Generative AI in Healthcare: Current and Future Applications

Summary

Dr. Eric Bricker's video discusses the current applications of generative AI in healthcare, emphasizing its utility in documentation, reducing physician workload by creating notes through ambient listening devices. He also highlights AI's role in streamlining prior authorizations with automated letters to health insurance companies and its potential in searching and summarizing complex insurance documents and medical records, with tools like Unriddle AI and Google's partnership with Mayo Clinic, enhancing efficiency in patient care.

Takeaways

• 😀 Generative AI is currently being used in healthcare to improve efficiency and support professionals in documentation and other tasks.
• 📝 Documentation is a significant burden for healthcare professionals, with 62% of doctors' time spent on electronic medical records rather than patient care.
• 🎤 Ambient listening devices combined with AI can transcribe doctor-patient conversations into medical notes, saving doctors time and improving workflow.
• 🏥 Companies like Deep Scribe are already implementing AI for medical note transcription, saving up to three hours a day for doctors.
• 📄 Prior authorizations for medical procedures can be streamlined using AI to generate letters to insurance companies, saving time and effort.
• 💡 AI can be used to search through dense documents like insurance plans and medical policies, making information more accessible and understandable.
• 🔍 Unriddle AI is a tool that can read and summarize PDFs or answer specific queries about documents, useful for understanding insurance coverage.
• 🤖 AI's ability to search medical records can assist clinicians in quickly understanding patient histories, especially in complex cases or emergency situations.
• 🤝 Partnerships like the one between the Mayo Clinic and Google are exploring how AI can analyze and summarize vast amounts of patient data.
• 🚑 In emergency rooms, AI could potentially provide quick summaries of patient medical records, aiding in faster and more accurate diagnoses.
• ⚠️ While AI offers many benefits, it's important to note its limitations, such as the potential for inaccuracies and the need for fact-checking, especially with information post-2021.

Q & A

What is the main topic of Dr. Eric Bricker's training video?

The main topic of Dr. Eric Bricker's training video is the application of generative AI in healthcare, focusing on current uses rather than theoretical future applications.

What percentage of doctors' time is spent on electronic medical records according to the video?

According to the video, 62 percent of doctors' time is spent on electronic medical records rather than with patients.

How does generative AI assist with documentation in healthcare as mentioned by Dr. Bricker?

Generative AI assists with documentation by using ambient listening devices that listen to the conversation between healthcare professionals and patients, then create notes in the style and format that physicians use, which can save doctors significant time.

Which company is mentioned in the video that is currently using AI for documentation in healthcare?

DeepScribe is the company mentioned in the video that is currently using AI for documentation in healthcare.

What is the potential time saved for doctors by using AI for documentation as stated in the video?

The video states that using AI for documentation can save doctors three hours a day, which is a significant improvement over traditional documentation methods.

What is the role of AI in prior authorizations as described in the video?

AI can assist in prior authorizations by generating letters to health insurance companies, justifying the need for specific medical procedures, and referencing supporting scientific literature, which can save doctors time and effort.

What limitations does Dr. Bricker mention regarding the use of AI for prior authorizations?

Dr. Bricker mentions that AI's knowledge is limited to pre-2021 data, and it may not include newer studies or articles. Additionally, there have been instances where AI generated incorrect references, so human verification is necessary.

What is Unriddle AI, and how can it be used in healthcare?

Unriddle AI is a software tool that can read and interpret PDF documents, such as insurance policies and medical policies. Users can ask it questions or request summaries, making it easier to understand complex documents and find specific information quickly.

What is the significance of the partnership between the Mayo Clinic and Google mentioned in the video?

The partnership between the Mayo Clinic and Google is significant because it involves using Google's generative AI to analyze de-identified patient data, which could greatly improve the efficiency and effectiveness of searching through medical records.

How can AI improve the process of searching medical records, as discussed in the video?

AI can improve the process by quickly summarizing patient states, finding specific events or conditions within medical records, and providing answers to queries, which can be particularly helpful in emergency situations where time is critical.

What is the slang term used by clinicians for reviewing complex patient medical records, as mentioned in the video?

The slang term used by clinicians for reviewing complex patient medical records is 'chart biopsy'.

