Video-Enhanced Advance Care Planning: Joshua Lakin and Kei Ouchi

Eric 00:12

Welcome to the GeriPal Podcast. This is Eric Widera.

Alex 00:17

This is Alex Smith.

Eric 00:18

Alex, today we’re going to be talking about the use of video to enhance advance care planning. Who do we have with us today?

Alex 00:24

We have 2 repeat guests today, Eric. We have Kei Ouchi, who is an emergency medicine and internal medicine doctor and researcher who practices This is at the Brigham and Women’s Hospital in Boston. Kay, welcome back to the GeriPal Podcast. It’s been a minute.

Kei 00:38

Thank you so much for having me.

Eric 00:41

Oh, I love the energy.

Alex 00:42

And we have Josh Lakin, who’s a palliative care doc and researcher who did his palliative care fellowship and residency here at UCSF. He’s now at the Dana-Farber Cancer Institute, also in Boston. Josh, welcome back to the GeriPel Podcast.

Josh 00:55

Good to see you, gentlemen.

Eric 00:57

So we’re going to be talking about 2 papers that came out that included some type of video-enhanced advanced care plan. One published in Journal of General Internal Medicine, JGIM, and another in JAMA Network Open. We’ll have links to both of them in our show notes. But before we do, we always start off with a song request. Who has a song request?

Josh 01:16

I do.

Eric 01:17

Josh, what song?

Josh 01:20

It’s the only one I could find that fit this one. It’s called Video Killed the Radio Star, which I love, by the Buggles.

Alex 01:28

And we have done this one a while ago. But it’s been a while and, uh, I think my, uh, you know, I tried to change it up a little bit. Why did you choose this song, Josh? Is it just obvious?

Josh 01:39

Because we’re talking, because we’re talking videos. That’s all I had. Plus, anytime you can bring the ’80s into a conversation, you should probably do that.

Alex 01:46

And we’ll see if you truly believe that video killed something like an old way of advance care planning conversations.

Kei 01:53

All right.

Alex 01:54

Here’s, here’s a little bit.

Alex 02:00

(singing)

Eric 02:48

That was awesome.

Alex 02:50

That was a fun one. This is a blast from the past. So one-hit wonder band. What was it called? Buggles or something? The Buggles.

Speaker 5 02:58

The Buggles.

Alex 02:58

That’s a good trivia question. Everyone knows the song. I doubt that many people know the name of the band.

Josh 03:08

Maybe for a little bit after this podcast.

Eric 03:11

Okay. So we got a lot to talk about. I’m going to turn to both of you first. Just why are you personally, or got interested in this, this topic of improving advance care planning? Through things like video. Josh, I’m going to turn to you first.

Josh 03:29

Yeah. So I’ve spent most of my career, both in training and then afterwards, really thinking about clinician interventions for improving communication. And there was something about considering ways of empowering patients outside of clinicians through these video decision aids that interested me. And I also, I hadn’t thought a lot about, like, we spend so much of our time thinking about the words we’re going to say and how we talk to people, how we are going to use different professionals to do different aspects of care.

But the idea of empowering people with images and videos in their own languages that have been interpreted into so many different languages was interesting to me. So it just felt like a very different tool. And throughout my career, I’ve been lucky enough to just get to play with a bunch of tools for advance care planning and serious illness communication. And this one was just unique in such a way that it was interesting. And Angelo Volandes, who built these along with his wife and the company ACP Decisions, are just so thoughtful about the way they deploy them. It was just an interesting new way of getting at what matters most to patients. And so that’s how I got into it.

Eric 04:35

It’s been a while since we’ve had Angelo on. We had him on.

Alex 04:38

We had him on. I think that’s probably the last time we sang the song Video Killed the Radio Star after his book came out. Yeah. We, and his company again, ACP Decisions.

Eric 04:49

And we’ll have a link to that podcast as well. Okay, Kei, how about you? How’d you get interested in this? You’re in the emergency room, right?

Josh 04:55

You’re —

Kei 04:56

Right, right. So as you can imagine, I clinically work as an emergency physician and I meet a lot of older patients who have actually never had these conversations before. And this is just one study among my body of work of trying to figure out how can we leverage the emergency department to improve serious illness communication, right? And all my other studies, we use a clinician to actually kind of get this going, do motivational interviewing or something like that, couple that with serious illness conversations.

And when Angelo told me about this study and idea, I just thought like, how can we make this more scalable? You know? And video is just one way of making things a little more scalable.

Eric 05:45

Yeah.

Kei 05:46

And now that we combine both of our studies, it combines the video, which is like totally reproducible, high intervention fidelity, with a very shortened human-driven conversation after that. And I think that really attracted me to try to execute this study because I’m always looking for scalability. I know what is good. I’m just looking to see what we can reproduce.

