Beyond the Algorithm: Technology, Trust & Patient Care, part 1 | Ep 82

WEBVTT
00:00:00.880 --> 00:00:03.600
Yeah, welcome back, everybody, to the Valor Flight Crew Podcast.
00:00:03.680 --> 00:00:05.839
I'm Blake Schreider here with Mike Fidelity.
00:00:05.919 --> 00:00:06.480
We're your hosts.
00:00:06.559 --> 00:00:07.759
Mike, how you doing?
00:00:08.080 --> 00:00:09.279
I'm tired.
00:00:10.480 --> 00:00:12.480
It's been a long few weeks, buddy.
00:00:12.720 --> 00:00:13.919
I am definitely tired.
00:00:14.160 --> 00:00:15.839
I'm sorry for the for the long face today.
00:00:15.919 --> 00:00:17.120
I'm recovering from the weekend.
00:00:17.280 --> 00:00:22.879
And no, it was not a drinking weekend, but uh still had a good time anyway, uh playing with event-dependent kids.
00:00:22.960 --> 00:00:26.239
So had a great time, but I'm I'm exhausted.
00:00:26.399 --> 00:00:29.920
Definitely ready for a good night's sleep in my own bed tonight.
00:00:30.239 --> 00:00:37.679
Man, well, you know, I can I did not do anything nearly as fun, um uh but equally as tired.
00:00:37.840 --> 00:00:41.679
However, a conversation today, though, I uh I'm pretty excited.
00:00:41.759 --> 00:00:42.880
We get to nerd out.
00:00:43.119 --> 00:00:47.679
I I don't know if I've ever nerded out in my life, but I'm I'm about to and I'm looking forward to it.
00:00:48.960 --> 00:00:50.880
This is gonna be uh a great conversation.
00:00:50.960 --> 00:00:57.600
We're gonna be talking uh with our dear friend Christian Greengrass, who's been such a long-standing member part of Echo.
00:00:58.159 --> 00:01:01.039
And Christian has worn many hats for us.
00:01:01.200 --> 00:01:07.680
But uh, and I think uh Christian, I think this is your fourth, fourth time on our on our podcast.
00:01:08.000 --> 00:01:12.000
Future flight crew, future flight leaders.
00:01:12.480 --> 00:01:12.879
Yeah.
00:01:13.120 --> 00:01:13.519
Education.
00:01:13.840 --> 00:01:15.359
So am I gonna get like flyer miles for this?
00:01:16.159 --> 00:01:17.760
Is this gonna go towards my American Express?
00:01:18.159 --> 00:01:22.159
I think our partners might have uh you know start to think that you're one of the co-hosts here soon.
00:01:22.319 --> 00:01:23.599
So we've got to be a little bit careful.
00:01:24.560 --> 00:01:27.040
We're not gonna be able to invite you back for a little while.
00:01:28.239 --> 00:01:31.760
You know, probably not a bad thing because you know you gotta take me in doses.
00:01:31.920 --> 00:01:33.120
You know, that's what people do.
00:01:33.200 --> 00:01:36.239
You take me in doses, why the conference is great, come once a year.
00:01:36.640 --> 00:01:40.079
You know, it's like, oh yeah, I'm worn out, I'll see him next year, you know.
00:01:40.560 --> 00:01:45.200
We would never say that about you because one of my favorite people.
00:01:45.599 --> 00:01:52.400
But I think one of the best topics to have such a well-versed expert in.
00:01:52.560 --> 00:01:56.239
Uh, today we're gonna be talking about all things artificial intelligence.
00:01:56.319 --> 00:01:57.920
So we're gonna talk about AI.
00:01:58.239 --> 00:02:06.719
And and Mike, when we when I kind of pitched this idea, I know we were kind of like, well, what could we do with this?
00:02:06.959 --> 00:02:18.159
For me, I just really wanted to see how what this is gonna look like for us in healthcare particularly, and maybe more on the side of the hems world and how it's gonna impact us.
00:02:18.719 --> 00:02:19.439
It's everywhere.
00:02:19.599 --> 00:02:26.960
It's everywhere we look, it's built into, I mean, now everything is advertised using AI, every web browser, your phone.
00:02:27.280 --> 00:02:28.000
It's everywhere.
00:02:28.080 --> 00:02:31.599
And I don't know where this is gonna take us, especially in healthcare.
00:02:31.680 --> 00:02:39.840
And to be honest, I'm kind of nervous because I it's the unknown, I think, that that kind of makes me feel a little uneasy.
00:02:40.639 --> 00:02:41.840
Yeah, I think so.
00:02:42.000 --> 00:03:00.159
Uh and so now, Christian, I don't know if you're aware of this or not, but I'm gonna bring in on a little secret we have here as far as whenever we've had other podcasts, there may or may not have been a drinking game, even though we weren't talking about AI, that every time you mentioned AI, we we may have had to take a shot.
00:03:00.560 --> 00:03:04.479
I I know this game because I was playing too.
00:03:04.719 --> 00:03:08.240
Just to be fair, I flask under the table here.
00:03:08.319 --> 00:03:11.520
Like I was, you know, just like, oh yeah, I just need to look at something real quick.
00:03:11.599 --> 00:03:15.120
Yeah, I was just saying it just to make you drink more.
00:03:15.360 --> 00:03:25.039
You know, I and truly, Christian, you are the person that introduced me to some of the abilities of AI at one of the conferences and being able to sit down and talk about that.
00:03:25.439 --> 00:03:32.400
And I've always been one that has resisted the AI, you know, um frontier.
00:03:32.719 --> 00:03:35.520
And it's shall we call it that?
00:03:35.599 --> 00:03:40.319
But you know, we made the joke for Mike to just bust out his typewriter to bring what he's gonna do.
00:03:40.639 --> 00:03:41.199
100%.
00:03:41.919 --> 00:03:42.240
100%.
00:03:42.719 --> 00:03:46.000
You really should have like phenomenal.
00:03:46.560 --> 00:03:56.240
Now, now I I literally use it every single day, and I'm seeing it more and more integrated in all the things that are around us and every single thing that we do.
00:03:56.560 --> 00:03:59.840
And I'm kind of getting excited about it.
00:03:59.919 --> 00:04:10.960
I'm interested in seeing where this goes, and and I'm really looking forward to this discussion today because it's only a matter of time, and I've I've stopped resisting at this point.
00:04:11.120 --> 00:04:14.240
I've I've recognized that there is no way around this.
00:04:14.400 --> 00:04:16.079
It's happening whether I like it or not.
00:04:16.240 --> 00:04:27.040
So I just need to embrace it and I just need to figure out how it's going to best work for me, my program, my care, and how it can advance the services that we provide.
00:04:27.199 --> 00:04:39.920
And and I think it's really uh going to be a great discussion about that because you have such a broad insight into where we could potentially go with this with your knowledge of AI.
00:04:40.000 --> 00:04:46.720
And so uh I that's what I think is gonna be so cool to see how much of what you predict is going to come true and when.
00:04:47.199 --> 00:04:49.120
Well, that's the interesting part, Mike.
00:04:49.279 --> 00:04:52.560
It we really don't know what the next five to ten years look like.
00:04:52.639 --> 00:04:56.639
You know, I was looking at the stuff before the podcast, it was one of the things mentioned.
00:04:56.800 --> 00:05:01.040
It we really don't know because it it really is gonna depend on what we do now.
00:05:01.279 --> 00:05:04.800
Now, I don't consider myself to be the most technical of experts.
00:05:04.959 --> 00:05:11.120
Like, like I'm not in the models looking at the internal stuff, the calculus, the the integral stuff.
00:05:11.279 --> 00:05:12.720
I understand the basics of it.
00:05:12.800 --> 00:05:16.560
I definitely consider myself to be more of a of an AI strategist.
00:05:16.800 --> 00:05:32.639
And and really, more or less, I'm someone who I I want to prepare us for taking on these tools and deciding when a tool is right and when we maybe we should use something else first before we graduate to using the these AI tools.
00:05:32.800 --> 00:05:43.040
I think a lot of leaders and a lot of clinicians alike we like these things, and it's like, well, we want AI, and that's cool, but what are we looking at that comes out?
00:05:43.199 --> 00:05:44.319
It has to come first.
00:05:44.480 --> 00:05:54.639
There are things that we're gonna have to do first, and of course, I know we'll talk about this more, but I I really think that I like how you both exercise a cautious optimism.
00:05:55.040 --> 00:05:57.279
I think that's balanced, and I think that's a good way to put it.
00:05:57.439 --> 00:05:58.319
Good way to put it, yes.
00:05:58.480 --> 00:05:58.959
Yeah.
00:05:59.439 --> 00:06:22.000
Christian, can you for people that haven't listened to this uh podcast before, can you just give uh a little introduction about yourself, Christian, and kind of your background and your flight experience and how that's kind of helped you get into some of these roles where you've used AI for professional, personal, and kind of how that's led you to to kind of how we're having this conversation today?
00:06:22.399 --> 00:06:22.959
Yeah, sure.
00:06:23.120 --> 00:06:28.480
I mean, so I've been an EMS uh for about 17 years this year, unfortunately.
00:06:28.560 --> 00:06:29.680
It makes me feel very old.
00:06:29.839 --> 00:06:32.399
Um, the paramedosaurus.
00:06:32.480 --> 00:06:33.120
I like it.
00:06:33.360 --> 00:06:33.600
Yeah.
00:06:34.000 --> 00:06:36.800
Yeah, I was a paramedic first, and my then my wife.
00:06:36.959 --> 00:06:42.560
Um I've been EMS for a while, but my wife, bless her, she got me into the nursing world.
00:06:42.639 --> 00:06:43.759
I I fell in love with that.
00:06:43.920 --> 00:06:49.920
And of course, you know, I was working in ICU nursing and and of course jumped right back into EMS on the air medical side.
00:06:50.000 --> 00:06:55.360
And um I this is I right at my sixth year of being in the air medical profession.
00:06:55.519 --> 00:06:56.800
I've worked as a flight nurse.
00:06:56.879 --> 00:07:00.319
I've been an educator for a flight program, a new developing flight program.
00:07:00.480 --> 00:07:05.920
I've been a uh data analytics and integration manager for um care flight in Dallas Fort Worth.
00:07:06.000 --> 00:07:06.639
I was just there.
00:07:06.800 --> 00:07:09.360
We just moved back down to Houston to be close to family.
00:07:09.519 --> 00:07:11.360
I work for Memorial Herman Life Flight.
00:07:11.439 --> 00:07:15.120
I'm one of the flight line flight nurses, and let me tell you, best job in the world, right?
00:07:15.360 --> 00:07:16.959
Just flying doing the job.
00:07:17.120 --> 00:07:24.560
Um, but on the side, I've also really have dove into just in my own interest into the data science, into quality.
00:07:24.720 --> 00:07:28.319
I've also taken a dive into AI, which all those things are kind of married.
00:07:28.560 --> 00:07:37.680
I've also owned some small technology companies that we've worked with AI, and I I currently do some work on the side uh for AI, and I also do a little consulting.
00:07:37.759 --> 00:07:42.000
So I do a little bit of that stuff, just kind of all synergistically working together.
00:07:42.160 --> 00:07:51.759
And I really want to help people to understand what these technologies are and what they are not, and then what we need to do to be successful with using them in our care.
00:07:52.160 --> 00:07:54.879
I think you obviously uh we've known you for a long time.
00:07:55.120 --> 00:08:11.120
We know that you have uh quite a bit of experience and uh you're uh uh an amazing educator that we've been blessed to have part of the Echo family for some time now, and really were kind of the catalyst for or the driving force behind creating the Future Flight Leaders program within Echo.
00:08:11.600 --> 00:08:18.800
Uh and I think you've always had kind of a maybe a little bit of a different perspective on the industry.
00:08:18.959 --> 00:08:37.440
And I think that that is something that has been always one of those things that I've loved when I sit down and have a conversation with you because of the perspective that you bring, I think is just one that uh are is oftentimes things I have never really processed in that particular way.
00:08:37.519 --> 00:08:40.080
And I think that's what's so cool about having those conversations.
00:08:40.320 --> 00:09:03.200
And then when you started talking about AI and all these other podcasts and interactions that I had with you and kind of talking and seeing the excitement that you have about the possibilities that are out there, uh and I obviously, you know, I've I've long said I'm one that's been resisting this movement, uh, and have been very, very cautiously optimistic.
00:09:03.519 --> 00:09:10.639
But I I am getting on board with it, but I have to ask the question are we ready for it?
00:09:10.879 --> 00:09:11.200
Right?
00:09:11.360 --> 00:09:19.759
Like, are we as an industry ready to take that step and to integrate this in and and where are we already seeing it and maybe not even realizing it?
00:09:20.159 --> 00:09:21.200
That's a good question.
00:09:21.440 --> 00:09:23.759
I'm gonna say it like this yes and no.
00:09:24.159 --> 00:09:25.840
Do we have a use case for it?
00:09:26.000 --> 00:09:27.840
Yes, we do, very much so.
00:09:28.240 --> 00:09:29.679
Plenty of use cases.
00:09:29.919 --> 00:09:32.879
Think about what you do when you're a clinician on the flight line.
00:09:32.960 --> 00:09:34.240
You're very busy, right?
00:09:34.480 --> 00:09:47.360
You're busy with paperwork, you're busy with processes, you're busy with patient care, you're busy with scheduling, you're busy with looking at the finances, especially Mike, you being in the position you are at your program, uh, positions I've been, we're very, very task saturated.
00:09:47.519 --> 00:09:53.519
Most organizations are quite lean, with the exception of a few that are much larger organizations that that provide services.
00:09:53.759 --> 00:09:58.480
We also have increased billing uh issues that are coming about that are creating issues.
00:09:58.639 --> 00:10:00.879
So there's a business use case for it.
00:10:01.120 --> 00:10:09.120
Now, where I say kind of no is is the organization ready to take on these products?
00:10:09.279 --> 00:10:14.799
Because think about what at the very basis, the basics of what this is.
00:10:14.960 --> 00:10:18.320
If I think about large language models, there's several different types of AI.
00:10:18.480 --> 00:10:22.559
There's Agenic AI, there's RAGs, there's LLMs, there's all kinds of different things.
00:10:23.039 --> 00:10:26.799
LLM, what we commonly use, it's a large language model.
00:10:26.879 --> 00:10:29.840
And pretty much all it's doing is predicting the next word in the phrase.
00:10:29.919 --> 00:10:33.679
So if I was to say Mary had a little, you guys would say what?
00:10:34.000 --> 00:10:34.320
Lamb.
00:10:34.639 --> 00:10:34.960
Lamb.
00:10:35.360 --> 00:10:35.919
There you go.
00:10:36.159 --> 00:10:38.000
Fidelity GPT, there it is.
00:10:38.080 --> 00:10:41.679
You know, like uh we'll go and get you on the, we'll get Sam Altman to sign you on.
00:10:41.919 --> 00:10:43.120
But that's kind of what it does.
00:10:43.919 --> 00:10:44.720
We started here soon.
00:10:44.799 --> 00:10:45.679
Don't worry.
00:10:46.000 --> 00:10:50.559
But for those things to work, you need good context.
00:10:50.799 --> 00:10:53.120
You need an understanding of yourself.
00:10:53.360 --> 00:10:55.039
You need to understand your program.
00:10:55.200 --> 00:10:56.720
You need to understand your mission.
00:10:56.879 --> 00:10:58.480
You need to understand the medicine.
00:10:58.639 --> 00:11:02.480
You need to understand the literature, you need to understand your data, your finances.
00:11:02.639 --> 00:11:07.200
You have to know those things before you start integrating these tools.
00:11:07.519 --> 00:11:13.440
And once again, it may seem like I'm not excited, because I am, but I'm excited when it's done correctly.
00:11:13.759 --> 00:11:23.279
Programs are going to have to start getting on board with understanding their data environment, making those investments, understanding even their charting forms, how they're being built.
00:11:23.440 --> 00:11:27.519
I've been a part of that process several times with with charting software vendors.
00:11:28.000 --> 00:11:36.559
You have to understand how your forms are built, the psychology of how your clinicians document things, the how things actually get transformed on the back end.
00:11:36.720 --> 00:11:40.000
Like a data dictionary is also something I think many of us are not aware of.
00:11:40.080 --> 00:11:45.039
Like if I was to say, for each and every program, what do we defined as a missed flight?
00:11:45.279 --> 00:11:47.600
What do we defined as a lift time?
00:11:47.759 --> 00:11:50.639
What do we defined as a good airway?
00:11:50.879 --> 00:11:55.519
What do we define as a patient that needs a crike or a chest tube?
00:11:55.679 --> 00:11:58.080
There's always going to be a different definition.
00:11:58.399 --> 00:12:03.440
Where we're not ready is we're not ready because many programs don't have these things down pat.
00:12:03.679 --> 00:12:05.360
You have to start documenting these things.
00:12:05.519 --> 00:12:06.480
You have to define them.
00:12:06.639 --> 00:12:08.799
Everyone has to speak on the same terms.
00:12:09.120 --> 00:12:16.960
Because when you're putting data into these systems and you're not providing it good or even conflicting definitions, the outputs are not going to be good.
00:12:17.120 --> 00:12:19.759
Garbage in as uh garbage out, as they say, right?
00:12:20.000 --> 00:12:23.200
That's where I think and also policies too.
00:12:23.440 --> 00:12:31.759
The the clinicians have to be protected because, for example, uh, I've heard stories where crew members are using Chat GPT to write narratives.
00:12:31.919 --> 00:12:36.240
They're using ChatGPT to with patient information, uh, and that's not okay.
00:12:36.399 --> 00:12:38.480
You need to be aware of what a BAA is.
00:12:38.559 --> 00:12:40.639
You have to be aware of those things.
00:12:40.799 --> 00:12:41.759
You have to have good tools.
00:12:41.840 --> 00:12:48.159
You also have to have a good relationship with those people that annoyingly tell you to turn off your computer and turn it back on again.
00:12:48.799 --> 00:12:52.559
Your IT team, you have to be on good relationship terms with them.
00:12:52.639 --> 00:12:53.759
You have to communicate with them.
00:12:54.000 --> 00:12:55.200
So, yeah, is it sexy?
