AB: All right, so our next speaker, we’re really excited to introduce you to him. His name is Dr. Nahar, he is a pediatric cardiologist and associate professor of pediatrics at Children’s National. And so Dr. Nahar, if you’re on, would you come on and give a brief introduction and background, and I’ll pull up the slides for you while you’re doing that.

JN: Thank you, Ainsley, and thank you to Stan and the HITLAB team for having me, and it’s my pleasure. This is my first presence here at the HITLAB platform and meetings, so welcome to everybody. I’m a pediatric cardiologist, so basically I’m a clinician. I’m totally patient-facing. I work at Children’s National Hospital in Washington, D.C. My interest is also in advanced digital technologies, including the artificial intelligence and upcoming the whole light language models, conversational interfaces, and how can we use technology to make the lives of providers, meaning the clinicians and our supporting staff and the patients and their families better. So that’s where I am, and my topic actually is related to it and fits in the broad theme of health equity. The topic is current and potential applications of ambient artificial intelligence promoting Techquity. So next slide, please. The agenda for the talk today is actually, first of all, I’ll give a brief introduction about technology. So technology is the ambient AI and its clinical applications, and then next, integrating technology with the health equity. That’s the Techquity. That’s from the Techquity lens, what do we see? And I’ll end up with some barriers, challenges, and call to action in future directions. Next slide, please. So Stan and I co-authored this paper, which was recently published in Mayo Clinic Proceedings, Digital Health, which gave us a little incentive for this conversation. So the topic is basically current and potential applications of ambient artificial intelligence. Next slide, please. So first of all, going back to the basics of technology, what is ambient intelligence? Cook has described the ambient intelligence as a presence of digital environment, which is sensitive, it is adaptive and responsive to presence of people. And when I say ambient intelligence, it uses – it’s basically the technology which lies at the interface of IOT sensors and the whole gamut of digital sensors, variables, and remote sensors, which are patient or end user-facing, machine learning, artificial intelligence, knowledge graph technologies, human-computer interaction, and pervasive computing. Next slide, please. So this slide actually is a high level of a description of what are the clinical applications of ambient AI in healthcare delivery. Out of these three categories, versus the outpatient, inpatient, and the home health, in each of these categories, I’ve highlighted one point with red font, and basically those are the things which are important for promoting health equity. So let’s dive in deeper. At the point of care, for patients, especially those who have English as a foreign language, for them to communicate with the provider, for them to get best benefits of the visit at the point of care. Be it clinic visit or be it a virtual visit, sometimes you have a language as a barrier. And what can we do to promote a smooth interaction with the patients and the families? So nowadays with ambient AI technology and integrated GPT-4 and the newer advanced large language models, there is integration of seamless foreign language translation and interpretation services, which can help with the promoting one aspect of the disparity which can happen. So next is the hospital journey. The most important thing which I see here is, and again, in the upcoming era of generative AI is discharge planning. When the patients leave the hospital, especially for complex admissions, there is too much being told. Many people come and talk to the patients and their families, and there’s a lot of – sometimes there’s a lot of apprehension and patients might forget. So at that time, if there are some generative AI tools and solutions available to carve out simple discharge planning in patients’ level of understanding, culture, the economic status, maybe it could be a voice recording as well, to help them in discharge planning process and so that they can carry it home in a multi-modal fashion. Next is home health optimization. That’s actually a big space to promote health equity, especially for people who have disabilities, people who are older, and people who cannot work temporarily. So in the home health optimization, ambient AI from the equity perspective, there are multiple uses. Number one is in monitoring of activities of daily living. Especially for people who are older, it might help out for detection of risk for fall. And if the fall happens, these can be picked up promptly or prevented so that the caregiver is activated and emergency services are being called. Home rehab is actually also an upcoming field in which patients who are recovering from stroke and other motor disabilities can have help of ambient sensors, which can be used along the interface with the artificial intelligence and help these patients to fulfil a good home rehabilitation program. This is good for patients who are in remote areas, rural facilities, or patients who have problems with transportation. Same again, how can we promote equity for these patients? Next and upcoming field is also very important is early warning systems and disease detection and surveillance. There has been work going on in Parkinson’s and other diseases in which, again, patients who do not have access to the care or who are at risk can be detected by the use of ambient sensors. And that can help in actually improving the health equity and their health outcomes. Next slide, please. If you look at value creation of the ambient AI on the two squares, one on the left upper, ambient AI by actually – by promoting health equity can promote better outcomes, as I mentioned earlier. Ambient warning systems and prevention of adverse events, those are very, very important. On the extreme right lower is improve patient experience. Again, if you provide patients with technological tools at their level of understanding and what is feasible and assist with daily living, working, and support of the healthcare, this would improve patient experience as well. And ultimately, help in lower cost and good life – health and life balance. Next slide, please. So next, coming to the lens of Techquity. So Techquity is a strategic decision development and deployment of technology to advance health equity. And the reference for this definition is written below. Next slide, please. So let’s look at the adoption barriers in relation to Techquity, looking at the barriers and challenges. First and most important is the access. There might be people who might not have internet, broadband internet access. People might not have internet access to this fancy sophisticated sensors and gizmos, the health tech. There might be lack of digital literacy, and tech literacy, digital divide, affordability. They cannot – maybe they don’t have capacity to purchase the computational devices. And what about the insurance? People who are underinsured, who do not have insurance, or those who are marginalized, there is absolute access and a barrier to the Techquity. Next is user interface and experience. This is a good one as well, especially for people, again, who are not tech savvy, or people who are older. We need to make the solutions which are at their level of understanding and which they can operate smoothly. Patients might have problems typing, so maybe a voice interface might be better. And ultimately, if the patients can use it, they’ll be able to sustain engagement with these solutions. The other one is financial, which is a system level cost. Who is going to pay for these solutions? Who is going to pay for all these tech solutions? How are the reimbursement models going to be carved out? And data management and logistics, who will deal with all the data coming out of this? Do we have a pipeline? Who is the clinical staff? And naturally, that’s going to be a problem as well. That’s a big burden. And last but not the least is legal liability. What happens if there is a problem within an AI system or the data is not looked properly and that opportunity is lost? Next slide, please. So since we’re in the Health Equity Symposium, I definitely felt very pressured to put this slide is actually assessing bias and fairness, especially when we talk about AI solutions and ambient AI. The biases, as we all know, are the data sets, biases in the input data, algorithmic biases, and two important biases which we tend to overlook is human cognitive bias. Actually, this includes development, design, and deployment of AI solutions. And last but not the least is systemic bias to the level of healthcare organizations and enterprise, which can sometimes be adverse to the marginalized or under-served populations. Next slide, please. Future directions and call to action. So there are a few initiatives which can be taken. First is development of collaborative communities and partnerships, especially which can help with people who lack resources, who lack internet access, who lack digital or tech literacy. Those are the types of partnerships multi-stakeholders could help. And I see HITLAB as actually a very noble platform to bring to foster these partnerships and communities. Next is human-centered design. For a tech to stick, for an end-user to use the tech, it has to be human-centered like the end user-centered to enable a good user experience, good user interface. Innovation in reimbursement models is very important. Ultimately, we can talk all day about the tech and how fancy these things can be and what they can do, but until they are reimbursed or there’s a good reimbursement model which are lacking behind, there won’t be enough traction or adaptation. Next is data and AI governance, which is last and most important, is actually, again, having safeguards against the biases and ensuring a fair and inclusive and equitable solutions for the healthcare. And I think that was the last one. Yeah, thank you for your time, and I’ll be happy to entertain any questions.