Eric 06:12

Because your study — so let’s talk about your studies. called the Effective Emergency Department Initiative Video-Enhanced advance care planning on Documentation and Goal Concurrent Care Randomized Controlled Trial, published this year in JGIM. We’ll have a link to it in our show notes. It sounds like, again, you wanted to tackle something in the emergency room, right?

Kei 06:32

Yes.

Alex 06:33

Yeah. Why is the emergency room a good place for advance care planning conversations?

Kei 06:39

Oh, you have stuff there?

Alex 06:40

Okay. Yes, I can explain that. Like, why the emergency department? Like, as you pointed out in the intro, like, there’s not much done there, but that’s probably because most people think it’s not the place for it. Why do you think that’s the right —

Eric 06:50

Yes. Yeah.

Kei 06:50

You know, that’s —

Eric 06:50

In and out medicine, right?

Alex 06:52

Right, right.

Kei 06:52

Yes. Most of my mentors and friends think that that’s not the right place to do everything in emergency department. And at the same time, uh, this idea about intervening in emergency department actually came from Alex’s paper when I was in training that showed You know, most older patients visit the emergency department in the last 6 months of life. And also, a lot of people have never had these serious illness conversations before. I still remember this, and those numbers are still true because I see this in clinical care.

You know, one-third probably of my patients are older adults with serious life-limiting illness. When I talk to them about non-acute care stuff, they have no idea. They’re like, what do you mean I have this, like, really serious life-limiting illness? And I know that I can make this better by doing just a little bit of emergency department to kind of kickstart this for these people. So, I mean, this is all based on Alex’s paper as well as my clinical experience that emergency department only sees patients that you guys miss, you know, all the outpatient providers miss. And those are the most vulnerable patients that we can Actually moved the needle on.

Eric 08:11

So let’s talk about the most vulnerable. Who, who did, did you study all emergency room patients or did you just pick a specific population that were most vulnerable?

Kei 08:20

Yes, we chose to study older adults with serious life-limiting illness. You can define that however way you want to define it, but we use it based on literature review and then also include people without serious illness. If they’re older than I think 70, I can’t remember exactly what the number was, but I think 70 years of age and 70 or 80, I can’t remember actually.

Eric 08:44

It’s all right.

Kei 08:45

It’s in the paper.

Eric 08:46

Just older. Yeah. We’ll have a link to it on our paper, in our show notes.

Kei 08:50

Yes. And we catch them in the emergency department while they’re receiving their care, or if they’re boarding in the emergency department waiting for the beds to open up. Yeah.

Eric 09:01

And then, so, and then you randomize those 2 people to just usual care in the emergency room. Is that right?

Kei 09:09

Yes.

Eric 09:10

Versus what?

Kei 09:12

Versus, uh, versus the, the intervention. So the intervention is, um, I think, you know, a long time ago when, uh, Angelo Valendez first started with the video, it was just a video. And then now it’s kind of combined with Just very short components of serious illness conversation guide, like asking about patients’ preferences as well as patients’ hopes and values. That’s kind of tagged along after watching the video. And then patients are kind of … a lot of people get admitted.

Most people get admitted, actually. And they just kind of go on with their care. And our team, the research team, will summarize, uh, this is before large language models, so we actually manually summarize what was discussed in the video as well as what the patients had shared and message that out to inpatient and outpatient teams.

Alex 10:08

So there’s kind of like 3 parts. In the first part, they view the video, and it, it, is there any way our listeners could view these videos or are they like firewalled off at ACP Decisions?

Kei 10:19

Oh, I thought the, the patient-facing videos are all open.

Alex 10:22

They’re all open. Okay, great.

Kei 10:24

I think so. Um, I think if we’re, uh, if you’re, uh, I think the patient ones I thought were open and, um, the, if you were to use it in a health system or something like that, that, mm-hmm, that is kind of blocked off.

Eric 10:37

We’ll have a link to it. Right. Listen, if it’s available, but, and the, the videos, they could choose multiple different videos, right? Uh, or, or is there only one video?

Kei 10:46

One video for this study.

Eric 10:47

For this study, it was just one video. Yeah. And that video focused on what? advance care planning?

Kei 10:51

Yes.

Eric 10:52

Okay. And it was short. It was like 5 minutes, right?

Kei 10:55

Yes.

Alex 10:56

Do you need a video for advance care planning? I mean, I know that’s the topic of today’s podcast, but I understood, like, I remember when we had Angelo on, he was like, you know, when you have a patient who’s, um, really sick and, you know, would consider being in the ICU if he got sicker and they’re on the floor. I remember he wheeled the patient over to the ICU and it was like, that was transformative. And he realized, wow, if they could only see it, I can show it to them. Like, what do you show them in that advance care planning video?

Like, do you show them people, a per — person talking with a surrogate and assigning a surrogate? Or do you show them like CPR or intensive care unit? Like what? Because the choice, the question is, you know, like what goes into that video? Because you could imagine it might persuade people or bias people. Could be, you know, some people might use those words in one way or another. Nudge would be another word, uh, depending on what video you showed them. Yeah.

