For Caregivers Making Over-the-Counter Medication Decisions Alone, Help May Be on the Way
INFO Researcher is using AI to turn dense OTC drug facts labels into personalized guidance
Photo licensed by Adobe Stock
It’s late. The pharmacy is closed, the doctor’s office is closed, and you’re standing in the medicine aisle trying to figure out whether the pain reliever in your hand is safe for your mother. You flip over the box. The drug facts label is right there, technically. Dense as a legal brief, organized for no one in particular. People often don’t actually read it, says Eun Kyoung Choe, an associate professor at the University of Maryland College of Information.
“People glance at it. Even in a structured task where they are asked to make a decision, they miss things. They read something, but it doesn’t register.”
Choe is the principal investigator behind Aidara, an AI-powered mobile app designed to do what the drug facts label doesn’t: translate complex medication safety information into clear, personalized guidance—for older adults, for people with low health literacy, and for the informal caregivers making medication decisions on behalf of someone else, often without much clinical support.
An FDA Researcher Brings a Problem to UMD’s Door
The project started in 2024, when Jeffrey Cox, a researcher in the FDA’s nonprescription drug division, reached out to Choe after learning about her work in health informatics. Cox’s team had been studying how consumers—particularly those with low literacy—interpret over-the-counter (OTC) drug labels. The picture wasn’t encouraging.
“He brought me this problem space,” Choe says, “and I was immediately fascinated. Because as an OTC medication consumer myself, I admit I don’t pay a lot of attention to the drug facts label either. I was surprised to learn how much decision-relevant information is actually in there.”
The two secured a small FDA-supported grant to get started, and a year later, Choe expanded the work with funding from MassAITC, the Massachusetts AI and Technology Center for Connected Care in Aging and Alzheimer’s Disease, through the National Institute on Aging. That’s where the caregiver piece came in.
“Certainly this tool could help people with low literacy or older adults,” Choe says. “But informal caregivers who need to administer OTC medications on behalf of someone else—especially those caring for a loved one with Alzheimer’s Disease—they’re an incredibly underserved group. Many of them are older adults themselves, often navigating those decisions without much guidance or support.”
How the App Works—and What Testing Revealed
The app is still a prototype. To evaluate it, Choe’s team recruited adults 60 and older alongside informal caregivers 50 and older and walked them through medication decision scenarios in the lab.
Aidara guides users through a structured process: enter the drug by typing, speaking, or photographing the pill bottle; the app pulls from the drug facts label and generates screening questions—checking for allergies, drug interactions, age-related risks, and disease contraindications—before delivering a recommendation: safe for use, consult a doctor, consult a doctor/pharmacist, do not use.
Each recommendation comes with a personalized rationale—not a generic warning, but one tied to the user’s own health profile and the specific medication in hand. For those who get the all-clear, the team also plans to surface directions for use in a more accessible format. Notably, the lab study found that participants began paying closer attention to safety information and wanted to know more—asking about drug interactions with their other medications.
Automating the Hard Work of Reading Drug Labels
The prototype works for a small set of medications, but there are tens of thousands of OTC medications on the market. To scale up, the team is using large language models to automatically extract safety information from drug labels and convert it into decision trees—without losing anything important.
“We really want to lower the false negatives,” Choe says. “If we leave out key information in the extraction process, it won’t be part of the questions, and that’s a safety problem.” The team has been working with researchers at Yonsei University in South Korea, annotating drug labels to build a benchmark dataset and evaluate the model’s outputs.
The ultimate goal is an app that works across a wide range of OTC medications and helps people make safer medication decisions, reducing the risk of adverse drug events.