News

AI Powered Chatbots in Search Engines Show Promise but Pose Risks for Patient Drug Information, Study Finds

• By John K. Waters
• 10/21/2024

AI-powered chatbots integrated into search engines can provide largely accurate drug information to patients, but they also present significant risks due to readability issues and potentially harmful advice, according to a new study ("Artificial intelligence-powered chatbots in search engines: a cross- sectional study on the quality and risks of drug information for patients").

The cross-sectional study, published in BMJ Quality & Safety, examined the quality and potential safety concerns of drug information provided by Microsoft's Bing AI copilot.

Key Findings:

• Readability: The study found that chatbot answers were generally difficult to read, with an average Flesch Reading Ease Score of 37.2, indicating a college-level reading proficiency is needed to understand the information.
• Completeness and Accuracy: Overall, the chatbot provided complete and accurate answers, with median completeness of 100% (IQR 50.0-100.0%) and median accuracy of 100% (IQR 88.1-100.0%). However, inconsistencies were noted in 26% of the answers.
• Scientific Consensus and Potential Harm: In a subset of 20 chatbot answers evaluated by experts:
• 39% (95% CI 25% to 55%) were found to oppose scientific consensus
• 66% (95% CI 50% to 85%) were judged as potentially leading to harmful outcomes
• 22% (95% CI 10% to 40%) could result in severe harm or death if patients followed the advice

"While AI-powered chatbots show promise in providing drug information, our study highlights significant concerns regarding patient safety and the need for caution when using these tools," said lead researcher Dr. Wahram Andrikyan, in a statement.

The study analyzed responses to 10 frequently asked patient questions about the 50 most prescribed drugs in the U.S. outpatient market. Researchers evaluated readability, completeness, and accuracy of the chatbot's answers, comparing them to information from the pharmaceutical encyclopedia drugs.com.

Experts in the field emphasize that despite the potential benefits of AI-powered search engines in helping patients find drug information more quickly, the technology is not yet reliable enough to replace professional medical advice.

"Healthcare professionals should exercise caution when recommending AI-powered search engines to patients until more accurate and reliable alternatives are available," Andrikyan said.

The researchers suggest that future developments in AI-powered search engines for patient drug information should focus on improving readability and ensuring that information is sourced exclusively from validated patient information databases.

As AI continues to evolve in the healthcare sector, this study underscores the need for ongoing research and development to address the challenges of providing safe and accessible drug information to patients through digital platforms.