Kei 11:49

Yes. So the video showed, basically, there’s a narrator physician who is just kind of explaining about the process of advance care planning as well as different interventions and divide the care largely into 3 buckets, full care, limited care, and comfort care. Comfort care meaning like just comfort only, and limited care meaning you can choose However way, whatever you limit is kind of up to you, but it doesn’t actually define what you limit, like intubation or something like that. And full care is full care, everything.

And basically divide, kind of define the care into those 3 buckets. And then also explains about CPR and intubation. And it does try to educate the viewers In understanding what the probabilities that they’re facing with each of those interventions, not in a specific number, but in general sense, like most of the time when you do CPR, it does not work type of situation.

Eric 12:57

Mm-hmm.

Kei 12:58

And then the viewers after the video are asked to choose, like before and after the video, what do you think about all these different types of care and what would you choose is one question that follows the video. And then we go into talking about some values like hopes and stuff like that.

Josh 13:15

All right.

Eric 13:15

So you get the 5-minute video and then you have this structured kind of ACP conversation based on, at least with clinicians who are trained on Vital Talk, right?

Kei 13:28

Yes. So how was that done is that Angelo, well, Angelo did most of it, but I do the coaching afterwards. So we trained the clinicians on, like, a —

Eric 13:41

What kind of clinicians? Physicians, RNs?

Kei 13:44

Yes. There was no physicians. And there were some social workers and RNs and medical students. And we trained them in the same way as just VitalTalk Remap using the same type of actor-based training. And it’s very, it’s, it’s not very long. It’s, it’s, it’s only like 3 hours long. Just like if you kind of think about like the Vital Talks training in one, one shot and remap, but it’s not remap. It’s basically responding to emotion and asking key value-based questions like, what’s important to you type of situation.

Eric 14:25

So they get the, the clinician intervention generally with a social worker, a nurse, or a med student. That one, the interaction with the patient lasts usually how long? Do you have a sense?

Kei 14:37

Uh, 5-minute video and probably like 5 to 7-minute conversation. And then, uh, the person just writes them all down and then put it in REDCap and summarize later.

Eric 14:48

And that gets, that gets sent out through the EHR, but not in the patient’s notes, but through like messaging?

Kei 14:54

Not in the patient’s notes. Yes.

Eric 14:55

Why’d you choose not in the patient’s notes?

Kei 14:58

Well, so to do something differently is probably my answer, but, um, so I have another randomized trial, very similar, but without the video and using the human, uh, interventionist. And I document them in a medical record because it’s, you know, I would think that’s the best thing to do.

Josh 15:15

Yeah.

Kei 15:16

And, uh, it kind of complicates my analysis because I have to like de-identify those notes. to kind of blind the outcome assessment. It kind of complicates my outcome assessment. It can be done, but it kind of complicates it. And the clinician, inpatient clinicians are receiving the same message anyways. So that’s just a, it’s just trying something different.

Eric 15:41

Okay. So they get randomized to these 2 groups and then the outcomes that you look at, the first one is, did advance care planning happen? How did you know if advance care planning happened?

Kei 15:52

Based on documentation. So, um, whenever there’s a new documentation after the ED visit, and well, I used to do this in like a, like a chart review type of situation, but this one is done by NLP by Charlotte Lindvall.

Eric 16:09

Natural language processing. So like —

Kei 16:11

Yes. So it’s more like, um, so you can do high number of volume. as well as more accurate reporting of it. So, that’s one outcome. That was the main outcome and the increase in that. Because, you know, you would think that if there’s a new documentation of someone, some clinician discussing with the patient, it’s probably likely that there has been some type of conversation that occurred. And we’re not talking about like full code presume type of documentation, like actual clinician notes.

Eric 16:43

Yeah.

Kei 16:44

So, uh, that’s sort of the assumption under why this was, uh, determined to be an outcome.

Alex 16:50

Mm-hmm. And the outcome, just to be clear, is not the clinician who had the conversation in the emergency department writing a note. That doesn’t count. That, that’s, that’s not what happened. They did not write a note. Uh, this is —

Kei 17:05

In this trial, they did not write a note. So there is no such thing. Yeah.

Alex 17:09

So it was a, It was another clinician who wrote a note that counted as an advance care planning conversation as determined by a natural language processing algorithm, basically like a computer program algorithm that went, combed through the electronic record and identified this is advance care planning conversation. And we had Charlotta Lindvall previously on GeriPal talk about this sort of stuff.

Eric 17:34

Yes.

Kei 17:35

Great.

Alex 17:35

Okay. So that was, uh, the first, that was the primary outcome. What else did you look at?

Kei 17:40

We also looked at the patient’s knowledge in ACP, which Angelo likes to do all the time.

Josh 17:46

Uh-huh.