AB: Thank you so much. And I have a question actually. So to follow up on that and the overall theme with health equity, so how do you think the AI can actually help increase health equity? Do you have any thoughts on that?

JN: So good question, actually. Depending on what type of AI solution you’re looking at, the first and most important is actually, which is a topic near and dear to me is natural language processing. Especially when we have the – as clinicians we use the electronic health record. So, there is a vast treasure of information in the electronic health records. When we are dealing with a patient who has silent social determinant of health gaps, how can the AI help up – natural language processing can target those, can actually highlight those gaps and bring us to our radar so that we can have a care management team help out with appropriate resources for these patients when they go out in the communities to kind of take care of the health for preventing any adverse events and preventing any undue hospitalization. Second also is actually about the AI solutions. As I was talking about the generative AI, how can you be creative with the technology actually, especially for people who are being discharged, who are apprehensive, who do not have enough resources. Come up to the level, maybe a low tech touch, but give them something which they can carry on, be their life partner, be their intelligent digital concierge. That’s where I feel AI can help. There are a lot many possibilities, but these are some of the important use cases.

AB: Well, thank you so much for sharing your insights Dr. Nahar. It was really great to have you on our symposium and thank you so much for taking the time today to share this with our audience.

JN: Thank you so much. My pleasure and thanks everybody for being on