00:12:55.279 --> 00:12:55.840
Is it cool?
00:12:56.000 --> 00:12:56.720
Not always.
00:12:56.879 --> 00:13:02.399
Uh, but the very beginning to get us to yes and yes, we have to get that other part from no to yes.
00:13:02.559 --> 00:13:05.840
And some programs are getting closer to that, and others are not ready for that.
00:13:06.080 --> 00:13:08.000
And that's just kind of the blunt answer.
00:13:08.240 --> 00:13:12.559
Now, to answer the second part of your question, like, where are we using it now?
00:13:12.960 --> 00:13:18.720
So I think in air medical, particularly, I think not many places.
00:13:18.960 --> 00:13:25.679
I think there are certain dispatch CAD systems that are starting to use it to look for closest asset.
00:13:25.919 --> 00:13:30.799
I think the models are getting stronger, as I've read, to the few systems that actually do that.
00:13:31.039 --> 00:13:46.639
I think in radiology, in the in the inpatient setting, that is one of the biggest things that I'm observing now, is that in radiology, they are using AI to assist radiologists on certain reads so they don't miss certain things on scans, MRIs, pet scans, you name it.
00:13:46.799 --> 00:13:48.399
The AI is getting involved with that.
00:13:48.799 --> 00:13:51.120
The big thing is though, is it's not perfect, right?
00:13:51.200 --> 00:13:54.399
And and I mentioned here it assists the radiologist.
00:13:54.879 --> 00:14:05.039
Human in the loop is going to be a massive point of discussion that I think we'll have today is AI will not take over what you're doing and it shouldn't.
00:14:05.200 --> 00:14:09.039
It should assist you with what you're doing, it should take over mundane tasks.
00:14:11.919 --> 00:14:12.799
Data science.
00:14:12.879 --> 00:14:19.759
We're using it for data normalization, business insights, uh, in the more advanced side of things, uh, geospatial.
00:14:19.840 --> 00:14:22.879
Uh Kyle Danielson from he was at Airlift Northwest there for a bit.
00:14:22.960 --> 00:14:24.639
He's a good person to talk to about that.
00:14:25.039 --> 00:14:28.639
Geospatial, being able to find like, hey, where's the next base gonna go?
00:14:28.960 --> 00:14:30.399
Where are we gonna put our next base?
00:14:30.639 --> 00:14:32.480
And is this what's best for the community?
00:14:32.559 --> 00:14:34.159
And what kind of asset do we need there?
00:14:34.240 --> 00:14:35.759
And what kind of people do we need there?
00:14:36.080 --> 00:14:36.720
That's important.
00:14:36.960 --> 00:14:37.519
Research.
00:14:37.679 --> 00:14:51.279
I do literature reviews with with certain AIs constantly because instead of having to look up keywords in a library, I can pull a hundred papers in a matter of minutes, and then I can actually read the papers.
00:14:51.360 --> 00:14:54.960
Now, what I'm not allowing it to do is read the paper for me.
00:14:55.279 --> 00:15:02.000
I'm gonna use this, my brain, to read it and make those inferences from that paper and go, yeah, this study's not good.
00:15:02.080 --> 00:15:06.480
And here's the reasons why I'm gonna use this paper over this paper because of this context, right?
00:15:06.720 --> 00:15:13.120
Clinical documentation is also uh, I know you this is also on the list for us to talk about later, but clinical documentation is gonna be huge.
00:15:13.200 --> 00:15:24.639
And of course, administratively, even things like the IDR process for out of networks claims for the air medical stuff that's going on, no surprise, uh like NSB and all those other companies that are out there doing that stuff.
00:15:24.720 --> 00:15:30.000
They're they're doing well with it and they're helping with the claims to improve reimbursement so that we can continue our mission.
00:15:30.080 --> 00:15:32.559
But um, that's just a very small list.
00:15:32.799 --> 00:15:34.000
It's a lot, right?
00:15:34.320 --> 00:15:38.320
But very small, very small list of what's actually uh going on out there.
00:15:38.480 --> 00:15:40.240
So I hope that answers your question, Mike.
00:15:40.559 --> 00:15:43.679
I think it just generates even more questions, to be honest.
00:15:43.840 --> 00:15:48.399
And so because yeah, like how long have we talked about the human factor, right?
00:15:48.480 --> 00:15:55.679
So if we kind of relate to like, is healthcare prepared for how quickly AI is is evolving and growing, right?
00:15:55.840 --> 00:16:02.080
So if AI is dependent on what we feed it, right?
00:16:02.159 --> 00:16:05.600
You said the context, the information, the the things that go into it.
00:16:07.120 --> 00:16:09.440
And then we also define the human factor.
00:16:09.519 --> 00:16:26.639
And I think we we've found in a lot of cases that the human factor was actually a causation of maybe something that happened, you know, it doesn't have to be super bad, it can be super bad, but maybe different aspects, the human factor was maybe something in it.
00:16:26.879 --> 00:16:39.919
So where do we roll that balance where like we know that uh a mechanical ventilator as long as you is dependent on the information that the human gives it?
00:16:40.080 --> 00:16:47.600
But the information you give it, that ventilator is going to be hands down better at ventilating that patient than I would be on my own, right?
00:16:48.399 --> 00:17:10.319
And so when we look at AI, how do we balance that shift of will the machine learning or this learning process will evolve to be almost as good as a human, or will it be better than humans, or will a human still always hold that hierarchy?
00:17:10.559 --> 00:17:12.480
I'll still be one above the machine.
00:17:12.640 --> 00:17:15.440
Like, how is that gonna evolve for us in healthcare?
00:17:15.839 --> 00:17:18.319
I love what you're the direction you're heading in here.
00:17:18.480 --> 00:17:20.160
So there's a few things to unpack.
00:17:20.240 --> 00:17:20.559
Okay.
00:17:20.799 --> 00:17:28.640
So the first things first, it let's say a very, very, you know, we all know the ventilator that we all commonly use in our transport field.
00:17:28.720 --> 00:17:30.319
It gives us a lot of feedback, right?
00:17:30.559 --> 00:17:32.559
It's gonna tell us a whole lot of important things.
00:17:32.640 --> 00:17:42.960
So if I see like the C steps out of compliance, if I see the RCE XP, I look at the text battory time consum, look at the minute volume ventilation, it can use data to give you some information.
00:17:43.039 --> 00:17:45.680
And I'm sure that they're that company's probably working on some things.
00:17:45.759 --> 00:17:46.559
I'm sure I'm not sure.
00:17:46.720 --> 00:17:48.079
I'm just guessing here, okay.
00:17:48.319 --> 00:17:48.559
Right.
00:17:48.720 --> 00:17:52.960
But one thing it can't do, it can't look at your patient on the stretcher.
00:17:53.599 --> 00:17:55.200
It can't look at the monitor.
00:17:55.440 --> 00:18:03.119
It can't look, I mean, and maybe in the future there'll be interoperability through what's called APIs, which are data connections between different softwares, how they communicate with each other.
00:18:03.279 --> 00:18:13.680
Maybe in the future there will be things that will communicate with your transport zone monitor, what have you, and it will say, like, hey, your SATs are this, you know, this is what I'm getting on the vent.
00:18:13.920 --> 00:18:15.920
You should do something about this, right?
00:18:16.240 --> 00:18:19.920
But your physical assessment, your visual assessment will never go away.
00:18:20.079 --> 00:18:22.160
Your critical thinking will never go away.
00:18:22.400 --> 00:18:24.480
The human touch will never go away.
00:18:24.720 --> 00:18:30.880
And so what these things are gonna do for you in the future is it's gonna feed you more concise information.
00:18:31.119 --> 00:18:45.680
So instead of me getting 50 pieces of information and converging it to one decision, it's gonna be maybe 25 or even 10 pieces of information to converge that decision, which means my decision point shrinks in time, which means I'm more efficient with my time.
00:18:45.839 --> 00:18:49.039
In other words, I'm faster with doing life-saving things.
00:18:49.200 --> 00:18:53.359
We don't want to take away the human factor because as we know, AI hallucinates.
00:18:53.440 --> 00:18:56.319
Uh, and obviously it's getting less and less and less and less, right?
00:18:56.799 --> 00:18:57.680
Oh, it totally does.
00:18:57.759 --> 00:19:04.720
Uh, you think that thing is eating mushrooms or something on the side, but um, with some of the outputs of gun, I'm like, whoa, well, that's that's not true.
00:19:04.960 --> 00:19:06.559
Um, are you my four-year-old son?
00:19:06.640 --> 00:19:09.839
Are you just telling me a lie here that the sky's purple?
00:19:10.319 --> 00:19:12.160
But that that's to one part of it.
00:19:12.319 --> 00:19:20.640
But as for how can we, you know, human decisions that make end quote bad decisions or maybe a clinical decision that maybe wasn't the best, right?
00:19:20.799 --> 00:19:22.799
And that's where good CQI comes in.
00:19:23.039 --> 00:19:24.880
CQI creates data.
00:19:24.960 --> 00:19:27.359
Um, there are some charting systems that do really good job of this.
00:19:27.519 --> 00:19:39.039
In fact, I I work on the back end of one that does a very good job of this, has rules that are not AI that funnel things to people to help look at that chart and go, hey, we need to do something about this.
00:19:39.119 --> 00:19:45.680
Let's do a root cause analysis and let's go ahead and make some good decisions to help improve patient outcomes in the future.
00:19:45.920 --> 00:19:50.240
And so where I think the AI will help with that human factor is not when you're on the call.
00:19:50.319 --> 00:19:51.759
It's gonna be beforehand.
00:19:51.920 --> 00:19:55.839
It's gonna be when you're in in the classroom, when you're at the base.
00:19:56.000 --> 00:20:01.680
Like, imagine this being able to sit on your phone at the base, which most of us do.
00:20:01.920 --> 00:20:03.279
We're all addicted to our phones.
00:20:03.359 --> 00:20:06.799
Unfortunately, we're all dopamine depleted um individuals, right?
00:20:06.960 --> 00:20:07.119
Yep.
00:20:07.279 --> 00:20:12.400
And being able to just have like a chat bot or something that you can go, hey, you know what?
00:20:12.480 --> 00:20:17.680
I was thinking about it hadn't had a bad Impella page, like bad sit crashing Impella in a while.
00:20:17.839 --> 00:20:20.079
Throw me a case, give me, give me this.
00:20:20.240 --> 00:20:21.839
Uh, I want to practice with this.
00:20:22.000 --> 00:20:39.599
And then you and that machine learning model or rag or what have you, or a combination of both, will allow you to have that simulation ongoing so that you can work through those problems, or even having educators go, hey, we've had a rash of these calls or these things have happened.
00:20:39.920 --> 00:20:54.160
Let's work together, let's create some scenarios with our LLM friend here, and then let's create some scenarios in which the team can work through uh so that when they encounter this in real life, that they're gonna be a lot better prepared.
00:20:54.319 --> 00:20:58.799
And and that and and just having custom simulations, you know how hard it is to write a sim.
00:20:59.119 --> 00:21:05.279
I've been there having to write SIMs for uh your quarterly education, what have you, it's a pain in the butt, right?
00:21:05.359 --> 00:21:07.519
And then you're just like, well, how do I make this thing accurate?
00:21:07.759 --> 00:21:16.960
But imagine the future where data from actual calls and CQI can feed into a model or a rag as we call it.
00:21:17.039 --> 00:21:24.319
So a rag, so you guys are aware, is like if I give my protocols to this AI, for lack of better terms, and say, you only know this.
00:21:24.480 --> 00:21:25.839
This is the context you have.
00:21:26.000 --> 00:21:27.839
Do not be pulling stuff from outside.
00:21:28.000 --> 00:21:29.519
That's kind of what I'm referring to.
00:21:29.680 --> 00:21:31.759
So if you hear me say rag, that's what I'm saying.
00:21:32.000 --> 00:21:41.359
Whereas the LLM can be from lots of different sources, so like book, internet books, internet magazines, archives, you name it, documents, uh, and just Google searches, what have you.
00:21:41.440 --> 00:21:42.400
You can pull that for the LLM.
00:21:42.480 --> 00:21:44.799
So if you hear me say those, that's that's what that means.
00:21:45.039 --> 00:21:52.079
But imagine being able to feed a rag of some kind, like it to an agent AI where it's like, hey, here's all the calls we've had.
00:21:52.160 --> 00:21:53.599
It's all being fed in here.
00:21:53.759 --> 00:21:56.480
Here's our guidelines, here's our CQI.
00:21:56.640 --> 00:21:59.759
Okay, based on the data you have, let's generate 50 cine.
00:22:00.160 --> 00:22:03.359
Scenarios on the calls that we have not been performing the best on.
00:22:03.519 --> 00:22:06.240
So that we can where's our biggest place of improvement?
00:22:06.400 --> 00:22:12.079
And so instead of having to look through hundreds of hours of all these things to find that out, you can find out in minutes.
00:22:12.319 --> 00:22:19.920
And then once you find that out, then you can educate your team so that when those decisions come up, Blake, you're not making a poor decision, right?
00:22:20.079 --> 00:22:24.160
You're going to make as a human a better decision with the data you've been given.
00:22:24.319 --> 00:22:27.200
And so once again, long-winded answer, right?
00:22:27.440 --> 00:22:29.680
But that that's that's the reality of it.
00:22:29.759 --> 00:22:31.200
Um I think that's how that's gonna go.
00:22:31.359 --> 00:22:32.079
That's my opinion, though.
00:22:32.160 --> 00:22:35.039
And maybe something will change, but that's where I that's what I think about it.
00:22:35.359 --> 00:22:41.200
I think you mentioned a couple of things that kind of want to uh touch on and maybe circle back to a little bit.
00:22:41.440 --> 00:23:10.000
I I vaguely recall in prior conversations in prior podcasts with you or discussions when we've kind of talked about, you know, there's the possibility of, okay, well, one day maybe all of us are going to be wearing cameras, and that camera is now going to document that, okay, I'm giving, you know, four milligrams of Zofran or I'm titrating my Norepi or what have you, and then it can automatically import some of those things into whatever charting system you're using and and vice versa, or it's watching the ventilator changes that you're making or those kind of things.
00:23:10.160 --> 00:23:14.400
And and I think that aspect or that concept of it is pretty cool to me.
00:23:14.559 --> 00:23:23.279
Uh, and then you talk about you just uh a couple minutes ago with regards to you know some crew members that are utilizing AI to generate narratives.
00:23:23.440 --> 00:23:33.759
And I know that there are some charting uh programs that do have, you know, AI generators built into their charting system already, and and others are working towards doing that as well.
00:23:34.000 --> 00:23:42.480
And and I I think obviously within those systems, that's obviously protected health information that is stored only within those particular areas.
00:23:42.720 --> 00:23:48.079
But I have to still get worried about, I mean, anything that's on the internet can still somehow be discovered.
00:23:48.240 --> 00:24:10.160
Uh, and so that that's I think where my biggest like hang up or concern is with a lot of this stuff right now, whether we're doing a recording and it's recording everything I'm doing and the times and the adjustments and the medications or the ventilator changes or what have you, as well as looking at like these narratives that we're writing on all of these particular patients.
00:24:10.319 --> 00:24:20.640
Like, how secure is this information and and how easily accessed is it by whoever wants it?
00:24:21.680 --> 00:24:24.720
So this is really going to become an administrative problem.
00:24:24.880 --> 00:24:34.240
Okay, so this is one of the reasons why So Mike that's just this is this is where I'm a giant man, this is where I'm a giant brake pedal.
00:24:34.559 --> 00:24:38.000
I'm telling you, I'm putting pedal to the floor on this.
00:24:38.640 --> 00:24:48.160
So the question is really, when you're buying this stuff and you're using it, number one, do you understand the terms of your BAA?
00:24:49.039 --> 00:24:51.440
Are those terms accepted by the vendor?
00:24:51.839 --> 00:24:54.960
Are you protecting yourself legally speaking?
00:24:55.039 --> 00:24:59.680
Is your legal team versed in the these technologies?
00:24:59.920 --> 00:25:04.079
Make sure that whoever your league, your legal team is, they are they are versed in this.
00:25:04.319 --> 00:25:04.880
Ask the question.
00:25:05.039 --> 00:25:07.359
Hey, listen, to AI technologies are coming out.
00:25:07.519 --> 00:25:11.839
We're gonna be getting contracts and things like that with BAAs for these things.
00:25:11.920 --> 00:25:13.519
Uh, are you prepared for that?
00:25:13.680 --> 00:25:14.640
I would ask the question.
00:25:14.799 --> 00:25:19.359
I mean, maybe it's because I'm on spectrum and I'm not afraid to ask questions, but that's a question I would ask.
00:25:19.519 --> 00:25:22.480
Um, get with your legal team, be like, hey, are you ready for this?
00:25:22.640 --> 00:25:24.960
Number two, do you have a policy in place?
00:25:25.200 --> 00:25:27.759
Do you have an AI ethical use policy in place?
00:25:27.920 --> 00:25:35.839
You need to have AI policies in place if you plan on using it, because you don't want to buy or advocate for technologies that go against your own policy, right?
00:25:36.000 --> 00:25:36.880
That's that's number two.
00:25:36.960 --> 00:25:38.640
You got to put those safeguards in place.
00:25:38.880 --> 00:25:44.240
Like, for example, like if you have a big issue with chat GPT on your network, have your IT ban it.
00:25:44.400 --> 00:25:51.759
Just go ahead and and and and block those sites because what you don't want to do, well, and and for good reason, right?
00:25:51.839 --> 00:25:57.279
Because a crew member might think, oh, well, I pay the$21 a month for Chat GPT, open AI, or whatever.
00:25:57.359 --> 00:25:58.720
It's secure, it's safe.
00:25:58.960 --> 00:26:03.759
The reality is you're still feeding data into the LLM, right?
00:26:03.839 --> 00:26:08.559
Right now, when you have like a BAA in place, and and I understand you don't always have to have a BAA.
00:26:08.799 --> 00:26:11.119
I I'm one of those people I'm a little more rigid with that.
00:26:11.279 --> 00:26:17.839
Like, especially right now, since this technology is new, you're gonna need that documentation when you purchase uh that software.
