You are here

IGF 2020 - Day 9 - WS252 Connected Health in the Post-Covid-19 Era

The following are the outputs of the real-time captioning taken during the virtual Fifteenth Annual Meeting of the Internet Governance Forum (IGF), from 2 to 17 November 2020. Although it is largely accurate, in some cases it may be incomplete or inaccurate due to inaudible passages or transcription errors. It is posted as an aid to understanding the proceedings at the event, but should not be treated as an authoritative record. 

***

 

 

>> BRIAN SCARPELLI: Thank you, everyone. I think we are at the starting time. Hello, welcome. Welcome to this IGF2020 session, connected health in the post COVID‑19 era.

I think as IGF staff has noted in the chat, I can briefly say that everyone please be informed that the session is recorded and hosted under the IGF Code of Conduct and U.N. rules and regulations. And the chat feature is for social only. Only the Q&A feature should be used to ask questions. Both of those are enabled for us here today.

Again, my name is Brian Scarpelli. I'm senior global policy counsel with an organization called the Connected Health Initiative, a digital health multi stakeholder policy and legal advocacy organization. Very, very excited to be here today to talk about this extremely timely topic. We have a great diversity of experts here.

Thank you for your attendance, your interest. We look forward to an engaging discussion here. I can provide just a brief panel overview and then I can briefly note who our speakers are before turning it over to the speakers for opening remarks. So to give you an idea, a baseline about why we are here, the COVID‑19 crisis has really been a catalyzing agent for very rapid change in a number of sectors worldwide, but probably most of all the health care sector.

As the world has had to move to do things like produce travel and even reduce nonessential health care visits health care providers have adopted digital modalities, telehealth, remote patient monitoring solutions to handle a greater load of cases, and to incorporate social distancing policies and other policies to mitigate the spread of the coronavirus. These technologies can provide benefits with continued advancement in areas like AI which we'll address here. But these changes come with their own set of challenges and those most in need of connected health care can be the ones who are least able to access it for a number of reasons I'm sure we'll discuss here. Some of these impediments can be lack of smartphone access and internet coverage and capacity. Sometimes there are regulatory regimes and policies geared toward health care, in‑person care. That may inhibit the ability for these tools t greater provision of health care over the internet is really an incredible opportunity. But again there are new vectors for cyber security and therefore risks. How do we address this? That's what the panel is here today to discuss. There is no better place to discuss it than the internet governance forum. Thanks to the United Nations and IGF for having us on the panel. We are looking to explore the prominence of health care. It will be a lasting one that can include well‑being beyond the coronavirus pandemic, the current crisis or looking toward benefits of the move to telehealth, remote monitoring and other drawbacks ‑‑

[ Audio disruption ]

‑‑ expected outcomes that have been shared in the information for this panel. I believe we should make sure those are clear for all of you attending.

One, we are looking to understand the spectrum of opportunities and challenges that telehealth will bring to bear on communities during and after the COVID‑19 pandemic. And how the challenges are mediated by socioeconomic factors.

Two, we are looking to learn about what the IGF community can do to further ‑‑ for action and collaboration to realize the potential and challenges.

[ Audio disruption ]

In this conversation.

Three, we are looking to share perspectives ‑‑ I hope my audio is okay. Apologies. Oh, I think ‑‑ sorry.

Third, we are looking to share diverse perspectives on priorities and challenges for the IGF community. Is my audio okay?

[ Audio disruption ]

Not at all?

>> SVEATOSLAV VIZITIU: I don't understand nothing from what you were saying.

>> BRIAN SCARPELLI: Now? Now it's better?

>> SVEATOSLAV VIZITIU: It's better.

>> BRIAN SCARPELLI: Goodness. Well, to be efficient with our time here, what I was just doing was giving a brief overview of the abstract of the description of our session. I won't repeat that. It is listed, you know, on the IGF website. I presume everyone who has attended here hopefully has read it. I won't repeat it. I was going over that and expected outcomes. Many apologies for my audio issues. They seem okay now.

I would leave it to each of our panelists in their opening remarks to share anything about themselves they would like to share rather than take up more time. We are joined by a great slew here of Sveatoslav Vizitiu, the cofounder of well OD AI. Jelena Malinina with the European consumer organization, Subbarao Kambhampati from Arizona state. And the chief information officer with the minister of health from Buenos Aires, Jelena Malinina and Geralyn Miller with AI for good and Microsoft.

Again, we can turn it over to Sveatoslav if that's okay for some of the ‑‑ for your opening remarks.

>> SVEATOSLAV VIZITIU: Hello. I think I will present myself. That's more easy.

I'm Sveatoslav Vizitiu. We are helping the families with artificial intelligence systems to be more healthy. We have good results here in Romania. We also have summer schools. We have ‑‑ our children will be more healthy. Also we have a company that's developed a mobile application and games for children.

>> BRIAN SCARPELLI: Great. Thank you. I think Jelena is next to share anything about yourself, opening remarks.

>> Sure. My name is Jelena. I'm a digital health expert working at the European consumer organization. We are based in Brussels and represent 44 consumer organizations from 42 countries, mostly in Europe.

>> BRIAN SCARPELLI: Great. Rao?

>> SUBBARAO KAMBHAMPATI: I work in artificial intelligence which has gotten much of public interest because of the connections and societal impacts AI can have. Of particular interest that's relevant to this panel is how AI can be helpful in health and in particular connected health. So I'll share some of that information.

>> BRIAN SCARPELLI: Great. Analia?