Kei 17:47

And then also the engagement, ACP engagement survey.

Alex 17:50

Uh-huh.

Kei 17:51

Um, Rakesh Dora’s ACP engagement survey. And then, uh, finally, uh, goal-concordant care.

Eric 17:57

Goal-concordant care. Holy smokes, Alex. Oh, we’ve had a lot of podcasts about this. And if I remember, you’re always…

Alex 18:04

The holy grail. Why is it the holy grail? Validated care research. Because it’s so hard to measure, right? Did the treatments – I mean, that’s part of the idea behind palliative care, right? This is what Diane Meyer said when some senator asked her, like, what is palliative care? It’s aligning treatment with patients’ goals, like at its essence, right? But then it turns out that measuring that is really difficult. How did you measure it?

Kei 18:31

So we decided that we will have 2 physicians who are blinded to intervention or control receive a medical record number and patient’s name of all the people who are enrolled and died. So these are all decedents. And go through the chart and try to decide on a 0 to 1 scale, 0 being like totally discordant and then 10 being perfectly concordant, Like where this patient belongs based on like my own chart review, basically. So I’m gonna do a chart review before they died, and we’re gonna try to figure out like, is there anything that was discussed about this patient about either their code status or their values and goals or whatever it is? Like we do like, like a manual chart review.

Josh 19:21

Yeah.

Kei 19:22

And then we look at the care near death. or whatever that is documented at the time of death. And basically if the patient said, you know, I would never wanna die in ICU, or, you know, I would, you know, most important thing for me is to be at home at the end of life or whatever. And if the patient died in ICU, that would be considered not concordant.

Eric 19:47

What if they changed their mind? Like in the emergency room, they didn’t wanna die in the ICU, but like, 3 days later in the hospital they did, and then they end up dying in the ICU.

Kei 19:58

Yes. So, um, as much as we can, we try to, starting from like 6 months before death to like the time of death, as much as we can, we try to check these time points and see if anything felt like, um, it was different. And then right at the time of the last hospitalization and when they died, We try to come up with, like, is it concordant or discordant?

Alex 20:24

Ah, so it wasn’t based on the conversation in the emergency department?

Speaker 5 20:30

No. No.

Alex 20:31

It could have been based on the, the ACP conversation that happened after that on the floor or in the ICU, for example.

Kei 20:38

Because these are all people who are enrolled. So, like, you know, our intervention conversation already happened, but that’s never documented in the EHR. And then, uh, there are more documentations of the intervention patients in the EHR after the intervention date. And then we follow them until they die. Well, not everyone, but, you know, these are the patients who died, like 90-some people died during the follow-up period.

Alex 21:04

Yeah. So it’s a subset. It’s not the full study population.

Kei 21:07

Not a full. It’s not a full. Yes.

Eric 21:09

Is this the holy grail, Alex, that you’ve been looking for?

Alex 21:11

Well, I, I do like this measure. I mean, It, it, uh, it’s limited in that it’s only in patients who died. And of course we want to follow people’s goals whether or not they die, right?

Eric 21:23

And you gotta document everything.

Alex 21:24

And some people who get goal-concordant care end up surviving, right? They might get full care, all sorts of treatment, like ICU care and such. And I really appreciated that in your figure, like the number 10, like the strongest agreement example was a patient who said they wanted to be full code and receive all care possible. All interventions possible, and then they ended up dying in the ICU after a prolonged stay.

Kei 21:48

Exactly. So it could be highly intensive care, dying in the ICU as 10, and person who wants to die at home, dying home hospice for a long time, like, you know, weeks, and then die at home, that’s a 10 too. So 10 is best, and patients can choose what that means.

Josh 22:09

Yeah.

Eric 22:10

Okay, so results. Let’s quickly talk about the results. For the results, at 3 months, I think was your primary outcome, right? And you showed a significant increase in the intervention patients. 46% had advance care planning documentation versus 31% in the control group. So big difference there. Any difference in goal concurrent care?

Kei 22:35

Yes. Yes. So that’s the whole point. And, and this is probably the most exciting trial that I have done in my career up until this point because of this secondary outcome. You know, the primary outcome, I have shown the same thing about the same magnitude in my prior trials already.

Eric 22:56

I think Angelo has shown the same thing too. Like, right, right, right. Uh, videos improve. advance care planning.

Alex 23:02

Right. If you focus on advance care planning, you will have more documentation.

Eric 23:07

Of advance care planning.

Alex 23:08

Of advance care planning. Like, that’s okay.

Eric 23:09

Does it change anything?

Kei 23:10

Right.

Alex 23:10

So what?

Eric 23:11

Did it change anything? You said yes.

Alex 23:13

What did it change?

Kei 23:13

Well, it sounds like it, it really looked like when I was reviewing these charts that it did change something. And that’s why I was so excited about it. But we didn’t really know because we were blinded.

Eric 23:23

Yeah.