00:26:18.000 --> 00:26:26.400
The other thing is you become friends with your IT people, uh, become friends with your database administrators, become friends with your data analytics people like me.
00:26:26.480 --> 00:26:29.359
You know, it's like I know people either love or hate data people.
00:26:29.440 --> 00:26:31.599
Uh, it just depends on what data you're given, right?
00:26:31.759 --> 00:26:35.759
But you need to be friends with those people and go, hey, listen, we're looking at this technology.
00:26:36.079 --> 00:26:37.680
Let's interrogate this together.
00:26:37.839 --> 00:26:39.599
Let's ask the hard questions.
00:26:39.839 --> 00:26:42.319
You need to ask them things like, how is your data stored?
00:26:42.480 --> 00:26:44.319
How is it protected on your end?
00:26:44.640 --> 00:26:46.480
How are you securing your servers?
00:26:46.640 --> 00:26:48.079
Where are you getting your models from?
00:26:48.240 --> 00:26:49.839
How are you training your models?
00:26:50.000 --> 00:26:51.519
Those are questions you have to ask.
00:26:51.680 --> 00:26:54.960
Now, to help you, there are several frameworks out there.
00:26:55.039 --> 00:27:01.920
Now, the one I personally advocate for uh is the National Institute of Standards and Technology or NIST.
00:27:02.079 --> 00:27:07.839
They have a uh a guide for uh AI technologies being incorporated into your business.
00:27:08.079 --> 00:27:13.759
And I highly recommend looking up that or similar frameworks when you purchase these systems.
00:27:14.000 --> 00:27:16.160
That's the only way you're gonna know it's secure.
00:27:16.400 --> 00:27:17.119
The only way.
00:27:17.359 --> 00:27:19.119
And there's never a guarantee, right?
00:27:19.200 --> 00:27:23.519
Because we cybersecurity is a growing field right now.
00:27:25.759 --> 00:27:42.240
I think that there's also the other aspect of it that I think a lot of crew members and and and myself included uh was this was my thought process of like, you know, if I'm inputting something into us, I can say, be sure to exclude all patient information, like identifiable patient information kind of thing.
00:27:42.400 --> 00:27:47.359
And and I think that that's all well and good, but that's still you're still inputting some of that information.
00:27:47.519 --> 00:27:49.279
So that's still there.
00:27:49.440 --> 00:27:57.440
And and I guess that's what my concern is with all of this is like how secure are we really talking with this and and making sure that things are non-discoverable?
00:27:57.519 --> 00:28:10.960
And and I mean, I I think at the end of the day, like let's be real, any one of our charting systems, if somebody wanted to truly get into those, there probably would be a way that that could potentially happen, you know, and uh there where there's a will, there's a way.
00:28:11.039 --> 00:28:29.039
Uh, but yeah, I don't want to see any of our crew members, and I and I think it's I think it's becoming a great tool, and I'm seeing this more useful in all the things that I'm doing on a daily basis, and I'm seeing some of my crew members and teammates get excited about the possibility of what AI can bring to the table.
00:28:29.440 --> 00:28:36.880
Uh, I'm just still super cautious because of all of that input into the system and and how secure are we really talking at the end of the day?
00:28:37.200 --> 00:28:37.440
Yeah.
00:28:37.519 --> 00:28:43.599
Now, some of the systems that are out there, I will say for like the charting system companies that are once again, I don't work for them.
00:28:43.680 --> 00:28:46.319
I have no financial affiliations with any of these companies.
00:28:46.559 --> 00:28:50.799
But one thing I will speak for on that is number one, they have to keep up with the rules, right?
00:28:50.880 --> 00:28:54.160
They have to have SOC2 encrypted servers, they have to have BAAs.
00:28:54.480 --> 00:28:56.799
They are a healthcare information technology company.
00:28:56.880 --> 00:28:58.400
They're very highly regulated.
00:28:58.720 --> 00:29:01.279
So there are strict rules on these things, right?
00:29:01.440 --> 00:29:11.599
Now, what I will say is I I'm not really worried necessarily as much for a breach on these like chart, the charting software companies specifically.
00:29:11.839 --> 00:29:17.279
What I worry about is the crew member side of it not being a human in the loop.
00:29:17.440 --> 00:29:23.119
For example, like let's say I use OCR technology and I like scan a face sheet or I scan this or that.
00:29:23.279 --> 00:29:25.839
I pull the information in, I just click apply.
00:29:26.160 --> 00:29:28.559
Didn't look at it, just apply.
00:29:28.799 --> 00:29:37.039
And then I look at the, you know, the now, granted, I I'm not a big fan of narrative generate narrate uh narrative generators with AI.
00:29:37.440 --> 00:29:46.160
I am good with the rule-based templates where you can put a template in there, like on Epic, where you can pull over stuff using dot phrases, what have you.
00:29:46.319 --> 00:29:52.319
That's okay because all it's doing is pulling that data and it knows that's a rule-based type application.
00:29:52.400 --> 00:29:53.519
It's not really AI.
00:29:53.759 --> 00:30:04.400
Where the AI is great, though, is where, like, for example, I've used systems where if I click a button, it will find things in my narrative and go, hey, um, you don't have these in your interventions.
00:30:04.559 --> 00:30:06.160
Like, do you want to plug these in there?
00:30:06.400 --> 00:30:11.119
And then sometimes there'll be duplicates, so I'll just go and click in and go, okay, I forgot I forgot to document that IV.
00:30:11.200 --> 00:30:12.480
I forgot to document this intubation.
00:30:12.559 --> 00:30:14.559
Well, I wouldn't forget that, but it happens.
00:30:14.799 --> 00:30:16.559
Click apply and then it moves it over.
00:30:16.640 --> 00:30:17.519
So then guess what?
00:30:17.759 --> 00:30:21.119
My chart is now lined up with my narrative.
00:30:21.359 --> 00:30:22.000
Vice versa.
00:30:22.160 --> 00:30:27.759
That whole double charting thing that we always worried about, it's not nearly as much of a concern because of the consistency piece.
00:30:27.839 --> 00:30:29.759
So it's good for those things.
00:30:29.839 --> 00:30:32.079
Um, and also like the AI suggestion tools.
00:30:32.160 --> 00:30:40.799
I've seen with some of these uh products where it'll tell you, like, hey, you put that you intubated, did you um put them on entitled CO2?
00:30:40.880 --> 00:30:42.880
You probably should document that intervention, right?
00:30:43.039 --> 00:30:43.759
And what have you.
00:30:43.920 --> 00:30:52.880
But where the issue is gonna lie with the accuracy of these things, Mike, is and not to jump down too far down a rabbit hole because I'm pretty good at that, as you know, uh, is Nempsys.
00:30:53.039 --> 00:30:55.119
Okay, so like Nempsys 3.5.
00:30:55.359 --> 00:30:57.839
Um, we're not in 3.5.1 or anything like that yet.
00:30:57.920 --> 00:31:08.880
One of the biggest things that me and and several of the data folks in the industry have observed is that a lot of the data sets that are found for the national data sets or even state reporting data sets is not consistent for critical care.
00:31:08.960 --> 00:31:10.160
They're custom fields.
00:31:10.319 --> 00:31:21.519
Um and I do want to say out's the world that we should work together to work with Nempsys or other organizations to be able to define data sets for our industry and critical care at least.
00:31:21.839 --> 00:31:23.680
First of all, I'm having PTSD.
00:31:23.759 --> 00:31:27.440
Uh just talking about the Nempsys rules.
00:31:27.599 --> 00:31:32.880
And you know, we talk about all these must-haves in our chart, right?
00:31:32.960 --> 00:31:41.279
And I think for anybody listening to this, you you're probably having that sudden heart rate increase, and you're like, oh shoot, gotta get all these in there, right?
00:31:41.440 --> 00:31:43.759
Because you and and that's where I guess we're moments.
00:31:44.000 --> 00:31:45.599
My boy, they're not driving and getting road raid.
00:31:46.000 --> 00:31:46.160
Absolutely.
00:31:46.720 --> 00:31:47.759
Please pull over, be safe.
00:31:47.839 --> 00:31:48.079
Yeah.
00:31:48.319 --> 00:31:48.480
Yeah.
00:31:48.720 --> 00:31:49.039
Pull over.
00:31:49.200 --> 00:31:51.599
But like I feel like that's my hope with AI.
00:31:51.680 --> 00:32:01.359
And it and you know, since we're talking about documentation, and you know, I I've used it, like what you said, Christian, to to make like a rule-based template, right?
00:32:01.440 --> 00:32:19.839
And I've used that to kind of guide and make sure that I'm including the necessary data, but it's still incumbent on me to be the person that verifies not only what the data is that I'm inputting and that it's also correct and what's in there.
00:32:20.000 --> 00:32:29.440
And so I feel like I've always tried to keep it as that kind of almost rule-based or framework, especially when it comes to documenting.
00:32:29.759 --> 00:32:31.759
Because I mean, I'll be honest, 3 a.m.
00:32:32.000 --> 00:32:37.599
and you've been up for a while, and you know, a lot of our legs back to base can be at about an hour.
00:32:37.839 --> 00:32:49.200
And so then you're stuck with I have four different sheets of all types of signatures and billing and all this stuff that I have to then upload.
00:32:49.359 --> 00:32:55.759
And then once that gets put in, then I have to start this chart, and then now I have to go back in and put in all these procedures.
00:32:56.079 --> 00:33:30.400
So I feel like there is a very good intended use for AI that my hope is that will help us as clinicians, whether you're in the back of an ambulance, uh an airplane, a helicopter, whatever it is, and even on the administrative side, there's got to be my hope is that there's got to be a way that AI will really help us kind of uh trim the fat, I guess, if so to speak, you know, because I I know uh in a previous charting system, I won't name names, but it was just drop-down button click heavy, right?
00:33:30.480 --> 00:33:38.319
And and I could easily shave off 15 minutes of time for things that can be auto-populated, right?
00:33:38.480 --> 00:33:46.559
And and then I think back to years, like how long did physicians and still do dictation, talk to text, right?
00:33:47.119 --> 00:34:09.840
So I mean, there's gotta be a way going forward that we can use AI to help us as clinical providers at patient bedside on that line to really help us cut down on the amount of time that we spend documenting, but also like helping assist us to make sure our documentation is where it needs to be, but like making it a little bit easier for us.
00:34:09.920 --> 00:34:18.639
And I I I don't know if I'm sound lazy saying that, but I kind of I don't really feel like it should take me three hours for certain charts, you know.
00:34:18.960 --> 00:34:20.159
I agree completely.
00:34:20.320 --> 00:34:23.039
And but here's the I'm gonna tell you the secret, right?
00:34:23.199 --> 00:34:29.440
So AI does serve a place, but maybe not as heavy of a place as you might think.
00:34:29.599 --> 00:34:43.440
Um, because I work in the back end of charting systems now, and tell you what, there's actually a lot of tools that are already in place that accomplish so many of those things that you're already mentioning, it's it's almost kind of hilarious to me, right?
00:34:43.519 --> 00:34:47.679
Because, like, for example, the system I use, there are like preset values.
00:34:47.760 --> 00:34:50.639
I can click a button and it will auto-populate certain things for me.
00:34:50.800 --> 00:34:53.360
I gotta go back and I gotta double check them naturally, right?
00:34:53.440 --> 00:34:56.239
Because every patient's it's a big gray area we live in.
00:34:56.320 --> 00:34:56.639
Yeah.
00:34:57.599 --> 00:35:23.519
But getting your people who build the charting form, getting them educated, whatever certifications and stuff exist for them, and having them understand the culture, having them understand the business, the quality, the CQI, being involved in all those processes to understand and really focus in on how you chart as a clinician and understand your program.
00:35:24.400 --> 00:35:28.320
You don't even really need AI to accomplish 80% of that, 90% of that.
00:35:28.559 --> 00:35:30.880
Because if you think about it, a lot of it's rule-based.
00:35:31.039 --> 00:35:36.719
It's like if, for example, if I fly an ECMO patient, I'm not gonna walk them to the stretcher.
00:35:37.039 --> 00:35:39.119
They're gonna be sheet transferred, right?
00:35:39.280 --> 00:35:39.679
Right.
00:35:40.239 --> 00:35:43.679
They're if they're intubated, their GCS is going to be A3.
00:35:43.840 --> 00:35:45.440
They're not gonna be able to sign for themselves, right?
00:35:45.599 --> 00:35:48.719
There, it's implied consent or family has to sign, right?
00:35:48.880 --> 00:35:54.000
So we can use visibility rules, validations, we can use all kinds of things to reduce that.
00:35:54.159 --> 00:35:57.360
But I think a lot of people get into these systems, they're overwhelmed.
00:35:57.599 --> 00:36:01.119
Uh and and and like I said, I'm not gonna speak to any particular charting system.
00:36:01.280 --> 00:36:03.119
Sometimes the support's not always there.
00:36:03.199 --> 00:36:06.880
Um, sometimes you need to buy extra support or you need to help with that support.
00:36:07.280 --> 00:36:08.320
Or it's outsourced.
00:36:08.719 --> 00:36:09.599
Or it's outsourced.
00:36:09.920 --> 00:36:11.199
Or we just call Christian.
00:36:11.440 --> 00:36:11.679
Right?
00:36:12.000 --> 00:36:13.840
Or yeah, no, it's supposed to be.
00:36:13.920 --> 00:36:20.639
There's there's some people out there in our industry um who who I could call by name who who would just who are just experts in this stuff.
00:36:20.880 --> 00:36:27.519
And when you talk to them and you think about it, it really comes down to culture, it comes down to process.
00:36:27.679 --> 00:36:33.599
Um, having a team of people at your organization who are informatics, I think that's critical.
00:36:33.760 --> 00:36:39.199
If you don't have an informatics group of clinicians that look at your form, if you're in charge of that form.
00:36:39.280 --> 00:36:41.760
Now, granted, there are some companies out there, they're huge, right?
00:36:41.840 --> 00:36:43.440
We got some of our national companies.
00:36:43.599 --> 00:36:51.119
Their charting forms are that that's a whole nother animal, but they have people that work on those things because they have billing and everything else involved on a national scale.
00:36:51.280 --> 00:37:06.880
But if you have a smaller program, like the program I fly for, uh the past programs I've flown for, we have a lot more control over that charting form, which means on the charting side of it, you can reduce your clinician's workload, but you have to put in the effort, you have to put in the time, you also have to test things.
00:37:07.039 --> 00:37:08.960
You can't rush things to market either.
00:37:09.199 --> 00:37:11.039
You have to plan to study act, right?
00:37:11.119 --> 00:37:16.719
And I'm sure all the nurses that are listening to this are probably like, dude, this guy, like, what is his problem?
00:37:17.440 --> 00:37:19.119
They're all twitching right now.
00:37:19.679 --> 00:37:22.159
Believe it or not, it actually works if you do it.
00:37:22.400 --> 00:37:36.639
If you actually do it and focus in and you you work on asking the question, if I'm a crew member charting on this particular patient and I'm the biller and the quality person and the billing person, like, how can we make this work to where it's more efficient, right?
00:37:36.719 --> 00:37:44.639
Because if we build, if if we document more efficiently, we're gonna have better quality, which means better data for models that help us with the AI sign a bit.
00:37:44.880 --> 00:37:47.440
So as you can see, we need to crawl, then we walk, then we run.
00:37:47.599 --> 00:37:54.320
And in this case, you crawl by looking at the things you already have in place that maybe you don't use to the fullest capacity.
00:37:54.719 --> 00:37:56.400
Make those things really good.
00:37:56.800 --> 00:38:02.639
Make sure your policies and process are good, and then move on to the stuff that'll board you up for AI.
00:38:02.960 --> 00:38:05.679
Well, and that's what I was just, you know, hearing you talk about all this.
00:38:05.760 --> 00:38:18.320
I think that's one of the things in my new role that I'm most excited about is I I essentially am also functioning as like, you know, the quality manager side of things and and looking at some of the data and how we can improve the program from the quality side of the.
00:38:18.880 --> 00:38:40.880
And I but but having, and I think a lot of people share this similar uh challenge out there in those smaller programs in that you have multiple roles and there's just not enough hours in the day, the week, the month, the year in order to be able to truly dedicate to those roles what you would like to.
00:38:41.599 --> 00:38:53.119
And so I think the integration of AI into those roles to help streamline some of that process and make us more efficient is something that I'm super excited to see how that all works out.
00:38:53.679 --> 00:39:05.199
But I think on the certainly on that QA side of things or on that quality metric side of things is really going to be at least what I'm hoping for, the forefront of it.
00:39:05.280 --> 00:39:13.360
But you're almost replacing a quality manager position by utilizing some of these AI integrations.
00:39:14.079 --> 00:39:19.039
For me, if I'm in a smaller program and have multiple roles, that's a great thing.
00:39:19.519 --> 00:39:30.320
But you know, that might not be if that's your sole job, maybe you might definitely sitting over here, maybe like a little bit worried about what that's gonna look like.
00:39:30.480 --> 00:39:46.719
Uh and and so I I think there are so many really cool avenues to go down with this, but again, like you still need to have the time to officially analyze or or work on the input so that you can appropriately analyze the output.
00:39:47.039 --> 00:40:04.320
Like and I think that's that's the biggest task of all of this is like finding that time to be able to and and look, I get it, like the entire healthcare system nationwide, whether you're talking about in-hospital, out of hospital, I don't I don't care what role you have within healthcare, you're under some constraints right now.
00:40:04.400 --> 00:40:06.559
There's a lot of changes that are coming down the pike.
00:40:06.719 --> 00:40:16.000
You're constantly trying to figure out a way to be able to make this whole system work and still be able to save a little bit money, more money.
00:40:16.079 --> 00:40:18.400
And I get it, like I totally do.
00:40:19.360 --> 00:40:29.119
But this is the I I don't know that I see another option without the integration of AI to be able to make some of those things happen.
00:40:29.199 --> 00:41:01.920
Uh and so like I'm excited but terrified at the same time to see what this is all gonna end up looking like because I think that there are a lot of possibilities out there uh that can make my job a heck of a lot easier, but I still need to be able to have dedicated time as a human to look at all of those things and to be able to make sure that I can modify what I can modify to improve the front end user experience for my team and to maximize the back end output uh for the analytics.
