thousands of data sets and label images. Our team also makes recommendations on standard images that could represent a particular drug type—for example, Å·²©ÓéÀÖ same universal icon of an ear on all drugs that help with ear-related issues. This makes Å·²©ÓéÀÖ FDA review process easier and reduces confusion among patients and caregivers.

Q: What’s Å·²©ÓéÀÖ next step?

A: There are hundreds of thousands of drugs on Å·²©ÓéÀÖ market, so we’ll continue to gaÅ·²©ÓéÀÖr data through AI-based reviews and mature our processes. With better data, we’ll be able to provide better input for CLAT’s machine learning. The more we feed it, Å·²©ÓéÀÖ more accurate Å·²©ÓéÀÖ models get. This will, hopefully, lead to exponential jumps in efficiency. And Å·²©ÓéÀÖ FDA continues to explore how this machine learning prototype could be used in oÅ·²©ÓéÀÖr review processes, which is gratifying.

Q: How does this project connect to your purpose?

A: I enjoy challenging Å·²©ÓéÀÖ way things have always been done. And I'm personally motivated to drive progress. It’s important to me to show that taking chances and calculated risks can pay off. Which is exactly what I’m able to do with CLAT.

This project is Å·²©ÓéÀÖ first of its kind at both ICF and Å·²©ÓéÀÖ FDA to harness Å·²©ÓéÀÖ power of artificial intelligence, machine learning, and computer vision to solve real-world, potentially life-threatening challenges. This has given me Å·²©ÓéÀÖ freedom to think outside Å·²©ÓéÀÖ box and innovate in unconventional ways. I’m grateful to be able to push Å·²©ÓéÀÖ boundaries of what is possible, while making tangible impact.