>> First I would like to thank you for the invite to the forum. It is my first time here. It is a pleasure to share the health information strategies implemented in Buenos Aires city. I'm a specialist in health informatics. Also a professor of health information systems in universities. For more than 15 years I have been implementing health care institutions. Currently I'm the chief information officer at the ministry of health of Buenos Aires city. Since 2016 I have been working on digital transformation which means implementation of electronic records according to standards. We started working on the outpatient setting and the integration and during the pandemic we focused on in‑patient settings, telehealth and patient monitoring. I will give you more details later. Thank you very much.

>> BRIAN SCARPELLI: Geralyn?

>> I drive our strategy for our health investments. A program called AI for health which is a five‑year investment that's geared at bringing artificial intelligence, machine learning, analytics and visualization to some of the world's greatest challenges in health care. Thank you for inviting me today.

>> BRIAN SCARPELLI: Great. Thank you all for being here. So I guess we can get into it here. Again, we encourage those watching here to engage with the Q&A function in Zoom. But definitely have some questions that I can ask here, open ended. Please, anyone jump in. Respond to each other if you like.

Again, I hope my audio is okay now.

>> Audio is fine.

>> BRIAN SCARPELLI: Okay. As the public health crisis evolved we have seen telehealth monitoring solutions adopted to a greater degree than ever before what about these use cases is encouraging and what do you think are the major challenges that still need to be addressed? Do you think that the more rapid uptake during the COVID‑19 crisis will be lasting or do you see something of a reset, hopefully soon, not that we see a reset on use of digital health tools but hopefully the end of a handling of the COVID‑19 pandemic soon.

What will the impact of moving on past the pandemic be? Anyone.

>> SUBBARAO KAMBHAMPATI: I can start. I was just thinking as I was bicycling here about this "House" episode in 2008 called the frozen where House is stuck having to diagnose what's wrong with this lady that's a scientist in Antarctica. The whole thing is done in the old Skype. We have shifted obviously from Skype to Zoom now. What looked like a futuristic thing has become normal day‑to‑day scenario across many of these ‑‑ for many of us during this pandemic lockdowns.

I think it is very interesting that ‑‑ you asked what's going well and what are the things that we should be worried about and what might stay longer.

What's going well to some extent is at least speaking from the technology side we essentially had the ability to connect. I mean compared to something like 1918 pandemic, right now we can do things like Zoom and we have diagnostic technologies from AI, looking at x‑rays, chatting technology for triaging of patients and so on, all of which are being used as far as I know. My sister is a pediatrician and for several months during March she was actually seeing patients only through Zoom. So that has worked, you know, to some extent. The worries, of course, is there is also ‑‑ I think AI ‑‑ we can get to it later. But as of now AI technology is well suited as a triaging technique, but there is a tendency for people to let it take more role than just a triaging technology. That can lead to significant concerns, both ethical and possibly health oriented concerns. In terms of whether we will continue having this, I think there is no complete going back, it seems to me. We will get used to being able to do many things as a remote way. I think it is certainly going to have advantages in terms of reach and access the people in remote areas before the pandemic started. They didn't quite have the level of health care that city, for example, had.

So I think those things might continue and will be hopefully for good. I'll stop there and we can get back to any of the issues.

>> GERALYN MILLER: I would love to join in on those comments. What I think has gone well are opportunities like this for people in different areas of focus and cross sector and cross domain to come together and have an honest conversation about the state of things. That extends to the collaboration I'm seeing on the immense research that's happening. From a Microsoft AI for health program perspective we did an open call over the course of the summer this year. We accepted over 500 proposals and funded 150 of them which is just an amazing response in terms of people who are focusing on COVID‑19 research at the intersection of health care ranging from models with x‑rays and CT scans to understand potential disease to people looking at dissemination of behavior of information. Just to see the collaboration has bye‑bye amazing.

I want to also echo the sentiment that I think what the pandemic has done is laid bare places where there are inequities. So I think as responsible citizens of the global community that's something we need to take forward with us. We are now aware and that's one of the things to take forward.

>> JELENA MALININA: I will jump in next if I may. I think with COVID‑19 crisis at least in the EU we have seen a massive encouragement to use the technologies because if we compare Europe with the U.S. or with Asia, we are a little bit more conservative when it comes to adopting new technologies. It is also quite challenging in the health care sector, I would say, because despite we have the common legislation. We still have 27 different member states with very different rules on health care. So whatever is happening at the national level is happening at the national level. With the major health care service we are forced to move a lot of services online including health care. So now there are quite a lot of opportunities to speak with your GP, even to get some mental care possibilities. But when it comes to looking at what is next, I would say the next one is adopting our legal systems into the technology world.

I can give you an example for Poland for example, at least until. The act only considered a medical act when two people are present in the same room or more than two people. Basically talking about medical consultations in Poland and any rights for the patient or person and liability. If we speak about cross‑border health care service when say someone from Romania decides to call the polish doctor this would be illegal.

Then the patient who might receive any kind of harm through this service would receive nothing in return because the legal system doesn't cover it. There are plenty of examples when it comes to telemedicine, when it comes to diversity of products, AI which is also for technology advancements.

>> ANALIA BAUM: In Buenos Aires the issue of health is installed in the public sector. Patients, health care professionals and policy makers realize the benefits. So it is an issue of public interest. So this means it requires political commitment. The second good news is the health emergency introduced the concept of remote monitoring patients. This is new for public health in Buenos Aires City because home care services didn't exist and I think this strategy is here to stay. What still needs to be addressed is implementation research to understand accessibility, inequity in telehealth services and patient needs. Also regulatory frameworks like Jelena said.

Although we published temporary regulations now we need to work on telehealth and prescriptions. This is a trend that will last beyond the crisis. I think that it will. At the moment there are no agreements between professional associations on this issue and in the private sector there are no agreements with care providers and the insurance services either.