00:41:02.159 --> 00:41:09.360
Uh and and and that's that's that's the biggest challenge I I find right now uh is really putting those big pieces together.
00:41:11.760 --> 00:41:15.280
Well, and and that's where good project management's gonna come into play, right?
00:41:15.440 --> 00:41:21.360
So you you have limited resources and you want to use these AI tools because they can help you, right?
00:41:21.599 --> 00:41:25.360
So uh for example, like imagine yeah, there are systems that exist.
00:41:25.440 --> 00:41:35.760
Once again, I won't mention any names, but there are systems that will take your charts, run custom prompts through it for quality questions, and then it will answer them.
00:41:35.840 --> 00:41:37.920
And then you can look at your dashboards and whatnot and see that.
00:41:38.000 --> 00:41:39.360
And that's a great tool.
00:41:39.679 --> 00:41:44.880
The biggest thing from that is gonna have to be it's the initial upfront investment to get that working.
00:41:45.039 --> 00:41:51.119
That's the biggest thing is that it's not these systems are never really out of the box, ready to go.
00:41:51.360 --> 00:41:51.599
Right.
00:41:51.760 --> 00:41:53.760
Because once again, it's a contextual problem.
00:41:53.920 --> 00:41:55.679
And I think that's always gonna be the issue.
00:41:56.000 --> 00:41:56.159
Sure.
00:41:56.320 --> 00:42:02.559
And every program's different and they have different, you know, standards of care and different ways that they accomplish those things.
00:42:02.639 --> 00:42:04.559
And so yeah, I totally get that.
00:42:04.639 --> 00:42:05.840
That makes perfect sense to me.
00:42:06.079 --> 00:42:15.119
But yeah, how you learn that or how you help the the AI tool learn that, I think is where a lot of people may end up struggling.
00:42:15.519 --> 00:42:16.559
I would agree with that.
00:42:16.639 --> 00:42:26.480
And I think the way that you start off is is doing basic things, like answering questions like, hey, let's define what something is.
00:42:26.719 --> 00:42:28.239
Like, you know, I I love the question.
00:42:28.320 --> 00:42:32.559
I took a course on quality once and they asked, what is quality?
00:42:32.960 --> 00:42:34.639
What a great question, right?
00:42:34.800 --> 00:42:37.119
It's like, what is like I don't know.
00:42:37.360 --> 00:42:39.119
There's so many like quality in what?
00:42:39.199 --> 00:42:41.920
Like if I'm at the Doritos factory, is it crunchy chips?
00:42:42.000 --> 00:42:43.199
Is it is it more flavorful?
00:42:43.360 --> 00:42:43.920
I don't know.
00:42:44.639 --> 00:42:50.480
But I think that when it you have to define the basic things first, it it takes time.
00:42:50.800 --> 00:42:53.679
And I know we're very much that group of people.
00:42:53.840 --> 00:42:56.480
We want things as quick as we can because we take care of patients.
00:42:57.360 --> 00:42:57.760
Yeah.
00:42:58.159 --> 00:42:59.519
Absolutely.
00:43:00.000 --> 00:43:05.760
But at the end of the day, these systems, you know, you can buy them, they're ready to go out of the box to some extent.
00:43:05.840 --> 00:43:07.119
Um, you're gonna run them.
00:43:07.199 --> 00:43:21.119
But here's the it's a paradox in the sense that yes, it might be easy when you first get it, but when you start running stuff through it, what'll happen is if you don't have the the foundation built like anything else, it's gonna sink, it's gonna fracture.
00:43:21.280 --> 00:43:28.880
And so what'll happen is you're you'll you'll start making decisions on outputs, and then those output decisions will harm your program uh down the line.
00:43:29.039 --> 00:43:34.639
And it will also allow us to not make good decisions when it comes to you know clinical care, et cetera.
00:43:34.880 --> 00:43:40.960
So where we have to start, if you don't have a lot of resources, the first thing I would tell you is a few things.
00:43:41.199 --> 00:43:45.440
Look at your number one, look at your policies and make sure they're consistent before Kynes comes around, right?
00:43:45.599 --> 00:43:47.119
Everyone's like, oh, Keynes is here.
00:43:47.199 --> 00:43:48.480
Let's look at our policies, right?
00:43:49.039 --> 00:43:50.639
Look at them all the time.
00:43:50.800 --> 00:43:52.079
Do they make sense?
00:43:52.320 --> 00:43:53.360
Do they flow?
00:43:53.599 --> 00:43:55.840
Do they do are there contradictions?
00:43:56.000 --> 00:43:58.000
Clinical guidelines are also very similar.
00:43:58.079 --> 00:44:04.800
I work for multiple places and And you'll always find one or two hanging out in there, making sure that consistency is the key.
00:44:05.039 --> 00:44:07.519
Uh I'll I'll take I take that from Dave Lebesk.
00:44:07.599 --> 00:44:11.440
Um consistency, you know, um, consistency is key here.
00:44:11.599 --> 00:44:30.239
So if you're consistent in your policies, your procedures, you're consistent in your training, you're consistent in your documentation, you're consistent the consistency in things, even without the AI, that's a great place to start because even if you do have limited resources, it doesn't take a lot of resources to be just be consistent, you know, with what you're doing, right?
00:44:30.400 --> 00:44:32.000
If everyone's consistent, cool.
00:44:32.159 --> 00:44:34.880
And once that's done, then it's like, okay, well, we're consistent now.
00:44:34.960 --> 00:44:38.719
Now let's define in that consistency what makes us consistent.
00:44:38.880 --> 00:44:40.960
What is important to us?
00:44:41.280 --> 00:44:43.599
Then you define your own definition of quality.
00:44:43.679 --> 00:44:48.159
And once you have that definition of quality, you're like, okay, these are the five most important things.
00:44:48.480 --> 00:44:57.280
You only need three to five measures to start and go, what are the five most important things right now to keep my operations safe and take care of my patients effectively and make the money I need to keep on going?
00:44:57.519 --> 00:45:04.000
Or to the crew members, you know, give them five, like, hey, what are the five things as crew members that hey, you guys find to be the most important?
00:45:04.159 --> 00:45:13.280
And then get really good at those things and then surround the AI tools you want to use, which each of one of those five things, and focus on the data points there.
00:45:13.360 --> 00:45:15.440
And that's you that's when you're starting to walk, right?
00:45:15.519 --> 00:45:18.400
Because now you're like, okay, now I define what I want to look at.
00:45:18.480 --> 00:45:21.119
And so because if you look at everything, it's paralysis by analysis.
00:45:21.199 --> 00:45:23.199
You're gonna get, I mean, it's overwhelming.
00:45:23.599 --> 00:45:24.400
Where do you even start?
00:45:24.639 --> 00:45:24.880
Yeah.
00:45:25.119 --> 00:45:25.519
Yeah.
00:45:25.760 --> 00:45:28.480
It's like I think it's funny, like lift times, right?
00:45:28.800 --> 00:45:32.639
I I kind of laugh about that one because it's like, okay, lift times, that's great.
00:45:32.880 --> 00:45:35.599
You know, we do need to move quickly, right?
00:45:35.760 --> 00:45:37.840
But it's like, what are the implications of that?
00:45:38.000 --> 00:45:38.159
Right?
00:45:38.320 --> 00:45:40.480
Some programs, it matters way more than others.
00:45:40.639 --> 00:45:43.360
Like if you're an IFR program, like I fly at an IFR program.
00:45:43.599 --> 00:45:47.039
There's always gonna be a little bit of a time delay because you have to flight plan your flight, right?
00:45:47.199 --> 00:45:48.079
That's just how it goes.
00:45:48.320 --> 00:45:50.719
Move with a sense of purpose, everyone's mission driven.
00:45:50.960 --> 00:45:52.800
That's also your culture, too, right?
00:45:52.960 --> 00:45:56.159
Is your culture that, you know, once again, consistency.
00:45:56.239 --> 00:45:58.400
Is your culture consistently mission driven?
00:45:58.559 --> 00:46:02.320
Are they consistently practicing safety, uh, safety policy?
00:46:02.639 --> 00:46:05.840
Are they consistently documenting things like near misses?
00:46:06.000 --> 00:46:06.639
All those things.
00:46:06.719 --> 00:46:15.840
Because then, like, for example, imagine an AI tool that could look at all your near misses over time and go, hey, Mike, your guys are doing your people are doing this.
00:46:16.000 --> 00:46:17.360
This is this is not safe.
00:46:17.519 --> 00:46:20.400
And it's gonna lead to something down the line that could be catastrophic.
00:46:20.559 --> 00:46:22.480
Like, let's take care of that right now, right?
00:46:22.639 --> 00:46:23.840
That's the prediction.
00:46:24.000 --> 00:46:31.519
And so that's where context meets the models, which meets the outputs, which means consistency with the outputs.
00:46:31.679 --> 00:46:37.039
If you're consistent, it will too be consistent, and thus you're gonna get the things you want.
00:46:37.280 --> 00:46:39.199
But unfortunately, it does take time.
00:46:39.280 --> 00:46:43.440
But that's why we crawled and we walked and we run, especially when you're resource limited like that.
00:46:43.599 --> 00:46:44.639
So I feel you though.
00:46:44.800 --> 00:46:45.119
I do.
00:46:45.199 --> 00:46:47.199
I I know that that's a that's a tough one, right?
00:46:47.360 --> 00:46:49.519
So let's move to the run part.
00:46:50.880 --> 00:46:54.079
We've crawled a little bit, we walked a little bit impatient over here.
00:46:54.159 --> 00:46:55.199
He wants it yesterday.
00:46:55.440 --> 00:46:56.239
He wants it now.
00:46:56.320 --> 00:46:57.760
It's my money and I need it now.
00:46:57.920 --> 00:46:58.159
No.
00:46:59.039 --> 00:47:00.320
So let's talk about this, right?
00:47:00.480 --> 00:47:13.679
Okay, so we've under so we've talked about some data processes, we've talked about how to use AI to really identify some things that maybe we do good, maybe we do bad, and we should improve on.
00:47:13.760 --> 00:47:14.639
So we have that.
00:47:14.880 --> 00:47:18.800
We've talked about using it into building good scenarios, right?
00:47:18.960 --> 00:47:21.679
So let's say you discovered you had a bunch of near misses.
00:47:21.760 --> 00:47:23.760
Okay, how do we train and recorrect that?
00:47:23.920 --> 00:47:30.639
So A, so you used AI identified you had a bunch of near misses, so now you do some training, you write it writes some scenarios, right?
00:47:30.719 --> 00:47:42.079
So you practice as a crew member, and I I've often heard of AI starting to be that digital crew member, right?
00:47:42.239 --> 00:47:46.239
Like that third wheel.
00:47:47.119 --> 00:48:12.480
So what would be the probability that so if it's capable of learning and predictive software, again, it's based off of trends, how much can, could or would AI help us treat our patients like live in the field based on data feedback, right?
00:48:12.559 --> 00:48:16.239
So you were talking before about let's say two data points were able to talk to each other.
00:48:16.320 --> 00:48:27.280
Let's say in the future the ventilator was getting real-time data back on from the monitor, more particularly end titles and spot2s.
00:48:29.519 --> 00:48:39.039
What's that possibility look like of either suggestive or actual real-time management of the patient?
00:48:39.599 --> 00:48:42.159
Autonomous based on data.
00:48:42.719 --> 00:48:43.920
Is that a possibility?
00:48:44.079 --> 00:48:50.559
Like, would that be something that would help us in the future or be part of how we take care of patients in the future?
00:48:50.960 --> 00:48:56.400
I think to your point, you know, that that's a very, very loaded question, but a great question.
00:48:56.719 --> 00:49:00.559
So, yes, I think it very well could assist us, right?
00:49:00.800 --> 00:49:02.079
So a couple things.
00:49:02.239 --> 00:49:04.960
Number one, business can't take that out of this, right?
00:49:05.119 --> 00:49:05.360
Sure.
00:49:05.519 --> 00:49:10.079
You know, you have competing equipment device providers out there.
00:49:10.320 --> 00:49:13.119
The question is number one, are they willing to share data with each other?
00:49:13.360 --> 00:49:18.400
Are they willing to communicate with each other and go, okay, hey, listen, we're gonna do this for the betterment of the patient.
00:49:18.559 --> 00:49:22.239
You know, are these device companies gonna allow each other to communicate?
00:49:22.480 --> 00:49:24.639
And is the data gonna be consistent on the back end?
00:49:24.719 --> 00:49:26.400
And that's that's kind of the big deal, right?
00:49:26.480 --> 00:49:28.480
Is how we allow that stuff to communicate.
00:49:28.639 --> 00:49:34.320
What I would love is a world where my IV pumps, my monitor, and my vent talk to each other.
00:49:34.480 --> 00:49:42.639
My ECMO pump talks to my monitor, talks to this, talks to that, talks to the other, because that's all data points at the end of the day, right?
00:49:42.800 --> 00:49:43.039
Yeah.
00:49:43.199 --> 00:50:03.440
So, like, for example, you know, if let's say I've got a PEEP of 15 on my ventilator and I'm getting a lot of art line variability, and my blood pressure's dropping, and and my SATS are still okay, the monitor could say, Hey, could we suggest doing this with the ventilator?
00:50:03.599 --> 00:50:04.719
Hey, let's give some volume.
00:50:04.800 --> 00:50:12.079
Hey, we have intrathoracic pressure increases, like alarming or warning or advising in real time.
00:50:12.239 --> 00:50:16.239
I could see that being a future uh in the future, I could certainly see that.
00:50:16.400 --> 00:50:21.599
I think the first barrier, once again, is going to come down to are the device companies willing to communicate with each other?
00:50:21.760 --> 00:50:23.679
Are they willing to allow the devices to do that?
00:50:23.760 --> 00:50:24.480
That's number one.
00:50:24.719 --> 00:50:26.960
Number two, we also need to communicate what's important.
00:50:27.039 --> 00:50:31.039
Like I was telling Mike, uh, we have to define what is important in that moment.
00:50:31.199 --> 00:50:39.360
We have seen in the literature and in my studies, I have found that paralysis because that alarm fatigue is a big deal, right?
00:50:39.519 --> 00:50:41.039
We need to think about what's important.
00:50:41.199 --> 00:50:46.400
Once again, if we can use the AI to take those 50 data points, reduce them to 25, and maybe even 10.
00:50:46.800 --> 00:50:48.400
So that's a lot less for me to look at.
00:50:48.480 --> 00:50:52.559
So when something actually alarms at me, it's not, uh, it's the it's the O2 thing alarming again.
00:50:52.719 --> 00:50:56.239
Yeah, because we're at night, you're at 100% FIO2, it's alarming again.
00:50:56.320 --> 00:50:57.360
Can you change that, dude?
00:50:57.519 --> 00:51:00.480
Whatever the case may be, we eventually start writing it off, right?
00:51:00.559 --> 00:51:05.679
Because as humans, that that's where we we we falter is that alarm fatigue.
00:51:05.840 --> 00:51:09.519
So we would have to find what we actually look at as being important.
00:51:09.599 --> 00:51:10.719
Now, how do we do that?
00:51:10.960 --> 00:51:13.760
The way we do that, it it's not it's not fun.
00:51:13.920 --> 00:51:14.880
It's it's research.
00:51:15.039 --> 00:51:22.880
We we're going to have to conduct research to find out what is going to be the most important and how effective is that going to be.
00:51:22.960 --> 00:51:27.679
And once we have that research, in our field, we we borrow a lot of knowledge.
00:51:27.760 --> 00:51:33.599
You know, I was saying this to someone recently because they're like, you know, why are you doing you're in school, you're doing all these things.
00:51:33.760 --> 00:51:36.079
Like, what's the point of doing like research air medical?
00:51:36.159 --> 00:51:42.239
I'm like, well, the point is that when you borrow knowledge from other professions, it's their context, it's not yours.
00:51:42.559 --> 00:51:45.440
So our industry, we need to start conducting more research.
00:51:45.599 --> 00:51:47.280
I'm very glad that we have the aeromedical journal.
00:51:47.360 --> 00:51:51.119
I think it's a, like I said, no affiliation, but I I will give a shout out to them.
00:51:51.280 --> 00:51:52.719
They are doing amazing work.
00:51:52.960 --> 00:51:57.920
And I think that that journal needs, in my opinion, we need to start giving it as many papers as possible.
00:51:58.000 --> 00:52:02.400
We need to start really finding good research questions, conducting research on behalf of our industry.
00:52:02.559 --> 00:52:12.400
Because if we can find measurements, if we can find things that matter, we can find essentially bundles of alarms or bundles of parameters that are evidence-based in transport that work together.
00:52:12.639 --> 00:52:20.960
Then in the future, once the devices start talking to each other, we can go and the AI is saying, Hey, Blake, you why don't you reduce your title volume, man?
00:52:21.039 --> 00:52:22.320
You should probably come down on that.
00:52:22.559 --> 00:52:26.800
Um in this case, my guy sounds like the big Lebowski, but that's me.
00:52:27.039 --> 00:52:28.719
I'm struggling I'm struggling over here right now.
00:52:28.800 --> 00:52:32.320
I'm I'm I'm really I'm really struggling over here right now.
00:52:32.559 --> 00:52:46.239
And and so I have to say this that maybe this is just because, you know, I I'm I've been in this industry for a little while now, and and every single one of these scenarios that you're talking about, I can think of 10 other patients that I've had.
00:52:46.400 --> 00:52:48.880
Like the patient, this is all well and good.
00:52:48.960 --> 00:52:50.480
This is great, this is fantastic stuff.
00:52:50.639 --> 00:52:52.400
Have them all communicate together.
00:52:52.719 --> 00:52:56.960
But I think the negative effects of this have the potential to be so great.
00:52:57.199 --> 00:53:01.840
I mean, we all know that none of now never once has our patient followed the textbook.
00:53:02.000 --> 00:53:02.960
Never once, right?
00:53:03.039 --> 00:53:03.199
No.
00:53:03.360 --> 00:53:12.960
And especially when we talk about critical care medicine and we talk about transport and in the hems industry, you add in all these other factors of the sound, the vibration, the movement, all of these other things.