Kei 23:24

We didn’t know who was in intervention or control. But when the blinding came off, It was kind of exciting. I felt like this is doing something.

Eric 23:34

Was it a big change, little change? Hard to quantify because it’s on these 10-point scales. Like —

Kei 23:39

Yes. So, it’s hard to quantify, sure. But I guess the difference was 10 versus 7 is what we found. And is that big? Is that clinically significant? Is what people always ask, right?

Eric 23:49

Yeah.

Kei 23:50

Is that clinically significant? And I mean, please imagine 10 is like the Best, like the best possible, right? You are great. Amazing.

Eric 23:58

You have great concordance at 10.

Kei 24:00

Amazing.

Eric 24:01

And 0 is like, oh my God, totally discordant. Like you guys aren’t even, yeah.

Alex 24:06

Okay.

Kei 24:07

And so, um, is it a, is it clinically important change? Well, I’m, I’m certain that this is in the right direction for sure. And if it’s hard to say what that 3-point difference means, But sorry, I, I can’t remember what the difference was actually. But that point difference is a lot by like 3 or 4 points, I think. And I think that’s really important and really meaningful to patients.

Eric 24:37

All right.

Josh 24:38

I suppose it also comes based on how that happened, right? Like if you took 2 people from 0 to 10 and everyone else was a 7, that seems like a win to me, right? Like, If you dramatically change 20% of the population’s experience, is that not a huge win? I don’t know. I guess if you take everyone from a 7 to an 8, that’s debatable. But if you take just a couple people outta 10 from a very low number to a very high number, and that’s where your 3 points come from, that’s avoiding what we would call never event kind of stuff in palliative care, isn’t it?

Eric 25:08

I think that’s just like a pharmaceutical trial, right? Like we’re, hey, you know, these Somebody should get paid then, right? Well, I think this is the challenge.

Kei 25:18

Okay.

Eric 25:19

I’m going to ask you this before we go to Josh’s article is that, um, like how outside of a research study, outside of a large academic institution, like, is this practical to do? It feels like, cause you’re all about scale, right? It feels like the ACP video is easy. Shooting somebody a message is easy, but the hard piece is that additional step of having a clinician, you know, spend 5 to 10 minutes asking these additional questions.

Alex 25:54

Yeah. In the emergency room where you’re so pressed for time. Yeah.

Josh 26:00

Yeah.

Kei 26:01

So I think I’m sure you guys can all feel it and believe it. The world will change in like 10 years because of AI and large language model. And I really think that it’s going to be scalable. My next — our next grant is already in, but something more scalable, it sounds more scalable in that study. And because video is reproducible, that’s one component, it can be incorporated into just the regular emergency department procedures that doesn’t include me, the physician or nurses.

Right. And then the piece about conversation can also be done by a trained palliative care trained AI speaker. And I’ll come up with that later, but we’re doing that too. And then the documentation and summary can be done by the large language model. And in essence, the really difficult part of this whole process is the final recommendation, right? And then confirming all the values and goals. The physicians or clinicians can’t do that. You know, the rest of like information gathering and preparing the patients and understanding what’s going on, that can be scalable automatically.

Alex 27:15

Ah, so VitalTalk-trained AI will have that conversation that follows the video.

Eric 27:20

I mean, they can also do the video too.

Kei 27:22

Yes.

Eric 27:22

The, it could be an AI-based video personalized for —

Alex 27:26

Oh, personalized. Right, right. So instead of giving them a generic this is what happens to people who are seriously ill on average. You could give people like you —

Eric 27:36

Yeah. So somebody with COPD —

Alex 27:37

These are the outcomes. Right. Yeah.

Kei 27:39

With these medical problems, you know, coming into the emergency department.

Josh 27:43

Mm-hmm.

Eric 27:44

This will have to be the 3rd time we sing Video Killed the Radio Star, but it will be about technology advancing to AI.

Alex 27:51

Right. Yeah. Killing off the poor med student jobs of doing these conversations.

Eric 27:56

Yeah. Video’s gonna do it.

Kei 27:58

I actually really think that, like, you know, All these technologies and all these videos and AI speakers and all that stuff and large language model will allow us, the clinicians, to focus on the parts that only we can do, you know, make us more humane in that respect, I think.

Alex 28:15

Yeah, maybe. I think we’ll learn some surprising things along the way. There’s already backlash against AI. I wouldn’t be surprised if these patients are like, no, I’m not. I’m not. I’m in the emergency department. I’m not talking to your AI. I don’t —

Kei 28:29

I’ll show you that next time.

Eric 28:31

Okay.

Alex 28:31

You look forward to that.

Eric 28:32

I wouldn’t trust UnitedHealth AI to tell me what to do. That’s right. All right, Josh, let, let’s go to your article, Intervention to Increase advance care planning Among Older Adults with Advanced Cancer: Randomized Controlled Trial, published in JAMA Network Open amongst great authors. But I wanna know from you, like, why you told us a little bit about why you did this study. Like, what was the really interesting part of this for you?