Also the national ministry of health needs to finish up the regulation frameworks of the recent laws in order to implement telehealth services.

With remote patient monitoring we need to discuss which strategy the minister needs to address that will benefit from connected health. During the pandemic the necessity was clear although the resources ‑‑ to stopping the circulation of the virus. Now we are expecting about the future policy decision and investment. But surely it would be great if we could continue offering the services to all our patients. Thank you.

>> SVEATOSLAV VIZITIU: I'm the last one on this question. For myself and for our company it's really advanced but worth the crisis of COVID. It is not a good result about health and people dying, but it's given like a rocket for digital health. Also for telehealth. For example in Romania with telehealth it is popular now. Results are showing that people are addressing now to the doctor more than before and the results are better. It's lower than last year with COVID which is amazing from my opinion.

Before COVID Germany made the first steps toward digital health and because of COVID every country was forced to employ the digital health and help people online and so on. In my personal opinion about the COVID, it's positive for families because each family stayed home at least two months and they stayed with kids. It is an advantage for kids and families because they were very united. They did many activities together and so on.

We have some minuses but also we have some pluses for this crisis. In my opinion it will give us big advances in the future because before everyone say said, oh, digital is complicated. Doctors were saying I'm too old for digital. Now they are forced to be digital and be more happy with digital.

>> BRIAN SCARPELLI: That's interesting. For the comments you just said, Sveatoslav and also made Analia, too. I'm curious what you see the ‑‑ what maybe the top practical ‑‑ for people utilizing these ‑‑ and also for anyone else on the panel actually. I would be curious to know.

The practical barriers that you see to the uptake in use of the tools, maybe the top ones. Is it capacity? I throw that out. Capacity. Like infrastructure. Like broadband infrastructure. I throw that out and I don't know if that's the answer. Everyone who has robust enough broadband to support a video call with a physician, it doesn't always exist. I was curious if that's a challenge. I have heard also, Jelena I heard about liability which is a very important question. Not only just for what I would say is basic digital health interactions, and I don't mean to downplay the importance of just simply a video call between a patient and a physician, but also I think there are implications for AI obviously when you talk about the use of AI in health care and liability issues.

But across borders as you were mentioning, what do people see as the biggest challenge practically you run into? I would be interested. So many different perspectives here. It would be interesting to hear them.

>> SVEATOSLAV VIZITIU: Here in Romania, I think we are in the top four above the internet over the world. But with these results also we have only 60 person from all the countries that have access to internet. The rest don't have access to internet.

Also we have issues about the schools because for example we have online school online now. And many of them don't have any devices to connect with the schools. The big problem is when weather channel more children you can't connect more children because you only have so many devices to connect on and so on.

With telehealth for example and with our partners in application it works very well. Like I said, it is less appointments and more healthy than before. They have issues with their weight of course because everyone is staying home, but it's advanced now.

>> BRIAN SCARPELLI: Anyone else?

>> GERALYN MILLER: I'll take that from an AI perspective. I want to echo a sentiment I heard earlier about caution from a technology perspective. I think there is a part of this that's about literacy and people understanding in a context that makes sense for them the technology, pros and cons and making sure that also we have the concept of inclusive design so we are not designing for. Rather we are designing with people. Who will be recipients in receiving the technology. This touches on some of the things happening globally and to prevent possible bias in artificial intelligence. Making sure we understand what's in the data that it is a diverse sample. Making sure teams who build these solutions are diverse in terms of their make‑up. That's super important. All of these are things we need to be aware of and keep an eye on.

>> BRIAN SCARPELLI: Absolutely. Speaking about literacy is crucial. It's not just about literacy of everyone. If we speak about doctors especially the older generation of doctors, their views on technology are quite different. Also the way they think about, for instance, if we speak about electronic health record how we see many hospitals, the electronic health records being approached we need an education about privacy rights to doctors, manufacturers and more. They do need to comply with certain regulations but in the health care sector where the data is representative there is a need for education both at the level of health care professional education, it should be taught at the university I think. We cannot expect every doctor to be an IT person and understand how deep learning is working. How to access the data, speak about it with your patients and so on. Same goes for any regular citizens because I think it is time to teach these things in school. About digital rights, consent, data sharing and other things.

>> BRIAN SCARPELLI: I know Analia went into this in some earlier remarks. It would be great to hear more if you would like about how Buenos Aires has dealt with the public health crisis and how your office dealt with the crisis. It sounds to me like you were describing how both live telehealth and patient monitoring. The idea that some physiologic characteristic, blood glucose reading, heart rate, something, could be collected and then transmitted to the caregiver for review in an asynchronous way so it is always feeding data in.

There are health systems in the world and I would love to hear if this is an issue for you or not because whether it is or not, it will ‑‑ I think your experience will be interesting. There are some health systems in the world where payment is enabled for live interaction and not enabled by law for an asynchronous interaction such as a remote patient monitoring scenario. Maybe that's an example of what Jelena spoke to earlier. I would love to hear more about your experiences and some more detail on the practical reflections you might have.

>> ANALIA BAUM: Okay. I will share my screen. Firstly, I would like to tell you that Buenos Aires is the capital and largest city in Argentina. There are around 3 million people. The health system has three sectors ‑‑ private, social security, and public. The public health system has 117 health care centers. 34 are hospitals and the rest are primary care centers.

In the last four years the primary care centers and 20 of the 34 hospitals have implemented electronic records and more than 2.5 million people have already benefitted from this.

We had our first confirmed case of COVID in March. The government declared the lockdown as you can see. Appointments went down. Immediately the minister of health of Buenos Aires city declared a mandatory telehealth services and digital prescriptions to guarantee the continuity of health care. This regulatory framework from the minister given the pandemic were very important. Because this regulation made these practices available. Under the emergency conditions. We have a national law that says the medical prescriptions have to be signed by manuscript and the rules for telehealth services weren't clear until now.

