Louisiana hospital uses AI to transcribe doctor visits | Louisiana Health

Ambient artificial intelligence, a type of AI that solves problems without human intervention, is being used across the United States in hospitals to turn patient conversations into complete, accurate notes with real-time insights for doctors.

The AI transcription tool looks to enhance the patient-physician relationship and improve patient engagement as well as increase the amount of time a physician can interact with a patient instead of a computer during the visit.

On average, a doctor will spend over 15 hours a week taking notes and inputting patient information into a secured online database. Time many practitioners would prefer to give back to their patients.

This new technology enables physicians to be fully present during the patient’s visit while reducing the burden of documentation for doctors, according to Dr. Jason Hill in New Orleans.

Hill, a board-certified internal medicine doctor in New Orleans, has spent a greater part of his career “leveraging tools for physicians and patient care.”

In November 2023, Hill took the role of innovation officer at Ochsner where his responsibility is to implement novel innovations in the hospital system.

“In 2024, new innovations is basically all AI,” Hill said.

The new AI transcript tool is currently in use with 350 doctors at Ochsner in the southeast Louisiana region primarily with primary care providers and a small number of orthopedics specialists and more.

Why is an artificial intelligence transcription tool necessary in hospitals?

Documentation burden is a really big deal, and there have been some studies that show that doctors spend over half their day interacting with the electronic health record.

If you’ve ever been a clinic in a doctor’s office, you see a doctor hunched over at their computer while talking to you. It doesn’t make a great experience at either the doctor side or the patient side.

I think that’s one of the key criteria: How can we transform the clinical encounter?

As a doctor, we have to see over 15 patients a day. How do I remember everything that’s happening with everybody at every meeting? I have to write accurate, complete notes down either in the appointment (and we have a weird patient encounter); or I squeeze in note taking in between visits; or I take my notes for the day at 10 o’clock at night after I put my kids to bed.

Those things are real challenges doctors have been facing. And documentation has only increased in its complexity. 

As a doctor, I’d much rather be hanging out with my patients than interacting with the electronic health record.

How have patients and doctors reacted to the new technology?

Our patients, by and large, really liked it. We did have people who did not want to have their visit recorded and felt like that was an intrusion of their privacy — that was definitely a very small number of people.

The vast majority of people really liked it, and what we found is that the patient satisfaction with the encounter went up significantly. Doctors loved it. This was one of the very first solutions that when we put it into practice, the doctors loved it.

They started sending me videos of themselves, telling me how great the product was and attest to the AI product we use.

We realized we are on something really big. This is something that could both improve the documentation as well as give the doctors and patients a better appointment that they really wanted and deserved. It all goes back to that clinical encounter, being able to have that connection between doctor and patient is really, really important.

I think we have, unfortunately, because of the intrusiveness of electronic health records over the last decade or so, we probably eroded that experience a little bit. This is a technology that helps to get it back.

Are there limitations to this new technology?

I think it’s important to not set our expectations high on the initial rollout. We know that language models hallucinate — putting things in the notes that may not reflect the actual conversations.

We’re working with an artificial intelligence network, for a quality and trust framework where we can help evaluate notes — along with our own doctor’s evaluation — to make sure that the transcription we are creating accurately reflects the conversations that are taking place.

Quality and trust for all of our language models and transcription tools are important. Good documentation is how we judge how sick our patients are.

We want to be able to create a tool that standardizes the whole system, that can be exchanged between all doctors within the health system to treat a patient.

We’re working hand-in-hand with DeepScribe using multiple layers of technology to make sure that transcriptions are accurate.

How are hospitals making the new technology secure for patients and their data?

I would say we are hyper-focused on the privacy of patient information — we go above and beyond HIPPA. We make sure that we have secure connections backward and forward.

We have a lot of work in what we call data governance. Data governance is a way of assuring that if you’re giving your data to a software company, that that software is being responsible with that data.

Those databases are some of the most secure cloud databases in the world.

Security and safety is something we will continue to have to reaffirm. We can’t ignore the benefits of this technology in the health care space. Our patients are aware at every step of the journey. We will give them options if they want or not want to be recorded.

Overall, we are working through this as a better platform for health care delivery and trying to make sure patients and doctors alike understand that the benefits are well worth the risks.