How AI Is Transforming Communication in Emergency Care

Aurelio Muzaurieta (00:08)
Hi, everyone. Welcome to Care Culture Talks here with No Barrier. My name is Aurelio, and I'm going to be hosting today's talk with a very esteemed guest from Brown Emergency Medicine, Dr. Yosef Berlyand. I hope I'm saying that correctly. Sure. So we're really excited to chat today and get to know Yosef. He comes from

an interesting background of caring for patients who don't particularly use English as a first language. So, Yosef if you could kind of start us off, tell us a little bit about your journey into emergency medicine. I'm particularly interested in hearing this just as an emergency resident at Stanford myself. I kind of am always interested in hearing about different career paths within emergency medicine and what gets people excited about it. And then we can kind of go into...

that talks about caring for patients with different language needs.

Yosef Berlyand (00:59)
Thank you, first of all, Aurelio, for having me. So I guess to tell you a little bit about myself, my parents initially came actually as refugees from the Soviet Union. So they came in the early 90s. And we were Russian-speaking family. My grandparents moved to the United States also in the mid-90s, also Russian-speaking. And I think I have that very similar background to a lot of people who are interested in caring for patients with who

from different cultures who speak different languages where I myself grew up speaking a different language and found myself as a kid in that position all the time where I was the medical interpreter with absolutely no medical knowledge. My clinical background now is that I work at Brown University where most of my clinical care is at Rhode Island Hospital. Rhode Island Hospital is the safety net hospital for the state of Rhode Island. So what that means is we care for the vast majority of our patients who are underinsured, uninsured, who have no other access to care, no primary care physicians. And as you know, that often ends up correlating, unfortunately, just the way it is in our country with patients who don't speak English very well, who may speak no English, who come from a variety of cultural backgrounds. And that is just a large part of my practice. And our emergency department, unfortunately, serves as their only access to care.

And so I find myself on a typical shift caring for a very large fraction of patients who have limited English proficiency and needing to find ways to be able to speak with them and communicate so that we can actually give them the care that they're looking for.

Aurelio Muzaurieta (02:24)
Hmm.

Absolutely. What are some of the methods that you and your team use currently at the Rhode Island Hospital to communicate with patients? Could you talk a little bit about what cultures and languages these patients are typically representing?

Yosef Berlyand (02:39)
Yes, so a lot of our patients are Spanish speaking. That is like the large chunk of sort of single language. Then we have a large swath of people who are Portuguese speaking, specifically Brazilian Portuguese. And then from there, there are just a variety of languages. We see Haitian Creole quite a bit, which is actually quite difficult to find interpreters for.

And from there, just a variety of languages, Mandarin, Cantonese, et cetera. Occasionally, I get to see a Russian speaking patient, which is very nice, but unfortunately not the majority.

Aurelio Muzaurieta (03:06)
Right. Sure.

Yosef Berlyand (03:08)
We have, at Rhode Island Hospital specifically, we have excellent in-person Spanish interpreters available for us 24-7 in the emergency department. And we find when we're speaking with patients who speak Spanish, we actually have really excellent ability to communicate with patients. And it's for every other language where we don't have readily available in-person interpreters, where we find that a lot of things, at least in my experience, fall through the cracks or are just harder to communicate and harder to do. We rely, which I think most places do, on interpreters on wheels, which have the iPad screen that has a selection of different languages. It dials into somebody and that has variable availability of languages. Sometimes a language is listed but no interpreter is available. Sometimes there's a very long wait for an interpreter.

Sometimes the iPad isn't loud enough for an interpreter, for a patient to be able to hear, especially patients who are hard of hearing, who are older. And sometimes there's just simple connectivity issues. We're in a corner in the ER and the wifi isn't very good. And I'm trying to use the Haitian Creole interpreter, which we don't have an in-person for in the middle of the night. And the iPad won't stay connected. And I can't speak to the patient, even though we pay for a service and have a service, but it's these like little things that all add up, that create barriers to communication.