The fact that we have this culture extended the medical records and mandatory social distancing helped with rapid adoption of telemedicine. Currently the public system, 20% of appointments are telehealth services. Most of the people that use these services are young women and residents of the Buenos Aires city.

This map shows that all the areas of the city are covered especially in slums that are in the south of Buenos Aires city.

It is important to mention that in Buenos Aires 8% of the people don't have access to the internet and 20% of them don't have a computer either. However, 90% of people do have a cell phone. The most effective population in this are elderly people. Those with no access to education. These worsen in slums and metropolitan area of Buenos Aires.

To stop the circulation of the virus we developed software for contract tracing which helped us to create heat maps to monitor the most affected areas around the city. And for remote monitoring COVID patients and their close contacts we used a chat bot in WhatsApp that's the most used social network in Argentina. 3.5 million conversations took place up to date.

COVID‑19 has accelerated adoption of digital tools like chat bots and telehealth in Buenos Aires. However, we have to face some challenges. First it is necessary to analyze accessibility of telehealth and solve privacy and security issues. We also need to continue working in the regulatory frameworks in telehealth and digital prescriptions. As regarding social we need to focus on skills not only for patients but also for health care workers and policy makers. Thank you very much.

>> BRIAN SCARPELLI: Thank you for the insights there. Very interesting.

Anyone else want to make a comment or ask Analia a question, please just unmute yourself.

>> JELENA MALININA: I wanted to ask about ‑‑ you mentioned privacy and security issues. What did you encounter and in what tools? Also, separately I want to ask about your Corona tracing app? Did you use Google and Apple notification exposure? Thanks.

>> ANALIA BAUM: Thank you. At first we developed all our applications using guidance standards at the national level published. For security threats we used the center for people to call. It's useful for different transactions in the Buenos Aires city. We use it for patients to call. And people can see the appointments for telehealth and the doctor on the computer can see the list of patients for telehealth consulting. The patients can choose if they want to get with the physician by telephone or by video calling.

Most of the consultants were by telephone because the links to the video, we sent it by e‑mail and we have a lot of problems for patients to access their e‑mail and then for the video calling.

Patient access to the health system in their own cell phone and applications. So for the doctor, if they ‑‑ if the physician didn't know about the patient because it was a first ‑‑ at the other side of the telephone line. So we need to improve the government.

>> BRIAN SCARPELLI: Thank you so much. Well, something that I think got teased out a little bit. And I know there is a number of experts here for the audience with great insights. The role of artificial intelligence and the future of health care.

Geralyn, while there is a lot of talk and admittedly probably a lot of hype about the role of AI in addressing this public health emergency, how is AI being supported and leveraged by you all during this crisis? I know you have mentioned a few things already you mentioned 500 ‑‑ that's significant. That's impressive. We'd love to hear more.

>> GERALYN MILLER: It's hard to decide what to talk about. There are so many different areas. There was a conversation about dissemination of accurate information. I think that's an area where we have seen people building models more on the behavioral side. We have seen it applied actually even to scenarios where people are doing basic research and the nature of the virus itself building models, doing drug discovery, trying to pore through existing databases of compounds and small molecules to find out what might be relevant for the SARS‑CoV‑2 virus.

Also in machine learning, AI to do things like track the trajectory of the disease, we have work that we do with a global map. Also working with the institute of health metrics and evaluation hear in Washington state. So very many different ways. And then from a direct to the in‑person perspective there are things like we have what's called a convalescent plasma health bot where people who have had COVID can find out if they are a candidate to donate plasma. It's a cell phone‑based app.

So many different ways. There is also a framework, I think, that extended ‑‑ that's been created before COVID and will likely extend afterwards. Just looking at how you use artificial intelligence not to replace the physician but to augment a physician. Examples of work we are doing under AI for health, one example is diabetic retinopathy, one of the leading preventable causes of disease. Hundreds of millions of people around the world have it. Only 210 physicians qualify to do the diagnosis.

This is a place where you can use artificial intelligence and models at the edge to be able to make sure the people who really need to get in, get screened, have a referral have access to the technology.

>> BRIAN SCARPELLI: Great. It's really exciting to me that you mentioned the diabetic retinopathy use case. It is very powerful. You're absolutely right about the reach of it. I couldn't agree more. Something maybe worth mentioning and this is a domestic development in the United States right now, but one thing I have been watching with great interest has been a proposal within the Medicare program ‑‑ basically the government‑provided health care program mostly for people 65 years of age and older and people with disabilities, et cetera, to provide payment to physicians with an algorithm to assist in evaluating the diabetic retinopathy data protected. In other words, the picture of the back of someone's eyeball.

Maybe folks out there ‑‑ I would love to know if that's the first of such a proposal or if ‑‑ has that kind of capability been provided, endorsed or incentivized in other health systems? It's the first proposed incentive like a direct payment to physicians for using an AI tool that I know of. Very interesting. Is that the only one? Are there others out there? Gosh, it may be the only one. That's really interesting.

Sveatoslav, could I maybe turn it over to you? I know your company utilizes AI to provide nutritional advice and plans to parents and children and other uses you described earlier. Maybe you could talk a little bit about the benefits that AI brings to the table for you and the approach your company takes, kind of running with this theme.

>> SVEATOSLAV VIZITIU: Yes. Using machine learning and AI, it helps for each family. It helps us to be trained with this system and to provide specific diet for each children. For example, if the child is moving more in the morning, maybe it's moving more in the evening then he needs more proteins in the afternoon. Also maybe he's lazy, staying home. It helps us with so much data from families to provide more information to them. Also some children have some allergies and so on. Machine learning if we are training with more information growing up it will be more easy to have so many nutritional experts to provide more exact information to our families. That's how we are using. Also some parts for Microsoft. Interesting stuff.