00:53:13.039 --> 00:53:20.159
And that textbook patient that has been sitting in that ICU bed playing by all the rules, suddenly decides, nope, not happening today.
00:53:20.400 --> 00:53:20.719
Right.
00:53:21.199 --> 00:53:24.719
And and it it just becomes very frustrating in that situation.
00:53:24.800 --> 00:53:25.840
We're constantly changing gears.
00:53:25.920 --> 00:53:30.159
And it's like, okay, well, you know, you get to the receiving facility, and they're like, well, why do you have to change all these things?
00:53:30.320 --> 00:53:32.159
Well, see what happened was.
00:53:33.599 --> 00:53:39.440
You know, and so and so I gotta play devil's advocate here because I'm like, I'm sitting here listening, okay, great.
00:53:39.599 --> 00:53:52.159
If if my vent and my monitor can talk to each other and my vent can suggest that I should make changes based off of my entitled CO2 and my pulse oximetry, well, that's not what needs to be changed right now, is because my patient needs to date it.
00:53:52.320 --> 00:54:05.039
Like it's those little things that that I what I worry about, what I truly worry about in this situation, is that while AI is a great tool, is it going to make us as providers dumber?
00:54:05.360 --> 00:54:13.679
Are we only going to be able to look at this input from the AI and say, oh, well, AI is smarter than I am.
00:54:13.920 --> 00:54:15.519
It suggested I do this now.
00:54:15.679 --> 00:54:17.360
Like that, that to me is not okay.
00:54:17.440 --> 00:54:21.039
And and I'm gonna stand on my soapbox.
00:54:21.119 --> 00:54:23.280
I am absolutely gonna stand on my soapbox here.
00:54:23.519 --> 00:54:32.800
Look, it's got a time and place, but I really struggle with how this is gonna integrate into the clinical side of things because there are so many variables.
00:54:32.880 --> 00:54:39.599
And then again, this is this is what I worry for our newer clinicians, who I think that our newer clinicians are very brilliant.
00:54:39.760 --> 00:54:44.559
And I think I love their passion and drive and desire in the industry.
00:54:45.519 --> 00:54:55.519
But they're also more technologically advanced and I think much more apt to tie into what's going on with them technologically.
00:54:55.760 --> 00:55:06.960
And that gut instinct from 27 years of medicine, in my opinion, for me, is never going to replace what a machine is telling me about that patient.
00:55:11.840 --> 00:55:24.960
And I struggle because we have a lot of clinician, you know, we we as as programs, we look at those clinicians that have been in as from nursing side of things working in an ERICU for three to five years.
00:55:26.239 --> 00:55:26.960
That's great.
00:55:27.119 --> 00:55:27.599
That's good.
00:55:27.760 --> 00:55:37.119
I think that for most people who have the passion and and drive and desire to do this job, that they've probably got that solid gut instinct at that point.
00:55:37.519 --> 00:55:39.280
But that's not everyone.
00:55:39.679 --> 00:55:53.360
And I I I I'm sorry, but I struggle to believe that this is going to make us necessarily smarter as clinicians if you don't have that gut to back up what is in front of you.
00:55:53.679 --> 00:55:56.480
Mike, you have no idea how much I love you right now.
00:55:56.960 --> 00:55:59.599
As we say in the South, preach, brother, preach.
00:55:59.840 --> 00:56:01.920
Listen, no, this is good.
00:56:02.079 --> 00:56:02.639
This is good.
00:56:02.800 --> 00:56:04.480
This is exactly what I'm saying.
00:56:04.719 --> 00:56:08.559
What you're saying right there, don't ever lose that.
00:56:08.800 --> 00:56:09.920
Because you're right.
00:56:10.159 --> 00:56:10.880
100%.
00:56:11.599 --> 00:56:14.239
This is why I'm a brake pedal about some of these things.
00:56:14.320 --> 00:56:16.960
And now you see you see exactly what I'm saying, right?
00:56:17.119 --> 00:56:18.719
Like we have a lot of possibilities.
00:56:18.960 --> 00:56:22.079
Yeah, I totally get it on like the admin or the charting or the other things.
00:56:22.159 --> 00:56:26.480
But when it comes to patient care, yes, I I'm I'm okay with with feedback.
00:56:26.559 --> 00:56:31.039
I'm okay with suggestions, I'm okay with, hey, this is a possibility.
00:56:31.280 --> 00:56:38.079
But if if I don't have that gut instinct to know, this isn't because it's a ventilator issue, this is because my patient needs sedation adjusted.
00:56:38.239 --> 00:56:38.559
That's right.
00:56:38.719 --> 00:56:44.800
Like that, it it that's that's where the gut comes in that I feel as though we're gonna we're gonna struggle.
00:56:44.880 --> 00:56:50.480
And I don't want to see us as clinicians lose our ability to think on our own.
00:56:50.880 --> 00:56:51.599
That's right.
00:56:51.840 --> 00:56:55.119
And that's where, Mike, the education is helpful.
00:56:55.599 --> 00:57:07.440
So that is why if I had to pick between education using these systems and clinical, live clinical decision support, clinical decision support is always gonna lose.
00:57:07.760 --> 00:57:17.440
Because for me, quality, CQI, and education are gonna be the first places that we need to really be looking at these tools for.
00:57:17.599 --> 00:57:17.840
Yeah.
00:57:18.000 --> 00:57:18.400
Right?
00:57:18.639 --> 00:57:25.280
Because that's where we can take our experience, we can take our knowledge, we can build scenarios, we can educate better.
00:57:25.360 --> 00:57:31.599
You know, your your your trainee can sit on the couch because they're tech driven anyways, because this generation, and they can ask questions, right?
00:57:32.239 --> 00:57:35.119
I can't even imagine that day for me personally.
00:57:35.280 --> 00:57:37.119
Like that is not how I learn at all.
00:57:37.360 --> 00:57:37.519
Yeah.
00:57:37.679 --> 00:57:39.440
But being a typewriter.
00:57:40.159 --> 00:57:41.039
Typewriter.
00:57:41.280 --> 00:57:42.719
I want to see this thing, by the way.
00:57:43.039 --> 00:57:43.599
Maybe offline.
00:57:43.760 --> 00:57:44.960
You just pull it out, man.
00:57:45.119 --> 00:57:45.760
Pull out the typewriter.
00:57:46.000 --> 00:57:49.599
Well, instead of a soapbox that he stands on, it's the typewriter box that he stands on.
00:57:49.760 --> 00:57:49.920
Exactly.
00:57:50.480 --> 00:57:51.440
But that's good though.
00:57:51.679 --> 00:57:59.360
That's that's no, I love that because as someone who is an advocate for these tools, I'm also an advocate for using them correctly.
00:57:59.840 --> 00:58:02.000
And for example, right now I'm on orientation, right?
00:58:02.159 --> 00:58:03.840
I've got an amazing FTO.
00:58:04.079 --> 00:58:06.559
Guy has thousands of flights under his belt.
00:58:06.800 --> 00:58:07.519
You know what?
00:58:07.760 --> 00:58:11.280
And I'll I'll say this on a podcast, a national podcast.
00:58:11.360 --> 00:58:12.000
I'll say this.
00:58:12.159 --> 00:58:16.880
You know, the other day, he was like, hey man, get your head out of the paperwork and look at your patient.
00:58:17.599 --> 00:58:18.480
He's right.
00:58:18.719 --> 00:58:19.840
Absolutely right.
00:58:20.079 --> 00:58:23.199
And I had, I had, I had to get my head out of the paperwork.
00:58:23.280 --> 00:58:29.360
And you know, I work tomorrow night, and that's all I'm thinking about is I'm gonna do a better job at focusing on my patient.
00:58:29.519 --> 00:58:32.719
I'm not gonna be focusing on paperwork, I'm not gonna be focusing on charting.
00:58:32.880 --> 00:58:36.320
You know, the way my brain's wired, you know, neurodivergently wired.
00:58:36.400 --> 00:58:38.639
Like I want to look at the paper because I want all the answers.
00:58:38.719 --> 00:58:45.199
But at the end of the day, you're right, Mike, that's where the human stays in the loop in the in these technologies.
00:58:45.360 --> 00:58:47.920
That's why humans should always stay in the loop.
00:58:48.079 --> 00:58:52.000
Do not let these technologies replace you because it can't.
00:58:52.079 --> 00:58:52.400
Yeah.
00:58:52.639 --> 00:58:57.599
At least not in the foreseeable future, um, it from what I've observed.
00:58:57.679 --> 00:59:08.719
But I think that is if I if anyone could take anything home from this podcast today, it would be if you feel free to let these tools help you, but not replace you.
00:59:09.039 --> 00:59:14.239
Because at the end of the day, you, you know, the patients need that that touch, right?
00:59:14.400 --> 00:59:17.840
Uh, there's a lot of evidence that actually shows that you know the human touch is helpful.
00:59:18.000 --> 00:59:23.679
Yeah, that's why someone was uh someone who's not medically related was telling me it's like, oh man, robots could replace nurses one day, no, they're not.
00:59:23.760 --> 00:59:24.719
No, no, they're not.
00:59:24.800 --> 00:59:31.519
No, no, no, I wasn't, I'm not sitting here saying that like you were saying, oh, we should we should absolutely go down this route.
00:59:31.599 --> 00:59:34.239
I know that you're talking about it as these are possibilities, right?
00:59:34.320 --> 00:59:35.760
Like these are things for the future.
00:59:36.000 --> 00:59:58.079
But I could just feel like my blood starting to boil over here because I I feel like we as clinicians have worked so hard to get to the point where we can independently care for these critically complex patients and make all of these physician-level decisions on how we are managing that particular patient or population or what have you.
00:59:58.239 --> 01:00:01.760
And and that is that is the part I don't want to see us lose.
01:00:01.920 --> 01:00:05.519
We have worked so hard as an industry to get to where we are.
01:00:05.599 --> 01:00:20.800
And I look at, I look at when I when I started in in Hems, when I started in healthcare, period, uh, you know, and again, uh back in that day, we were throwing everybody on 15 liters down our breather because that was that was what the book said we're supposed to be doing, right?
01:00:20.880 --> 01:00:35.280
Like, and and look where we've come in that time period and all the things that we have advanced as an industry, and I don't want to see us lose that because we have a suggestion from AI that that's what we should, the route that we should be going.
01:00:35.679 --> 01:00:37.119
That's where I struggle with it.
01:00:37.360 --> 01:00:39.199
Now, Mike, if I may add this.
01:00:39.360 --> 01:00:44.800
So, you know, I was watching um it was it was like a TV show or something, right?
01:00:44.960 --> 01:00:47.280
You know how we all sit and watch reels and whatever.
01:00:47.519 --> 01:00:50.400
And there was a really, really good point that was made.
01:00:50.480 --> 01:00:53.039
It was like, oh, it was one, it was actually an influencer, Dr.
01:00:53.119 --> 01:00:53.840
Glauck and Black.
01:00:53.920 --> 01:00:54.800
I mean, he's hilarious.
01:00:54.960 --> 01:01:04.880
But he was saying to the senior physician, he's like, Well, back in my day, you know, we we would work 20-hour call days, we didn't have work, you know, work hour limitations for residency.
01:01:05.280 --> 01:01:08.400
And these two, you know, medical students go, oh, let me get this straight.
01:01:08.639 --> 01:01:20.000
So, you know, you're expecting us who have to learn three times as much information as you, you know, medicine is advancing fast because of these technologies, right?
01:01:20.079 --> 01:01:24.159
Like we're doing protein mapping, you know, cancer treatments are getting more customized.
01:01:24.239 --> 01:01:25.119
It's incredible, right?
01:01:25.360 --> 01:01:25.599
Sure.
01:01:25.760 --> 01:01:26.079
Yeah.
01:01:26.400 --> 01:01:29.119
And the knowledge base is getting huge.
01:01:29.519 --> 01:01:32.400
Like how we treat things, or there's so many pathways, right?
01:01:32.480 --> 01:01:37.199
Like I always go down my my methylene blue, you know, presentation pipeline, right?
01:01:39.280 --> 01:01:42.159
Presentations you could probably ever give, just saying.
01:01:43.199 --> 01:01:47.760
I thought it was good, but you know, it's just it is a little bit rabbit hole, right?
01:01:47.840 --> 01:01:50.000
But it I don't give those to you, though.
01:01:50.400 --> 01:01:52.559
It incites the mind to think.
01:01:52.719 --> 01:01:53.760
That's that's what it's there for.
01:01:54.000 --> 01:01:57.519
But but the point I'm that that proves the point, like right here.
01:01:57.760 --> 01:02:00.960
Like, Mike, that like that proves that point, right?
01:02:01.039 --> 01:02:12.320
And and so of why the human factor shall and always must be part of the healthcare and how we treat patients, right?
01:02:12.400 --> 01:02:17.440
I I'm thinking back to what you were saying earlier about let's say using AI as an assist, right?
01:02:17.760 --> 01:02:19.039
Helping radiology.
01:02:19.519 --> 01:02:19.840
Okay.
01:02:20.400 --> 01:02:25.360
So as the radiologist, if you miss something and AI catches it, right?
01:02:26.480 --> 01:02:28.719
Who's responsible for that, right?
01:02:29.119 --> 01:02:38.079
And you know, to to speak to Mike's point, like the the human element factor, it it can't go away, right?
01:02:38.159 --> 01:02:48.880
We train, we've advocated so hard for not only the field, just all of us in healthcare, like whether you're at a basic level, all the way on up to critical care, even physician level, right?
01:02:49.199 --> 01:02:59.119
But the human part of us, like we must always know that that we're all we're gonna make mistakes because we truly are that human, right?
01:02:59.679 --> 01:03:07.119
And and I think in a society where now there's almost this, like I don't know how to describe it.
01:03:07.280 --> 01:03:16.159
Maybe there's like this dependence that technology and tools and all these things will like essentially make us perfect.
01:03:16.320 --> 01:03:24.559
I I think that's the the wrong pathway or or or way to think that this will make us, you know, it to that level.
01:03:24.719 --> 01:03:31.199
But the the human element will always be there, and I think we have to trust that and and never let that go.
01:03:31.360 --> 01:03:34.159
Like, yeah, we're gonna make mistakes, like that, but guess what?
01:03:34.480 --> 01:03:38.159
Artificial learning, and like it's literally in the name artificial.
01:03:38.480 --> 01:03:58.880
So like it's going to make mistakes, and we know that that's that's a possibility, and and I wholeheartedly agree that that human element, because if it was that easy, like if it truly was that easy to take everything that we've learned and all this data and then create something that is so perfect, then why isn't it happening?
01:03:59.119 --> 01:03:59.760
You know, right?
01:03:59.920 --> 01:04:23.760
Which uh it isn't, and I think that really kind of rides this point where a lot of the reasons why I felt this discussion was going to be really beneficial is using it to help us and allow us to assist us as providers to get us to that next level, but we can't replace that human element to what we do and how we take care of our patients.
01:04:24.480 --> 01:04:27.039
Yeah, so Mike, don't lose your fire, my friend.
01:04:27.280 --> 01:04:27.760
That's great.
01:04:28.000 --> 01:04:31.840
Because but seriously, because you're you're protecting your people and your patients, and that's critical.
01:04:32.000 --> 01:04:32.559
And absolutely.
01:04:32.800 --> 01:04:34.559
And I and there's five points.
01:04:34.960 --> 01:04:45.119
If there's any leaders listening who have any influence on buying these things, there are five things that you have to think about whenever you buy an AI product for your team.
01:04:45.280 --> 01:04:48.400
Number one, is it safe for your teammates?
01:04:48.559 --> 01:04:49.599
Is it safe for you?
01:04:49.760 --> 01:04:51.119
Is it safe for your patients?
01:04:51.199 --> 01:04:51.920
That's number one.
01:04:52.400 --> 01:04:54.719
Number two, is it secure?
01:04:55.199 --> 01:04:57.599
Is it secure for your patients, your people?
01:04:58.239 --> 01:04:59.519
Is it secure for you?
01:05:00.320 --> 01:05:01.280
Data quality.
01:05:01.760 --> 01:05:04.480
Is your data quality good enough to start even using it?
01:05:04.880 --> 01:05:08.000
Is the data going into it even going to help you with good outputs?
01:05:08.320 --> 01:05:09.920
The fourth thing is your culture.
01:05:10.159 --> 01:05:17.679
Is your culture in a place where, to Mike's point, are you leaning too hard on that technology to make decisions for you?
01:05:17.840 --> 01:05:26.639
Or do you have a good group of clinicians who are very, very sharp, who are on the basics, who can take care of patients, who can care for them?
01:05:26.880 --> 01:05:31.280
And this will help offload mundane tasks so they can focus on that.
01:05:31.519 --> 01:05:37.199
And then the fifth thing, are your policies aligned to safely take on these technologies?
01:05:37.679 --> 01:05:39.360
So that those are the five points.
01:05:39.519 --> 01:05:49.280
Like you take those five points with you when you're buying these technologies, or when a salesman comes to you, if you can answer them correct in a way that's like, Yes, this would help me, then okay, go for it.
01:05:49.519 --> 01:05:58.159
But if you can't answer those questions correctly or properly, I I would highly recommend getting yourself in a good state before you take those on.
01:05:59.119 --> 01:05:59.280
Yeah.
01:06:00.079 --> 01:06:02.559
And I and I I I I love that you said that.
01:06:02.639 --> 01:06:06.400
And I'm I'm in agreement with you for the most part on all this stuff.
01:06:06.719 --> 01:06:19.679
Uh but what I keep going, you know, I I think where my fear is is stemming from in all of this is the the power of suggestion can be very compelling at times.
01:06:20.239 --> 01:06:31.840
And there may be that patient situation in front of you where you may not necessarily be 100% sure of what's going on, but you got a strong gut instinct of where you should potentially go.
01:06:33.679 --> 01:06:46.559
I don't want to see us rely on AI in those circumstances to suggest something that may not be in the best interest of our patient because not all of the contextual information is there.
01:06:46.880 --> 01:06:55.760
And that's what I fear for us as clinicians to be able to maintain that gut, to maintain the knowledge base that we have, to continually build upon that.