Aurelio Muzaurieta (04:26)
Absolutely. I understand kind of from the emergency perspective, it's such a fast-paced environment to be caring for patients and there are corners of the emergency department that don't have any Wi-Fi and you're frustrated. Oh, this is this it's working in this room and then not in another room. You know with the hard of hearing patients, it's so much nicer to have somebody who's there in person. They can try treat the volume and make it easier or even in a trauma situation I don't know if the Rhode Island hospitals a level trauma if you guys I assume so so even those can be particularly challenging as well you know when you're trying to assess neurologic function and status it's like can be really really hard and you know it's one of those things where when I think about the future of of our country and more multilingualism in patients particularly in older patients and new immigrants, there's so much kind of exciting technology coming out with respect to this advent of AI and being able to, literally large language models were kind of made perfectly for the interpretation space. But there are certain kind of areas where I've thought of like how it could add so much value to our patient interactions and other parts where we have to be a little bit more careful in terms of what we kind of let a robot do and what we want to delegate to our professionally trained interpreters, either in person or through electronic means. Are there any kind of situations where you feel like an interpreter that is kind of AI powered would kind of be sufficient and ideal in other situations where you feel like, particularly in our practice in emergency medicine, where it might not be so good?

Yosef Berlyand (06:09)
Yeah, that's a great question. I think the vast majority of the time, an AI-powered interpreter is probably the best solution because many of our problems are hinged around this timeliness in care. We have to be able to get to a patient and in the moment that we are available to see them, because we don't control our schedules, we don't have blocks of time when we know a particular patient is coming.

We have to be able to communicate with them in the moment and do it on demand. And the best solution for that is using AI because it's on demand without wait times. Let me give a few anecdotes. So I think a couple of months ago, I was working at one of our hospitals. So I work at a few hospitals. One of them is a specialty OB-GYN hospital where I work in the emergency department. But it doesn't have, pediatrics in the sense that we don't have a pediatric floor, although we as the emergency physicians do provide pediatric care. And a common misunderstanding by patients, especially those who are not native English speakers, are that they delivered a baby there. And so they assume that that's the right place to bring a baby for care if their baby is sick. And so I was working as the emergency physician. We have one of us on at a time there. It's a small hospital.

And a woman came in carrying her six-day-old baby. She spoke no English. She was Haitian Creole speaking. She could not tell us the name of the baby, the age of the baby, or what was going on. She just brought in an ill-appearing infant. And here we are all scrambling, trying to get an interpreter on the phone just to find out what is going on, what is happening, why are you here, how old is this child.

Yosef Berlyand (07:43)
Like did they fall? Was this a trauma? Like we have no idea. And it took us minutes. It took us 10, 12 minutes to get somebody of, you know, we used one of the excellent national services and we waited for 10 minutes and interpreter came on the phone, but 10 minutes is critical when you're caring for such a sick patient.

Yosef Berlyand (08:02)
And that kind of a moment is where you can see if I could just, I just need on demand interpreter services and that AI can fill that gap so well, so much better than any other thing else can. And that was just one example. And there are dozens of anecdotes like this where we need interpreter services in the moment. And you know, this is an emergency physician. We ended up relying on means that are far from ideal. Yes.

Aurelio Muzaurieta (08:25)
Suboptimal.

Yosef Berlyand (08:27)
...to try to communicate with our patients and try to reassure them and investigate what's happening. And once we had an interpreter on the phone, we got down to what was going on and figured it all out. But for 10 minutes, we couldn't even get the baby registered, which also meant when I tried to call a consult, they asked, who am I even seeing? I don't see anybody in the medical chart. So...

Aurelio Muzaurieta (08:49)
Wow. I can empathize with the fear that you kind of feel in those moments where you feel like you're a little bit desperate and communication is so important, yet you sort of are without the means to do that. And what can you do? How do you mobilize the important resources around you to make that interaction a successful one in care for your patients. And that's what is kind of, I think the most exciting part of this advent of AI in entering into healthcare interpretation is this major gap that we have, I think, particularly palpable in emergency situations. In these areas where patients come in speaking so many different languages and so many different ⁓ backgrounds. I feel similarly about the relatively good availability of Spanish speaking providers and interpreters that are able to kind of care for the large swathes of Spanish-speaking patients. Even one of the challenges actually that we face here in California is that the different dialects of Spanish actually kind of cause a little bit of an issue with patients, for example, majority are Mexican or Central ⁓ American here in California in a Spanish-speaking...