>> GERALYN MILLER: Brian, you raise an interesting point because we are in the middle of a pandemic. Needs to bring nutrition to people still exists. It was interesting. I was on a panel yesterday. There was somebody who was representing a bunch of work in maternal and infant mortality. They said just because we are in the middle of a pandemic, don't forget childbirth is still a thing. We didn't lose sight of some of the other initiatives that are just as important to drive forward including the example you gave.

>> BRIAN SCARPELLI: Rao, thank you. Yeah. I wondered, Rao, I could ask you. I want to make sure to bring you in there. Some of your work at as state centers around explainable AI and discussing if and when there needs to be a period where humans need to be in the loop and able to intervene when AI makes a particularly sensitive decision. For example, AI chat bots have gotten a lot of attention. Can you talk about the risks and what you're working on and stuff?

>> SUBBARAO KAMBHAMPATI: Sure. First of all, AI has become popular in catching the public imagination, as we talked about. There's already been this issue of AI can be used for everything sort of mentality. Obviously when COVID came along there was significant interest in seeing all sorts of ways in which the technology can be used. In fact, I put in a link to an article that summarizes some of the different ways AI technologies were being used and were planning to be used in managing the war on COVID.

In general, the issue of AI and health care AI, you brought up the issue of the liability and there is also this ethical issue aspect. The pendulum is somewhere between doctors don't believe this stuff to full fanboys and groupies where they basically assume AI technology is pretty much more than they are able to do. As Geralyn was pointing out, too, there are significant applications out there about diagnostics and chat bots especially being used for mental health and so on.

The tricky part, I think and this is the academic in me speaking. People understand the success more of AI technologies. Those are things we celebrate. We don't have any clue on their failure modes because the technology is not robust in the way we assume other intelligent technologies would be. If you see a surgeon you don't question whether they will know about elementary age store things. You can have a diabetic retinopathy classification system like one more recently, a melanoma classification system which basically is only optimizing just to be correct on predicting that. Sometimes it can be easily taken by spurious correlations.

There was a study last year in Nature where the system found that surgical marks put next to the melanoma were actually ‑‑ it was basically focusing on that. It's like saying a dog is an Husky because there is snow on the ground. We realized they must have snow on the ground. This can be a big issue in terms of even the mental health scenarios.

Some of us say this is the best natural language completion system that's out there. And recently last week there was this news item basically saying some people tried to evaluate it in simple chat bots. And using GPT‑3 as a chat bot is almost using sledgehammer AI technology and even that sledgehammer AI technology has pretty brittle failure modes. For example, the thing that the patient asks, I'm feeling depressed, I'm having suicidal thoughts. This is a fake patient. GPD3 says I'm sorry, I can try to help you with that. He says should I then commit suicide and GPD3 says, sure, why not? This is a level at which the technology to some extent is odd.

It's important as Geralyn is pointing out, extremely important to view these technologies as augmenting rather than replacing in any sense human doctor oversight because we have no idea how brittle they are. We take for granted common sense which the systems don't currently yet have.

They are extremely useful for experts but that gets to the point of explainability. Even experts. I'm actually worried about paying doctors to use diabetic retinopathy thing. It looks good on one level. There is a variation for using north point compass system to decide whether people should be allowed what was used in Broward county, Florida, where judges had to take the system's recommendations. But most of the time they wound up taking the recommendations anyway. There is an auto makes bias we tend to have. After a while people just start blindly following the system's recommendations.

So the issue of explaining the rationale becomes important. Technically, that's a huge area right now. Many of the AI systems don't have vocabulary common with what we have. So they really can't explain. It's like a cat trying to explain what it's doing. That's the level of which we are ‑‑ we can have very smart cats which can do useful things. But the explanation is an interesting issue. Those are very important for miracle scenarios, in particular because just to make sure that the doctors aren't taken by spurious correlations and the past performance to just believe the current diagnostics and so on.

So that's something we need to keep an eye on. This technology will be used because it's huge amounts of advantages. It is really trying not to be just too much of a fanboy and trying to make ‑‑ have a caution in the back of our heads and that's important, I think.

>> BRIAN SCARPELLI: Analia ‑‑ oh, go ahead, Geralyn. I didn't mean to interrupt you.

>> GERALYN MILLER: That's okay. I also want to point out that there is a level of education that needs to make sure the data scientists on the grantee sidewalk away with skills. They have an understanding. There have been cases where we have come into projects and they have had to tell people, you need to really sit and look at what the model is doing. It's awareness and a training thing. We can get there. So those are points that are well taken in terms of the caution.

>> SUBBARAO KAMBHAMPATI: I want to add that it's something that the medical community is saying that much of medicine is about what is right or wrong for white males because that was the data on which basically people focused on mostly and it has happened before. You can put it on steroids now because the kind of data is more directly available. So you wind up training your models on that. Even when COVID came along most of the earliest ‑‑ COVID is an extremely interesting, fast, forwarded history in progress in front of our eyes.

If you go back only to March and see what people were saying about what this thing does versus what we now know ‑‑ we already know now for example ‑‑ we don't know the causality, but we already know for example African‑Americans are disproportionately affected. These were things we wouldn't have ‑‑ it would be great if, in fact, from the beginning as Geralyn was saying you take into account a diverse set of data and then you make sure that you actually train your models and test them. Something as simple as checking gender of a person from their face it's been shown last year most commercial grade systems just fall flat on their face if you go out of white males.

