Once used primarily to interpret human communications, natural language processing algorithms now “serve as the backbone for translation and clinical transcription services,” according to John Halamka, MD, and Paul Cerrato of Mayo Clinic Platform, and look to play an even bigger role in the future.
One way to improve the “widely contrasting results” often garnered by ChatGPT is through prompt engineering, according to Mayo Clinic Platform researchers. In this piece, Dr. John Halamka and Paul Cerrato discuss how digital tools can supplement decision-making.
The impact that large language models have had both on the medical community and IT developers is undeniable. But are they being trained with the relevant data and the right kind of self-supervision? These are questions that need to be asked, according to Mayo Clinic Platform researchers.
Allowing stakeholders to access data while protecting patient rights remains a significant challenge. But there is a potential solution, according to Mayo Clinic Platform researchers, who describe a “federated approach” in which data are stored in encrypted containers.
When leveraged for their original purpose – to “create a skeleton note for humans to augment and edit, reducing administrative burden – generative AI can help provide more time for patient care and clinical decision making, according to John Halamka, MD, and Paul Cerrato of Mayo Clinic Platform.
Large language models offer great potential for improving care and efficiency, but “that must be done within guardrails and guidelines so that we do no digital harm,” said Paul Cerrato and John Halamka, MD. In this piece, they discuss the key components for successful development and deployment of AI tools.
If digital tools don’t solve the problems they’re designed to solve – or worse, “do more harm than good” – they aren’t much good to healthcare organizations, said John Halamka, MD, and Paul Cerrato. The key, they believe, is in constant validation and earning the trust of stakeholders.
Through partnerships with health systems, payers, medical device companies, and academic medical centers, Mayo Clinic Platform_Connect aims to find better ways to diagnose, treat, and prevent disease, according to John Halamka, MD. How? “By connecting clinical data through a federated, secure architecture.”
Large language models and GPT may not be the apocalyptic nightmare some critics fear, but they certainly require due diligence by thought leaders, health care professionals, and the general public, wrote John Halamka, MD, and Paul Serrato. “We need a framework to manage them responsively.”
Does the industry really need another set of guidelines for implementing AI in healthcare? Yes, said Dr. John Halamka, noting that the latest publication from CHAI offers “a holistic approach to assessments of AI-based clinical systems for health systems, consumers, and end users,” he wrote in a recent piece, along with Paul Cerrato.