⁓ California, yet you you might be in in New York where more are Dominican or Puerto Rican speaking or Cuban speaking and Jamaica Plain by Boston and Portuguese has a similar issue with Cape Verdean Creole versus European Portuguese and Brazilian Portuguese. I those people and I say this because I've studied these languages and looked at some of these places where I've found a difficulty as a Brazilian Portuguese speaker having patients in Boston who were from Cape Verde, know, many words and different like difficulties. We could figure it out, but it took a lot of time and actually somebody who's more trained in that probably would be better. And some of the AI tools, actually, No Barriers is one of the main ones, have multiple versions of Spanish and dialects of different languages that we're training on different models. ⁓ And it's pretty impressive to see that, you know, we might have a Spanish interpreter but not a Spanish interpreter for this particular dialect of Spanish and how do we kind of work with that. We're trying to fill those gaps for different forms of the common languages. I had a situation in the surgical ICU where I had a Turkish speaking patient and he was extremely tachycardic and his pain was poorly controlled and he had a femur fracture and I'm like, um, you know, like it's, is kind of looking like PE and kind of concerned or fat embolism rather. Uh, and, know, a little concerned about this. Uh, but it was really hard to, every time we wanted to talk to him on the phone, it was with the interpreter, was like 10, 15 minute wait time to do that.

And he initially presented the ER, it was even worse. We had very difficult times connecting. So I do feel like the value is very high there and it's exciting. I also think about times where it's a little bit more dangerous maybe to rely on AI or an interpreter. Like one of the ones that you had mentioned that I don't know if you would agree with or not is potentially the older hard of hearing patient ⁓ that could be potentially misinterpreted. The other one that I think of is goals of care and end of life care. And I don't know, do you have any thoughts around those areas, especially in emergency where these conversations are probably had often?

Yosef Berlyand (12:24)
So that's interesting that you bring up the hard of hearing patients because that is who I struggle with for all forms of interpreter services, whether it's in-person interpreter services, although that probably is the best, versus the iPad ones with AI. Similarly, you're relying on an iPad speaker, which can be quite difficult. But I think a benefit of AI is you can put text on the screen too.

And even if a patient is hard of hearing, you actually might be able to get a solution here. If a patient has their reading glasses and can read very fine, which we see all the time in older patients, they can't hear very well, they left their hearing aids at home, how often does that happen when the ambulance picks them up? Hearing aids are at home, but their glasses are on them. And you can still be able to communicate with them using AI. Whereas...

Aurelio Muzaurieta (13:01)
You're right. That's true.

Yosef Berlyand (13:08)
End of life care, I agree with you. I think when you are making a life and death decision and making a decision to comfortably allow somebody to pass and having that conversation with their family with them, that's where it does matter to have somebody there physically in person to make sure that there are absolutely nothing that is misunderstood.

But I think, you know, those conversations, while we have them, of course, in the ER, that is not what the vast majority of our conversations are. Most of our conversations are taking a history. And for that taking a history portion, there's so much room for AI to be able to step in and improve. I, again, agree with you for sort of that common dialect. So at Rhode Island Hospital, most of our Spanish speaking patients are from the Dominican Republic.

Yosef Berlyand (13:55)
It is, but by happenstance, all of our interpreters are also largely from the Dominican Republic. So we're able to communicate very well. And that to me is the gold standard. That's fantastic. But it's as soon as somebody else is coming in who is not from the Dominican Republic, who speaks a different dialect of Spanish, who speaks a different language, who speaks a rare language. That's to me like this big gap in care that we can improve right now.

Aurelio Muzaurieta (13:59)
Okay, that's great

Yosef Berlyand (14:21)
And there's a huge opportunity for that.