Beyond white males the gender becomes close to 50/50 toss‑up. That was a great call. Then people like Microsoft or companies like Microsoft, IBM, Amazon, they actually improved training and now we are in a much better situation. It is not at that level of problems in terms of the ability to do it across the populations. That's a day‑to‑day problem in health in particular. Many things are common in humanity, but also enough things that are different from the different racial backgrounds and so on. It's very important to make sure these things will be checking that in the training part.

>> BRIAN SCARPELLI: Thank you. Analia, I was going to ask, a chat bot is part of what ‑‑ that's part of your system, correct? Am I correct about that? I think you're muted.

>> ANALIA BAUM: The chat bot is a service of WhatsApp to the government of the city.

>> BRIAN SCARPELLI: Okay.

>> ANALIA BAUM: It was one of the first cities to prove that. We use it for people ask about different issues in the government. You know? And we introduce them during the pandemic and at the beginning of the pandemic, every people who work from Europe because from here, people after the holidays came back to the country and our telephone center collapsed. So we introduced the chat bot to people to make questions. Then if those questions were similar with COVID symptoms medical attends this chart and decides this person needs to be isolated in hospital or in a hotel at the beginning of the pandemic.

Then we introduce more algorithms and use for people that we know for our internal record that the person has COVID. We send the text by telephone and invite them to interact with the chat bot to monitor the symptoms. Something similar in their record. All those informations, we send it to the contact tracing software.

>> BRIAN SCARPELLI: Thank you, great. Appreciate that. So I had a topic just to bring it back to something you had initially raised was the worry about additional threats when services that would be transacted in person are now over the internet. I was curious about some best practices that you might know of or recommend or maybe where there is a gap. Maybe that's part of where we are identifying a need for further work or all of the above. But best practices that could mitigate some of these concerns and how would you recommend that policymakers either with a government policy maker or even as an organization. Could it be like a large company, for example? Ensure that patients have privacy and security assurances as health care is digitized.

>> JELENA MALININA: Thank you. There is certainly a lot of work to be done. I think one of the greatest solutions and one of the most complicated as well is to have legally enforceable rules. We already have in Europe the general data protection regulation which some claim that it's a little bit preventative of innovation. At some point it's true. I would say preventative of innovation which doesn't serve for people. One of the things that must be taken into account is what's done with the data? For now many people do not fully realize what data means. The data we produce that it has value, a lot of value for certain entities and certain entities sometimes acquire this data in quite illegal ways. It can have implications personally and collectively. The personal implications for instance could be that especially for famous people or politicians that it can be personal bullying or cyberthreats. We have seen it recently in Finland. There was a massive leak of the data of patients from a psychotherapy center. The person who initiated the cyber-attack started to blackmail patients because you can imagine what kind of personal data is there.

He started to threaten them that he will leak certain aspects about their private life and so on. So it's a huge gap as well. Not only having the rules but enforcing them. With respect to IT security rules by design. It's not there yet unfortunately in many developments in many countries as well. Also very developed digitally as Finland we have seen that.

Collective threat can be discrimination. There was already a mention for instance of the U.S. example of a black community that's discriminated and there are examples especially from AI in hospitals where the algorithm is programmed in a way that it is discriminated by design. It can happen anywhere else.

Depends how we program the algorithms.

There is a huge need to establish the borders, how much we would allow the technology to come into our lives.

What are the new ethics? If we speak about the doctors having the Hippocratic oath and certain norms and values which are established in practice and they are very long term evolving with the societal needs as well then we have to speak about digital technology having the same norms and having also enforceable norms because just saying somewhere this is ethical or not is not enough.

For now there are many ethical codes especially about AI produced by different entities even at the country level. But all of them are different. If we speak about a hundred ethics codes we speak about a hundred different ethics.

So how to come up with universal ethical norms, how we approach technology and how much we allow it in our lives is a huge question.

>> BRIAN SCARPELLI: Geralyn, I believe AI for Good has recommendations, correct?

>> GERALYN MILLER: Yeah, all from a Microsoft perspective we have a framework for security and privacy from an infrastructure perspective when people put data in the cloud. As a software and technology person, I was leaving the Microsoft genomics team when we were going through the prep for GDPR. It was all hands on deck for Microsoft. Every single product had to be touched and looked at. I think it made for better products. It forces you to take a new lens to what you build and really look at every place where data is, every data flow. Not only in the data that's coming in that might be somebody's PHI but where could the data leak out in logs or crash dumps. It makes for better products.

I think it is the flip side. We are better because of that. Then I also want to quickly touch on privacy preserving technologies. There is innovation in this area and we need to continue to push that forward in the U.S. We have HIPAA and high trust. Some of the things that are HIPAA and high trust are almost like hitting a sledge hammer with a nail. It is a pretty macro level way to look at this. We are doing investment work on privacy preserving technology called differential privacy. This is work we are doing with Harvard IQSS. It goes back about 13 years to some work done by Cynthia Davorka and we are doing work beyond research and the Azure team is using it to add statistical noise to data

There are people looking at things like synthetic data generation using GANS so a bunch of techniques we need to push forward and invest in research in these areas as well.

>> BRIAN SCARPELLI: Thanks. That's very helpful. Sveatoslav, I'm curious to ask if kind of from the start‑up angle how you all have tackled privacy and security? Just to throw it out there, something that I know a lot of people talk about, it's easier to talk about than to do. Is privacy by design, the idea of building in some ‑‑ measures to enable for some kind of scalable risk management, you know, not even just for privacy and X by design, right?

The real world perspective. I would love to hear your thoughts.

>> SVEATOSLAV VIZITIU: For example in Europe we have privacy because we have GDPR and each application we are doing or for example the applications we are forced to apply this.

