AI roundup: EHR search functions, hands-free charting, and patient outcome estimations

It’s ChatGPT this, ChatGPT that. But the artificial intelligence (AI) revolution may already be here, as more healthcare executives implement AI in their clinical and administrative settings.

AI has the potential to improve both the clinician and patient experience amid ongoing staffing shortages and high burnout rates—and healthcare isn’t the only industry to turn to tech for solutions. Sectors like finance, retail, and technology have developed AI models that healthcare organizations could replicate to make the technology more commonplace, Baptist Health South Florida Chief Data Officer Sha Edathumparampil told Healthcare Brew.

“The ability for those industries to take advantage of AI and improve their business models, as well as their consumer experiences, has suddenly laid sort of a blueprint for the rest of us from a healthcare perspective to use and follow quickly so that ramp-up is much easier,” Edathumparampil said.

We rounded up some of the AI healthcare developments you may have missed this week.

AI-enabled search function: Google Cloud announced a partnership with the Mayo Clinic, allowing providers to use generative AI to more easily search across disconnected health databases and electronic health records (EHRs).

Hands-free charting: San Francisco-based Carbon Health is integrating an AI-enabled notes assistant into its EHRs. During an appointment, the generative pretrained transformer 4, or GPT-4, AI model will record audio of the patient-provider conversation and generate a visit summary that includes patient history, diagnosis codes, and any other notes the provider adds.

This GPT-4 model can, on average, fill out a medical chart in less than four minutes compared to 16 minutes for a manual chart, according to the health system. The technology can give doctors more time to treat patients: Providers typically spend around 34 minutes with each patient, but about half the appointment is dedicated to administrative work like charting.

The “AI doctor” is in: A new large language model (LLM) developed by researchers at the NYU Grossman School of Medicine can analyze EHRs to estimate a hospitalized patient’s length of stay, risk of mortality, and chance of readmission, among other factors, according to a new study. The AI tool, called NYUTron, was able to predict about 80% of patient readmissions within a month, the study found.

“Our findings highlight the potential for using large language models to guide physicians about patient care,” study lead author Lavender Jiang said in a statement. “Programs like NYUTron can alert healthcare providers in real time about factors that might lead to readmission and other concerns so they can be swiftly addressed or even averted.”

NYU Langone Health hospitals are using NYUTron to help predict readmission risks, according to the statement.

AI is making its way into every aspect of healthcare, but for now, the technology can’t fully replace providers.

“It’s hard to think of any area that will be spared, so to speak, the impact of this new technology,” Edathumparampil told us. “I do think the actual treatment itself of a patient unassisted will probably be the last to go for AI. That we will need a lot of confidence and a lot of data and a lot of background to be able to confidently and conclusively say we trust AI to take care of humans.”