Outlines

00:00

🚀 Generative AI in Healthcare: Applications and Benefits

Dr. Eric Bricker introduces the topic of generative AI in healthcare, emphasizing current applications rather than theoretical possibilities. He highlights the significant time doctors spend on documentation and how AI can alleviate this burden. Ambient listening devices with voice recognition and natural language processing can generate medical notes in real-time, saving doctors hours daily. Companies like Deep Scribe are already implementing this technology, which not only improves efficiency but also reduces costs compared to human scribes.

05:03

📄 AI Assistance in Prior Authorizations and Document Search

The second paragraph discusses the utility of AI in streamlining the prior authorization process, which can be time-consuming for doctors. AI can draft letters to healthcare providers, referencing scientific literature to support requests for procedures. However, Bricker notes the limitations of AI, such as the inability to access information post-2021 and the potential for inaccuracies, as seen in a case where AI generated non-existent sources. He also introduces Unriddle AI, a tool that can search and summarize PDF documents, which is particularly useful for understanding complex insurance policies and medical records, as demonstrated by its application in the Mayo Clinic's partnership with Google.

10:04

🔍 Enhancing Medical Record Review with AI

In the final paragraph, Bricker focuses on the potential of AI to improve the review of medical records, especially for complex cases with extensive documentation. AI can quickly summarize patient histories or locate specific events within medical records, which is invaluable in time-sensitive situations like emergency rooms. This capability can significantly enhance the efficiency and accuracy of medical professionals in understanding patient backgrounds and making informed decisions.

Keywords

💡Generative AI

Generative AI refers to artificial intelligence systems that can create new content, such as text, images, or audio, based on learned patterns. In the video, Dr. Eric Bricker discusses how generative AI is applied in healthcare, particularly for documentation and note creation, which is a key theme of the video.

💡Healthcare Documentation

Healthcare documentation is the process of recording patient information, medical history, and treatment details. It is crucial for legal, billing, and communication purposes among healthcare professionals. The script highlights how generative AI can automate this process, saving doctors significant time.

💡Electronic Medical Record (EMR)

An Electronic Medical Record is a digital version of a paper chart that contains all of a patient's medical history, test results, and treatment information. The script points out that doctors spend a significant amount of time interacting with EMRs rather than with patients, emphasizing the need for AI to streamline documentation.

💡Ambient Listening Devices

Ambient listening devices are tools that can capture and process audio in a room, often used for transcription or monitoring purposes. In the context of the video, these devices are used in conjunction with AI to transcribe doctor-patient conversations into medical notes.

💡Natural Language Processing (NLP)

Natural Language Processing is a branch of AI that focuses on the interaction between computers and human language. The script mentions NLP as a technology that allows AI to understand and differentiate between the voices of doctors and patients during a visit, enabling accurate note generation.

💡Deep Scribe

Deep Scribe is a company mentioned in the script that uses AI to automate medical note transcription. It represents a real-world application of generative AI in healthcare, saving doctors time and improving efficiency.

💡Prior Authorizations

Prior authorizations are required by insurance companies before certain medical treatments or procedures can be performed. The script discusses how AI can assist in generating these authorization requests, simplifying a time-consuming process for doctors.

💡Doximity

Doximity is a professional network for healthcare professionals that, according to the script, has a beta version of AI technology to assist with prior authorizations. It exemplifies another practical application of AI in streamlining healthcare administrative tasks.

💡Unriddle AI

Unriddle AI is a software tool mentioned in the script that can read and analyze PDF documents, answering queries or summarizing content. It is used as an example of how AI can help interpret complex healthcare policy documents, making them more accessible.

💡Medical Policy

Medical policy refers to the guidelines set by insurance companies that determine coverage for specific medical procedures or treatments. The script explains how AI can be used to search and understand these policies, aiding healthcare professionals in navigating complex insurance coverage decisions.

💡Chart Biopsy

A chart biopsy is a colloquial term for the process of reviewing a patient's medical records in detail. The script suggests that AI could automate and improve this process, particularly in emergency situations where time is critical.

Highlights

Introduction to generative AI in healthcare by Dr. Eric Bricker.

Focus on current applications of AI in healthcare rather than theoretical possibilities.

AI's role in reducing documentation time for healthcare professionals.