01:06:55.920 --> 01:06:57.760
And I think that AI is a great tool for that.
01:06:57.840 --> 01:06:58.800
I really do.
01:07:00.960 --> 01:07:22.559
I just I'm very, very cautious about how it's going to be able to have the insight into everything that's going on with our patients as well as we are with the years of experience and the gut that we have and the tools at our disposal.
01:07:22.800 --> 01:07:34.079
I mean, that that doorway assessment that we do is like one of the most important things that that we can ever do for our patients, I think, and gathering all of that information.
01:07:34.159 --> 01:07:48.239
And I can't tell you how many times I have walked into this extremely critical situation, transferring an ECMO or an Impella or what have you, and just stop for 30 seconds or a minute and just look around the room and take it all in.
01:07:49.519 --> 01:07:56.639
And then I'm like, okay, this is how this is what helps my helps me prioritize and organize and gives me my game plan of everything that's gonna happen.
01:07:57.039 --> 01:08:06.880
And and I I I just I don't want to see us relying on AI to take us down a path that maybe isn't where we should be going.
01:08:09.440 --> 01:08:10.800
I agree completely, Mike.
01:08:10.880 --> 01:08:16.000
And this is where now everything you said, we and we had talked about this on other podcasts.
01:08:16.159 --> 01:08:19.600
To me, this is where the future flight leaders program comes into play.
01:08:20.000 --> 01:08:22.399
And this is where good FTOs come into play, right?
01:08:22.479 --> 01:08:23.840
Think about the FTO I was just telling you about.
01:08:23.920 --> 01:08:27.439
My FTO right now, he told me, Christian, get your head out of the paperwork, man.
01:08:27.680 --> 01:08:29.920
Look at your patient, take care of your patient.
01:08:30.159 --> 01:08:32.319
You know, that's the key.
01:08:32.560 --> 01:08:37.840
Having good FTOs, having your your experienced people who are like, they know better.
01:08:38.000 --> 01:08:43.039
They like to if you were my FTO, Mike, I'd love it because you do think about that stuff, right?
01:08:43.279 --> 01:08:44.560
We have to keep that alive.
01:08:44.640 --> 01:09:00.079
And that's gonna come through training, that's gonna come through um having those good CQI discussions, having good FTOs that are trained on how to train people and understand how people learn and those concepts that you can't just quantify.
01:09:00.239 --> 01:09:04.239
Because in my world in data, I can't quantify everything, right?
01:09:04.399 --> 01:09:08.399
Intuition is not necessarily quantifiable, despite what some textbooks might say.
01:09:08.479 --> 01:09:13.680
They say, Oh, well, you know, intuition is based on data that you have that you just don't necessarily fully understand.
01:09:13.840 --> 01:09:21.439
I agree with that, but I also disagree with it too, because I think sometimes the gut feeling is going to help you say, you know what, I think this patient needs this right now, and I've had those moments.
01:09:21.680 --> 01:09:24.960
And so to your point, I think it's once again, it's a balance.
01:09:25.279 --> 01:09:27.199
And you've seen it in the news too, right?
01:09:27.359 --> 01:09:34.880
Have you guys seen the news in the last few days about some of that stuff that's been coming out with AI and being concerned with the dangers of it?
01:09:35.039 --> 01:09:35.279
Yep.
01:09:35.520 --> 01:09:36.479
Good, good.
01:09:36.880 --> 01:09:38.079
I'm glad they're doing that.
01:09:38.239 --> 01:09:42.640
Because as much as I am as an advocate for it, I'm also a cautious optimist about it as well.
01:09:42.800 --> 01:09:57.439
That's why I'm learning about it and wanting to become a strategist in it, because and I think every leader should, in their own way, become a strategist in AI because uh you're gonna have your crew members coming up saying, Hey, get this tool, get this, get that, buy this, buy that, let's do this.
01:09:57.760 --> 01:10:03.439
And that might be great, but we also have vibe coded garbage that's out there too that people are gonna bring you.
01:10:03.680 --> 01:10:07.600
It's vibe coded, and they're gonna go, Oh, yeah, this is this is excellent.
01:10:07.680 --> 01:10:09.039
And you're like, Yeah, okay.
01:10:09.199 --> 01:10:10.239
No, yeah, sure.
01:10:10.399 --> 01:10:17.279
Um, let's look at the the safety and securities piece of it, and then you tear it apart and you find it's the most unsecure, dangerous thing you've ever seen in your life.
01:10:17.439 --> 01:10:20.560
So no, I mean, I I I yeah, I think that's great, Mike.
01:10:20.640 --> 01:10:21.760
I love your insight on this, my friend.
01:10:21.840 --> 01:10:22.640
I think that's a good idea.
01:10:22.880 --> 01:10:32.239
Sorry to get so passionate there, but please I don't I don't feel like I don't feel like even and and I'll apologize to the listeners too because I don't feel like you've heard that kind of passion come out of me in a while.
01:10:32.560 --> 01:10:39.199
But that is definitely it's just one of those things that I I I I'm that's where I struggle the biggest with all of this.
01:10:39.520 --> 01:10:43.439
But that's the whole point of like that these are conversations that we should be having, right?
01:10:43.520 --> 01:10:45.359
And and I know that we have these at base level.
01:10:45.439 --> 01:10:46.079
I have them all the time.
01:10:46.239 --> 01:10:49.199
And and Christian and Mike, I'm gonna throw this example out there, right?
01:10:49.279 --> 01:10:58.960
And this kind of leads me into kind of our like that next section when you're talking about how this can really apply into leadership, the organizational responsibility of using AI correctly.
01:10:59.119 --> 01:11:00.239
I I will never forget this.
01:11:00.319 --> 01:11:01.760
This happened to be about a year ago.
01:11:01.920 --> 01:11:07.039
I had one of my dear friends uh ask me about a patient encounter that they had.
01:11:07.359 --> 01:11:21.279
The the parameters that they used, they they they ended up going with like this, this higher peep initially, because of uh what they did as part of a I know this is strange, a hands-on physical assessment, right?
01:11:21.439 --> 01:11:36.079
Like, you know, using a stethoscope, touching the patient, doing a full assessment, listening to lung sounds, then getting them onto non-invasive, then uh going through a drug-assisted airway, and then continuing that on.
01:11:36.239 --> 01:11:53.279
And I'll never forget because the person, the the people that are responsible for reviewing those types of calls completely came back with like this whole uh well, why did you go to to that level of PEEP, you know, to start off with?
01:11:53.840 --> 01:12:06.319
And it was very clear in the documentation based on what the provider encountered and what they ended up with and then transitioned to and kind of kept that going.
01:12:06.479 --> 01:12:10.239
And the patient acted great with that therapy.
01:12:10.640 --> 01:12:15.680
And I just remember they were they were asking me, hey, what do you think of this?
01:12:16.239 --> 01:12:22.640
And before I even read the the the QA note, I was like, Yeah, that's pretty good.
01:12:22.720 --> 01:12:28.079
I feel I feel like you transitioned well, you continued what you you started to a position where it was working.
01:12:28.319 --> 01:12:32.079
The patient improved, but yet they still had a lot of work of breathing.
01:12:32.159 --> 01:12:34.720
So you know you did the things and they did better.
01:12:35.039 --> 01:12:38.399
And I was like, I feel like you're what you documented supported that.
01:12:38.479 --> 01:13:10.319
And I'll just never forget the reply from the the person, and then my bias in knowing the person that did this is someone that is experienced, and their response and everything why you can read something and totally tell when it's AI written completely, and it just floored me only in the sense of like you didn't even take into consideration reading what physically had like what was documented, right?
01:13:10.399 --> 01:13:28.399
Because that's the only info that they had, they weren't on the call, and just reading this um reply, there was no human, like empathetical, like it was just all AI, like, well, based on the uh protocol or based on this, and it was just like, who are you?
01:13:28.560 --> 01:13:29.840
And like, why would you do that?
01:13:29.920 --> 01:13:36.000
And I just remember feeling for my friend and being like, Man, that sucks.
01:13:36.239 --> 01:13:59.359
Like, that's not even a way to have like a true actual like socratic discussion about what happened, and it was just it was totally reliant on an artificial like reply of well, these might be my points, but say this, and it didn't even speak to what was really going on with the patient.
01:13:59.439 --> 01:14:12.319
And I think for me that was like, wow, as someone as part of like training and like leading people and with experience, completely not something that I would have expected for from someone in that position.
01:14:12.479 --> 01:14:13.840
That's why I'm just wondering.
01:14:14.000 --> 01:14:32.479
So using that type of example, and then and Mike, that that also points to I think not only are we uh reliant on us, that human element in taking care of patients, that also transitions when you become a leader, when you start teaching and using education.
01:14:32.640 --> 01:14:35.840
I think it's the the podcast that we just did with Dr.
01:14:36.000 --> 01:14:43.279
Uh Jens The only way to get good and better at medicine is doing medicine and practicing.
01:14:43.359 --> 01:14:45.439
So that experience means something.
01:14:45.600 --> 01:14:45.840
Yeah.
01:14:46.000 --> 01:15:02.000
And I'm just curious how we can use like and not lose that part of like the AI reliance, but the assistance to help us have a true foundational leadership and organizational responsibility to do it the right way as we go forward.
01:15:02.319 --> 01:15:07.520
Your situation, Blake, it violated two of my rules culture and policy.
01:15:08.000 --> 01:15:08.399
Right?
01:15:08.640 --> 01:15:08.960
Yeah.
01:15:09.119 --> 01:15:15.439
Like number one, you should have policy that says for any sort of CQI interactions, AI is not permitted, right?
01:15:16.000 --> 01:15:17.520
Number two, culture.
01:15:17.680 --> 01:15:23.039
If your culture's not in a place to use the tools correctly, then you don't need the tools, right?
01:15:23.199 --> 01:15:24.239
That's that's number two.
01:15:24.399 --> 01:15:33.039
So for me, that's two out of the five rules we violated in that, and for me at least, you know, I have to make sure my team is in a place where it could use the tool properly.
01:15:33.199 --> 01:15:38.399
Like, are is our culture in a place where we can accept using this, or is this going to become a a a lever?
01:15:38.640 --> 01:15:42.079
For example, I once got an email, cracks me up, man.
01:15:42.399 --> 01:15:46.479
It said, hello, comma, in brackets, teammate.
01:15:47.760 --> 01:15:50.880
I was like, you can't even put my name on the email.
01:15:51.119 --> 01:15:52.079
You can't even write it out.
01:15:52.319 --> 01:15:52.960
Come on, man.
01:15:53.119 --> 01:15:54.720
Like, you remember that sports channel?
01:15:54.800 --> 01:15:55.279
Come on, man.
01:15:55.600 --> 01:15:56.079
Come on, man.
01:15:56.319 --> 01:15:57.119
Yeah, come on, man.
01:15:57.359 --> 01:15:59.039
Yeah, from ESP Sports Center.
01:15:59.279 --> 01:15:59.520
Yeah.
01:15:59.680 --> 01:15:59.920
Yeah.
01:16:00.399 --> 01:16:07.520
And then the other bit is uh the patterns they pick up on are like M dashes, those long dashes, yeah, run on sentences.
01:16:07.760 --> 01:16:20.239
You sound like a Nobel laureate and you're from like Texas, when you say y'all, but you're like, oh, you guys, and then it's like words that you know, I was like, uh is this from a uh like a Nobel paper or something?
01:16:20.319 --> 01:16:25.359
Because it's or my favorite is it's not just a job, it's a calling.
01:16:25.520 --> 01:16:29.279
Like when you say it's not just this, it's that, like those statements too.
01:16:29.439 --> 01:16:31.840
So it's really it's funny when you when you see that.
01:16:31.920 --> 01:16:39.600
But it to your point though, I think it hurts culture because when you're using these tools for those reasons, like for example, it once again, human in the loop.
01:16:39.680 --> 01:16:44.000
If I write an email and being on spectrum, sometimes I don't come across right.
01:16:44.159 --> 01:16:48.079
Sometimes I come across a little Sheldon Cooper like and it frustrates people, right?
01:16:48.159 --> 01:16:50.000
Because we have social rules in society.
01:16:50.239 --> 01:16:55.119
I will put it in Chat GPT and go, hey, listen, this is the tone in which I think it is.
01:16:55.279 --> 01:16:58.880
Help me understand a better way that I could write this in a tone that works for me.
01:16:59.039 --> 01:17:03.680
You are a blank, like you are a writing expert.
01:17:03.760 --> 01:17:05.680
Um, I use a persona prompt, right?
01:17:05.840 --> 01:17:07.600
And so prompt engineering can help with those things.
01:17:07.680 --> 01:17:10.640
And then it would give me an output and I go, okay, I like this.
01:17:10.800 --> 01:17:17.520
Then I go over on my other screen and I write the email myself using my words, not copy and pasting.
01:17:17.600 --> 01:17:22.960
I do not copy and paste outputs typically because I want to make sure that it's me doing it.
01:17:23.199 --> 01:17:23.439
Sure.
01:17:23.680 --> 01:17:26.560
This just suggested ways that I could do it better.
01:17:26.640 --> 01:17:34.079
It's like having that person, like, you know, if I call Mike Fidel and go, hey, listen, man, I got, you know, this, this, I took care of this kiddo, he's trached on a on event.
01:17:34.319 --> 01:17:35.199
Can you help me out with this?
01:17:35.279 --> 01:17:37.199
I know you have a lot of experience with it, right?
01:17:37.359 --> 01:17:39.680
I would look at it as a similar interaction.
01:17:39.840 --> 01:17:46.000
Um, but the one thing that you would have that the GPT or the LLM would not have is that experience, the context.
01:17:46.159 --> 01:17:50.079
Once again, context is everything when it comes to these tools.
01:17:50.239 --> 01:17:57.119
And that's where as data gets more prominent, hence all these massive controversial data centers are popping up everywhere.
01:17:57.359 --> 01:18:00.239
That data that's being fed in there is going to provide it more context.
01:18:00.319 --> 01:18:01.600
So, yes, it will get better.
01:18:01.760 --> 01:18:04.079
But once again, that's why we're talking about the five to 10 year thing.
01:18:04.159 --> 01:18:05.199
That's why I say, I don't know.
01:18:05.359 --> 01:18:08.640
It just depends on what the context of these, of these, of these models are.
01:18:08.800 --> 01:18:11.920
Like, are they gonna be good enough to where they can create some context and nuance?
01:18:12.000 --> 01:18:16.079
Not really for medicine, but I think overall, I think it has a place.
01:18:16.159 --> 01:18:18.800
But once again, your culture has to be right for it.
01:18:18.960 --> 01:18:24.319
Your policies have to be right for it, you know, and it that's just what it's gonna come down to.
01:18:24.479 --> 01:18:32.159
And at the end of the day, you know, we have to have good precepting, good people are gonna hold people accountable to good clinical care.
01:18:32.319 --> 01:18:34.000
Like there, there's no excuse for not.
01:18:34.079 --> 01:18:38.079
Um, you can't be lazy and just rely on these on these technologies completely.
01:18:38.479 --> 01:18:42.079
But but that's where this this is something I and I get it.
01:18:42.159 --> 01:18:47.520
The technology world is gonna expand at an exponential pace that none of us will ever be able to keep up with.
01:18:48.000 --> 01:18:50.159
Totally understand that aspect of it.
01:18:50.399 --> 01:18:54.960
But this is one of those things where it's like we do need to pump the brakes a little bit.
01:18:55.119 --> 01:19:06.159
And just because this is a cool tool that can do lots of fun stuff doesn't mean that this is supposed to be the end all be-all for every industry that's out there.
01:19:06.399 --> 01:19:21.760
And and and that's where I think when something grows too fast and all of we we don't we don't slow down for a second, think about all the things that yes, it could benefit, but also think about all the things that it could potentially hurt.
01:19:22.000 --> 01:19:26.640
Yes, that is where I fear we get into more trouble.
01:19:26.880 --> 01:19:31.760
And and I I I maybe that's where my resistance was coming from.
01:19:31.840 --> 01:19:43.439
Maybe that's where I was struggling with maybe accepting some of this right now, is I see how how exponentially this is growing, and I don't want us to do this the wrong way.
01:19:43.760 --> 01:19:45.520
Well, I don't I don't think you're resistant, Mike.
01:19:45.600 --> 01:19:49.600
I think it's it's not cautious, and and and rightfully so.
01:19:49.680 --> 01:19:59.439
I I think that's just a natural, good, ethical way to to really look really kind of look at it, at least from my perspective.
01:19:59.520 --> 01:20:01.199
I look at it as being cautious.
01:20:01.600 --> 01:20:01.840
Yeah.
01:20:02.159 --> 01:20:03.279
I agree completely, Mike.
01:20:03.439 --> 01:20:08.640
You uh if every leader was like you in that perspective, I think we would be much safer with using these tools.
01:20:08.800 --> 01:20:10.239
Once again, it's a tool, man.
01:20:10.399 --> 01:20:11.039
You know what I mean?
01:20:11.199 --> 01:20:13.920
You don't use a hammer to hit to do everything, right?
01:20:14.239 --> 01:20:15.840
I've been a leader for three seconds.
01:20:16.079 --> 01:20:18.720
Like I it literally counts.
01:20:18.960 --> 01:20:20.880
I've been a clinician for a heck of a lot longer.
01:20:20.960 --> 01:20:26.720
And so that's where I think my my most of my fear is coming from is from the clinician side of this.
01:20:29.119 --> 01:20:29.520
Right.
01:20:29.760 --> 01:20:35.680
But that's where leaders are gonna have to step up and understand the strategies on these things.
01:20:35.760 --> 01:20:44.159
And that's why I do hope some leaders listen to this because I want them to understand, like, hey, you know, on the business side, okay, you know, that's analytics.
01:20:44.319 --> 01:20:46.000
That's where this stuff came from.
01:20:46.239 --> 01:20:48.720
In fact, a little history, all right.
01:20:48.960 --> 01:20:51.119
So I can put some notes up here on my screen.
01:20:51.279 --> 01:20:54.880
So so this actually started in 1950.
01:20:55.039 --> 01:20:56.960
Alan Turing, are you familiar with the the name?
01:20:57.039 --> 01:20:57.920
The movie Enigma.