It is very interesting that we know that global rules for example about if children, any kid wants to create an account they have to be 30 years old. But we know children do and we don't accept this one to be public and to create.

For example, we have cases in summer school where we tested some interesting stuff. Everyone had Facebook. Everyone has Google account. And everyone is less than 30 years old.

It is important I'm pretty sure. But from the perspective about for example start‑ups, so a company that's early stages, in my opinion it's important and also an innovation. If some user wants to try it and wants to be involved here, I think we need to open these doors. But many rules like Apple store, Google store, they force this one. It's really difficult to innovate in this direction. That's my personal opinion.

And we have cases about the children when they create it because the application is divided into two parts ‑‑ one for the parents and one for the kids. Kids want to create the account but we can't create for them the account because the application will be banned in the app store and Google store. It's simple.

>> BRIAN SCARPELLI: Thanks. Okay. It doesn't look like anyone else is coming in here. I don't want to cut off anyone of course. There was one other thread that I think might be worth picking up on here that I know has been raised by several people which is it can't be overstated enough ‑‑ am I saying that right? It's a very important angle is what I'm trying to say.

Inclusivity, accessibility. I'm really glad a number of you raised this so far. You know. What do you think ‑‑ and really this speaks generally to, think, a shared priority across the idea of community about bridging the digital divide, if that's the right phrasing there, I believe it is. What do you all think ‑‑ anyone ‑‑ who is the best way to bridge the digital divide that threatens to leave some folks out of the positive advances being made in digital health and some of the fast forwarding is where I was describing it which I do think is a pretty apt way of putting it.

>> JELENA MALININA: I think it is very important to forget about the existence of the average consumer and to start thinking about society as a whole and about all the groups which are in this society if we speak for instance about the technology developed for a particular country or region or maybe worldwide.

What we will see if we look in the society, there are many people with very different needs and very different desires. Some are more open to use telemedicines. Others will be more willing to have their medical devices fully digitalized connected to their doctors. Other people would want to go to the doctor because that's part of their social life.

I'm coming from Eastern Europe and for my grandparents it was part of their socializing to go to their GP and chat. That also must be respected. Then we should not forget there are vulnerable groups which may not have access to the right technology. Again, I'm getting back to COVID apps. For instance in Belgium it is not possible to download an app on an iPhone which is three years old which is a little bit ridiculous. Not everyone is supposed to have the latest model of iPhone to use the tool.

So that's one part. Another part we should never forget about people with some disabilities. If we think of people with eyesight, hearing problems or anything else which would prevent them from using the technology.

So to really tackle that inclusivity we need to think about the variety of people and the variety of their needs.

Of course simultaneously we have to raise the level of digital literacy and I want to emphasize that it is very important simultaneously to use health literacy. Not only digital but in the context of health care, many people still don't know by the statistics where the heart is located on the right or left. How can we speak about understanding algorithm if we still don't know where our heart is. Many things to tackle. But we need to as a whole the society as a whole should be considered.

 

>> BRIAN SCARPELLI: Thank you. Anyone else? Analia?

>> ANALIA BAUM: Yesterday's panelist said inclusion is the concept where everyone can be included. This implies access to the internet, access to digital technology and also the possibility to use it in a way that it's empowering. It is worth saying it is ‑‑ to low income families. We need to find the right partner that helps us bring expensive networks into slums to promote capacity building inside the community.

Meanwhile, the government policy consists in implementing community networks in public schools, libraries and working in digital location for all the education program.

>> GERALYN MILLER: I would love to talk about vulnerable populations. We need to do a better job of accurately reporting out on data in some of these areas. Some of the work our team has done in the U.S. has noticed that when we report COVID positive status sometimes race data isn't included.

How do you fix a problem if you don't understand the problem? It is difficult to try to target interventions appropriately. It's a resource allocation problem, right? There is a limited pool of resources we can allocate. We need to know where the problems are. Some of it comes from being open and honest about this from a data perspective.

So I just want to raise that comment as well. On the accessibility front in addressing people who have accessibility needs I really think this again goes back to the concept of designing with people an inclusive design.

>> BRIAN SCARPELLI: Absolutely. I couldn't agree more with what you just said there.

One thing I don't think I had ‑‑ I definitely haven't asked about myself yet, but is a really important point you just mentioned I think was data interoperability. How much of a challenge is that there are ‑‑ like HL7 is an organization trying to do this to create specifications for ‑‑ fields, the fields that you would include so that race data for example is always included as information is shared across health systems and things like that.

I'm curious, Sveatoslav, Analia, Jelena, anyone. Health care interoperability. Has that presented challenges to you in practice? It's one that's talked about quite often. You know, what do you recommend how it be dealt with?

>> JELENA MALININA: It is a challenge. I can give you a positive example of what was done in the EU because many countries are now introducing electronic health records and because we move a lot from one EU country to another, it would be useful to have, for instance, my own record when I move somewhere or even for holidays if I need the medical care, for instance.

They are introducing a system allowing interoperability from different systems of the country. They needed to agree on how they encrypt the data, which standards they use and actually HL7 is the standard used for this infrastructure in the EU.

Of course it is only one fraction of data when we speak about electronic health records. There will be others. How do we agree on interoperability and the standards used for the others? Who knows.

Another problem in the area of less structured data. So there is using the real world data as a complementary source for any kind of decision making and also it is popular to speak about it in terms of new medicine developments. There will be new initiatives in Europe to apply the data as a complementary source for clinical trials. Then the big challenge is how do we ensure that this data is of good quality. If someone picks up a comment on Twitter about their reaction to a medicine, I turned red like a lobster. What does it actually mean? Is it good quality? Complementary source? Did the patient eat something else which caused the redness on the face? There are many questions to answer and how do we approach it.