62% of doctors' time spent on electronic medical records instead of patient care.

Ambient listening devices and generative AI for creating medical notes in real-time.

DeepScribe's implementation of AI in medical note creation at Stanford and Texas Medical Center.

AI's potential to save doctors three hours a day by automating note-taking.

The use of AI for prior authorizations in healthcare.

AI's ability to write clinical justification letters for medical procedures.

Doximity's beta version for AI-assisted prior authorizations.

Limitations of AI in accessing post-2021 medical literature and standards of care.

The importance of fact-checking AI-generated references and clinical justifications.

AI's application in searching and summarizing insurance policy documents.

Unriddle AI's capability to read and answer queries from PDF documents.

Free and premium tiers of Unriddle AI for document analysis and query responses.

AI's potential in improving medical record search efficiency in complex cases.

Partnership between Mayo Clinic and Google to utilize AI for analyzing de-identified patient data.

AI's role in enhancing the speed and accuracy of medical chart reviews in emergency situations.

Transcripts

hello good morning everyone this is Dr

Eric Bricker and thank you for watching

a healthcare Z today's training topic is

generative AI in healthcare so

so let's spend the rest of this

conversation talking about Healthcare

applications for AI and I wanted to

focus on applications that are literally

being done like right now or starting to

be done right now I mean you could

probably think of all sorts of Pie in

the Sky things that you can do so this

is not meant to be an exhaustive list

but these are things that are being done

right now that in my opinion are

incredibly helpful uh in healthcare so

one is documentation okay so there is

gods and gobs of documentation that

Healthcare professionals have to do in

the doctor's office in the hospital you

name it so the doctors says the

occupational therapist the respiratory

therapist amen it's like you just spent

so much time

reputation it's like

unreal so with that

62 percent of doctors time is spent with

the electronic medical record not with

the patient Okay so

there actually is today a way for the

generative AI so like you've got the

note right the note has to communicate

you know it's a legal record it has to

communicate to other Healthcare

professionals or other doctors like

what's actually transpired during the

office visit

um it can be shared with the patient so

that the patient can see it's used for

billing purposes to justify the billing

that's performed now

today like literally

We compare ambient listening devices so

like I can just I Dr Bricker can have a

a visit with Joe and Jane and that you

can have just a microphone

sort of like an Amazon Echo in the room

and it can just like listen to the

conversation and then has voice

recognition and natural language

processing so it knows that like my

voices the doctor's voice and that the

patient's voices the patient and then it

can use generative AI because that has

learned how to create a note in the

style and the format that Physicians

create notes it has to have like a

history has to have a chief complaint it

has to have a history it has different

sections his chief complaint history of

present illness past medical history

family history social history

Etc and it can just listen to the office

visit and it can create the note so you

don't have to spend all this all this

time writing the note how incredibly

helpful is that that is like the coolest

thing ever now

this is done today deep scribe is a

company that literally does this today

they're doing it at Stanford and at the

Texas Medical Center down in Houston now

I have no connection with deep scribe

but it literally saves

three hours a day day and it only cost 1

6 the cost of a human scribe I mean

there are some Physicians that use

scribes like literally a person all they

do is is come with a laptop and they

just sit there and type the note so that

the doctor doesn't have to do that and

so

you can you can save three hours a day

for a doctor I mean that is unreal okay

next up prior authorizations this is one

of the first things that came out

um when

Chan gbt came out is that a doctor in

Florida who's a rheumatologist said huh

what if I'd put in a prompt of write a

letter to United Healthcare asking them

to approve an echocardiogram on a

patient with systemic sclerosis which is

a aromatic disease that causes your skin

and your soft tissues to harden

and make reference to supporting

scientific literature and list the

appropriate articles and guess what it

did it in like a second

so doctors today have to do this by hand

they've got to write a letter dear

United Healthcare I would like to have

an echocardiogram for John or Jane Doe

they've got systemic sclerosis this is

why I wanted to be done here's why

echocardiograms are clinically Justified

here's some references I mean it takes a

ton of time and

he just said hey could you just do this

for me and then of course everything in

healthcare has to be done by facts so we

take this amazing new technology and

then we have to transmit it to the

health insurance company via fax machine

but the point is is that it can do it in

fact doximity has a beta version of this

that they are using right now uh and

again I have no connection to doximity

now