Josh 28:58

For me, to be honest, was one of the questions you’re already getting at, which is the pragmatic nature of it. I’d never done any pragmatic study before or a stepped wedge trial. And I was pretty curious to see if there would be any impact, clinically meaningful or not, by just kind of tinkering with the system without much control. Right. So it was a randomized controlled trial, but it was randomized in a stepped wedge kind of way. And so, what’s a step wedge?

Eric 29:26

Is that type of shoe?

Josh 29:27

Cheese?

Alex 29:29

It sounds like a shoe. A shoe? Okay.

Josh 29:33

We can make it a clog, a cheese shoe or something. It means that you randomize clinics over a time period. So, these oncology clinics would get an intervention, which was the VitalTalk training, similar to what Kay described, 3 hours of communication skills around responding to emotion and talking about goals of care. Delivered by James Tulsky and the Vital Talk team. And then they’d get teaching about the deployment of the decision aids, and then they were on their own. The researchers were not there to make them do it. They did not randomize patients within those clinics. They just had to run the intervention with some guidance from the research team. So that’s way more real world than —

Eric 30:10

And the intervention is, hey, these are some videos. And it’s a, this is your article was a collection of videos, right? You can choose.

Josh 30:17

Yeah, exactly.

Eric 30:18

Feeding tubes or goals of care or advance care planning, hospice. You got this collection of videos and then providers who may or may not have been trained, but they had the option to train around communication skills, vital topics.

Josh 30:33

That’s right. And then they were left to the real world and then the computer went in and assessed documentation. We can get to that later, but yeah. So that’s what I thought was interesting because my hypothesis was that won’t change anything.

Eric 31:14

And, and why not just, why stepped wedge? ‘Cause that introduces time bias, right? Especially you guys did your study during COVID Like, is it, it just changes with time that you’re seeing, or is it like if you, if you randomize individual providers, you can say, you can actually see, is there a difference, not pre-post, but with this intervention with particular providers? Yeah.

Josh 31:39

Yeah. So in retrospect, especially since COVID happened to happen, that made it a very difficult kind of trial design to run. I don’t actually know why Angelo and the team made that decision early on in the research. I came onto the team afterwards from an implementation standpoint.

Kei 31:55

Yeah.

Josh 31:56

But I think it has to do with feasibility of being able to get this intervention out to a lot of places, 3 different large health systems with many, many different oncology clinics. And so it was to get the intervention into as many places as possible. But it definitely introduces that timeline bias, Eric, which is a huge problem, especially during COVID no doubt about it.

Eric 32:16

And especially from a pragmatic trial, like actually you guys used 3 different large groups. There were tons of patients.

Josh 32:23

Thousands.

Eric 32:25

Yeah.

Alex 32:26

Yeah. Step wedge, which step wedge was the cutting-edge science at the time. It was like the, yeah, true. You know, it was, you can’t really randomize the patients because they’re located within clinics. And you’d like to have all the clinics have the opportunity to participate in the intervention eventually. So you start by training some clinics, but not others. And then over time, as you move along the wedge, you end up training all the clinics. And then for any individual clinic, you can compare it to other clinics that didn’t have, you know, let’s say they have the intervention at the time you’re comparing it to other clinics that didn’t have the intervention at that time.

And you can compare it historically to a time at which they didn’t have the intervention previously. So there are ways of — there are advantages to the stepped wedge. It’s now somewhat fallen out of favor for a variety of reasons, including that it’s very power hungry. And like, if something happens temporally that shifts things in that time period, then it kind of becomes, you know, there’s that before-after element that’s, you know, weaker design.

Eric 33:28

And importantly, pragmatic study. So you didn’t actually look like — did the providers actually didn’t have to do the training or the patients didn’t have to use the video? Do you know if they did?

Josh 33:41

Uh, there’s data on that. I don’t think Angelo has published any of the secondary outcomes. The data usage was, or the video usage was quite high at 2 of the organizations and lower at one of them. But the video use was high and the provider training uptake was pretty high too. So they put a fair amount of work into the, into both of those things. So again, that. From my standpoint on a, as an implementer, I think those are 2 things that were probably higher than would happen in the real-world setting. And I think there was pressure to get that done.

Eric 34:09

Yeah.

Josh 34:09

Um, so both, both were high, but it wasn’t required and, and the older patients weren’t flagged. Right. So it could be any patient watching the video.

Eric 34:16

Some difference between the sites though. So 2 sites with higher use than another site.

Josh 34:23

Correct.

Eric 34:23

And we don’t know why yet.

Josh 34:25

No, we don’t.

Eric 34:26

And did it do anything?