Aurelio Muzaurieta (14:23)
Absolutely. It's interesting to think about where those opportunities lie. Are there any areas in your practice where you guys have to do sort of discharge instructions. We know that a lot of patients will come to the emergency department, get care, but not quite understand what the aftercare is like or what's required or what's necessary, often because of a language barrier. And sometimes even in English, you know, we have difficulty with after-visit summaries and discharge paperwork and things like that.

Do you all have a practice where physicians sort of go through the paperwork with patients or nursing or some other person in the staff before somebody leaves? And what is it like when there is a language need intertwined in that interaction?

Yosef Berlyand (15:17)
Yes, so our policy is to make sure that each patient at discharge, somebody has to go through their paperwork, whether it's a physician or nurse with the interpreter. Often that is an interpreter on wheels if it is not Spanish. If it's Spanish, then it's often an in-person interpreter. The actual written documentation, it's a bit variable.

For those of us who speak only English, usually the customized instructions that we write are only in English, but there are often some pre-made attachments. For example, say you have a sling, there might be a pre-made how to use a sling attachment that's available in a number of languages.

I serve as our Associate Director of Quality and Patient Safety. And one of my jobs is to help make sure that our discharge instructions in general are less variable and contain sort of the same information across different physicians. And so I've been slowly working with my team here on building up a larger repository of instructions in different languages. But unfortunately, we can only account for sort of standardized things. We can't customize.

It's not necessarily safe to use out of the box like Google Translate or something else. And I do think that this is where AI can help, especially if it's integrated into Epic, which is where a lot of things are trending. That is where you can absolutely create excellent discharge summaries and make sure that they are correctly translated.

Aurelio Muzaurieta (16:39)
Potentially even improve the quality of the language interface that a patient would have in their MyChart or MyHealth epic interface that is on the patient's ⁓ seeing side.

I know that there's more technology coming out where there are actually new AI models that are helping to interpret some of these tests and prompt when somebody should see their doctor a little bit earlier than later. And if that can be integrated with language concordance care in mind, I think that there could be real measurable improvements in health outcomes. You've probably seen it. People with language needs don't generally get the same quality and timely care that primarily English-speaking patients do receive. And I don't think it's based on ⁓ any intentional bias from the provider, but as much as it has to do with the literal the linguistic and cultural barrier that is formed there and then how to kind of get over that.

I think that they've done some studies in wait times in emergency departments with English speaking patients versus a language need patient and they are longer than language need patient and it's not a surprise. I think that there's a long way to go to bridge that gap.

This particular health equity issue. But there's some good hope as technology improves and gets integrated into our health systems. Do you feel like there are particular...

Yosef Berlyand (18:01)
I agree.

Aurelio Muzaurieta (18:05)
risks involved in this technological development that's unraveling before our eyes? Maybe some people take a more optimistic look at it, or maybe like a cautiously optimistic rather. Do you have any thoughts around that within our field of emergency medicine or the other areas you practice?

Yosef Berlyand (18:20)
I would say I am generally a slightly more risk tolerant person because I very much believe that you can't let the perfect be the enemy of the good. And there is a perfect solution in mind, which would be that we have 24-7, every language, 180 in-person interpreters available in-house, just walking around the hospital who can come down at a moment's notice. But we can't do that. That's just not feasible.

Yosef Berlyand (18:45)
And so we have to figure out what the next best alternative is. And I think certainly like everything else we do in medicine, we can't predict what the medical legal risks are. Like how things will be described in the future, what kind of misinterpretations, if they occur with AI, they probably occur less frequently as has been studied by No Barrier than with some of the...

Yosef Berlyand (19:07)
national telephone services. And I think the risks are the same, essentially. There can be a mistake. That's true with services. But we can't let that stop us from providing the best care that we could possibly provide in the moment. So if this means that we can make our care better and there is some risk, then it's probably worth it if we can make our care better.

It's better than not using an interpreter or waiting 15 minutes or 20 minutes or 30 minutes to get someone on the phone.

Aurelio Muzaurieta (19:32)
I agree with that perspective. Well, wonderful. Thank you so much. I think we touched on a few really salient topics in this area. And thank you for your time.

Yosef Berlyand (19:40)
Yes, thank you very much for having me.