it's important to note that there are

certain limitations here so like chat

GPT is not actively connected to the

internet in fact chat GPT was

pre-trained on the internet

pre-2021 so there's a lot of like

standards of care and a lot of things in

the medical literature and a lot of

things that can be justified for prior

authorization where information as of

pre-2021 is totally okay but just note

that if there are like newer studies

that come out or newer articles or what

have you if it's from 2021 or later it's

not going to show up because chat gbt

was not trained on that now

um the rheumatologist in Florida also

said look when you see those references

you also need to double check them as

well

um so that

um there was actually a case of an

attorney

um that used chat GPT and apparently the

um AI actually even made up this made up

the sources uh it made up the cases

um these cases did not exist so

certainly it's not perfect so you're

going to want to um but but when he read

the clinical justification for like why

it should be done he's like oh that's

totally right

now another thing that AI can be used

for

is to search and this is a personal

favorite of mine is to search

and plan Dot

so

everybody so a lot of people here work

in health and

and like the planned documents I mean it

I mean it's obviously all in PDFs these

days but you know back when they used to

print them out I mean they're like this

thick and then also

um and that describes like what your

insurance covers and what it doesn't

cover and how it covers it like what's

your deductible what's your coinsurance

like how does it cover durable medical

uh equipment and it's got gobs and gobs

of fine print in there now there's also

medical policy which is even more layers

of fine print which is hey your health

insurance plan will pay for this medical

procedure in this situation but it won't

pay for the same medical procedure in a

different situation and I've uh I have a

whole nother uh training session on

medical policy and its source of denials

and it's incredibly long and complicated

me as a physician I have a super hard

time understanding this so like your

typical like a lay person or plan member

they would have no clue how to read and

interpret this plus you'd have to go

through these pages and pages and pages

that are not easily searchable to find

the information that you need but

you can have ai

you can just ask AI a question about a

document and it'll go through the

document and it'll answer that question

for you and there is a software tool

that's called Unriddle AI that reads

PDFs and you can just type in queries or

you can just ask it to summarize it for

you and you can use it for free for a

document up to 120 pages and you can do

four queries a month for free now if you

want to upgrade it you can do up to four

thousand page documents and you can do

250 queries a month for fifty dollars a

month again I have no connection to

under Unriddle AI but if you were

working as a let's say an account

manager as a brokerage firm and let's

say your client is like hey all right

you know I've got or even if you work in

benefits in HR and you've got an

employee that's like hey you know

can I do you know my out my health

insurance is saying it doesn't cover

this or whatever you're like I don't

even know if our plan does or does not

cover that instead of you having to

Rifle through the planned document to

the medical policy you could just use

Unriddle AI to be like hey for our Blue

Cross United policy does it cover you

know echocardiograms for 17 year olds

who had one episode of palpitations

while playing soccer and like it'll

shoot out an answer for you

um now interestingly

um it can also search medical records

and so there's actually a partnership

between the Mayo Clinic and Google that

was just announced on June 7th of this

year 2023 where uh Mayo is going to take

you know de-identified patient data and

they're going to allow Google's

um generative AI to then look through

those medical records and that would be

extremely helpful as well because

doctors and you know clinicians we have

to do what's we call it um sort of um

the slang term for it is a chart biopsy

where you know you've got these complex

patients that have gobs of gobs of

previous visits or their Hospital stay

might already be a month long and you're

just coming on to be consulted as the

cardiologist or what have you and you're

like oh man I gotta spend all this time

looking through their medical record

whereas literally you could just use the

AI to be like hey could you summarize

their state for me or could you tell me

specifically what happened with this

particular like I know the guy had had a

cardiac calf but can you just find that

in there and let me know what happened

in the cardiac half so the effectiveness

and the accuracy of the chart biopsy of

actually looking through those charts

could be tremendously better especially

in the ER right because in the ER you

don't have time to look through all this

stuff right so the patients would come

into the ER and the ER doctor would look

at their medical record that I was

already in the system and they're like

oh man I can't keep this straight

there's no way I can figure all this out

it's an emergency whereas they could

actually just use the AI to like very

quickly answer like what they would need

to know from the huge medical record I

mean that would be unbelievable