Josh 34:29

Uh, it did. So it changed, it changed documentation in the medical record as assessed by the tool that Charlotta, uh, designed again, that Kay described, right? So there, I think it was a 25% to 20% roughly difference between intervention periods, again, not clinics, as Alex was, Alex was saying. So intervention time slots had more documentation than non-intervention time slots. And the documentation was assessed as, um, natural language processing, looking for clinician documentation of goals of care, kind of conversations of treatment preferences of, um, hospice or palliative care.

So it was looking for specific things. And again, that tool is a human-assisted NLP, which I think Charlotte probably talked to you about. So there’s a human coder going through the notes, but the NLP takes those tens of thousands of data points and distills them down to small quotes, chunks that the human reviewer can look at, who is blinded to whether the person is control or intervention timeframe.

Alex 35:31

And is that, what do you make of that difference? 20% versus 25%. Like it’s not —

Eric 35:38

Yeah, 5% difference.

Alex 35:39

It doesn’t blow you away. It’s also still pretty low. It’s also like low at baseline. You’d think patients with advanced cancer, we’d be doing better at baseline. So there are a lot of stories there. What do you think?

Kei 35:54

Yeah.

Eric 35:54

Is the 5% worth the squeeze?

Josh 35:57

Yeah. Good question. So I think as someone who’s done implementation outside of research around palliative care and communication interventions, it’s a bit impressive actually, considering how big the institutions are. I don’t, I don’t think this number is an easy number to move amongst large institutions. So I’ll say that I was impressed, right? I told you my hypothesis was this is not going to do anything. So I was a little surprised that even at an absolute increase of 5%, it was impressive, but a relative increase of 25% or whatever that is, my mathematics are not going to be good today. I was impressed.

Eric 36:37

You are starting to sound like a pharmaceutical company focused on the relative risk increase, but not the absolute.

Josh 36:44

Yeah. But again, is it high enough? Oh man, I would’ve liked to see a lot higher. I would’ve, uh, I personally would’ve liked to see a lot higher. And I think it begs the question of kind of which of these interventions and in what way are we gonna get the biggest impact?

Eric 36:59

And it also sounds though, like, you know, is it the juice worth the squeeze? Like the lemon wasn’t very big here, right? Like it was just access to videos and, uh, some training, but there was nobody saying you gotta do this, gotta do this. There’s, there, you’re not hiring additional people.

Josh 37:15

That’s right.

Eric 37:16

This was, you know, as far as getting this up and running in someone’s clinic, actually fairly low investment for a 5% absolute risk increase, which is actually quite large when you think about kind of what you’re getting out of this.

Josh 37:36

I agree with that, Eric. Like, I think we’ve seen a number of studies that have shown that training alone don’t move the needle. So here you had a combination of an elective training plus these videos, and we move the needle a little bit, right? Which is something, I suppose. I don’t think there’s a magic bullet here that takes the number from 20% to 60%. I think it takes work. I think it takes a number of different interventions. And I think from an ACP study standpoint, this is a pretty low, a low control, low push thing. And this is maybe what we can expect if you don’t do a big, big push. Right.

Eric 38:11

What’d you think about the 5% difference?

Kei 38:13

I mean, you’re talking about, so now we’re only talking about my, our study of only like a few hundred people. These are like thousands of people, right? So, in absolute number, even the 5% is a lot. And I don’t know if my family member or myself was in that patient population, I would really like that. I think it’s just the bigger scale.

Josh 38:38

Yeah.

Kei 38:38

The absolute number is really large.

Alex 38:40

Well, one of the things I love about both of these studies is we’ve had people on talking about about, you know, prepare for your care, talking about Vital Talk, talking about video advance care planning. And we’re always like, you know, how does this one intervention compare to the other? And what you did here is you married 2 of the interventions. It’s the videos and it’s the Vital Talk training, right? Like, so it’s like Vital Talk in Kay’s case was like used to train the RE:MAP framework, was used to train these clinicians. And in Josh’s case, they had the opportunity, the clinicians had the opportunity to view these videos, the Vital Talk. I’m sorry, the Vital Talk, is it videos or interactive website, right?

Kei 39:27

Mm-hmm.

Alex 39:27

To learn the skills. So, it was like a kind of a marriage of these 2 innovations. And that’s, I think, one place the field needs to go is like often You know, we have these studies designed where there’s just one intervention and you wonder, you know, and it has, it doesn’t work so well, right? But now we’re seeing these multi-component interventions that have these different parts and it’s, you know, we’re starting to see some positive results and maybe that’s what it takes. Cause we’ve had a lot of negative, we’ve had a lot of interviews with, with people who’ve conducted negative trials. Right? Studies that just didn’t work out for one reason or another. And one of the things that I think is really interesting about this is that marriage of different interventions at the same time.

Eric 40:16

Okay. I got a question for both of you, and especially when it comes to K’s, this change in goal concurrent care, does this settle the argument whether or not advance care planning works or not?