It's never easy.

At the EU, once again there are many health care systems and national rules on how it is approached, especially in health research. Rao, I see you're unmuted. I thought you were trying to talk. No worries at all

Anyone else? I'm curious about the data interoperability.

>> GERALYN MILLER: Interoperability is a challenge. It's always been a challenge. I think there is great work happening in the HL7 community to enable that. There is a subtly though. The challenge isn't always technology. Sometimes it's policy. By policy people may not report out certain sets of information. Policy is a little harder thing. Understanding where people are in terms of their state ‑‑ like lockdown or not. Where are they in the spectrum? That's information that's hard to gather. There is no consistent way to capture that and represent it. Most people don't report out on it. It is super important information if you want to track, for example, progression of disease and understand what has been most efficient at curbing the spread. So it's not all technology. There is a policy component to it.

>> ANALIA BAUM: I'm sorry, it is in Spanish of course because at the national level. They have been working with a lot of interoperability standards for problems. We are a federal country, so each province can decide to adapt their system to this interoperability standard.

In Buenos Aires City we did it. Our electronic records accomplished with these objectives.

Technically we can exchange information between the public sector or another that accomplished the same objectives

So we need to define the politics to do that because at the national level there is an application that they can put if they want to share or not that they have information between different province or different sectors ‑‑ health sectors. But the problem is that each government decides to share that information between different provinces. We understand that it's useful for better quality of health but generally they are afraid they use it for political issues. So we are discussing if the formation ‑‑ the patient is the owner of the information we shall know that it's changed. But it's not happening yet.

>> BRIAN SCARPELLI: Well said. Sveatoslav, I saw you unmuted earlier. Did you have anything? I don't want to skip over you.

>> SVEATOSLAV VIZITIU: I agree with all that they said. I think the main point is the policy and rules of each country.

For example, in Europe it is a really big advance but Europe has so many countries but it is from the same umbrella.

At the same time each country has rules about policies and so on. When you want to extend to a new market you have to start with the policies. Only after you can go with technology and go there about the health. Yes, I totally agree that it's very important, the policy to be united or simplified about innovation.

>> BRIAN SCARPELLI: Well, I know that ‑‑ I had maybe one more ‑‑ just looking. I do not see any audience questions yet here. We are closing on the end of our session here. There was one I can kind of ask a two‑prong question that you can answer both questions if you would like or just one of the two maybe.

But, you know, one is, you know, is the IGF as an org has told us they are interested in what voluntary commitment or commitments us as the session participants are undertaking or will undertake during the next year to address some of the challenges that we have raised.

So that's one. The other maybe part of this is, you know, if there was a way ‑‑ one policy that you could change that would responsibly advance digital health uptake and improve outcomes for patients what would it be? Maybe these are linked. These were two questions I wanted to make sure and try to fit them in here.

So, please, anyone, go ahead.

>> ANALIA BAUM: Okay. Promoting trust and security in digital environment because it was the most important issues discussed during the telehealth services. So we need to work on that. We are beginning on biometrics and personal ‑‑ the public health sector for better security and privacy for patients and health care workers.

>> GERALYN MILLER: I want to echo that. I think that's the overarching umbrella. When you look at AI you can't be successful without trust and security and privacy. That would be our commitment. If I was queen for a day I would hope that we can bring broadband to all, frankly, to enable people who may be marginalized and realize what we learn in the pandemic, that access ‑‑ internet, broadband access is a social determinant of health and affects the way we work, the way our children learn. It affects our health care. It affects the way we play. We are not out of the weeds yet from a vaccine standpoint. Though there are vaccines and phase three trials it will be a long period of transition before we get back to normal. A lot of that depends on equal access.

>> SVEATOSLAV VIZITIU: I agree about the umbrella and I hope the next step when COVID will end that everyone will start to invest in health care. Because we see that each country is not really about the health care. We need to be more in the health care. This is important steps in my opinion after the COVID crisis.

>> SUBBARAO KAMBHAMPATI: I tend to be a realist. I'm obviously not on the ground like many of my copanelists are. The great Pfizer COVID vaccine requires minus 94 degrees Fahrenheit maintenance. Tell me exactly where outside of the richest places and richest counties can that be used.

Having said that, I think digital connected health, I see it as a great positive force in reducing inequities. It would be easier to transport it. In fact, one of the interesting facts is Google did a study in Thailand to see whether or not the super accurate system they developed, how is it working on the ground. They learned a lot of interesting lessons, not the obvious ones in particular that do have broadband. But they actually wound up saying the real use on the ground actually changes the kind of accurate images they might be expecting and they are to go back to.

I see it as a huge positive thing. I didn't really see that in the old days without connected health or digital health. You would be sort of ‑‑ I mean private companies don't have that much interest in making sure what they are doing would be helping the countries that don't have the sort of resources.

One nice thing is the cell phone penetration is extremely high, even in some of the poorest countries in the world. There is a certain level of health care support that will be brought with respect to this sort of connected health because of those kinds of things. There is an interesting article I read somewhere about this guy sitting in the U.S. is helping folks in Nigeria or somewhere where they basically used the cell phone very bad pictures but that's still enough to give them useful diagnostic health they never had before.

So I think of this overall we talked about the worries we have. But it's also a question of I think this is probably helping very much the sustainable development in the sense that it actually brings what is happening in the U.S. and countries much faster to be of use to countries which don't have that kind of resources. So, yeah. That part of me is actually less cynical on the connected health side.

 

Contact Information

United Nations
Secretariat of the Internet Governance Forum (IGF)

Villa Le Bocage
Palais des Nations,
CH-1211 Geneva 10
Switzerland

igf [at] un [dot] org
+41 (0) 229 173 411