01:20:58.319 --> 01:20:58.960
You familiar?
01:20:59.119 --> 01:21:02.640
So Alan Turing wrote a paper called The Imitation Game.
01:21:02.800 --> 01:21:06.800
This is uh he broke the the code for Nazi Germany against Nazi Germany and the British.
01:21:07.039 --> 01:21:07.279
Yes.
01:21:07.439 --> 01:21:07.680
Yes.
01:21:07.840 --> 01:21:09.840
So that's where the Turing test came from.
01:21:10.000 --> 01:21:12.720
Like, is this human or is this a machine, right?
01:21:12.880 --> 01:21:17.680
And then in 1956, John McCarthy actually coined the term artificial intelligence.
01:21:17.840 --> 01:21:20.720
So this was at uh in Dartmouth at the Dartmouth conference.
01:21:20.880 --> 01:21:22.960
So that's when the field was founded.
01:21:23.119 --> 01:21:27.439
And in 1960, we actually had our first chatbot, believe it or not, named Eliza.
01:21:27.840 --> 01:21:30.880
Um, and then in in 2022, that's when we had ChatGPT.
01:21:31.039 --> 01:21:35.279
So this technology has been building for generations, actually.
01:21:35.600 --> 01:21:39.439
It it just goes to show that, like, you're right, Mike.
01:21:39.520 --> 01:21:46.479
I think what happened was is in 2021, 2022, we hit this point and now we're exponential, in which that should be alarming, right?
01:21:46.560 --> 01:21:51.199
It's like when you look at Enron's stock prices and in the early 2000s, they're rising very fast.
01:21:51.439 --> 01:21:51.840
Why?
01:21:52.079 --> 01:21:52.399
Right.
01:21:52.560 --> 01:21:54.800
And it turns out there are some issues.
01:21:54.960 --> 01:21:56.800
But I think the same thing applies here.
01:21:56.960 --> 01:22:00.000
That, you know, yes, this is a expanding and growing technology.
01:22:00.079 --> 01:22:01.439
It's going very, very fast.
01:22:01.600 --> 01:22:03.439
Leaders are gonna have to understand it quickly.
01:22:03.520 --> 01:22:11.199
Like they need to go and read on this stuff and understand what they are, what they do, what they can do, what they can't do with it, what the strategy is, where it should be safe.
01:22:11.439 --> 01:22:19.279
That's why I say, you know, uh it's gonna be uh if you really want to take on these technologies, you're gonna have to do a heck of a lot of work up front, you know?
01:22:19.600 --> 01:22:24.640
And if you can't get the basics down of your program, then these tools are not for you.
01:22:24.800 --> 01:22:28.399
Um, you need to have the basic stuff done first.
01:22:28.800 --> 01:22:37.359
Like your pol, you need to be a high-performing program with a good culture, and then you have to find safe niches where this will apply.
01:22:37.520 --> 01:22:40.880
For example, I like I said, charting is I think is a very good one, right?
01:22:41.039 --> 01:22:42.960
Because we get tired, we get fatigued.
01:22:43.039 --> 01:22:46.079
That can help with fatigue, risk, it can help with a lot of things there.
01:22:46.239 --> 01:22:49.840
Um, it can help with, you know, even overtime and other little things like that.
01:22:50.000 --> 01:23:01.119
But once again, the clinical decision and support, I think that is, I think it is best used probably for the inpatient settings and in the academic settings, um, where you have more time to express these things.
01:23:01.199 --> 01:23:13.039
But we mean the nature of our business, Mike, as you know, as long as you've been doing Sam Blake, we're a very time-sensitive bunch, um, which means that I'm not always gonna have time to pull out my up to date, you know, when I'm in flight, right?
01:23:13.119 --> 01:23:18.159
I'm gonna have those patients that are sick and I need to put down the freaking paperwork and take care of my patient, right?
01:23:18.479 --> 01:23:20.000
So I love it.
01:23:20.159 --> 01:23:20.479
I do.
01:23:20.960 --> 01:23:24.319
I'm just really wondering then, five to ten years from now.
01:23:24.640 --> 01:23:29.279
I mean, I'll just be honest, five to ten years from now is uh retirement, you know, for me.
01:23:29.920 --> 01:23:38.000
And maybe just from one aspect of the job, not stopping work completely, but moving on to something else.
01:23:38.239 --> 01:23:49.199
And I can't even imagine what younger, newer clinicians are having to to learn and grow through and do so much that are coming into this field.
01:23:51.279 --> 01:23:56.560
What does this process really look like for us in that five to ten years from now?
01:23:56.800 --> 01:23:59.760
What how is this how is AI helping us?
01:24:00.239 --> 01:24:09.119
Like what should we expect for future flight clinicians five to ten years from now as you're looking to to get into this field?
01:24:09.439 --> 01:24:15.760
What should we what should they expect as they start looking and evolving and growing?
01:24:15.840 --> 01:24:26.159
Because I look at people that are like brand new today that are gonna be looking at in that five-year mark to start embarking on getting flight jobs, breaking into the hems industry.
01:24:26.319 --> 01:24:33.680
But what does the role of AI and and helping people adapt, what does that look like in the future?
01:24:34.319 --> 01:24:36.239
So, very good question once again.
01:24:36.319 --> 01:24:38.880
And I think there's two pathways that we can take.
01:24:39.119 --> 01:24:49.840
Pathway number one is the pathway I would hope we take, which is we as a as an industry start looking at ourselves and going, okay, we need to define our terms.
01:24:50.079 --> 01:24:54.079
We need to make sure that we're consistent in all aspects of our program.
01:24:54.399 --> 01:24:58.079
We need to make sure that the business use case for it is real.
01:24:58.640 --> 01:25:07.600
We need to make sure that the tools that are available to us, we're using them to their fullest extent before we start looking at things like AI tools.
01:25:07.760 --> 01:25:11.680
And then when AI tools are being used, there's a process for it.
01:25:11.920 --> 01:25:15.520
It's you, your IT team has vetted it to be safe and you begin to use it.
01:25:15.680 --> 01:25:21.760
Now, for what applications, I think it's gonna be where we see it now, which is uh EPCR products.
01:25:21.920 --> 01:25:26.000
I think it's a great and safe place for it for the most part, depending that your culture is good.
01:25:26.079 --> 01:25:27.039
Remember my five points?
01:25:27.199 --> 01:25:28.319
Culture is one of those.
01:25:28.479 --> 01:25:29.920
Your culture has to be good for it.
01:25:30.079 --> 01:25:37.119
Are your clinicians just gonna hit accept and submit and then be like, oh yeah, I just I just let the AI do it, but I didn't review it.
01:25:37.279 --> 01:25:38.560
That's that's part of it.
01:25:38.720 --> 01:25:42.159
Um, we're also gonna see it, I think, in things like quality tools.
01:25:42.239 --> 01:25:44.560
Uh the big I think that's the big stuff is quality.
01:25:44.720 --> 01:25:56.640
Uh so like running AI prompts against your charting system, be like, hey, you know, like dash one a uh metrics or or or your your CAMES metrics, your your CAMES PIF when you're plugging in that data, right?
01:25:56.720 --> 01:26:00.319
Sometimes you have to go, well, what's the definition of dash 1A metrics?
01:26:00.399 --> 01:26:02.960
What is the definition of a first pass success rate?
01:26:03.119 --> 01:26:09.439
And using AI tools to be able to help run those numbers for you based on the definition that you plug in through proper prompt engineering.
01:26:09.600 --> 01:26:12.319
Um I will see that coming into market, I'm sure.
01:26:12.399 --> 01:26:16.640
Uh, we're gonna begin to seeing uh the testing of different recording devices.
01:26:16.800 --> 01:26:25.520
So, like, you know, in doctors' offices, there's a lot of like plot and all these other tools that will record encounters so that charting and documentation is more accurate.
01:26:25.680 --> 01:26:35.680
We're also gonna see it in education, clinical uh decision tools to help with uh education, being able to run literature reviews, being able to help develop research, all these things.
01:26:35.920 --> 01:26:46.479
Now, 10 years from now, I could foresee there being more of the push into the clinical environment, uh, such as what Mike's concern is, which once again, rightfully so.
01:26:46.640 --> 01:26:49.039
Like if it was to happen right now, I'd be very concerned.
01:26:49.359 --> 01:26:57.439
Um, but I think that as time moves on, um, we're gonna learn more about these technologies and how they work and how they work for our industry.
01:26:57.520 --> 01:27:05.600
And then we will begin to learn to say, okay, let's pump the brakes, let's be smart about this, let's, you know, answer the five questions.
01:27:05.920 --> 01:27:07.520
Does this work for our team?
01:27:07.760 --> 01:27:09.600
Will this actually work in practice?
01:27:09.840 --> 01:27:10.640
Yes or no.
01:27:10.880 --> 01:27:16.640
The other thing is it's not like just buying a medical device where you just go, okay, we'll do a little training toss on the aircraft.
01:27:16.720 --> 01:27:30.800
This is gonna have to be full work, like your workflow integration, which means you're gonna have to understand your workflow on calls to see how this thing could actually work uh within your program without causing patient harm.
01:27:30.960 --> 01:27:32.880
And then also we need to conduct studies too.
01:27:32.960 --> 01:27:48.159
So I foresee a lot of people like me and others, uh, myself and others, who want to conduct more research in this realm so that we can understand what it's doing, where it's going, and what we need to do to be keep our patients safe and keep our programs uh steady and doing the right thing.
01:27:48.399 --> 01:27:50.560
Um so once again, it's not a clear answer.
01:27:50.640 --> 01:27:52.640
It's definitely a it depends answer.
01:27:52.880 --> 01:27:59.600
Now, I will add the other way I could see it going is a large majority of programs taking these tools.
01:28:00.079 --> 01:28:04.239
Out of the box and trying to replace that clinical decision, right?
01:28:04.399 --> 01:28:14.000
That to Mike your point, that that that assessment across the room, hey, I did this because my gut was telling me that I know this is gonna work because I've seen this movie play out before.
01:28:14.159 --> 01:28:16.560
I need to take care of this patient and it worked, right?
01:28:16.720 --> 01:28:21.520
You can't always explain it with a clinical decision tool because it doesn't see what you see, right?
01:28:21.680 --> 01:28:29.359
So I could see some programs making some very bad decisions by saying, hey, we're just gonna buy this stuff because it's cool and well, it's AI, right?
01:28:29.520 --> 01:28:34.079
But they don't have a foundation that's a proper concrete slab with rebar and everything else.
01:28:34.159 --> 01:28:34.960
It's gonna be sand.
01:28:35.119 --> 01:28:36.720
You build a house on sand, what happens?
01:28:36.800 --> 01:28:38.000
It falls apart, right?
01:28:38.079 --> 01:28:42.079
Um, I hope that's not the case, but it is a real possibility, I think.
01:28:42.239 --> 01:28:55.119
So, not to all be an alarmist, but I think we should be realistic and go, okay, well, since we're humans, we know that we can either take the easy button path, which is the sand, or we understand this is gonna be a lot of work.
01:28:55.520 --> 01:29:24.560
But if we we we dial it in right, it can help reduce the noise that's out here to where we can focus on what we should be focusing on, which is exactly Mike what you outlined, which is using our our brains, our intuition, our experience, our caring hearts, the things that we care about most, which is our patience, our passion, our drive, things you can't put into a computer to be able to make good clinical decisions.
01:29:24.720 --> 01:29:34.960
To me, that's that's the good vision here is how can I offload the stuff that's frustrating so that my clinicians can have a good culture where they can do the thing that we we we pay them for, right?
01:29:35.039 --> 01:29:38.159
We we're paid as clinicians to take care of people, right?
01:29:38.399 --> 01:29:39.520
Not let machines do it.
01:29:39.600 --> 01:29:40.960
Otherwise, why would we be here?
01:29:41.119 --> 01:29:42.960
So, you know, that's that's how I see it.
01:29:43.119 --> 01:29:44.159
That's the future, right?
01:29:44.319 --> 01:29:48.880
We have one of two pathways, and I certainly hope we take that good pathway uh moving forward.
01:29:49.199 --> 01:29:50.399
Yeah, I certainly do too.
01:29:50.479 --> 01:29:55.600
And and I hope that you just come back and tell us when it's okay.
01:29:57.359 --> 01:30:01.279
I don't, you know, I don't know if it's ever gonna be a hundred percent okay.
01:30:01.439 --> 01:30:19.680
I think we're we have to build frameworks and other things to go, you know, that's what things I I want to do with my life's work one day is hopefully help build us some frameworks and work with other experts to understand what this stuff is and we can and do some research on where the need's gonna be and where it's safe and reasonable to do so.
01:30:19.840 --> 01:30:23.039
Um, because right now that there's a big gold rush ongoing.
01:30:23.279 --> 01:30:27.600
I work a little bit in the software development side of things, and I can tell you there's a gold rush.
01:30:27.760 --> 01:30:32.000
And people are in the rush to build the next best hottest thing.
01:30:32.159 --> 01:30:39.600
And um, you you have to look at frameworks like NIST and others and and ask the question, is this gonna be safe?
01:30:39.760 --> 01:30:41.439
And is this really gonna be useful?
01:30:41.600 --> 01:30:46.800
Or is a salesman coming after me and are they just making really good work of my of my psyche, right?
01:30:46.960 --> 01:30:49.279
Are they are they doing a good job as a salesman?
01:30:49.439 --> 01:30:49.920
Which is fine.
01:30:50.000 --> 01:30:52.000
I mean, there's great salesmen out there.
01:30:52.079 --> 01:30:55.680
Um, but personally I like the ones who are a little more blunt than those who are not.
01:30:55.840 --> 01:30:57.520
But just tell me what it can do.
01:30:57.680 --> 01:31:00.399
And so that's my big thing is don't be swayed completely by sales.
01:31:00.479 --> 01:31:03.199
Uh as they always jokingly say, trust but verify comrade.
01:31:03.359 --> 01:31:04.479
Make sure you're doing that.
01:31:04.720 --> 01:31:07.039
Um and make sure you know what you're looking for.
01:31:07.119 --> 01:31:11.039
Um if you do that, then you if you really care, you'll make good decisions.
01:31:11.359 --> 01:31:12.720
I I hope that we go that route.
01:31:12.800 --> 01:31:13.279
I really do.
01:31:13.359 --> 01:31:16.399
Uh and I am excited to see what this is going to look like.
01:31:16.560 --> 01:31:27.439
And I I think that there are many avenues within our industry, specifically in hems and public safety aviation, that AI tools can be a huge help instead of a hindrance.
01:31:27.600 --> 01:31:28.960
I really do feel that way.
01:31:29.119 --> 01:31:31.680
Uh I just have some significant concerns right now.
01:31:31.760 --> 01:31:36.640
You know, like if AI tells me to push the nuclear codes and like am I gonna do it?
01:31:36.800 --> 01:31:38.479
Like that that's right.
01:31:39.039 --> 01:31:41.359
I I I that's where I struggle.
01:31:41.439 --> 01:31:49.199
So I I you know, again, forgive me for my soapbox earlier, but I I I also want to look at ways that like we're we're in the now.
01:31:49.279 --> 01:31:51.199
Like this is happening all around us, right?
01:31:51.359 --> 01:31:59.119
And and I have many of my teammates that are actively looking at ways that this can be more integrated into everyday life.
01:31:59.279 --> 01:32:15.840
And so as clinicians, and I'm because I think that there's so many other, you know, ways outside of being a clinician that AI can integrate and and be a useful tool for us, uh, despite this the fact that we're putting all of our personal information out there for all of this AI to learn about us.
01:32:16.159 --> 01:32:19.439
But I digress, I'll go back off my other soapbox.
01:32:19.760 --> 01:32:26.800
But I think that there are ways that we can do it both responsibly and ways that we should probably steer clear of for right now.
01:32:26.880 --> 01:32:30.159
And and make can you give us some framework for what that should look like?
01:32:30.319 --> 01:32:34.560
So, as clinicians, how can we be responsible with this technology at the moment?
01:32:34.800 --> 01:32:40.720
And and also what should we kind of have more of that pause or that reservation with?
01:32:41.279 --> 01:32:44.479
So I think it starts with the boring stuff, like I said before.
01:32:44.720 --> 01:32:46.399
Start with the basics, right?
01:32:46.479 --> 01:32:54.560
Like, so number one, if you're a clinician who uses Chat GPT or Grok or any of the other tools to write your narratives, please stop.
01:32:54.960 --> 01:32:55.760
Don't do that.
01:32:56.000 --> 01:32:57.760
It's like that old Michael Jordan commercial.
01:32:57.920 --> 01:32:59.119
Please stop, get help.
01:32:59.279 --> 01:33:01.840
Like, don't don't don't do that.
01:33:02.079 --> 01:33:08.640
If you're doing that now, you need to stop immediately because I guarantee you, if you have a robust enough IT department, they know you're doing it.
01:33:08.720 --> 01:33:16.720
Um, because we track IP that I know for a fact most companies have IT that track your IP addresses and where you go and what you do.
01:33:16.880 --> 01:33:18.399
I stop doing that, right?
01:33:18.560 --> 01:33:23.039
Number two, be very careful whenever you're using LLM to ask clinical questions, right?
01:33:23.199 --> 01:33:25.600
There are approved models that are out there.
01:33:25.760 --> 01:33:32.479
Like, like uh, for example, open evidence, if you have an NPI number or if you're a student of someone, like I'm an advanced practice student, so I will have access to open evidence.
01:33:32.640 --> 01:33:35.359
But even then, once again, it's trust but verify.
01:33:35.600 --> 01:33:37.039
Human in the loop.
01:33:37.279 --> 01:33:39.920
You make the right decision, you verify.
01:33:40.159 --> 01:33:45.520
Anytime you look at any sort of AI output, check it before you before you sign your name to it.
01:33:45.680 --> 01:33:50.800
If you're using a charting system that has AI tools in it, once again, trust but verify.
01:33:50.960 --> 01:33:53.119
Like you need to look and make sure the output is good.
01:33:53.199 --> 01:33:55.600
That is the work that you're gonna have to put into it.
01:33:55.680 --> 01:34:00.159
Is and make sure it's correct because I guarantee you, you every now and then you're gonna find something is wrong.