Kei 40:32

So, I mean, I know, you know, I know you guys all know that many, there are many different ways to kind of measure global burden of care.

Eric 40:41

Yeah.

Kei 40:41

And this is just one way of doing it. And when I saw this, I was like, wow, this is great. At the same time, you know, a lot of people do question me, like, it’s one way you do it and there are many other ways to do it.

Alex 40:54

Yeah.

Kei 40:54

And also, do we actually care how they died? And don’t we care more about how they feel before they die?

Eric 41:02

There’s always something.

Kei 41:03

You know, like there are many different ways to kind of think about this issue, but I do think that we’re getting to, um, it is a stronger outcome compared to like, did the conversation happen?

Eric 41:14

Yeah. I think Josh, there was no goal concurrent outcome in your paper, right?

Josh 41:20

Not in my paper. No, not in that paper. It’s coming out. I mean, Eric and Alex, you’ll see it. So there is a very cool study coming out. Angelo, Michael Paciorno, who’s another one of the senior researchers on this, and Edith Burns at Northwell Health and Charlotta are using NLP and AI to try and do another step up on kind of what Kay described, where there’s the AI tells the story of what it sees in the chart. The human tells the story of what they see in the chart.

And then people decide if that’s goal concordant, um, a description of a concordance picture using a scale again, which I’m really, I have not been involved in that, but I’m super curious to see what that shows. So like, again, it’s supposed to be more nuanced. Does this story tell, or does this, does this story tell a story that is one of, of a patient and clinician being empowered during a difficult time? Right. Like, which I think is what goal concordance is about.

Kei 42:21

Hmm.

Eric 42:21

Okay. So you are My last question to both of you. You’re a busy hospitalist CMO. You got the emergency room, you got oncology clinics, you got everything. You have a limited budget, but you’re — you’ve listened to this podcast. You’ve convinced them that something works, but they really only have the budget for one thing. Could be the videos. It could be a trained clinician in VitalTalk.

Kei 42:43

Oh.

Eric 42:44

Could be something else that we didn’t talk about. Like, what would you, what would you tell that CMO? That they should invest in?

Josh 42:52

I’ll take that one first. Yeah, I want to take that one first. So I’m going to, I’m going to double down on what Alex said, which is I would get, I’m going to, I’m going to answer your question, Eric, but a little bit in a cagey way, which is I would get a list of the different tools that are out there. There’s a good number of them that are free that include conversation guides, templates to fill out, workbooks for patients, patient videos, uh, video declarations. Like there’s so many neat tools out there. I would give them a list of those.

And then I would tell them to invest in their clinicians in terms of preparing them and giving them time to have these conversations and carrying patients through it. And then provide a list of tools that are out there, because there’s a lot of tools that people can use for free. And I do think the clinicians need some help, whether it’s through training or through coaching and ways to create space to have these conversations and improve on it. So I’d put the money in the clinicians broadly, nurses, social workers, chaplains, everybody.

Eric 43:45

Yeah.

Josh 43:46

And then give them the tools that we have all created because there are a lot of cool ones out there.

Eric 43:50

Wonderful.

Josh 43:51

Is that fair?

Kei 43:52

I totally agree. I mean, there are many tools, and we already know which components actually work together to actually produce more conversations and more goal-concordant care, I think. And the challenge for the health system administrators is to figure out how to make that more scalable, not just like Let’s pay clinicians to do this, but how do we make this more scalable? And solutions exist. I think we have to invest in making this more scalable, like using AI and other things to kind of augment, like kind of reduce the time that clinician needs to spend.

Eric 44:31

Well, again, we’re gonna have to have both of you on podcast for the next rendition of Video Killed the Radio Star, talk about how AI has changed everything.

Alex 44:39

Yeah.

Eric 44:40

Probably like six months from now, but maybe we’ll get a little more more of that song, Alex.

Alex 44:46

The pace is pretty rapid. All right, here’s a little.

Eric 44:48

How’s he listening to lyrics? Just think about AI.

Speaker 5 44:55

I heard you back on wireless back in fifty-two, lying awake intent on tuning into you. If I was young, it didn’t stop you coming through. Oh, oh, oh. They took the credit for your second symphony, rewritten by machine on new technology. And now I understand the problems that you see. Oh, oh, oh. I met your children. Oh, oh, oh. What did you tell them? Video killed the radio star. Video killed the radio star.

Eric 45:39

I love that. You know, one of the authors of this song, I just want to say a quote. When they came up with this song, they were picturing that a vision of a future where record companies would have computers in their basement and manufactured artists. So we are going to have to have that AI. Yeah.

Alex 46:00

There we go.

Eric 46:02

Josh Kaye, thanks for joining us on this podcast.

Kei 46:04

Always a pleasure to join. Thank you so much.

Eric 46:07

And thank you to our listeners for your continued support.

This episode is not CME eligible.

For more info on the CME credit, go to https://geripal.org/cme/