In the US, Miami-based Mount Sinai Medical Center is among the first health systems to deploy a new AI-powered tool from Epic as the organisation looks to better serve its Spanish-speaking patient population.
The tool is a Spanish-language version of Epic’s Augmented Response Technology (ART), which uses generative AI to analyse messages sent through MyChart and draft suggested replies for providers to review, edit and send.
The rollout builds on Mount Sinai’s early adoption of the English-language version of ART in 2023, which was driven by a surge in patient portal messages during the COVID-19 pandemic. According to Tom Gillette, CIO at Mount Sinai Medical Center, message volume spiked fivefold during that time as patients increasingly turned to digital channels for advice, prescriptions and follow-ups.
“Layer on top of that, more than 60% of Miami-Dade County speaks Spanish, and about 30% of our patient population has indicated Spanish as their preferred language in Epic,” Mr. Gillette said in an interview with Becker’s. “So extending our ART experience into Spanish was a natural next step.”
Mount Sinai worked closely with Epic on the development and testing of the Spanish-language tool, beginning in December 2024. The effort included IT leaders, bilingual physicians and care teams.
“Our clinicians were intimately involved in shaping this,” Mr. Gillette said. “They helped ensure the messages not only translated correctly, but also made sense medically and aligned with our clinical voice.”
While the tool was technically straightforward to deploy, Mr. Gillette emphasised the importance of rigorous validation—particularly in ensuring that clinical responses in Spanish are not just grammatically correct, but clinically appropriate.
“Accuracy was critical,” he said. “We had to be sure the AI wasn’t missing anything or mistranslating medical concepts.”
Beyond translation, the project raised deeper questions about consistency in clinical communication. Mount Sinai’s team engaged in “prompt engineering” to define organisation-wide standards for AI-generated drafts.
“When a patient says, ‘I think I have a UTI,’ one doctor might say, ‘Come in and see me,’ another might suggest an e-visit, and another might just offer advice,” Mr. Gillette said. “So we had to ask: What’s the response we want to start with? What reflects our standard of care?”
Source: Becker's Health IT, 8 August 2025
Recommended Comments
Create an account or sign in to comment