01:34:00.399 --> 01:34:08.079
The other thing is, I would say as a crew member, uh talk to your leadership and say, hey, listen, we have our concerns about AI tools being used.
01:34:08.239 --> 01:34:09.920
Can we have a discussion about this?
01:34:10.399 --> 01:34:17.039
Have that open collaboration because it's going to incite things like, hey, maybe we should have a good ethical use policy in place.
01:34:17.199 --> 01:34:20.079
Maybe we should talk to our IT and ask IT.
01:34:20.239 --> 01:34:20.880
Ask them.
01:34:21.119 --> 01:34:24.159
A lot of times they're nerdy people who really want to help you.
01:34:24.239 --> 01:34:29.119
Um, as much as they'll tell you to restart your computer and turn it back on again and it magically fixes it every time.
01:34:29.439 --> 01:34:31.600
Like, seriously, ask your IT team.
01:34:31.680 --> 01:34:37.439
Like, hey, listen, like, I like these tools, I use them, you know, what are the limitations of them?
01:34:39.920 --> 01:34:42.880
A lot of them are very versed into it because they have to be, right?
01:34:43.039 --> 01:34:49.359
They're just as uh afraid of the dangers of it as I am, as Mike is, and Blake as you are, as all of us are.
01:34:49.520 --> 01:34:52.800
Um if you're not if you're not afraid of it to an extent, you're not doing it right.
01:34:52.960 --> 01:34:53.600
You know what I mean?
01:34:53.840 --> 01:34:57.920
So um that's that's something that as a crew member, those are the starting points.
01:34:58.159 --> 01:35:06.000
Now, the other things you can do are you can advocate for certain tools, bring them to your leadership, right?
01:35:06.319 --> 01:35:11.840
Talk with them about it, but understand that at the end of the day, the IT team also needs to take a look at it.
01:35:12.079 --> 01:35:13.600
You need to make sure that it is safe.
01:35:13.760 --> 01:35:15.680
Remember the five things I was talking about, right?
01:35:15.920 --> 01:35:22.079
Ask the question is our culture good in a place where it could handle this tool and not irresponsibly use it.
01:35:22.239 --> 01:35:25.119
Are our policies in place to where it's consistent?
01:35:25.359 --> 01:35:29.840
The other thing you can do also as a crew member is consistently document.
01:35:30.239 --> 01:35:32.880
Consistently document, follow your documentation standards.
01:35:32.960 --> 01:35:34.479
That's gonna help with quality of data.
01:35:34.560 --> 01:35:37.600
So when your leaders go, you know what, we're meeting those five things.
01:35:37.680 --> 01:35:39.119
Let's go ahead and buy this thing.
01:35:39.279 --> 01:35:41.760
It's gonna have good outputs that actually can help you.
01:35:41.840 --> 01:35:45.600
But once again, you're trusting, but verifying, human in the loop, right?
01:35:45.680 --> 01:35:47.119
We always stay in the loop.
01:35:47.359 --> 01:35:50.720
The other thing you can do as a clinician is just take care of your patient.
01:35:50.880 --> 01:35:55.520
Once again, I learned the other day from a very, very, very smart FTO.
01:35:55.680 --> 01:35:58.880
Hey, sometimes you get your head out of the paperwork and focus on your patient.
01:35:59.039 --> 01:36:19.039
Take good care of your patients, take good care of them because ultimately your documentation will reflect that good care you provided, which means the models will be filled with good, fed with good things and what you're actually doing so that eventually down the line, when that 10 years comes and we're getting more clinical decision support, it's going to be more accurate.
01:36:19.199 --> 01:36:21.039
But once again, human in the loop, guys.
01:36:21.119 --> 01:36:22.960
I said it a thousand times, I'll say it again.
01:36:23.119 --> 01:36:25.279
Don't just trust things that it puts out.
01:36:25.439 --> 01:36:30.800
If, for example, it's really funny when people look at papers, because I'm in grad school and I'll tell you it's really funny.
01:36:30.960 --> 01:36:40.960
I'll click on a discussion post or something, and I'll click on a DOI link and it doesn't exist, which means that the the the source was fabricated.
01:36:41.119 --> 01:36:45.920
It happens, and that's every university, not just where I'm going, it's everywhere, right?
01:36:46.239 --> 01:36:54.239
So you need you have to trust but verify these things because you're gonna look silly when you copy and paste hello, comma teammate on an email and look silly.
01:36:54.560 --> 01:36:56.880
I'm gonna address you that way every single time.
01:36:57.279 --> 01:36:58.800
Hello, teammate.
01:36:59.199 --> 01:37:00.000
It's so funny.
01:37:00.239 --> 01:37:02.640
You're not just Christian, you're a flight nurse.
01:37:02.800 --> 01:37:05.840
Yeah, it's like it's so funny with an M-dash in the middle.
01:37:06.159 --> 01:37:09.119
But I would say for crew members, that's the thing.
01:37:09.279 --> 01:37:17.039
You have to treat this stuff like you would treat anything, any medications, any therapies or treatments, you have to do it safely.
01:37:17.359 --> 01:37:19.840
This is not something just to be blindly leaned upon.
01:37:19.920 --> 01:37:26.560
As much of an advocate for it as I am, I also understand that like it's a tool that you have to really understand.
01:37:26.720 --> 01:37:34.800
It's a very, very complicated piece of technology that seems simplistic on the front end, but on the back end it's very complicated.
01:37:34.960 --> 01:37:37.600
So but yeah, that's what I'd say for crew members.
01:37:38.239 --> 01:37:45.840
I I think though that this is not gonna be the last conversation that we'll have about this.
01:37:45.920 --> 01:37:51.600
I I think that this this is just going to be part of our everyday future as as clinicians in healthcare.
01:37:51.760 --> 01:37:55.600
And I think not just in healthcare, it's gonna be in all walks of life.
01:37:56.319 --> 01:38:08.880
But what I feel, at least for myself, is I I feel like I'm at that point where I'm looking at where I started to where I'm at now.
01:38:09.439 --> 01:38:13.840
I'm finally that person that I met like 20 years ago.
01:38:13.920 --> 01:38:20.239
When you met that person, I was like, back in my day, you know, I feel like that's the point that I've that I've gotten to.
01:38:20.479 --> 01:38:30.720
And I could only imagine what when I was brand new 20 years ago, what my FTO was and seeing come down the line.
01:38:31.279 --> 01:38:37.600
So I feel like I'm now kind of or am seeing these things come out.
01:38:37.680 --> 01:38:42.000
Like it's like some of this stuff is just like unbelievable, you you know, for me.
01:38:42.239 --> 01:38:51.119
And it and it's not this, you know, resistance is futile, but I think caution caution is mandatory, right?
01:38:51.279 --> 01:39:23.359
Like you we have to uh I I want to embrace it and I want to believe that this can be something that will help, but I also have to remember that if I am going for me as Blake as an individual is going to be inherently a better person, a better clinician, uh, you know, all that, it's up to me to put in that work, do the things, do research, do study, do all these things.
01:39:23.439 --> 01:39:33.680
I can't rely on something fully autonomous to uh flash it in front of my brain and be like, boom, you got it.
01:39:34.239 --> 01:39:45.920
And until the until those futuristic uh movies start really coming more to life, you know, things that we saw, Terminator, Idiocracy, any of the any of these things.
01:39:46.319 --> 01:39:48.560
I still think in some places that could be a documentary.
01:39:48.640 --> 01:40:02.079
I don't know, statistics to me, but but I just my hope is that I as I'm aging and getting older and and moving through this world of healthcare and as a clinician and and just I'm still gaining more experience.
01:40:02.319 --> 01:40:08.720
You know, it just because it's been 20 years, I I I could still have 20 years to go to continue learning and doing more.
01:40:08.880 --> 01:40:23.840
So I I don't want to say that I'm like aging out, but I feel that when I look at the this process of AI, I want to be optimistic and I want to be like in in the right cheerleader corner for it, I guess.
01:40:24.079 --> 01:40:28.640
But I also need to be cognizant of like I want to see you there with your pom-poms.
01:40:28.960 --> 01:40:30.319
Oh, it could happen.
01:40:31.039 --> 01:40:32.640
It would be fully automated though.
01:40:32.720 --> 01:40:34.800
It would it would like tell me to cheer.
01:40:36.000 --> 01:40:46.000
But and and Blake and Mike, if I could offer a bit of sunshine here, imagine a place like let's say you have to do your monthly checks in the supply room, right?
01:40:46.399 --> 01:40:50.800
I will say there will be a time when there will be a system that really helps you with that.
01:40:51.039 --> 01:40:55.199
What usually was three to four hours, could be 20 minutes or 10 minutes.
01:40:55.279 --> 01:40:57.520
And then what can you replace that time with?
01:40:57.840 --> 01:41:14.960
You could sit with your trainee and you can go over scenarios, you can talk about experiences, you can develop that relationship with your trainee, you can work with your partner to plan, pre-plan for a flight, to practice your walk arounds, to talk about safety, to talk about the evidence and literature that's out there.
01:41:15.119 --> 01:41:29.359
For me, that's that's the the bright part of this, is that we can offload certain basic things that normally are very time consuming with the stuff that matters for the critical thinking parts of this, like being able to, you know, talk about, hey, like Mike, I don't fly Neos that often.
01:41:29.520 --> 01:41:32.800
Like, let's talk about like hypoplastic uh syndromes and things like that.
01:41:33.039 --> 01:41:34.000
I want to learn more about.
01:41:34.319 --> 01:41:38.000
Tell me your experiences of that, like those valuable conversations.
01:41:38.479 --> 01:41:41.680
We will have more of those, I think, if it's done correctly.
01:41:41.840 --> 01:41:43.359
And we and we can do that.
01:41:43.920 --> 01:41:44.399
Just wait.
01:41:44.560 --> 01:41:50.159
Next time we talk, we'll be talking about AI creating VR experiences for us.
01:41:51.359 --> 01:41:53.600
And in terms of our training.
01:41:54.880 --> 01:41:58.079
It's not it, I know those things are probably in development.
01:41:58.319 --> 01:42:03.680
Once again, I don't know every technology, but I mean we're in the phase where a lot of things are being developed.
01:42:03.920 --> 01:42:08.159
We're in a development phase and and experimentation's going to have to take place.
01:42:08.399 --> 01:42:09.920
Projects are gonna have to be done.
01:42:10.079 --> 01:42:21.039
But once again, I think the realm of research is going to help us with those things to help us understand, like I said earlier, you know, we should do research to find out where this stuff could be helpful, right?
01:42:21.359 --> 01:42:34.239
And if we know for a fact like where it could be helpful, and we've done, you know, up to a meta-analysis saying, like, hey, we've done looked over 15 papers about this one particular type of AI technology, and we notice that patient care is better when we use it.
01:42:34.479 --> 01:42:35.039
Okay, cool.
01:42:35.279 --> 01:42:38.319
Patient outcomes are better, we know it's better, okay, then we can use it.
01:42:38.479 --> 01:42:43.840
I I do see some bright spots, but once again, I think we have these, we have several camps.
01:42:44.000 --> 01:42:45.119
We have three camps.
01:42:45.279 --> 01:42:50.880
We have I we have the camp where people are way overly optimistic and they're just like, use it all.
01:42:51.039 --> 01:43:01.920
And then we have the middle camp, which is where I think a lot of us lie, which is where those who who are looking into this stuff is like, okay, it's great, it can do a lot of things, but where and how?
01:43:02.159 --> 01:43:03.439
We need to do this safely.
01:43:03.600 --> 01:43:04.800
We don't want to cause harm.
01:43:04.960 --> 01:43:07.680
And then we have another group, you know, that is staunchly against it.
01:43:07.760 --> 01:43:17.600
But once again, I don't knock the staunchly against it because the the the possible unknowns from this that could cause harm are quite real.
01:43:17.840 --> 01:43:20.079
Um and once again, we have two pathways.
01:43:20.239 --> 01:43:22.560
We're at the fork in the road, which which way are we gonna go?
01:43:22.720 --> 01:43:25.760
That's the could that to those people that are staunchly against it.
01:43:25.840 --> 01:43:30.720
Uh, I have no uh I'm not against what they're saying at all because I think that's what they see.
01:43:30.880 --> 01:43:38.960
They see the fork in the road and they really want to go one way, but they're not gonna jump on board with it till they know it's gonna be going to that right direction as compared to the wrong direction.
01:43:39.119 --> 01:43:40.800
So that's that's my assessment.
01:43:41.119 --> 01:43:50.960
I feel like if my boss hears this, they're gonna be like, Man, we just cut down the date checks from four hours to like 20 minutes.
01:43:51.119 --> 01:43:53.840
Man, we could fill that up with more OSHA training.
01:43:54.000 --> 01:43:56.399
We could find more fun stuff for you to do.
01:43:56.960 --> 01:43:57.279
Yeah.
01:43:57.439 --> 01:43:57.840
Oh, good.
01:43:58.079 --> 01:44:01.199
Outside the scope of this talk, but you know, there is that.
01:44:01.279 --> 01:44:02.560
Like that could happen too.
01:44:02.720 --> 01:44:04.720
But once again, we could fill it with good things, right?
01:44:04.880 --> 01:44:05.039
Yeah.
01:44:05.199 --> 01:44:06.319
And that's what I hope that happens.
01:44:06.399 --> 01:44:07.600
I hope we can fill it with good things.
01:44:07.920 --> 01:44:09.439
Well, thank you, Christian, for joining us today.
01:44:09.520 --> 01:44:11.039
I really appreciated the discussion.
01:44:11.199 --> 01:44:21.760
Uh, I certainly still have many more questions than answers, but I love being able to have these conversations and really kind of uh solve the world's problems, this time all about AI.
01:44:21.840 --> 01:44:25.680
So I have no doubt we could talk for hours and hours more about all of this.
01:44:25.760 --> 01:44:35.840
Uh, and I'm so grateful that you were willing to take some time out of your day to discuss with all of us and all of our listeners as well, because uh I I do think that this that we we can't resist.
01:44:36.000 --> 01:44:39.039
We just need to be cautious in how we move forward with it.
01:44:39.119 --> 01:44:45.119
And and I I appreciate you uh and all of your insight that you have been able to provide over the past couple hours for us.
01:44:45.439 --> 01:44:46.880
Yeah, no, it's it's my pleasure.
01:44:46.960 --> 01:44:48.880
And and I think it's good to get that message out there.
01:44:48.960 --> 01:44:53.439
Once again, not everyone's gonna agree with what I'm saying, and that's okay because I think we're all figuring this out.
01:44:53.600 --> 01:44:56.000
I I certainly don't have it all figured out.
01:44:56.239 --> 01:44:56.800
I don't.
01:44:57.039 --> 01:45:07.439
But I think that there is a number of places where we can get started on the back end of things on the basic level to help prepare us to tee us up to get us into a better position with these technologies down the line.
01:45:07.520 --> 01:45:11.920
So no, it was a great conversation, and and I love the I love I love every bit of it.
01:45:12.000 --> 01:45:13.279
So this is wonderful as usual.
01:45:13.439 --> 01:45:14.239
Thank you for having me.
01:45:14.800 --> 01:45:15.520
Always a pleasure.
01:45:15.680 --> 01:45:20.000
I'll just say again, Christian, fourth time uh on the podcast.
01:45:20.079 --> 01:45:21.840
We'll see if we can go for number five.
01:45:22.079 --> 01:45:24.880
We'll see what uh we'll we'll give you a little break.
01:45:24.960 --> 01:45:26.960
I know we're we don't want to overwork you over there.
01:45:28.800 --> 01:45:29.600
No, it's a pleasure.
01:45:29.680 --> 01:45:30.159
You're good.
01:45:30.399 --> 01:45:50.640
I I think a lot of this again, like you just said, that this this all has a great application and it w whatever camp that you're in, I I just appreciate Christian the ability to really kind of take that unbiased look and really allow us to see both sides of the playing field of what can potentially be coming down the road.
01:45:50.800 --> 01:45:52.319
So I'm super grateful.
01:45:52.479 --> 01:46:01.119
Chrisha is always your one of my best friends, one of my favorite people to uh to not only hang around with, um, but to just we could talk for this for hours.
01:46:01.199 --> 01:46:05.279
I know, I know that we could, and I'm just excited to really see what's in the future.
01:46:05.439 --> 01:46:16.399
I'm probably in that middle camp, but I I just uh whether it's AI or not, I'm just really excited to see where we're gonna be in another five to ten years as as clinicians.
01:46:16.640 --> 01:46:20.640
So I'm hoping that it will all obviously be good.
01:46:20.720 --> 01:46:23.760
I I think everything's been good so far, things that I've learned over the years.
01:46:23.840 --> 01:46:26.960
So I'm um I'm pretty stoked to see to see where we end up.
01:46:27.199 --> 01:46:28.000
Oh, for sure.
01:46:28.159 --> 01:46:39.039
I mean the possibilities are potentially endless, and it and once again, there's probably 15, 20 hours more we could probably discuss on the we didn't cover a we covered the surface today, guys.
01:46:39.279 --> 01:46:39.840
We really did.
01:46:39.920 --> 01:46:41.439
There's so much I didn't talk about on here.
01:46:41.600 --> 01:46:43.680
Some folks, well, why didn't they mention this or that?
01:46:43.840 --> 01:46:45.279
Because there's just so much to talk about.
01:46:45.439 --> 01:46:47.199
So, but like I said, great conversation.
01:46:47.279 --> 01:46:48.000
Thank you, Blake.
01:46:48.239 --> 01:46:49.760
Yeah, I have no doubt about that.
01:46:49.840 --> 01:46:52.079
We hope you enjoyed this conversation as much as we did.
01:46:52.239 --> 01:47:00.319
Once you're all caught up on the Valor Flight Crew podcast, make sure you head on over to verticalhelicast.com where you can check out the Hangers E podcast, helicopter podcast, and the vertical MRO podcast.
01:47:00.560 --> 01:47:05.680
Special thank you to all of our listeners out there for your continued support of the Vertical Hellcast platform and the Valor Flight Crew Podcast.
01:47:05.840 --> 01:47:07.119
Be safe and we'll see you next time.